Hermes-Lite 2 Update

[Steve Haynal]

Hi List,

I've been working on the HL2 layout. Attached is a picture. I am laying out islands, and the Ethernet, FPGA, Clock and front IO islands are about done. Of course there is inter-island routing and some placement tweaking to do, but I feel that these islands can be routed. There are notes on the picture to convey what different blocks are. I am working on the power supply island currently. After that I will work on the AD9866 and OPA2677 islands where I plan to follow the existing layout as much as possible. I am working out to the 8cm or 10cm line and will make some final decisions on what PA and filters to include then. Here are some ideas I am currently thinking about as the layout starts to take shape:

• It may be possible to have a clean cut at 10cm out that does not include any filters or PA. This would result in a 10x10 cm board appealing to those who don't want filters and PA. Also the filters and PA are the blocks I am most unsure of and a partition here would allow those to be changed while keeping the core of the design, which I am more sure of, stable. I still want to have a 10x15cm board and feel this is possible with a 10x10cm main board and 5x10cm extension with PA and some filters. These 2 board could still fit in the current target enclosure by abutting within the enclosure. The number of signals crossing this abutment is small (<20) and could be hand wired. This also allows for larger than 5x10cm "companion" PA/filter cards, such as a stacked 10x10cm configuration. This is going back to some of the original thinking, but I want to finish more layout to get a better sense of what might work easily.
• There is not much nice rectangular space for relay-switched filters. John's layout gives me a pretty good sense of what can work here, and it appears unlikely that 7 switched filters will fit. Even 4 could be a challenge. I am considering including only 1 good ~32 MHz LPF on a 5x10cm extension. Remember that a long term goal is to use predistortion to eliminate the need for filters. In the end, this is the configuration I want. If a partition of two boards exists, full filter/PA companion boards can be built and used in the interim. Also, any single filter I include will be exchangeable via a small daughter board.
• If the dimensions of a standalone unit may be 10x10cm, there are other cases with slightly less internal width than 10cm that I'd like to be able to use. I am considering designing for a narrower width, but including tabs on one edge that can be broken off. With the tabs, the width is 10cm. With the tabs broken off, the width is ~9.5cm, or whatever the exact width of the case I have ordered is.   
  

Any feedback is appreciated.

73,

Steve

KF7O

[Alan Hopper]

Steve,

I'm not keen on being limited to 4 filters, if a bigger board is needed for 7 filters then that would be my vote.  I see the attraction in splitting the board, not least in that there is already so much new stuff to get right without the pa.  If this reduces your workload then it is a good thing.

Only time (and a lot more work) will tell if predistortion can eliminate some filters but having a layout that is easily reused is great.  Something to be considered for both pure signal and harmonic cancellation is keeping rx/tx crosstalk low before the pa.

In looking at the 1.41 front end I noticed there is a lot of opportunity to improve the heat transfer from the op amp.

Thanks again for all the hardwork.

73 Alan 2E0NNB

[Steve Haynal]

Hi Alan,

How many filters (or bands) do you consider to be a minimum requirement? I appear to be out on my own regarding 4 or fewer filters. I usually can only transmit on 4 or less bands due to antenna restrictions, and feel this is acceptable provided I can easily swap one or two filters out when I want to operate on another set of bands. This is more than the softrock RXTX ensemble can do.

Some options for increasing the number of bands in a basic unit are:

• PIN diode switching. The area savings are nice. I'm still not sold given the need for higher voltage, introduced IMD, and issues with using TX filters for RX.
• Daughter boards. The flexibility of these solutions are nice. The cons of this are added assembly and wiring complexity and cost, as well as potential grounding/coupling issues. 
• Smaller relays. If we use some smaller but more expensive relays, we may be able to fit 6 LPFs which should cover all amateur radio bands 160-10M. I am doing some cost and area trade-off analysis versus a daughter board solution. Another advantage is we could consolidate more parts from common supplier like Digikey or Mouser. 
• Use SKY13374-397LF for switching. The advantages are ease of manufacturing, ease of use, low cost, low parts count. The disadvantages are hard to hand assemble, performance is untested at HF frequencies, and power must be limited to 5-6W. This is really the direction my heart would like to go as it is a bit different from what you traditionally see. I'm not that interested in just regurgitating what others have been doing for decades and enjoy trying something with a bit of risk. This is a CMOS switch. I may purchase or build an evaluation board to try it out. There are already so many small surface mount components in the HL2 that I think few will find it enjoyable build one entirely by hand. Most will want some preassembly.  

73,

Steve

KF7O

[in3otd]

Hello Steve,
manually swapping the TX filters does not seem an acceptable solution to me, with a modern radio one expects to be able to transmit on any band just by selecting the TX frequency - regardless if a proper antenna is available or not, hi.

PIN diodes may still be an option; I have some BAR64 which I'd like to try in a parallel of 2 or 4, maybe they will be enough for 5 W (and as you know, we can get 2 diodes in a single SOT23 package).
I've also looked at the available relays, have a couple here to try, but as you know their cost is relatively high.
With my latest Mouser order I have included also a couple of SKY13374-397LF, I may be able to find some time to test them next week.

73 de Claudio, IN3OTD / DK1CG 

[Alan Hopper]

Steve,

I'm with Claudio on this. I bought a softrock rx enssenble rather than the rxtx just because of the filter compromise and then picked up a Flex 1500 as it was the softrock that I really wanted.  My feeling  to the number of filters question is all or nothing,  I liked the idea of an all in one board as a better flex 1500 or cheaper Anan 10, if all bands filters don't  fit in the size you want I'd vote for splitting it or making it bigger over a compromise.  Walking down to the shed to change filters is not very SDR :)

I can see that just a 5w pa and a 30MHz filter on the main board with optional filter board might be an attractive option if it allowed direct connection to a hardrock  or similar amp as you would not be paying for surplus filter components and board space if higher power was wanted.

73 Alan 2E0NNB

[Graeme Jury]

Hello Steve,

Afraid I support the all bands camp. With a Hexbeam and 80/40/30 wire antenna I like to be able to use any band and work the warc bands under heavy contest conditions. I operate remote from my hardware and trotting out to change filters is not a good option for me. Regarding the filter arrangements I think you have laid out the pros and cons pretty well but I lean to Peregrine switches on the Rx filters and PIN on the Tx. Claudio's suggestion of parallel low power low cost genuine PIN's looks like a great idea and certainly should be investigated. There is no particular need to involve the Tx filters on Rx unless it is desired to have selectable wideband rception for skimming etc. At present none of the radio software support being abe to select the Hi pass and Lo pass filters on the fly but it would be easy to add particularly using the command window you referred to on an earlier post. I would really like to see the HL as a basic unit coming out as the present 1.42 version does on one board and the 5 watt PA and filters come out on another board with all bands. Whatever size is needed to accomodate this is the final size of the unit. For my personal part I would just want the HL board and would roll my own PA and filters as there seems to be too much variance of thought on how they should be.

73, Graeme ZL2APV

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Sid Boyce]

The 2 board solution is my preference as it still makes for a small unit.
5W or 10W but with a preference for 10W.
73 ... Sid.

On 10/09/16 22:47, Graeme Jury wrote:
>
> Hello Steve,
>
> Afraid I support the all bands camp. With a Hexbeam and 80/40/30 wire
> antenna I like to be able to use any band and work the warc bands
> under heavy contest conditions. I operate remote from my hardware and
> trotting out to change filters is not a good option for me. Regarding
> the filter arrangements I think you have laid out the pros and cons
> pretty well but I lean to Peregrine switches on the Rx filters and PIN
> on the Tx. Claudio's suggestion of parallel low power low cost genuine
> PIN's looks like a great idea and certainly should be investigated.
> There is no particular need to involve the Tx filters on Rx unless it
> is desired to have selectable wideband rception for skimming etc. At
> present none of the radio software support being abe to select the Hi
> pass and Lo pass filters on the fly but it would be easy to add
> particularly using the command window you referred to on an earlier
> post. I would really like to see the HL as a basic unit coming out as
> the present 1.42 version does on one board and the 5 watt PA and
> filters come out on another board with all bands. Whatever size is
> needed to accomodate this is the final size of the unit. For my
> personal part I would just want the HL board and would roll my own PA
> and filters as there seems to be too much variance of thought on how
> they should be.
>
> 73, Graeme ZL2APV
>
>
> On 11/09/2016 8:31 AM, "'Alan Hopper' via Hermes-Lite"

> <herme...@googlegroups.com <mailto:herme...@googlegroups.com>>

> wrote:
>
> Steve,
> I'm with Claudio on this. I bought a softrock rx enssenble rather
> than the rxtx just because of the filter compromise and then
> picked up a Flex 1500 as it was the softrock that I really

> wanted. My feeling  to the number of filters question is all or
> nothing, Â I liked the idea of an all in one board as a better
> flex 1500 or cheaper Anan 10, if all bands filters don't  fit in

> the size you want I'd vote for splitting it or making it bigger

> over a compromise. Walking down to the shed to change filters is

> not very SDR :)
>
> I can see that just a 5w pa and a 30MHz filter on the main board
> with optional filter board might be an attractive option if it

> allowed direct connection to a hardrock  or similar amp as you

> would not be paying for surplus filter components and board space
> if higher power was wanted.
>
> 73 Alan 2E0NNB
>
>
>
> On Saturday, September 10, 2016 at 5:13:27 PM UTC+1, Steve Haynal
> wrote:
>
> Hi Alan,
>
> How many filters (or bands) do you consider to be a minimum
> requirement? I appear to be out on my own regarding 4 or fewer
> filters. I usually can only transmit on 4 or less bands due to
> antenna restrictions, and feel this is acceptable provided I
> can easily swap one or two filters out when I want to operate
> on another set of bands. This is more than the softrock RXTX
> ensemble can do.
>
> Some options for increasing the number of bands in a basic
> unit are:
>

> * PIN diode switching. The area savings are nice. I'm still

> not sold given the need for higher voltage, introduced
> IMD, and issues with using TX filters for RX.

> * Daughter boards. The flexibility of these solutions are

> nice. The cons of this are added assembly and wiring
> complexity and cost, as well as potential

> grounding/coupling issues.Â
> * Smaller relays. If we use some smaller but more expensive

> relays, we may be able to fit 6 LPFs which should cover
> all amateur radio bands 160-10M. I am doing some cost and
> area trade-off analysis versus a daughter board solution.
> Another advantage is we could consolidate more parts from

> common supplier like Digikey or Mouser.Â
> * Use SKY13374-397LF for switching. The advantages are ease

> of manufacturing, ease of use, low cost, low parts count.
> The disadvantages are hard to hand assemble, performance
> is untested at HF frequencies, and power must be limited
> to 5-6W. This is really the direction my heart would like
> to go as it is a bit different from what you traditionally
> see. I'm not that interested in just regurgitating what
> others have been doing for decades and enjoy trying
> something with a bit of risk. This is a CMOS switch. I may
> purchase or build an evaluation board to try it out. There
> are already so many small surface mount components in the
> HL2 that I think few will find it enjoyable build one

> entirely by hand. Most will want some preassembly. Â

>
> 73,
>
> Steve
> KF7O
>
>
>
> On Sunday, September 4, 2016 at 10:50:50 PM UTC-7, Alan Hopper
> wrote:
>
> Steve,
> I'm not keen on being limited to 4 filters, if a bigger
> board is needed for 7 filters then that would be my

> vote. I see the attraction in splitting the board, not

> least in that there is already so much new stuff to get

> right without the pa. If this reduces your workload then

> it is a good thing.
>
> Only time (and a lot more work) will tell if predistortion
> can eliminate some filters but having a layout that is

> easily reused is great. Something to be considered for

> both pure signal and harmonic cancellation is keeping
> rx/tx crosstalk low before the pa.
>
> In looking at the 1.41 front end I noticed there is a lot
> of opportunity to improve the heat transfer from the op amp.
>
> Thanks again for all the hardwork.
> 73 Alan 2E0NNB
>
> On Sunday, September 4, 2016 at 6:42:21 AM UTC+1, Steve
> Haynal wrote:
>
> Hi List,
>
> I've been working on the HL2 layout. Attached is a
> picture. I am laying out islands, and the Ethernet,
> FPGA, Clock and front IO islands are about done. Of
> course there is inter-island routing and some
> placement tweaking to do, but I feel that these
> islands can be routed. There are notes on the picture
> to convey what different blocks are. I am working on
> the power supply island currently. After that I will
> work on the AD9866 and OPA2677 islands where I plan to
> follow the existing layout as much as possible. I am
> working out to the 8cm or 10cm line and will make some
> final decisions on what PA and filters to include
> then. Here are some ideas I am currently thinking
> about as the layout starts to take shape:
>

> * It may be possible to have a clean cut at 10cm out

> that does not include any filters or PA. This
> would result in a 10x10 cm board appealing to
> those who don't want filters and PA. Also the
> filters and PA are the blocks I am most unsure of
> and a partition here would allow those to be
> changed while keeping the core of the design,
> which I am more sure of, stable. I still want to
> have a 10x15cm board and feel this is possible
> with a 10x10cm main board and 5x10cm extension
> with PA and some filters. These 2 board could
> still fit in the current target enclosure by
> abutting within the enclosure. The number of
> signals crossing this abutment is small (<20) and
> could be hand wired. This also allows for larger
> than 5x10cm "companion" PA/filter cards, such as a
> stacked 10x10cm configuration. This is going back
> to some of the original thinking, but I want to
> finish more layout to get a better sense of what
> might work easily.

> * There is not much nice rectangular space for

> relay-switched filters. John's layout gives me a
> pretty good sense of what can work here, and it
> appears unlikely that 7 switched filters will fit.
> Even 4 could be a challenge. I am considering
> including only 1 good ~32 MHz LPF on a 5x10cm
> extension. Remember that a long term goal is to
> use predistortion to eliminate the need for
> filters. In the end, this is the configuration I
> want. If a partition of two boards exists, full
> filter/PA companion boards can be built and used
> in the interim. Also, any single filter I include
> will be exchangeable via a small daughter board.

> * If the dimensions of a standalone unit may be

> 10x10cm, there are other cases with slightly less
> internal width than 10cm that I'd like to be able
> to use. I am considering designing for a narrower
> width, but including tabs on one edge that can be
> broken off. With the tabs, the width is 10cm. With
> the tabs broken off, the width is ~9.5cm, or
> whatever the exact width of the case I have

> ordered is. Â Â

>
>
> Any feedback is appreciated.
>
> 73,
>
> Steve
> KF7O
>

> <https://lh3.googleusercontent.com/-6SMxCPui0P4/V8u0F5HVxwI/AAAAAAAAAlw/4eCBtQLuQdMePCYsuCMcr2IcowARPC_tgCLcB/s1600/hl2.png>

>
>
>
>
> --
> You received this message because you are subscribed to the Google
> Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it,

> send an email to hermes-lite...@googlegroups.com
> <mailto:hermes-lite...@googlegroups.com>.

> For more options, visit https://groups.google.com/d/optout

> <https://groups.google.com/d/optout>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send

> an email to hermes-lite...@googlegroups.com
> <mailto:hermes-lite...@googlegroups.com>.

> For more options, visit https://groups.google.com/d/optout.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[John Williams]

All,

I am still willing to layout a push-pull variant of my 5W board that
covers 160M to 10M for TX, as well as the next revision of the companion
RX board that uses the same switching arrangement. I would use the
amplifier circuit that we all think is best based on Claudio's excellent
research. I am not fixated on the RD16HHF1 mosfet (although I think it a
very good, robust device). The current 5W board is very stable and Glenn
and Graeme have done stellar work to sort out the issues on the next
pass of the RX board if we go there. To refresh everyone, the RX board
uses peregrine switches on 7 banks of BPF filters based on the MOBO
filter designs. The TX board currently has a single mosfet class A for
5W and 7 bands of LPFs. Graeme and Glenn have created filter pairs that
can allow me to go to 6 banks of filters for the same coverage so that
is also an option.

The biggest issue in my mind is convenient interconnect between the 3
boards. Perhaps I could even design a combined RX/TX/PA board to make
the overall radio a 2 board design.

Keep the ideas flowing...

John

[Heikki Ahola]

Steve,

I am in favor of using daughterboards to provide more space for filters etc.  Increasing the size of the main board is also quite acceptable to me, as there will be little space restrictions in my final setup. Using smaller (more expensive, more unreliable ???) relays for less space is not a good idea ! I strongly suggest that KISS technology is applied in this project. If this implies to larger boards, mechanical switches, relays etc., so be it.

To answer your question,  two or three bands is what I normally need (80, 40, 20 m). 

73 de Heikki (OH2LZI)

[James Ahlstrom]

Hello Steve,

IMHO filters for all ten bands from 160 to 10 meters is a feature everyone will want, but that almost no one will use, especially on the tail end of our sunspot cycle.  It is tedious and boring to make all the filters, and I question whether anyone will really populate all the filters even if the traces exist.  At my shack I put all the filters in two separate boxes and have used the same filters for several SDRs.  As Alan said, the low pass filters would not even be used except barefoot at 5 watts.  So for a single board design, three or four optional filters would be fine with me.

Sid suggested a two board design.  A single board with three filters is fine with me as I said, but a two board design has advantages.  The filters and power amp are a moving target.  We could make a HL2beta base board with everything but finals and a short length.  Then we have a way to test and work on the RTL now, and we have a real signal source for the finals.  A second board that fits the remaining space in the box and has three filters and finals is slid into the same slot as the base board and is connected by wire jumpers.  A second board with a third board connected by standoffs provides space for more filters, and is slid in and connected the same way.  When the design is final, we cut and paste the traces from the second board to HL2beta and make the full size final HL2v1 with a length equal to the base plus second board length.

I am back.  My wife and I went to a wedding in Ireland, a trip to Tbilisi, a birding trip to Republic of Georgia, and a five day trip to Barcelona to look at Gaudi architecture.  A long trip for us and great fun.  Nothing scheduled now until skiing.

Jim

N2ADR

[Graeme Jury]

Hello Jim,

I have also just got back from a weeks trout fishing in the Tongariro river, I think you will have enjoyed your trip also, as it seems a very spectacular program of events for you and your wife. I am very interested in your setup which is nearly the same as mine. The Tx LP filters in one box and Rx BP filters in another with a small decoder board to select the filters. I am considering rebuilding the Rx filters to be HP filters and including the appropriate Tx filter in the Rx chain to form a bandpass. The extra step I want to take is to be able to select the Tx and Rx filters individually or perhaps preselect them as a pair so that I can still have broadband receive for tasks like skimming. I wonder if there is any possibility that you may be considering the same thing and have plans for Quisk or if it will be necessary to write a widget (which would be a steep learning curve for me) but I would be prepared to give it a go.

73, Graeme zl2apv

[Steve Haynal]

Hi Claudio,

Thanks for investigating the BAR64 and the SKY13374-397LF. I look forward to learning what you find.

73,

Steve

KF7O

[Steve Haynal]

Hi John,

Have there been recent developments with the RX filter board? I haven't seen anything recently on the list. Last I heard Alan was actually seeing worse results with WSPR RX when using the RX filter board. It may be that such narrow RX filtering reduces the ambient dither so much that performance is impacted. Also, there are probably some losses through the filters, which can be compensated for some with the LNA. I am leaning towards some HPF filters, such as ~3 filters at 1.8 MHz, 5 MHz, 8 MHz, that work in conjunction with the TX LPF to provide some blocking of strong low frequency signals on active nights if desired.

That is great that you are still interested in doing some PA/RX/TX filter boards. It may be best to wait until the main HL2 is more firm before designing anything. I am working on the HL2 and have finished the AD9866 block and most of the power supply this weekend. I hope it will be firm enough in the next few weeks to have a good idea of what any companion cards should and could look like. Even if I put a PA and some filters on the main board, there will still be the option to not use those. Here are a few points that are starting to firm up in my mind for anyone interested in making a companion card:

• Unlike the v1.42 or other earlier experiments, the HL2 will not have relay drivers (ULN2803, etc.) on the main board. This is very specific to relays, and a companion card may use PIN diode switching or SkyWorks CMOS switches. There will be a I2C interface for a companion card, and the companion card will have to include an 8 or 16 bit bus expander and relay driver if required. A companion card can increase size to 10cmx10cm with typically no additional PCB expense, and this additional area can handle the additional ICs. 
• It is too difficult to plan enclosures and line connections up properly when using rigid interconnect between the two boards. Instead, the HL2 will use standard 0.1 inch spaced male headers that can be wired to a companion card with jumpers similar to these but perhaps shorter. For RF connections, I am also placing vertical SMA connectors that can be wired to a companion card with short coax. I recently bought a bag of 50 of these connectors and a bunch of short coax with connectors on E-Bay or AliExpress for very little.
• The set of possible connections will include something like the following, all of which do not need to be used by a companion card:
• Power in (12-15V), 3.3V, 2.5V (low current for some CMOS RF switches), grounds and possibly 9-10V.
• 2 wire I2C interface
• 4 low speed analog ADC inputs (4 single wires) for forward,reverse, and PA temperature sensing.
• TR switch signal
• PA on/off signal (I need to power down any PA at certain times, and also want to sequence TX/RX cycles)
• ~4 additional 3.3 IO signals available for stepper motor control (I would like to see a automatic Z-match in the future), or additional I2C bus, or some other yet to be determined purpose.
• There is no need to limit yourself to one transistor in the PA design. The test PCB that Claudio is testing is universal enough that it can be used with TO-220 devices and a couple of other packages using thermal adapter boards. See testpa.
• There will be support for adjustable bias via the I2C bus. See the latest HL2 schematics.

  

73,

Steve

KF7O

[Steve Haynal]

Hi Graeme,

Regarding your point about selecting mutliple filters, we have 7 bits that can be user defined and are sent for filter selection. These are currently "one-hot" and select one of seven filters. I would like to split these 7 bits into 2 or 3 words of 2 to 4 bits each. This way we can select more than 7 filters and we can select multiple filters. For example, suppose we divide the 7 bits as follows:

Upper 4 bits used for TX LPF filter selection. 4 bits means we can select 16 options.

Lower 3 bits used for TX HPF filter selection. 3 bits means we can select 8 options. 

And as exists now, this parameter can be defined for both RX and TX. 

There are also some unused option in the protocol that can expand the filter selection. There is also the possibility to use the side channel I described earlier.

73,

Steve

KF7O

[Steve Haynal]

Hi Heikki,

Thanks for the band preferences. The current relays we have specified for HL2 are pretty common and inexpensive ones made in China. I'm not sure if they are more reliable than others I have considered.

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

Glad you are back and had a good trip. I wished I had nothing scheduled now until skiing! I'm scheduled to try and keep customers like ARM and Intel happy during my day job, and then finish the HL2 and keep amateur radio operators happy with HL2 filter options during my free time... 

I've been thinking along the lines you suggested with two boards that can be combined with some simple hand wiring to make a larger final board. See the discussion on the 5 by 10 cm extension board from a week or so ago. I've also considered small filter boards, that fit tabs into slots on the main board vertically. This requires no connector, but just a little fancy PCB design. They would be soldered into place.

There are many good ideas for filtering, and I enjoy the challenge of trying to fit a complete set of filters on the main board, even if that might not end up being the case. If not, there will definitely be good connection points for a full coverage companion card. I am really focused on getting enough of the HL2 layout done so that we have a good estimate of space available for filters before committing to anything.

73,

Steve

KF7O

[Glenn P]

I'm also of the view that full band coverage should be available.   I am sympathetic to the view that many Hams won't use them all but I think it's overall a better idea to have them all available.

I lean also though to using Johns boards, Tx and Rx with expansion though to a push pull amp.  

Then, if the V2 board has the same IDC connector and pinout on it, its easy to add Johns boards to the V2 project with a daisy chain IDC cable. (or any others down the track if so desired.)

That way V2 can concentrate on the digital side of things.

glenn
vk3pe

[Graeme Jury]

Hi Steve,

Yes I agree that already we have sufficient bits to control just about any filter setup. I was thinking the same as you except with 3 bits for Rx and 3 for Tx leaving 1 for unspecified control or perhaps multiplexing the first few lines and using bit 7 as a Tx/Rx switch with the lines going to a latch. I was not particularly concerned about how these lines came from the HL board and a ULN2803 would have been fine as the voltage translation would have been on the filter board. Six filters are enough to cover all ham bands as 30/20, 17/15 and 12/10 can share filters so decoding to 3 of 8 is more than enough.

I am rebuilding my filters to a generic board which will contain the Rx Hi pass, Tx Lo Pass and an Arduino which will do the filter decoding latching and control. It will also have of course the I2C input plus the analog inputs for power and SWR etc. I am intending to have a lookup table to linearize the diodes on my SWR detector and want to set it up as a test bed on both my HiQSDR and the Hermes-Lite. I doubt that the final Hermes-Lite would be using an Arduino but maybe the chip could be on a filter board, at this stage who knows?

My previous post which you have responded to probably should not have been on this list but on the N2ADR Yahoo group as I was really looking at an enhancement to Quisk to allow variable HP/LP combinations. I visualized something like a right click on a band button pulling up a list of the Rx HP filters and allowing you to set the one to use with the current band button. Of course the ones above the band frequency would be grayed out but any of the lower ones would give the variable Rx bandwidth with the protection of not being able to transmit out of band as the Tx LP filter would belong to the band button in the normal way. I am aware that John Melton's designs have separate Tx and Rx filter selection although not quick and easy to change on the fly.

I am still not sure if a HP/LP combination on Rx and LP on Tx vs. Rx bandpass with Tx Lopass is the best way to go and I have decided that the best way is for me to just go ahead and try it and see what I think after using it for a while plus a series of whisper comparisons.

73, Graeme zl2apv

[Alan Hopper]

Steve, John,

I would not read too much into my test, whilst I believe it was reasonably fair it was just one day on one band at one QTH with one filter.  It would be interesting if anyone else with multiple HLs could duplicate the test, I do believe a proper splitter is needed, I tried with just a bnc T and got different results. 

 I did try using wideband fm in duplex mode as out of band dither, this was a highly scientific 5 min test that involved me whistling into the microphone and watching the panafall for changes:) I saw nothing and then got distracted before working out a proper test. This was without a filter as I don't have a setup where I can put the combiner on the radio side of the filter.  The antenna was connected so I might already have had enough dither. I guess the proper way to do this is with Claudio's IMD test setup.

73 Alan 2E0NNB

[John Williams]

Yes, I plan to wait to see more of the new main board prior to making any new efforts. I am quite happy with your direction on the list of differences.

On the RX, Alan does not have one of my RX boards so his measurements must be with some other filter. He has a superband that I built for 60/40/30/20 that has shared BPFs for filtering so perhaps that is the source of the filter measurements.

I have been busy with summer farming and thus note that there are extensive cobwebs on my workbench in the shack. Looking forward to cooler months and more time to tinker. Glenn and Graeme did all of the testing on the RX board. There is a list of circuit changes that Glenn provided that need to be made if we decide to do another pass, as well as documentation updates for the filters that were tested by both Glenn and Graeme.  For now it is on hold.

John

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.

To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.

[Graeme Jury]

Hi John,

Yes summer farming is a busy time particularly if you have crops but dairy and meat is busy too. There were some important points that came from your board John which I just want to keep in mind. The transmit filter was standard format but new ground was broken with the Rx filter so I am summarizing it below.

The Peregrine PE4259 switches work well with no detectable noise in my case and less than 0.5 dB loss in the 2 switches per filter.
The isolation across the board was excellent at around 60 dB.
The isolation between filters was > 60 dB
Strong signals > 2 volts was handled easily by the switches.
The standard E12 range of inductors was close enough to allow a filter design that was very good.
The passband loss of chip inductors was acceptable
Toroids gave an ~1 to 1.5 dB improvement in passband loss.
Toroids wound to formula worked well in filters designed for the same value toroid in each stage.
Pull-up resistors would be an advantage on the control lines as the Rx filter may be used without the Tx filter.
Glenn's filter test results are here

The dreaming, theorizing and designing are of course necessary stages in the production but a proof of concept build with careful measurements and subjective testing tell the final story and the Rx filter design is sound. You did a great job John.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme and Glenn,

What measurable RX advantage have the RX BPFs provided over no BPFs? Did lab measurement noise power ratio (NPR) or DR2/IP2 (2nd -order IMD) performance numbers improve? Are you seeing field advantages, such as better ability to work in environments with many strong signals, ability to increase the LNA gain without clipping and make more WSPR spots, etc.?

One concern around the PE4259 is low frequency operation, <10 MHz. What impact did the PE4259 have at 160M and the <1.8 MHz amateur radio bands once impact from the filters was removed? Although the noise from the PE4259 oscillator may be lost in the noise floor, it is possible to measure it as I did previously. What was the actual noise level, and did you see noise from all switches? 

73,

Steve

KF7O

[Graeme Jury]

Hi Steve,

The biggest measurable advantage for me was the reduction of strong AM broadcast intermod and other products. Unfortunately since retirement I live in a technical dessert and don't have good instrumentation for objective measurement and as stated my results are subjective. Yes I could run the LNA gain higher with Rx filters but can't be sure if the filter loss was a player in this. There is a perception that the general band noise is less and I ran the filters on my HiQSDR which I have run for some years with a 60 MHz LP filter and a 1.7 MHz Hi pass filter only. I definitely got less noise on 80 and 40 m bands.

I would be interested to know how you got the noise from the PE4259's as I could not find any at all. I terminated my Rx with a dummy load and searched from 160 m to 60 MHz using my HiQSDR and could not find anything extra. I also listened at 25 KHz? directly and still could not find any noise or signal so maybe my method is suspect.
I was using 2 switches linked together with no filter as my test bed.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

Do you think the less noise on 80 and 40 M is because strong out of band signals are now not mixing and producing products in 80 or 40 M? Is it possible to pick 2 strong out of band broadcast signals in your area and show the in band interference with and without the BPFs? That could be a very convincing practical measurement in my mind.

The PE4259 charge pump noise I saw was around 900 kHz. From correspondence with Peregrine they expect noise around 1 MHz.  See https://groups.google.com/forum/#!searchin/hermes-lite/pe4259%7Csort:relevance/hermes-lite/tUdCDAMf7Tw/ftC8TYy3BgAJ or my post on the  thread "PE425x switches noise at 900 kHz and harmonics (from the EMRFD group)" in this group from May 1. I didn't do anything beyond what is described in that post. If there is a reason I see noise with the v1.42 boards but you don't, I'd like to know why so we can reduce/eliminate PE4259 noise in the HL2.

[Graeme Jury]

Yes I am sure that strong out of band signals are now not mixing is the main reason and in particular due to my proximity to the 5 local AM stations I am particularly plagued by this. I guess that the filters are not strictly necessary if I have a sufficiently good Hi pass filter at 1.7 MHz but it is set to 30 dB so I can still effectively monitor LF. There are a lot of products from the broadcast signals which could be intermod from anything from the a bad joint in the antenna to non linearities in the Rx front end but filtering helps and the extra filtering on top of my permanently wired 1.7 MHz filter improves things.

Now for the admission. I quickly went back and re tested for switch noise and realized that I had been testing through my LP filter reducing the charge pump noise by 30 dB. After removing the HPF and going directly in I get a noise floor of -111.4 at 923 KHz and -99.3 with the switches i.e. 2.3 dB above the noise floor of the Rx. At 2nd harmonic of the charge pump (1.846 MHz) I get a noise floor of -112.2 and switch signal of 111.4 i.e. 0.8 dB above noise floor and disappears with any antenna connected. I'm really glad you picked up on my false measurements Steve and sorry for the false hopes. The spot in the 160 M band is a nuisance but the broadcast spots don't worry me too much but for a clean signal you can't beat relays.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

I'd like to try and quantify if a BPF is better than a good HPF with a HL. If it is hard to determine the sources/products of the noise, would it be possible to do A/B testing of some strong signal on 80M or 40M and estimate the signal to noise floor with BPF only and then with HPF only? Are your measurements with a HL or HiQSDR?

[Graeme Jury]

Hello Steve,

I am happy to do any tests for you that I can and will get right on to it. Do you want an HPF constructed from smd inductors or toroids and the same for the band pass filter? How far down do you want the skirts? If 40 dB is OK then 5 pole filters will do the job with reasonable pass band ripple and of course save on components. I have found that 40 dB is enough in practise but you may want to shoot for more. 40 M or 80 M bands are a good source of strong signals at night here so I will go for one of them at your preference. I am also wondering if you want to try the filters through PE4259's and PIN diodes to gauge their effect. Do you want to simulate the HPF in conjunction with the Tx LPF to get a bandpass or do you want a dedicated bandpass filter built? With 3 inductors for the bandpass it won't be as sharp as the 4 inductor HP/LP combination.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

Thanks for all the offers, but I had something pretty simple in mind. You already have a HPF to block the AM stations in your area, and the new BPF filters. If BPFs are providing more noise reduction, then hopefully we can quickly see it and at least estimate SNR differences with strong signals in 40 and 80 M when comparing the HPF with a BPF. Even if the attenuation of both paths are slightly different and the strong signals will vary, I'm hoping the SNR differences will be big enough to notice. Are you able to switch the different filters in and out easily with a manual switch? I wouldn't go to extremes, but would be curious if it is easy for you to try.

73,

Steve

KF7O

[Graeme Jury]

OK Steve, here are some simple test results.

The screenshot labeled Thru_HPF.png is the 30 M band with no activity and my normal gain setting of 6 dB. The broadcast stations are attenuated by about 30 dB and any other signals around 0.1 dB as there is almost no passband loss. Some wideband noise can be observed which will be a combination of the HF bands plus attenuated broadcast band.

The screenshot labeled Thru_BPF.png is with the 30/20 M filter via PE4259 switches and again the normal LNA gain setting of 6 dB. This time the response is a lot flatter with some cw signals rising out of the noise which were almost obscured in the previous shot.

The screenshot Labeled Direct.png has the antenna connect directely into the Hermes-Lite and is completely overloading the receiver producing a clipping level noise 30 dB above the Through filter benchmark. The readings have no value apart from showing that at my QTH a directly connected antenna flattens the receiver right across the spectrum.

The screenshot labeled Direct_Reduced_Gain.png is the result of changing the LNA gain from +6 to -9. Now the receiver is not overloading but the wideband noise is leaving a noise floor of more than 10 dB above the bandpass filter's noise floor and the ripples that were appearing in the HPF sample are very similar.

1. There is no doubt that the cleanest band was from using the bandpass filter but not hugely more than the HPF.
2. No filter is not an option for Hermes-Lite receivers exposed to high power nearby stations.
3. Unfortunately I currently don't have a 30 MHz LPF to use as a pair with the 1.7 MHz HPF to check full HF band coverage but the 60 MHz LPF I use with my HiQSDR definitely stops FM radio noise appearing on 10 and 12 Metres on that radio.

Hope this helps.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

That definitely looks like a plus for the BPFs. Was there no LPF or did you use your 60 MHz LPF with this setup? Do you have a v1.42 board with 30 MHz LPF? I wonder how much difference a LP reconstruction filter would make. Also, it is hard to estimate SNR without a very strong signal. I wonder if the BPFs cover some SW broadcast band, if the carrier of an AM broadcast would do. Software could help a measurement like this -- average the noise floor over a period of time, and compare it to the average strength of a tuned signal. Maybe a capture of the waterfall would better highlight the differences.

I am surprised that you use only +6 dB gain. According to the measurements by AB4OJ, the minimum detectable signal for the HL is not on par with other radios until the LNA is in the +18 to +20 dB range.

Thanks for the measurements.

[Steve Haynal]

Hi List,

Below is a picture of the developing HL2 layout. This one includes the AD9866 block and the power supply blocks. The power supply blocks will be packed into the location labeled power supply. It is looking more and more likely that we can have a clean cut at 10 cm out with no filters or PA. In fact, the original target of 8x10cm looks achievable without the originally unplanned for OPA2677 preamp and extra Cyclone IV power supplies. 10x10cm is a typical price break from Elecrow. Extra space on the 10x10 main board will be used for low speed ADC (forward,reverse power, PA temperature monitoring) and digital PA bias. I will create the cut in mind assuming a 10x5cm extension board can be abutted and handwired to create a single 10x15cm ensemble that fits in the target enclosure. The cut will also support connections for generic companion cards. The PA as being tested by Claudio can fit in the lower corner, and I have a relay-based filter circuit and layout in mind that should support 6 LPF and 3 HPF to fit in the remaining space.  

73,

Steve

KF7O

[in3otd]

On Tuesday, September 13, 2016 at 6:53:47 AM UTC+2, Steve Haynal wrote:
[snip]

I am surprised that you use only +6 dB gain. According to the measurements by AB4OJ, the minimum detectable signal for the HL is not on par with other radios until the LNA is in the +18 to +20 dB range.

Hello Steve,
at 10 MHz the noise floor due to the man-made noise in a 500 Hz bandwidth could be around -100 dBm (see e.g. page 10 here) if one doesn't live in a quiet zone so a high gain is not needed as already with 6 dB gain the H-L noise floor will be well below that (see my measurements here). And of course running a lower gain should help in reducing the IMD and/or overloading.
If the antenna is poor it may pick up less external noise, besides less useful signals and more gain may be useful.

[James Ahlstrom]

Hello Steve,

This all sounds excellent, and I am looking forward to my HLv2 board and to making companion cards.

I wasn't sure if you intend to connect FPGA pins directly to the companion card.  This is risky, as the FPGA is fragile, and the IO pins provide little current.  This new high side switch TBD62783 works at almost any supply voltage and provides 500 ma:

http://www.digikey.com/product-detail/en/toshiba-semiconductor-and-storage/TBD62783AFWG,EL/TBD62783AFWGELCT-ND/5514137

It may be useful, and there are other choices of course.

Jim

N2ADR

[James Ahlstrom]

Hello Graeme,

Trout fishing sounds excellent.  I am sure I would enjoy the scenery, but I am afraid I lack the patience for actual fishing.  But I could look for birds!

I saw your comment that the filters I am designing belong on the HiQSDR Yahoo group.  If this becomes too Quisk oriented, we can move it there.  But I think an initial discussion may be of interest here.

My test setup is my original Quisk hardware which is identical to the HiQSDR for this purpose.  My test antenna is a fan dipole for 60, 40, 20 meters.  My current filters are a box of low pass filters and a box of high pass filters in series.  There are seven low pass filters plus a short selected by two relays each.  There are six high pass filters plus a short also selected by relays.  On transmit, the signal goes through the high pass filters to reduce wide spurs, then the power amps, then the low pass filters, then an antenna tuner and the antenna.  Power is 100 watts.  On receive the signal from the antenna goes through the antenna tuner, the low pass filters, then the high pass filters, then an optional 20dB preamp (not used here), and the HiQSDR Rx input.

Bands are selected in the filter boxes by an AVR microcontroller running a home made bus.  The PC band select is sent to an Ethernet gateway, and the gateway can send the data on a one-wire bus, or on an RS-485 bus or an I2C bus.  I got tired of DB9 connectors and serial buses, and I always seemed to need a way to get a low speed signal to some box or other.  The gateway is cheap and the small microcontrollers in each box are even cheaper.  This is peculiar to my station, but I can provide more detail if anyone wants.

There is a lot of interconnection between the boxes, the filters are too reactive within the bands, I don't think the high pass filters are really necessary for transmit, and I am not really happy with the result.  So I am currently redesigning my filters so that they work with either my hardware or the Hermes Lite.  First I needed to see what filters are really needed.

I tested this setup on 40 meters today, and the band was noisier than usual.  With no input, the visual noise floor on the graph screen is -122 dB.  With the filter box in place it is -107 dB and there are multiple signals visible.  If I connect the antenna to the HiQSDR without any filters at all, the noise floor is -105 dB.  The two dB difference is hard to eyeball, but I think it is just the filter loss.  Most importantly, the waterfall shows the same signals at the same intensity with the filters or a direct connection to the antenna, and the noise floor remains flat.  The conclusion is that at my QTH I don't need any Rx filtering, not even for AM broadcasters.  This is not what others have found, but it emphasizes that the filtering needed depends on the QTH.  The Rx filtering is meant to reduce out-of-band signals so that they do not cause clipping in the ADC, and they do not produce IMD that may fall in-band and add to noise.  But if you directly connect the antenna, and the ADC does not clip, and if the ADC and all other circuits are very linear and don't produce much IMD, then you don't need Rx filtering.  Well, actually you need to reject higher Nyquist zones, but that is standard.

My new filter design is preliminary, but looks like this.  I am putting all filters in the same box to reduce all the annoying interconnects.  Tx will no longer use the high pass filters (pending a new search for spurs).  There will be the same 7 low pass filters plus short, but only three high pass filters plus short.  I plan to study all the stray capacitance and inductance from all the relay switching and figure out a way to adjust the filter values to compensate, so that the Tx return loss is good through the filters.

As far as HL is concerned, the Tx low pass filter issues are the usual ones, and I have little to add.  At issue are the Rx filters.  We always need a Nyquist filter to reject frequencies above 39 MHz.  If we leave the Tx filters in line for Rx, we get some free filtering, and we could switch in an AM broadcast band high pass filter at 1.7 MHz or so.  It is switchable so we can listen on low frequencies.

To go beyond this, we could switch in a few high pass filters along with the low pass, but it is hard to know what frequencies to use as it depends on the interfering signals, and that depends on the QTH.  If we get close to having a high pass per band, the high/low combination no longer makes sense.  We should change to a band pass Rx filter for each band because it has a better response shape.  But there are 10 bands from 160 to 10 meters, and we now have 20 switched filters.  Yikes!

As far as a recommendation, it is hard to say.  The 1.42 frontend has a Nyquist filter, and I think that it should remain. If a one-board design has room for maybe three bands, we could go the SoftRock route and leave the bands as a build option.  For example, for a 40-30-20 build, we would have three switched Tx filters and a single high pass filter at 6 MHz (plus the Nyquist filter at the AD9866).  A "short circuit" option for Tx and Rx filters is nice to provide general coverage.  Another viable (??) option is to have ten Rx filters and no Tx filters on the assumption that HL is a driver to an external power amp that has Tx filters.  In my case, I just need a way to connect to my new filter box, on the assumption that it ever gets finished (not everything does).  We may need to know how much board space we have to see what to do next.

I guess my main point here is that "testing" to determine the optimal Rx filters is not really testing, because it depends on the QTH.  But it does seem wise to add at least a single switched high pass filter at the frequency of our lowest band.

Jim

N2ADR

[in3otd]

Hello Steve,
if there are still unused GPIOs (...) could it be useful to have on the PA card a 3 or 4 pin connector, compatible with the typical fan with built-in speed-control commonly used for PCs? Since the H-L will sense the PA temperature it could also control the fan speed... just in case someone would like to include a big(ger) PA.
You mentioned using SMA for the RF interconnections between boards; if you need to save some space MMCX connectors could be used, AliExpress has also cheap MMCX jumpers. OTOH, they are more difficult to assemble, if one wants to prepare the cables by himself. Or maybe we do not even need a true coax connectors, just the plain flat cable with ground wires around the RF signal could be fine, if not too long.

73 de Claudio, IN3OTD / DK1CG

On Tuesday, September 13, 2016 at 8:22:45 AM UTC+2, Steve Haynal wrote:

Hi List,

[Graeme Jury]

Hi Jim,

You would have loved the bird life where I was, often there would be 50 or more swallows swooping right in front of me picking up flies as they hatched and emerged from the river, not to mention the prolific native bird life in the surrounding bush but I had better not turn this into a nature study in spite of it being as good as life can get.

Thank you for your very detailed description of your station filtering in particular and your station in general. It is amazing how close you are to my own setup and how you have identified all the issues that I am facing. I have read (avidly) all your QEX articles and have been carefully through your web site and understand your original Quisk hardware. My antenna is a fan dipole of 80, 40, 30 with 15 on 3rd harmonic of 40 and 17 on 5th harmonic of 80. I also have a 3 element tribander for 20, 15 and 10 metres. I am using Hermes-Lite with the BeMicro SDK and the 1.42 front end board barefoot and my HiQSDR with a 20 watt home brew amplifier and manually switched LP filters driving a 100 watt linear also with manually switched filters. I have a home brew Arduino controlled automatic antenna tuner. My antennas are remotely switchable between my shack and my xyl's (zl2ujt) TS-850 via coaxial relays.

As you have already done I am setting up an AVR controller using an Arduino nano so that I can autoswitch filters, initiate antenna tuning, measure transmit parameters and control antenna switching etc. all initiated from the PC that I am running the radio software e.g. Quisk from. To get a wifi connection I have just purchased some ESP8266 modules from China (very cheap) as I am trying to reduce 10/100 ethernet due to noise radiation although gigabit seems quiet. I would be very interested in the detail of your gateway setup and believe that it would be a good idea to keep as compatible as possible.

Interconnections have a lot of options with the Arduino and I have also got some very cheap RS485 modules from China which are excellent and give a long distance but are tied to wires whereas the ESP8266 wifi modules give a lot of flexibility. Of course the Arduino also offers RS232, I2C and SPI plus I also used modbus in another project. The ESP8266 module is very interesting in that it has the potential for stand alone operation with an inbuilt microprocessor and gpio pins. There different versions which offer more gpio and the Arduino IDE has an overlay which allows the ESP8266 to be programmed from there.

I require my filtering to support the Hermes-Lite as a stand alone clean hopefully 10 watts Tx but I can live with 5 and a filtered wide band receiver. As in your case having standard filtering in the Tx is a given and all the issues for me are about what to do for receive. I am in absolutely agree with you that the filter requirements on receive are QTH dependent and will vary hugely between countries as well as within any single country and strong signals may be clustered around 1 MHz or 7 MHz or 9 MHz etc. To have a subset of all the band filters would mean offering build options which would make kitting complex and the alternative of offering filters for all bands could lead to a lot of unnecessary filters. BTW the nyquist filter in the 1.42 frontend works brilliantly. Last night I built a really good outboard nyquist filter and added it in series with the 1.42 frontend and it made no difference i.e. the filter already there is doing the job and I get volts of 80 to 100 MHz FM on my antennas so it is a good test. With no top filter on my HiQSDR, 10 and 12 M are unusable.

I have collected all the parts and a PCB for a 300 watt diplexing transmit filter which will be attached to my new linear that I am building but is currently sidelined while I concentrate on the HL project. I mention this to point out that the HL and HiQSDR will become the drivers for this amplifier and I expect that this will be the case for many other builders so filtering, power level and external control signals will be a factor.

I had pretty much decided that I would have Hi Pass Rx filters for all bands, matching up two bands where possible like 30/20 for around 6 HP filters matching 6 LP transmit filters with the ability to alter the associated HP filter and both sets having a through connection. Now you have thrown a spanner in the works with your well reasoned arguments coupled with Steve's hints along the same lines and I am now leaning towards having the minimum filtering necessary in the HL box and keep it as small as possible and any "contest grade" filters belong external to the HL and could perhaps be associated with the external linear amplifier unit. I designed an LP/HP combination for 80 Metres last night with predicted path attenuation of < 0.3 dB and return loss of better than 30 dB. No chance to build and test it yet and will be tied up for next few days but will report on its effectiveness as soon as I can. It is a 5 pole Cauer for each filter with better than 40 dB attenuation and notches placed on harmonics.

I take your point that a dedicated bandpass filter can be constructed which will be superior to the HP/LP combination but as the LP needs to be there anyway I think it would be an acceptable compromise to match HP filters to them for Rx bandpass filtering as long as they are good enough. I also have a nagging worry that if looking for high grade filtering to perhaps operate 2 radios in close proximity on different bands like 80 and 40 M, we are connecting in series all the passband losses. Not really an issue on LF bands as I usually operate with 10 dB of attenuator in anyway but could be an issue on say 15 and 10.

Currently recommendations on my part are;
The nyquist filter should be part of HL main board.
The HL main board should have basically what is on the 1.42 frontend
The daughter board should have the 5/10 watt amp and switched LP filters
Some switched HP filters should be somewhere.
Perhaps only a switched 1.7 MHz HP filter could be on the PA/LP filter daughter board or even on the main board along with the nyquist filter.
There should be through connection on all filter banks to allow using the HL as a high power oscillator etc.
Filters should be switched with relays? PIN? Peregrine? I really don't know yet and testing needs to be done here.
Another option is to build a set of bandpass filters and share with Rx and Tx

It seems we are still some ways off a final design.

73, Graeme zl2apv

[Steve Haynal]

Hi Jim,

What I am doing is classifying if a signal connection is external (swtich input on front panel, PTT switch out on back, etc.) or internal (connection to another DB) and providing some protection for external signals. Even for external signals like the high speed LVDS and external clock input, I can't add much protection without impacting performance. For internal signals, people should be careful. If you look at other electronics in the class we are targeting, like the BeMicro CVA9, Arduino, Raspberry Pi, most of the Terasic FPGA boards and Digilent FPGA boards, there is little protection on IO. 

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

If I restate what I hear you saying, is that in noisy environments, +6 dB of gain is enough as the HL noise floor will still be below the higher ambient noise, and if you increase gain you are just amplifying noise. Maybe this is true for SSB signals decoded with a human ear, but for modes like WSPR, JT9 and JT65, where the signal can be 30 dB below the SSB noise floor, it seems more gain is beneficial so that weak signals even register with the ADC.

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

For slow control signals like a fan control, an i2c bus expander should work. There will be 1 or 2 i2c buses available to connect to. I am happy to work towards fan support via i2c in the firmware.

We have some boards at work with MMCX connectors. I looked at them briefly early on and recall that the number of insertions they support is lower than SMA, but not 100% sure. So far, the SMA footprints are not getting in the way -- they are in places where layout already needs a little spread and room for other reasons. I will keep it in mind if things get tight. We can use cables soldered to the SMA connector footprints too. All SMA connectors except those edge mounted on the front or back are for optional functionalities. 

73,

Steve

KF7O

[Steve Haynal]

Hi Jim and Graeme,

I am interested in how ~3 HPF work in practice, when also coupled with standard LPF TX filters for RX. It is possible that few PE4259 switched HPFs can be fit on the main HL board for RX. I would really like to understand what benefit, if any, narrower BPFs might have over wider LPF/HPF combinations in practice. I suspect that the BPFs are not adding enough to make the extra expense of 6 or 7 of them worth it, especially for a HL2 grade transceiver.

 73,

Steve

KF7O

[in3otd]

Hello Steve,

On Sunday, September 18, 2016 at 8:14:27 PM UTC+2, Steve Haynal wrote:

If I restate what I hear you saying, is that in noisy environments, +6 dB of gain is enough as the HL noise floor will still be below the higher ambient noise, and if you increase gain you are just amplifying noise.

yes, this is what I meant.
 

Maybe this is true for SSB signals decoded with a human ear, but for modes like WSPR, JT9 and JT65, where the signal can be 30 dB below the SSB noise floor, it seems more gain is beneficial so that weak signals even register with the ADC.

uhm, you mean here that a signal 30 dB below the noise floor will not be received because it's smaller than the ADC LSB ? But we do normally receive signals with an amplitude lower than the ADC LSB; compute the H-L ADC LSB size for a 6 dB RxPGA gain...

[Steve Haynal]

Hi Claudio,

I agree that there is processing gain, but MDS must mean something. I have you telling me that RX can't be improved by increasing gain in noisy environments, and AB4OJ saying in correspondence that he had to adjust the LNA to +19 dB to have comparable HL MDS with other receivers. I live in a noisy environment, and I definitely notice WSPR improvement when raising the LNA gain to +20 or more. How would you explain this?

73,

Steve

KF7O

[in3otd]

Hello Steve,
well, the MDS just indicate the receiver intrinsic noise floor; if the antenna presents to the receiver input a noise floor (from the external man-made noise) which is much higher than the MDS we are simply wasting dynamic range. As the AB4OJ presentation linked before shows, usually the external noise floor is comparable to the typical receiver MDS only in the upper HF and only if one lives in a not-so-noisy area.
If the antenna is not efficient (low gain) also the external noise it picks up will be attenuated and then a low MDS may be more important also at lower frequencies.

If the RX noise floor increases by several dB when you connect the antenna, more gain should not be needed for a good WSPR or SSB reception. If you notice an improvement with even more gain I may only suspect some non-linear effect may come into play (like the "incidental dithering" discussed before).

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Maybe we can find agreement if we consider dynamic range. The AD9866's effective number of bits is a little more than 10, which gives a dynamic range of around 62 dB. Processing gains in the firmware and software will add to this. If we don't increase the LNA to the point where the ADC is about to clip, we will sacrifice dynamic range at about 6 dB per bit. If we want to maximize dynamic range, we should increase the LNA gain to the point of just clipping. I am under the assumption that once the BPF was in place, Graeme could increase the LNA gain beyond +6dB without clipping. If we have other objectives in mind, such as reducing IMD, then I would agree that reduced gain can help.

[Graeme Jury]

Hi Steve and list,

I have had a play over the last couple of days with an HPF switched by ordinary 1N914's (picture attached) with pretty good results. The filter was designed with Elsie as a 5 pole Cauer optimised for standard capacitors with a measured minimum out of band attenuation of 40 dB and in band of 0.3 dB worst case at the band edge and generally 0.1 dB in band with a 25 dB or better return loss (not so important on Rx).

The switching circuit introduced around 0.9 dB loss over the input and output switch and better than 60 dB isolation falling to 50 dB at 30 MHz. I have not checked for intermodulation distortion but suspect that there is very little with on air observations as I can see nothing that changes between hard wired and diode switched. I used 20 mA forward bias and 12 volts reverse bias and altering from those numbers a reasonable amount (30%) did not change anything. Why did I even bother to do this? The Peregrine switches do introduce a spur on 160 M band and I was looking to eliminate this with the diode switching which is giving a comparable performance - perhaps around 0.3 dB more through loss but at the expense of more complicated switching circuitry in terms of needing Bias Tee chokes for supplying the switch current. The chokes were made from BN2402 cores with 7 turns of kynar wire wrap wire on them which fits OK but firmly. For those who can't live with the spur from the PE4259 switches it seems at first look that diode switching is a viable option.

Now to what HPF's are needed in conjunction with the HPF's. It is assumed that the Nyquist/roofing filter on receive is a firm design criteria and there will be an HPF for all bands with all bands up to and including 80 on their own filter and 60/40 30/20 17/15 plus 12/10 sharing a filter each. There could be some advantage in 12/10 being a roofing filter and in circuit all the time if it was designed with toroids for very low in band loss. The HPF's probably only need to be at 1.8 MHz, 3.5 MHz, 10 MHz and 18 MHz. I would be inclined to treat the 1.8 MHz filter as a flooring filter, i.e. designed with toroids and switched with miniature relays and unless wanting to operate on VLF bands always in circuit which means that there would be 3 HPF's to regularly switch with band changes. There could be some advantage in a fourth filter on 24 MHz in another 5 years or so when the propagation picks up but at this stage it does not give any advantage.

The choice of HPF of course can be modified by external circumstances. An example in my own case is where my XYL who usually operates on 80 or 40 is on air using the wire antenna which runs right past my beam can go on air the same time as me if I use a 10 MHz HPF and operate on 20 M up with the beam. I guess this would apply to hams who have another ham living nearby and an HPF can save the day.

A note on the HPF design. The picture of the prototype 3.5 MHz filter is attached. It was built from junk box parts and some of the capacitors are standard leaded ceramic with the toroids wound to the calculated turns and spread reasonably around the core. The filter was a carbon copy of the Elsie plots and required no adjustment at all. Apart from the tedious process of making a 22 and a 25 turn winding the toroids were straight forward and were not adjusted. This may not be the case for the higher frequency HPF's.

I took a look at the effectiveness of the bypass capacitors that I used on the bias tee chokes which were 1206 0.1 uF Y5V. I firstly connected 2 BNC sockets on the board and wired between them with the capacitor bypassing the middle of the wire. It dropped the signal around 40 dB with a resonance showing at 8 MHz and the effectiveness decreased going up in frequency from there. I tried the same experiment again with the board used by Claudio for his diode IMD testing which has a transmission line between the two coax connectors. The centre of the transmission line was bypassed and this time no resonance showed up plus the attenuation was better and much more slowly deteriorated as the frequency went up. I think that there is a message here that transmission lines for RF interconnects will reduce the effect of stray reactances in a very noticeable way.

The final step to come is to build an Rx HPF board with relays and compare it with one built with diodes and one with PE4259 switches, run it on air and decide from there. My expectation is that it won't matter much which topology is chosen for performance but IMD from the diodes is an unlikely possibility and the spur on 160 M from the PE4259 switches may be difficult to eliminate or even sufficiently reduce. The relays of course are big.

Finally, running the 80M HPF in conjunction with the 80M LPF gave excellent results and a noticeable reduction in wideband noise. It is a must have at my QTH

73, Graeme zl2apv

[Graeme Jury]

Sorry but I forgot to include the photo of my test jig. The bias tee choke construction can be seen on the BN2402 cores and the bias voltage was applied between the junction of the 150 ohm resistors and ground.

cheers, Graeme

[Alan Hopper]

Steve, Claudio,

I've been slowly working on a software rf agc, What I'm trying at the moment is :-

Continuously measure the lowest noise floor of all bands being listened to.

Pick a target gain that is an appropriate amount below the noise floor (what is an appropriate amount?) from either a calibration table or Claudio's graph.

Run agc with this as a maximum and reducing on some measure of clipping and peak wideband adc values.

I'm currently using only the 6db steps of the lna, to my mind the in between values reduce overhead without much snr improvement, am I missing something here?

I'll then try running this side by side with the  maximum gain strategy and see what happens.  I'm not expecting big differences, my previous attempts at optimising gain were inconclusive and it  took large gain differences to get significantly different results.

Could higher gains be making any 'incidental dither' more effective?

I think the WSPR snr values can be a bit misleading as they relate narrow signals to a much wider noise bandwidth, http://wsprnet.org/drupal/node/3853 hence -30db signals being visible on the waterfall.

73 Alan 2E0NNB

On Sunday, September 18, 2016 at 9:47:48 PM UTC+1, Steve Haynal wrote:

Hi Claudio,

[James Ahlstrom]

Hi Steve,

That is fine, and we will just have to document the need to be careful with the FPGA pins.

Jim

N2ADR

[James Ahlstrom]

Hello Steve,

I can not explain your WSPR improvement when raising the LNA gain.  But a very weak signal that is below the ADC least significant bit (LSB) can be received because of the dither provided by noise and other signals.  Consider a weak signal on a noisy band.  The noise is visible on the ADC output so it must be sweeping the voltage up and down enough to change at least the last few bits of the ADC.  Now add the weak signal with an amplitude less than the LSB.  On a statistical basis, it will cause some bits to change from what they would have been, and the result will be the received weak signal.

Jim

N2ADR

[Steve Haynal]

Hi Jim and Claudio,

I have and do understand what you are describing. My choice of words "weak signals even register with the ADC" was bad and confusing. What I had in mind was dynamic range. Do people disagree with my last reasoning that by adjusting the LNA so that the ADC is nearly clipping, you move the dynamic range of the ADC to utilize the complete dynamic range during RX, and that if you keep the LNA gain low, some of the dynamic range may never be utilized?

73,

Steve

KF7O

[Dani EA4GPZ]

El 19/09/16 a las 17:10, Steve Haynal escribió:

> Hi Jim and Claudio,
>
> I have and do understand what you are describing. My choice of words
> "weak signals even register with the ADC" was bad and confusing. What I
> had in mind was dynamic range. Do people disagree with my last reasoning
> that by adjusting the LNA so that the ADC is nearly clipping, you move
> the dynamic range of the ADC to utilize the complete dynamic range
> during RX, and that if you keep the LNA gain low, some of the dynamic
> range may never be utilized?

Hi all,

I'm not sure if I understand this correctly, but I think that if your
gain is set high enough that the noise floor raises significantly (say
more than 10dB) when the antenna is connected, then you can ignore
quantization noise.

73,

Dani EA4GPZ.

[in3otd]

Hello Dani,
yes, this is what we expect to happen, but Steve said that he saw an improvement by raising the gain at even higher values than the one needed to have the external noise floor well above the RX noise floor. This is not what I would have expected and I do not have an explanation for this...

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Steve,
ok, with a higher gain the signal "uses" more of the ADC dynamic range, but what improvement should this bring? I the ADC was ideal that should not make a difference, in practice some non-linear effect may come into play, as mentioned before, but at the moment I cannot see what this may be.
To simplify (maybe a bit too much), in my "linear system" view this will be the same as cranking up the audio level at the end of the chain, the output will be louder but the SNR will remain the same.

73 de Claudio, IN3OTD / DK1CG

[Robert Bennett]

Hello Steve

I instinctively support your view, but I cannot give any good 'scientific' reason.  We should not forget that the 'band noise' is not white noise but a complex mixture of large and small signals of all kinds which sounds to our ears like noise, but from which signals can be extracted.

To play a bit further on the 'cranking up the audio' analogy.............if you are listening to an orchestra and you crank up the audio, the signal to noise ratio stays the same, but you can likely hear better an instrument that is playing more softly in the background.

Perhaps a better question to ask is whether there is the downside from keeping the majority of the bits filled in the A to D? 

73

Robert G3WKU

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[in3otd]

Hello,
I have measured the BAR64 (single and two in parallel) and the SKY13374-397LF: in short, no good news...

As somewhat expected, the single BAR64 cannot handle much power; in the first test run I blew up a diode at some point, likely when trying to pass 10 W with 100 mA of DC bias... On the second run I lowered the test duty cycle and the diode did not get damaged, but of course its performances at that power level were bad.
Two diodes in (anti-)parallel work better but are good for maybe 1 W maximum.
The usual graphs are on my website at
http://www.qsl.net/in3otd/electronics/PIN_diodes/BAR64.html
and
http://www.qsl.net/in3otd/electronics/PIN_diodes/2xBAR64.html .

The SKY13374-397LF also seems good up to 1 W at 1.8 MHz; the 39 dBm specified in the datasheet are really probably valid only above 500 MHz, as specified there. Apart from that, it's a nice device, with a very good isolation for the "off" path. Making a PCB and soldering the device required quite some patience, though.
Maybe later I'll try to check for spurs from the internal voltage generator.
Details at http://www.qsl.net/in3otd/ham_radio/SKY13374-397LF/SKY13374-397LF.html .

73 de Claudio, IN3OTD / DK1CG

On Sunday, September 11, 2016 at 11:46:56 PM UTC+2, Steve Haynal wrote:

Hi Claudio,

Thanks for investigating the BAR64 and the SKY13374-397LF. I look forward to learning what you find.

73,

Steve

KF7O

[Graeme Jury]

Hello Claudio,

Many thanks for the very valuable information you have measured and collated. It has certainly been an eye opener. I would be grateful if you could possible do one more diode pitched at receive service and used as a switching diode rather than a PIN. I would be interested in the results of a 1N914 or 1N4148 run say up to 50 mA which I am pretty sure will be within ratings for these diodes. I realize how busy you are so no pressure, if you can fit it in would be great but no rush.

73 es tu, Graeme zl2apv

[Steve Haynal]

Hi Claudio, Robert and Dani,

The improvement I think increasing gain so that signal peaks are full scale on the ADC is that you've aligned the top of your fixed dynamic range with the peak signals so that you now weaker signals will be within your dynamic range whereas before they weren't. For modes like WSPR, JT9 and JT65 which extract weak signals from the noise, you now have a weak signal to extract. I readily admit I am no expert on this though. I poked around and found this paper and this paper. Both of these focus on the noise figure and discuss how an LNA in front of the ADC with low noise figure can be used to lower the overall noise figure. Intuitively, I am thinking that the noise from the ADC is fixed and on the not so good side given only 12 bits in the AD9866, but the LNA can amplify the signal above this noise without introducing more noise than the ADC. For very weak signals, you want to push this to the max. Please share a good short link or two covering any points you think I am missing.

73,

Steve

KF7O

 

On Monday, September 19, 2016 at 12:36:18 PM UTC-7, Robert Bennett wrote:

Hello Steve

I instinctively support your view, but I cannot give any good 'scientific' reason.  We should not forget that the 'band noise' is not white noise but a complex mixture of large and small signals of all kinds which sounds to our ears like noise, but from which signals can be extracted.

To play a bit further on the 'cranking up the audio' analogy.............if you are listening to an orchestra and you crank up the audio, the signal to noise ratio stays the same, but you can likely hear better an instrument that is playing more softly in the background.

Perhaps a better question to ask is whether there is the downside from keeping the majority of the bits filled in the A to D? 

73

Robert G3WKU

[Steve Haynal]

Hi Alan,

Sounds like interesting experiments. I think higher gain could help with incidental dither. As you point out, the additional processing gain for very narrow band WSPR signals lowers the noise floor to make the signal visible. I wonder if gain settings should be picked with that noise floor in mind, rather than the wider bandwidths traditionally used.

73,

Steve
KF7O

[Steve Haynal]

Hi Claudio,

Thanks as always for your measurements. Are you certain that the SKY13374-397LF can't handle 39dBm until the specified 500 MHz? I am wondering if the drop off actually occurs somewhere in the HF bands and if this device is usable for some HF bands. You could measure at 10M and see if there is any hope for that band and then if so, for lower bands.

73,

Steve

KF7O

[Graeme Jury]

Just wondering if the RF chokes cause some of the fall off. Have they got enough inductance with the current through them, is the core being saturated?

73, Graeme zl2apv

--

You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.

To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.

[Alan Hopper]

Steve,

I now have a basic software rf agc, like your firmware version it tries to maximise gain but only moves in 6db steps and has a fixed maximum gain of +30db.  This is means I can now see the gain being used and the display is scaled correctly.

I did a initial test on 20m jt65/9 with two receivers.  One had the sw agc on and spent most of the time at 24db or 30db with occasional overloads.  Eyeballing the panascope indicated a gain between 6 and 12 would put the band noise 10db above the radio noise (measured with dummy load).

With the second radio at 12db the results were neck and neck over an hour, I swapped gain settings over half way through to eliminate any other differences.

With the second radio at 6db it slowly fell behind and ended up 5% down.

I'll try some longer tests this week.  I'm open to test suggestions.

I have no idea where I am on Claudio's 'incidental dither' imd curves, maybe I have enough dither so imd is not affecting the higher gains significantly.  

As far as the narrow band noise floor is concerned, my understanding is the optimal gain works out the same as long as you use the same bandwidth for measuring the radio noise and the band noise but I'm happy to be put right on this.

73 Alan 2E0NNB

[in3otd]

Good point, Graeme;
the datasheet says their inductance should drop 5% with 175 mA of DC current, so there should be only a little effect at the 100 mA maximum bias used for testing the PIN diodes. It's true that a significant RF current will flow thru the chokes at 1.8 MHz and I always see some drop in the output power at the higher power levels which I'm not sure it's always completely due to the devices.
I'll do some high-power testing with a short circuit instead of a diode and see if the power still drops at the higher end, hi.

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Steve,
no, I didn't test at 500 MHz, mine was only an assumption, thinking that if the device was good for 39 dBm below 500 MHz they would have stated that, as they did for the gain, which is specified starting from 300 kHz.
I do not have a power source at 500 MHz, I can maybe get 5 W or so in the 70 cm band, or a higher power in the 23 cm band. I can anyway try at 30 MHz and see how the device behaves there.

73 de Claudio, IN3OTD / DK1CG

On Tuesday, September 20, 2016 at 5:50:03 AM UTC+2, Steve Haynal wrote:

Hi Claudio,

[in3otd]

Hello Alan,
thanks for the interesting experiments,

On Tuesday, September 20, 2016 at 8:36:24 AM UTC+2, Alan Hopper wrote:
[snip]

As far as the narrow band noise floor is concerned, my understanding is the optimal gain works out the same as long as you use the same bandwidth for measuring the radio noise and the band noise but I'm happy to be put right on this.

yes, even better if you leave out the bandwidth altogether just use the noise spectral densities, hi. The decoder (ear+brain or SW) won't discriminate between signal and noise and use the same bandwidth for both.

To (somewhat) answer your question abut the appropriate gain to use, if you set the gain to have the total received noise (amplified external noise plus internal RX noise) 'k' times the RX noise alone (in linear units, not dB), the SNR at the receiver output will be (k-1)/k the original SNR at the antenna. So if you set the gain to have the overall noise floor 3 dB above the RX noise floor the SNR will be degraded by 3 dB, if the overall noise floor is 6 dB above, the SNR degradation will be 1.2 dB, for 10 dB above the RX noise floor it will be less than 0.5 dB, etc. So it seems to me there is little to be gained going above 10 dB over the RX noise floor. Then the actual external noise is not constant so fading, intereference, etc., can change this a bit.

[Alan Hopper]

Claudio,

thanks for that, I now have a rf agc that tries to keep the total noise floor between 12 and 18db over the known radio noise floor.

A test against my max gain agc on 20m over approx 2hrs and 1000 spots resulted in an identical number of spots each. The max gain one spent most time on +24db and the new one between +6 and +12db. The reported snr for the JT65/9 spots were identical.  A test with the new agc set 6db lower was 4% down on the max gain agc.

It appears I have not gained much, maybe whatever prevented bandpass filters helping is at work here as well.  The new agc does have a headroom advantage so is less likely to clip. On a lower noise band the two agc end up working identically.

I'll give it a long run and see what happens with a wider range of conditions.

73 Alan 2E0NNB

[in3otd]

Hello Graeme,
I did a test at 1.8 MHz with the PIN test fixture with a short circuit instead of a PIN and with 100 mA thru the inductors; the results were identical as without the fixture, for both the gain/harmonics and IMD, up to the maximum power (10 W).

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Alan,

This is an interesting result. Were there any filters applied besides the reconstruction LPF? Did the receiver at +24dB gain clip excessively? Since results were the same at +24 dB and +6 to +12 dB gain, it makes me think you had at least 12 dB of dynamic range that was never really necessary. Maybe we should consider the 10-bit AD9865 little brother then for budget builds. I think it is footprint compatible with the AD9866. It is $10-$12 less from US distributors, and there are many hits in China in the $5 to $10 range.

It really bothers me that with lower gain settings, the upper ~2 bits are so seldom utilized. I know this gives us "headroom", but what about more "footroom?" I don't want to reserve this headroom for some rare event. Maybe I'm not understanding some fundamental facts about receiver noise floors and noise...

73,

Steve

KF7O

[Graeme Jury]

Hello Claudio,

Thanks for the test and the very conclusive result. Being wound solenoid style should help with the saturation of the cores. The BN2402 cores with 7 turns that I am using give a similar inductance to your choke and show no sign of saturation with 100 mA either. I do get some loss at the LF end due to the coupling capacitors and the shunt inductance as there are 3 inductors shunting the circuit for the LP filters but it is necessary to have 4 inductors shunting for the HP filters circuit with no continuity due to the series capacitors of the filter. an S21 sweep shows it just starting to curve down at 2 MHz so I think 47 uH is enough inductance.

73, Graeme zl2apv

[Alan Hopper]

Steve,

there were no filters and very little clipping, my 'max gain' agc is a bit less aggressive than yours.  On bands with lower noise all bits are used so a 10bit device would probably be at a disadvantage.  I admit my gut has a problem with the unused bits but it does make sense to me, the radio is just over specified on noisy bands without very strong signals.  I had hoped the lower gain setting would be better, I don't really know how to investigate why it isn't.

I have been lent an Orion board (as used in the Anan 200D). At my QTH I have never seen the overload flash and I'm not sure the top two or three bits are ever used!

73 Alan 2E0NNB

[in3otd]

Hello,
it may well be that the top bits are seldom used in some cases, but I think this is a "problem" of all radios and of many systems; you need to design for the peak performance/capacity even if the average can be much lower. It happens with power amplifiers, data networks, cars, etc.
There are lot of people using the DVB tuner dongles which have 8 b ADCs (and maybe around 7 b ENOB), everybody seems happy with the price/performance ratio but likely no ham contest station is using them, hi.

Having said that, I took a look at the AD9865 datasheet, the ADC (and DAC) actual performances seems not too different; for the RX it seems the noise floor is determined by the RxPGA, at the gain settings we are using and the THD is not much better in the AD9866. I guess we need to build an H-L with that and send it to Alan for a side-by-side comparison, hi.

73 de Claudio, IN3OTD / DK1CG

[Dani EA4GPZ]

El 22/09/16 a las 03:29, Steve Haynal escribió:

> Hi Alan,
>
> This is an interesting result. Were there any filters applied besides
> the reconstruction LPF? Did the receiver at +24dB gain clip excessively?
> Since results were the same at +24 dB and +6 to +12 dB gain, it makes me
> think you had at least 12 dB of dynamic range that was never really
> necessary. Maybe we should consider the 10-bit AD9865

> <http://www.analog.com/media/en/technical-documentation/data-sheets/AD9865.pdf> little

> brother then for budget builds. I think it is footprint compatible with
> the AD9866. It is $10-$12 less from US distributors, and there are many
> hits in China in the $5 to $10 range.
>
> It really bothers me that with lower gain settings, the upper ~2 bits
> are so seldom utilized. I know this gives us "headroom", but what about
> more "footroom?" I don't want to reserve this headroom for some rare
> event. Maybe I'm not understanding some fundamental facts about receiver
> noise floors and noise...

Hi Steve,

The problem with dynamic range is strong signals on the same band you
want to receive. When you have strong signals on the band you're
listening to, you have to lower the gain to prevent clipping (if they're
out of band you also have the option to use filters).

Lowering the gain has the effect of increasing the quantization noise.
If the RF noise from the antenna is higher than the quantization noise,
you won't notice this effect. However, at some point the quantization
noise will be stronger and the receiver noise floor will raise as you
decrease gain, wiping out weaker signals.

If you set the gain so that the ADC is barely clipping on the strongest
signal (which is the best you can do regarding quantization noise), for
10bit, the quantization noise is roughly 60.2dB below the strongest
signal. For 12bit, it's 72.24dB.

It's not uncommon to have an RF noise floor around S5 (in 2.7dB
bandwidth, S9 = -73dBm) in many bands in urban locations. In quite rural
locations the noise is even lower. With this RF noise floor, a signal
stronger than S9+36.2dB will raise your noise floor due to quantization
if you use 10bits. If you use 12bits, it's S9+48.24dB.

I think that signals which are S9+40dB are relatively common in some
bands (40m for instance).

That's my opinion: 10bit is not so good for general HF work. Might work
well for somebody depending on their RF environment. 12bit barely cuts
it, either, but it's probably the highest you can go without going too
expensive and/or running into dynamic range limitations due to the
analog parts of the system.

73,

Dani.

[Alan Hopper]

Dani,

I suspect we can get some improvement in effective dynamic range by allowing more clipping in some circumstances. As Steve has pointed out the current overload indicator is fairly crude and does not differentiate between single clipped samples or a whole block.  The overload permanently on might only indicate a few samples clipped per second.  Clipping is obviously bad but there is a balance between clipping and using too low a gain on a low noise band.  

Firmware changes are needed to be a be able measure clipping better, my vote would be for a simple count of the number of clips per packet. To make life easier for software agc and level display it would also be good to have adc peak magnitude and perhaps mean magnitude per packet.

Claudio,

given the apparently unchallenging nature of my QTH I think we should build a matched pair and send them on a world tour of QTHs :)

73 Alan 2E0NNB

[Steve Dick]

One other point: Don't forget that dynamic range of a receiver using an A/D
with decimation is affected by the amount of oversampling. I.E. the ratio of
A/D sampling rate to the bandwidth you end up with after decimation. The
higher the initial sampling rate, the more the quantization noise is spread
over a wider initial bandwidth. So you end up getting a 3dB improvement in
S/N for every doubling in sampling rate for a given number of A/D bits or an
improvement in S/N ratio of 3dB for every decimation by two, compared to the
initial A/D dynamic range which is approximately 6dB per bit. So receiver
dynamic range, in general, is much larger than the simple approximation of
6db per A/D bit due to decimation.

"Digital Steve", K1RF

--
You received this message because you are subscribed to the Google Groups
"Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

[in3otd]

Hello Dani,
as K1RF said, you forgot the "processing gain";

On Thursday, September 22, 2016 at 9:21:44 AM UTC+2, Dani EA4GPZ wrote:
[snip]

If you set the gain so that the ADC is barely clipping on the strongest
signal (which is the best you can do regarding quantization noise), for
10bit, the quantization noise is roughly 60.2dB below the strongest
signal. For 12bit, it's 72.24dB.

yes, but the quantization noise is practically spread over the entire bandwidth of the ADC (half of the sampling frequency). The quantization noise in an SSB channel will be a fraction of this, 10*log10(2*2400/80e6) dB below (assuming an SSB channel 2400 Hz wide and a sampling frequency of 80 MHz);
 

It's not uncommon to have an RF noise floor around S5 (in 2.7dB
bandwidth, S9 = -73dBm)

This in a 500 Hz or 2400 Hz bandwidth, much smaller than the ADC bandwidth...
 

in many bands in urban locations. In quite rural
locations the noise is even lower. With this RF noise floor, a signal
stronger than S9+36.2dB will raise your noise floor due to quantization
if you use 10bits. If you use 12bits, it's S9+48.24dB.

nope, you need to raise the signal from the SSB (or CW) channel noise floor, so 10*log10(2*2400/80e6) + 60 dB = 102 dB: you have then about 40 dB more headroom than the numbers above, so this will give S9+88 dB...

[Dani EA4GPZ]

El 22/09/16 a las 19:00, in3otd escribió:

> in many bands in urban locations. In quite rural
> locations the noise is even lower. With this RF noise floor, a signal
> stronger than S9+36.2dB will raise your noise floor due to quantization
> if you use 10bits. If you use 12bits, it's S9+48.24dB.
>
>
> nope, you need to raise the signal from the SSB (or CW) channel noise
> floor, so 10*log10(2*2400/80e6) + 60 dB = 102 dB: you have then about 40
> dB more headroom than the numbers above, so this will give S9+88 dB...

Hi Claudio and Steve,

Thanks for the correction. I forgot to account for the fact that the
quantization noise gets spread over all the bandwidth of the ADC.

That explains why these 8bit RTLSDR perform OK in many situations.

73,

Dani.

[James Ahlstrom]

Hello Alan,

On Thursday, September 22, 2016 at 1:59:21 AM UTC-4, Alan Hopper wrote:

On bands with lower noise all bits are used 

Can you really measure the maximum or average number of bits the ADC is using?   It would be fun to know the signal level.  At my station and with my early HiQSDR, I know whether the ADC is clipping, and I can see the level of signals in the band, but I can not measure the sum of all the signals in the ADC.  Some firmware can return a block of all the ADC samples so we could calculate this.  It may be useful for the firmware in the FPGA to return some measure of the ADC bits used.  My firmware does not do this.

Jim

N2ADR

[Steve Haynal]

Hi Alan and Jim,

It would be interesting to directly see the ADC bits and do some analysis. The current Hermes-Lite firmware does produce bandscope data, which are blocks of samples directly from the ADC. N1GP's port of cuSDR and John Melton's Android software use this data to present a complete spectrum view of activity which is updated periodically. You can find links to their software on the software wiki page:

https://github.com/softerhardware/Hermes-Lite/wiki/Software

73,

Steve

KF7O

[Alan Hopper]

Jim, Steve,

I use the wideband packet from the radio that, as you say, has a block of raw adc values. My software displays the peak of these values and I use them in my rf agc to guess if there is enough headroom to increase the gain.  They are a small sample so often miss the overloads so I use the overload flag as well. I agree some firmware measure would be useful.

I've had a play histogramming this data at low gain to try and get some idea as how to control/choose an acceptable level of clipping, I'll post the data sets later, some adc non linearity is quite visible.

 On my panafall display if you zoom out I fill in the sides of the display with the wideband data so you eventually zoom out to the full 30+ Mhz, It is far from seamless but it is my aim to make it so.

73 Alan 2E0NNB

[Alan Hopper]

Jim, Steve,

the attached file has a histogram of the adc data collected over 20 seconds or so. The adc overflow flashed a few times but was not caught in the data. The second graph is zoomed in and shows an obvious pattern every 4 samples. Either I have some extraordinary local noise or the bottom 2 bits of the adc are not the same size.  Both my radios show the same thing. I believe this normal, just interesting when it shows up.

73 Alan 2E0NNB

[in3otd]

Hello Alan,
I agree that in practice some clipping should not be too harmful; if you do a test just using a sinewave and make the ADC clip the error (the "clipped part" of the sinewave) will be periodical and show up as harmonics of the input sinewave and aliases. But in practice the input signal will be random, so the clipped part will also be and its power spread over the entire ADC BW, more or less as it happens with the IMD and the (incidental) dithering. IIRC, this was also the point of a presentation of Gerald Youngblood, K5SDR, at Friedrichshafen this year.

This reminds me of the NPR test, where random noise is applied to the ADC input until the clipped noise makes the noise floor rise above the quantization noise floor. But looking closely at the numbers, there is something I do not understand - I'd be glad if someone could check my math here:
for a 12 bit ADC the above linked paper says that the peak NPR is reached when the "RMS loading level" is -14 dB; this means that the noise standard deviation sigma is 10^(14/20)=5 times below the ADC peak input voltage, V0. So this means that the ADC clips the signal whenever the (gaussian) noise happen to have an absolute value > 5*sigma. Computing the probability that the noise is amplitude is > 5*sigma or < 5* sigma using the Cumulative Distribution Function for a normal distribution in GNU Octave with 2*(1-stdnormal_cdf(5)) gives 5.7e-7 which seems pretty small. If the ADC takes 80 Ms/s this means that in 1 second 80e6*5.7e-7=43 samples are clipped. May be a reasonable number, I have no idea.
The problem arises when looking at the higher-resolutions ADCs: a 16 bit ADC has a peak NPR for an "RMS loading level" of -15.47 dB which, repeating the computations above, means 0.23 clipped samples per second, or about one sample clipped every 4 seconds. It seems quite strange to me that in this case just one clipped sample every 4 seconds can cause the noise floor to rise... there must be an error in the calculations above...

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Claudio,

I've just started trying to understand clipping.  From the perspective of operating a ddc radio or writing agc rules it would be great to be able to relate some measure of clipping level to snr effect. I've found a few articles specific to very high speed optical transmission with modulation that has a high peak level to mean level that run at a high level of clipping down a private channel but nothing that relates to our situation.  As a first step at trying to measure clipping without modifying the firmware I have been  looking at the histogram from the adc  which looks like a normal distribution, by counting out from the middle of the histogram you can measure something resembling the standard deviation and hence predict the level of clipping i.e. if 68%(one standard deviation) of the samples fit in the middle third of the adc range then 0.1% of samples are probably clipped, I think this is a more robust measure than the the few clipped samples in the sampled wideband packet .  I've got no further than a displaying this measure and a short test that indicated that the overload indicator is over cautious.  My Weekend reading is to follow your link.

Thanks

73 Alan 2E0NNB

[Sid Boyce]

Hi Jim,
I haven't found the post where you wondered a while back if quisk also
supported the HPSDR Hermes.

The Hermes board is supported. The companion ALEX filters/eb104.ru BPF's
and LPF's obviously are not and their settings are different but both
use the ULN2003.

PTT activates.

That's all I tested so far.

73 ... Sid.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[Steve Haynal]

Hi Alan,

Thanks for the data! I had forgotten that you were looking at the bandscope data. Was this on a noisier band with the gain set to 6 to 12 dB? Also, what is you plan for multiband skimming? I assume you'd set your gain according to the quietest band, which will most likely always push the global gain setting up?

73,

Steve

KF7O

[in3otd]

Hello Alan,
interesting that some of the ADC non-linearity can easily be seen in the histogram; maybe it could be possible to correct the ADC transfer function to improve its performances. I've seen a few articles around, one is here, which shows that this seems possible. Just in case you are lacking an interesting project for the upcoming long winter evenings, hi.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Steve,

I think the histogram was grabbed with fixed gain around +24db as it was very occasionally clipping, the chosen band obviously does not affect the data.  As you say for multiband I take the quietest which often ends up being just like the max gain approach so it all comes to nothing:)

73 Alan 2E0NNB

[Steve Haynal]

Hi Alan,

Yes, the chosen band obviously does not affect the data. But your chosen band for RX listening has been affecting your gain setting. :) 

73,

Steve

KF7O

[Alan Hopper]

Claudio,

I had also wondered if it could be corrected, a fun fpga project for someone. With the dfc approach it might be possible to correct on the pc. Thanks for the link.

73 Alan 2E0NNB

[in3otd]

yep, a fun project for sure... Looking quickly at the article it seems just a look-up table after the ADC can give a significant improvement. Of course all the article is about how to determine the LUT coefficients... IIRC, the RX path after the ADC is still 14 bit, so we could easily add a correction of a half or quarter step.
Ah, so many interesting projects and so little time, hi.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Claudio,

I'm hooked, the harmonic cancellation code in the current firmware already has a mechanism to download the correction table, the lut is just like the harmonic stuff.  Just a simple matter of calculating the values :)  One advantage of the a 12bit adc over a 16bit one is these look up tables are a manageable size.

73 Alan 2E0NNB

[in3otd]

yes, this is what I was thinking about also, this looks quite similar to what you already put in place for the harmonics cancellation... just the "simple matter" of computing the right LUT values remains to be done, hi. I was wondering if a simple optimization running on the PC side, to e.g. minimize the IMD level with a two tone (or noise) input could do.
But, well, did not really want to distract you from the many things you are already working on, hi.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Claudio,

I wonder if a long term noise histogram  can simply be used, the ratio of expected bin count to actual gives a measure of the relative width of each adc value.  The obvious pattern in the lower two bits intrigues me.  Will we also end up correcting the pga as well or is its contribution not significant, if so might we need a table for each gain setting?  

73 Alan 2E0NNB

[Steve Haynal]

Hi Alan and Claudio,

This is an interesting observation. It would be great if you could clean-up the ADC transfer function. You are probably on the right track, but just wanted to throw out a few more ideas. If you look at the 2 least significant bits, they appear to want to stay at 0, especially the LSB. I am wondering if we are hitting some timing path here. The LSBs may be the most delayed bits and the path to the FPGA runs at 2x the clock frequency in full duplex and via some possibly crazy route through the connector. Do you see the same behavior with your CV and CVA9? Does heating the FPGA or ADC make any difference? Is there any difference with the half duplex firmware where the FPGA/ADC clock rate is halved?

I'm also wondering if this is some noise pickup from the interpolated clock or interpolation process for the TX in the same package, but don't see how that would lead to the pattern you are seeing.

73,

Steve

KF7O

[Alan Hopper]

Steve,

I did see a very similar pattern (just eyeballed) both on the non duplex CV and the duplex CVA9, I'm away this weekend so have not had a chance do any more tests.  You are right there could be a number of causes and I may be jumping to conclusions.  The errors do look to be in the right ballpark for just adc imperfection.  I'm interested to try the same on the Orion.

73 Alan 2E0NNB

[Steve Haynal]

Hi List,

Below is the latest picture of the HL2 pcb layout. Some comments:

• The routing is essentially done. Only 17 signals  (14 digital, 3 RF) plus some powers and ground leave and continue to the right.
• There is still silk screen cleanup and some double-checking and tweaking for DRC and EMI.
• This was a lot of work to do by hand! It is pushing the size limit of what I'd want to do with KiCAD.
• Because the increased number of voltage supplies (3.3, 2.5, 1.2, plus several islands of each voltage) on a single inner layer of a 4 layer board would cause many power plane discontinuities, and because I want to make best use of the decoupling capacitors on the back by routing power->capacitor->via->IC power pin rather than power->IC power pin->via->capacitor, I decided to route the power on the outer layers and have two inner ground planes. Rob KL7NA built a fork using this PCB stack-up and there were no detrimental effects from this change.   
• There may be too may decoupling and bulk capacitors in this layout, but I'd rather have footprints for them and leave them off later if not necessary.
• This week I will work on the OPA2677 layout, staying close to what Jim did.
• There are optional items for the middle space (slow ADC for SWR/PWR and temperature measurements, digital bias for PA, bus expander) that I am considering for the middle.
• Space for the 10V supply for the OPA2677 is reserved. I am considering making this a dual layout. One can build with a high current switching supply here that can then be used for the 5W AirFast PA that Claudio tested, or a simple LDO if only low current is required.
• I am experimenting with a filter layout that I believe will fit and provide 6 LPFs for TX on a 10x15 cm board. I will share this soon. My first choice is still to have everything on a single 10x15 cm board with good optional connections for external filters and PA if desired. (This option would require the builder to just leave off the onboard filters and PA.) I will forgo the cut at 10x10cm if I think I can fit everything I want on a 10x15 cm board.

73,

Steve

KF7O

[Alan Hopper]

Steve,

that's is looking good, I've only ever done trivial layouts and they have taken me ages so I have a deep respect for the time and effort that must have gone into this.  If you can fit 6 tx filters in that's great, if not, I have warmed to the idea of the optional rx filter board also having any missing tx filters and piggy backing on in such a way the main board rf connectors are still used for the external connection.

73 Alan 2E0NNB

[Graeme Jury]

Hello Steve,

You are doing amazing work. Like Alan I have only done relatively superficial boards and appreciate your effort to date. It will be no mean feat to get the PA and 6 filters onto the remaining 100 x 50 mm space and if not achievable I too am comfortable with continuing TX filters and placing the RX filters on a matching 100 x 150 board. I think that anyone wanting to build with an outboard filter board which may or may not incorporate the PA won't be worried about the unpopulated 100 x 50 mm portion of the main board and I definitely agree that the cut at 10 x 10 cm is not necessary as it would all be going in a case at least that size with extra height anyway.

I am wondering what your skimming requirements for RX filters are Alan. Are you wanting to run HP RX filters in conjunction with TX filters and do you want to be able to TX with any RX HP/TX LP combination plus would you want to work full duplex? These 3 questions greatly affect the filter control signals.

• To TX on the LP filter band with any lower frequency or no HP filter is fine with the present setup.
• To TX on a band lower than the LP filter is set to, means that the TX filter will need to be switched on the fly to provide the harmonic suppression. The same will apply it TX freq is above the LP filter's band but is an unlikely scenario.
• To operate in full duplex would either require separate RX LP filters or be limited to transmit only on the band the TX filter is set to.

Diode switching of the filters will allow on the fly switching.

I am wondering if you are still thinking of a 3 TX and 3 RX lines for 3 of 8 decoding to get 8 TX and 8 RX filter control lines off 6 "J16" outputs?

Great progress, 73 Graeme zl2apv

[Florin Tanasă]

Hi,
I saw on schematic în PA sheet exist 7 filters with 14 relays HK4100. I think is possible to use a half from this somethink like this http://www.tme.eu/ro/details/im06gr/relee-electromagnetice-miniatura/te-connectivity/2-1462037-3/
Or http://www.tme.eu/ro/details/rsm850b6112851012/relee-electromagnetice-miniatura/relpol/rsm850b-6112-85-1012/
This relays is bistable/latching so we can make some economy to power this and we powered only one coil and have double contacts în comparation with HK4100

73, YO8TFF

Trimis de la: YOGA Tablet 2-1050L

Pe 26 sept. 2016 09:01, Steve Haynal <softerh...@gmail.com> a scris:
>
> Hi List,
>
> Below is the latest picture of the HL2 pcb layout. Some comments:
> The routing is essentially done. Only 17 signals  (14 digital, 3 RF) plus some powers and ground leave and continue to the right.
> There is still silk screen cleanup and some double-checking and tweaking for DRC and EMI.
> This was a lot of work to do by hand! It is pushing the size limit of what I'd want to do with KiCAD.
> Because the increased number of voltage supplies (3.3, 2.5, 1.2, plus several islands of each voltage) on a single inner layer of a 4 layer board would cause many power plane discontinuities, and because I want to make best use of the decoupling capacitors on the back by routing power->capacitor->via->IC power pin rather than power->IC power pin->via->capacitor, I decided to route the power on the outer layers and have two inner ground planes. Rob KL7NA built a fork using this PCB stack-up and there were no detrimental effects from this change.   
> There may be too may decoupling and bulk capacitors in this layout, but I'd rather have footprints for them and leave them off later if not necessary.
> This week I will work on the OPA2677 layout, staying close to what Jim did.
> There are optional items for the middle space (slow ADC for SWR/PWR and temperature measurements, digital bias for PA, bus expander) that I am considering for the middle.
> Space for the 10V supply for the OPA2677 is reserved. I am considering making this a dual layout. One can build with a high current switching supply here that can then be used for the 5W AirFast PA that Claudio tested, or a simple LDO if only low current is required.
> I am experimenting with a filter layout that I believe will fit and provide 6 LPFs for TX on a 10x15 cm board. I will share this soon. My first choice is still to have everything on a single 10x15 cm board with good optional connections for external filters and PA if desired. (This option would require the builder to just leave off the onboard filters and PA.) I will forgo the cut at 10x10cm if I think I can fit everything I want on a 10x15 cm board.
>
>
>

> 73,
>
>
> Steve
>
> KF7O

[in3otd]

Hello,
I have measured the SKY13374-397LF switch at 10 MHz and 30 MHz and while it works a little better there, still is far to being able to pass 39 dBm without too much attenuation and distortion.
I have also measured the noise it injects while used in RX and it seems quite noisy; some spurs may be due to the power supply used, but for sure there are several spurs at low frequency generated by the switch and some are above - 100 dBm, so quite noticeable.
Usual graphs at http://www.qsl.net/in3otd/ham_radio/SKY13374-397LF/SKY13374-397LF.html .

73 de Claudio, IN3OTD / DK1CG

On Tuesday, September 20, 2016 at 9:00:53 AM UTC+2, in3otd wrote:

Hello Steve,
no, I didn't test at 500 MHz, mine was only an assumption, thinking that if the device was good for 39 dBm below 500 MHz they would have stated that, as they did for the gain, which is specified starting from 300 kHz.
I do not have a power source at 500 MHz, I can maybe get 5 W or so in the 70 cm band, or a higher power in the 23 cm band. I can anyway try at 30 MHz and see how the device behaves there.

73 de Claudio, IN3OTD / DK1CG

On Tuesday, September 20, 2016 at 5:50:03 AM UTC+2, Steve Haynal wrote:

Hi Claudio,

Thanks as always for your measurements. Are you certain that the SKY13374-397LF can't handle 39dBm until the specified 500 MHz? I am wondering if the drop off actually occurs somewhere in the HF bands and if this device is usable for some HF bands. You could measure at 10M and see if there is any hope for that band and then if so, for lower bands.

73,

Steve

KF7O

[Steve Haynal]

Hi Florin and Group,

The HK4100s are attractive for two reasons. First, they are very low cost. You can find them on e-bay or aliexpress.com for under USD $0.50 a piece. Second, they are SPDT which can result in some very clean layouts as seen in John's PA/filter design.

Thanks for the links. I've been looking at DPDT relays recently also to see how many filters can be fit on the HL2 board. In particular, the NA3W-K and EC2-3NU. These are available at DigiKey and Mouser for those who prefer, but are also readily available in China for a comparable price to two HK4100s. The DPDT relays take less space, but I haven't yet found a layout that will quite fit the space I have. Below is a picture of some layouts I am thinking about. These layouts have the filters in series. You activate the relay to insert a filter. The relay is a short to the next relay when inactive. This is different from the filter selection on the current PA/filter board where one of several parallel filters is selected.  

Also, instead of a single ULN2803 to drive all relays, I am considering something like the NUD3124 to drive the relays. The main reason is that is should be easier to find room for 6 or 7 small devices versus 1 large device.

I haven't thought through the control or expense of latching relays, but do agree they save power. Another advantage of the DPDT over SPDT is that there is lower power, although power is still required when engaged.

73,

Steve

KF7O

On Monday, September 26, 2016 at 10:55:11 AM UTC-7, Florin Tanasa wrote:

Hi,
I saw on schematic în PA sheet exist 7 filters with 14 relays HK4100. I think is possible to use a half from this somethink like this http://www.tme.eu/ro/details/im06gr/relee-electromagnetice-miniatura/te-connectivity/2-1462037-3/
Or http://www.tme.eu/ro/details/rsm850b6112851012/relee-electromagnetice-miniatura/relpol/rsm850b-6112-85-1012/
This relays is bistable/latching so we can make some economy to power this and we powered only one coil and have double contacts în comparation with HK4100

73, YO8TFF

Trimis de la: YOGA Tablet 2-1050L

Pe 26 sept. 2016 09:01, Steve Haynal 
>

[Steve Haynal]

Hi Claudio,

Thanks for the measurements. Too bad they appear unusable at any HF frequency. Did you ever do similar tests for the PE4259 on receive? Whatever happened with the AirFast 7.5V devices when running at 12 to 13.8V? Did you manage to fry one? :)

73,

Steve

KF7O