Version 2.0 Architecture

[Steve Haynal]

Hello List,

I have created a wiki page describing the Hermes-Lite version 2.0 architecture. Please post discussion regarding v2.0 to this thread. This thread is sticky and will be displayed at the top of the GUI interface. 

73,

Steve

KF7O

[f1v...@orange.fr]

Hi Steve.

It looks really great and love the form factor which makes it easy to case up.

I will definitely be looking forward to owning one.

73 Peter F1VKK

[Joe]

Hi Steve,

What about ADC input for SWR and power output as used on HPSDR?

Joe  wa9cgz 

[Simon Brown]

Here's a video of the software I hope to have working with the ANAN-10E and possibly Hermes-Lite later this year:

https://www.youtube.com/watch?v=cXOtYZymn3s

Simon, G4ELI
http://sdr-radio.com

[ik1xpv]

I like this clear and flexible architecture. It looks great. 

Bravo!

Maybe the LVDSIO could also be used to synchronize sampling in between two or more boards so that diversity/MIMO could be managed ?

73

Oscar

IK1XPV

[Rick Koch]

That looks very nice. Do you plan to support multiple rcvr slices?

[Simon Brown]

Yes,

 

I have an ANAN-10E here which I’ll get fully supported first, so I guess that by them Hermes-List should be quite straightforward.

 

Simon Brown G4ELI
http://sdr-radio.com

--
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.

[Kaj Wiik]

Hi!

Very nice design!

73,
Kaj OH6EH

[John Williams]

Simon,

Do you have transmit functional yet?

John

[Steve Dick, K1RF]

Hi Steve.  The architecture looks great.  I have been holding off purchase while waiting for version 2.0 to come along.  One question: In the configuration which uses the PA companion board in a sandwich, will there be a recommended case that fits the sandwich?  Many will want to use the sandwich configuration and a relatively low cost case that the two boards can be used in would be a plus.

"Digital Steve", K1RF

[Steve Haynal]

Hi Joe,

I'd like to use the TX->attenuator tap->RX option for power and SWR measurement. I did something like this with pysdrvna. It costs less than the HPSDR and is more versatile as you can do simple antenna analysis and spectrum analysis of your TX signal too. I will test these ideas with a new v1.22 frontend card design, which will be available to others too, before finalizing them for v2.

73,

Steve

KF7O

[Steve Haynal]

Hi Simon,

Very nice. Thanks for sharing. I look forward to possibly using your software with a Hermes-Lite.

73,

Steve

KF7O

[Steve Haynal]

Hi Steve,

I haven't found a case yet of the right size, but haven't looked too hard. I'd appreciate any pointers if your or someone else finds one. It would have to be a tall 100mm front by 80mm side extruded aluminum case.

73,

Steve

KF7O

[Rob Frohne]

Hi Steve,

Hammond has some cases that fit 100mm wide PCBs, but the standard lengths are not 80mm.  Usually if you want a really nice looking finished product, the PCB dimensions are driven by the enclosure, at least somewhat.  I didn't look for more than a few minutes myself.  The box can also be a significant cost.  You might want to think about looking for an inexpensive nice box a bit before finalizing on the PCB dimensions.  Looks on the version 2.0 are probably a more significant issue than the previous versions.

73,

Rob
KL7NA

--
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.

-- 
Rob Frohne, PhD, PE
Professor
EF Cross School of Engineering
Walla Walla University
100 SW 4th Street
College Place, WA 99324
(509) 527-2075

[Steve Haynal]

Hi Rob,

There is a 100mm by 80m case that I linked to on the wiki already that is perfect for the standalone unit. I have 3 of them already. The issue is finding cases for the v2.0+PA pair. Flat requires 100by160mm, sandwich requires tall 100by80mm. People can help by tracking some down. Look on www.alibaba.com.

You haven't responded to my private offer for a v1.22 upgrade. Are you going to take it or not?

73,

Steve

KF7O

[Simon Brown]

Hi,

 

Some info: http://sdr-radio.com/Software/Version-30

 

Simon Brown G4ELI
http://sdr-radio.com

 

From: herme...@googlegroups.com [mailto:herme...@googlegroups.com] On Behalf Of Steve Haynal
Sent: 11 August 2015 05:00
To: Hermes-Lite
Subject: Re: Version 2.0 Architecture

 

Hi Simon,

--

[Steve Haynal]

Hi Steve and Rob,

Element 14 has other cases by Box that will work for the flat and sandwich orientations. I've updated the wiki. The sandwich option is 1.77 inches tall and may be a tight fit as the PA is not designed, but it is a good target to shoot for. See:

http://www.newark.com/webapp/wcs/stores/servlet/Search?catalogId=15003&langId=-1&storeId=10194&st=categorynumber84657&pageSize=25&showResults=true&pf=811236420

73,

Steve

KF7O

[Rob Frohne]

Hi Steve,

I hadn't noticed your link.  It looks like you might be able to fit two boards in it with the sandwich setup if components don't stick up too far.  I'm sure you have looked at that, and it may not fit because of the Ethernet connectors, etc., but it looks like they have the "slots" there for the boards.  If not, there is one that is 1.7" tall instead or 1.18" that might work here.

There is also this B2-160 that looks about right for the side by side.

Thanks for the offer of a v1.22 board.  I think another more active developer would benefit more from it, as I explained a few minutes ago in a private email.

73,

Rob
KL7NA

[Heikki Ahola]

Dear Steve,

Really nice job ! Looking forward to the V2.0 kit, although I need to finish my V1.22 setup first, hihi. 

73 de Heikki (OH2LZI)

[Steve Haynal]

Hi Heikki,

How is the build going? Is it a matter of not enough time, or have you hit some problem?

73,

Steve

KF7O

[Steve Haynal]

Hi Oscar,

Thanks for looking at the design. Yes, I would like to be able to chain boards together, including older v1.2x with BeMicro CVs, via LVDSIO to have one synchronous clock and data transfer to do things such as diversity/MIMO and beamforming. I plan to test this with several BeMicro CVs I already have before finalizing v2 so will let the list know how that goes. I don't know how many independent receivers will be possible yet. Sorry, SDK owners, but I can't find any available LVDS IO on that design. If this works, people may be able to use the BeMicro MAX 10 with v1.2x boards as it has LVDS.

73,

Steve

KF7O

[Heikki Ahola]

Dear Steve,

Thanks for your concern !  It is still summer time (= temps above 20 C !) in Finland and my activities are quite different then. As the autumn winds/rains will arrive, I should have plenty of time for ham radio also. In case of some unexpected problems, I will notify this group.

73 de Heikki (OH2LZI) 

[John Williams]

Steve,

Will this allow the real time sensing of forward and reverse power downstream of the v2 board? The board setup that Joe is using senses the output of his 20W amp and feeds the voltage data to the F-R-G connector on the ADC daughtercard. It also feeds a separate connector with the direct signal from the Hermes Lite and the ADC daughercard processes the signal into a voltage. So it provides dual power readings - drive and final, and for final also provides reverse power. And it is a real time power display while operating. 

John

[Steve Haynal]

Hi John,

What I'd like to see happen is that a companion PA card use a similar tap to that which will be seen on the Hermes-Lite v2.0. This tap will have to attenuate more, but I think it can live permanently connected to TX with negligible impact. Once attenuated, the TX signal can be connected to the RX input via a small Peregrine switch. The firmware can then switch between the tap on the Hermes-Lite or the PA several times a second to take multiple readings. This is what the SPI ADC does -- time multiplexes several analog inputs to one actual ADC. Since we will be using the fast AD9866 ADC, we can take complex phase measurements for forward power, reflected power, drive power, final power, SWR, antenna analysis, VNA reflection measurements. I see this as a win as we get all the functionality you describe plus more at a lower price point.

I need to build a new frontend card that tests these ideas with a v1.22 first. I'm sure I will learn a lot in the process. If things work out horribly, there are the 4 extra FPGA signals sent to the companion PA card. These can connect to the SPI ADC you have been using. This requires the ADC to be on the PA card.

The PA companion card is the most undefined part of this project. Given the size of this relay and the size of the qrp labs LPFs, rough calculations indicate we can fit 6-7 filters plus the PA on a 100mm by 80mm board. This assumes a relay at each end of the filter and layout similar to this. We can't use qrp-labs style daughter boards with the sandwich configuration as I don't think there is enough vertical room. With the flat configuration, we can use qrp-labs style daughter boards to gain space. Also in the flat configuration, the PA companion board can be longer. If one really wants to make something fancy, they can define their own size of PA board (120x180 for instance) and have the Hermes-Lite be a daughter board.

I am also worried about heat sinking. I would like the RD16HFF1 to use the extruded aluminum enclosure as a heat sink. It needs to be placed to support this. This may heat up the whole radio too much and cause drift problems with the oscillator. Another option which would reduce the heat effects would be to have the Hermes-Lite and PA as two separate boards in separate enclosures.

What are your thoughts?

 

73,

Steve

KF7O

[John Williams]

Steve,

I now understand your approach on power sensing, and will help out as needed. Please feel free to propose a schematic/design concept and I can work on a board.

As to the second point, I am contemplating morphing the superband into a pa that can hold a single qrp-labs filter, or extend itself with a sandwiched QRP-Labs relay board for support of 6 filters. The RX side will have only 1 30MHz lpf. This should allow us to test the heat generation as a concept for a final board config. We can place it in a enclosure, and put thermal sensing in the enclosure to see the heating effects. I am going to drop the PCI-e connector and drive it with my Megabasic frontend that has rx/tx baluns and a relay driver for J16 signals for pre-v2 users, and allow it to mate up with a ribbon cable for v2 beta testing. 

This may prove to be a useful testbed.

On the filters, I have had some limited time to play with the VNA and my current filters. I find that they are not sharp enough to handle the second harmonic in a consistent manner. I think moving to the QRP-Labs design is a overall better idea. With the extra inductor and cap, the filter is much sharper on the slope. In a 6 filter configuration, you can pick and choose which bands to support easily.

My 2 cents...

John

[John Williams]

Starting new thread...

Steve - making a footprint - Are you wanting to use this relay - HK4100F-DC5V-SHG small signal Relay 5V 5VDC, or on the one used in PicAStar?

Following your direction on LPF configuration and relay wiring....

Do you want to use the 12V version of this relay so we only have to have 1 power rail, or do you want to bring 5V into the card in addition to the 12V already there?

John

[John Williams]

Steve,

Does this circuit embody your concept for the PA?

John

[John Williams]

Life of a hobby farmer ... as I was cleaning out the chicken coop it occurred to me that I messed up the relays. They need to be reversed to ground the filter,  not the TX rails. Big mistake.

Also, need a bypass if none of the bands are asserted and need an optional BC band highpass...

John

[John Williams]

Also, if we bring in 5V can replace the regulator with a transistor...

[ZL2APV]

Hi John,

Looking good and OK on the relay, you are not the first. Here I run the 30 MHz filter permanently connected to the Tx out and the filters for the other bands get switched in series. I get such strong Broadcast signals on AM that I also have to have a 1.7 MHz HP filter permanently in circuit also.

Regardless of whether or not the 5 V reg is on board, I switch the PA bias with an fet so the amplifier does not have any possibility of producing noise when on Rx.

I am currently working on switching my filter bank with GP10Y diodes which have a very long reverse recovery time and fairly low (relative) capacitance with encouraging results so far. Getting rid of the relays is good but other complications are being introduced so I am still not sure where this will go.

I am still not sure about single fet class A vs push pull class AB amplifiers. The extra complexity of the push pull amplifier is traded off against the lesser requirements for the filters in terms of second harmonic rejection. The half amp standing bias current of the class A does not figure as a factor as on Tx, as it is only a small part of the total current after a high power amplifier is added. It's just that the extra 20 dB of 2nd harmonic reduction is a powerful factor.

I go along with your thinking on the qrp labs filters. They are already there, they are cheap and they work. Pity about the head room and having to pay the price of an extra relay for each filter in terms of board real estate but there are advantages as the transmission discontinuities are less with the twin relays especially if the feed to the contact is via a printed transmission line. In my own case rather than a printed transmission line I used a piece of RG316 and (carefully) tapped it on to each relay's appropriate contact with the highest frequency filter connected to the furtherest end of the line so the shortest unterminated stub existed. It makes a difference.

73, Graeme zl2apv

[John Laur]

Hi Steve,

I have been carefully reviewing your 2.0 notes. I think you have an
extremely solid and achievable approach planned. I am very excited to
see the Pereguine switches enter into the design!

From a general overview it looks as if the design might end up being
pin constrained (Is that actually possible on an FPGA project!?)... Or
at the very least you may want to conserve LVDS pins, use a smaller
part, etc.

There are several opportunities where shift registers might be used to
conserve IO. Debug LEDs, DIP switches and even RF switching are all
potential candidates. In addition to being extremely cheap, register
outputs could be fed through to the built in RF board to handle PA
board rf switching, switch input (potential for PA board to indicate
supported bands), or debug LEDs on the PA board.

This might have several good benefits; one of the biggest is that J16
could remain an external interface instead of being required/consumed
entirely by the internal PA board. There are many experiments that
currently interface with J16: ALC, EER (Amplifier power supply
modulation via PWM), and CHIRP (GPS PPS input). I think ALEX is not
the best interface to control the PA, but I really like having J16
around as an external experimental interface.

Regarding the question of balun footprint, please consider a balun
footprint that is compatible with mass produced SMT transformers
(Minicircuits or similar). This would require adjusting the winding
procedure for hand wound transformers to have the primary and
secondaries emerge from opposite sides, but would allow for testing or
future manufacturability. It might improve mechanical attachment. as
well.

I was trying to do some research on how fast you can get a 0.100
header to go. I saw more than one suggestion that stated when you are
very careful with impedance/length matching they get all the way to
1GHz. It is probably best to try to get as many of the LVDS pairs out
to a connector as possible -- well at least 16.

Final question -- do you think there is any benefit to putting the
AD9866 stuff into a shielding can or building a footprint for such?
Are there even standard size (inexpensive in low quantity) RF cans?
This has always been neglected on the HPSDR boards, but some
experimenters have crafted such shielding for the ADC and DAC sections
and seen 3-6dB of improvement to MDS on Hermes and Angelia. Or perhaps
the internal noise of AD9866 makes that irrelevant?

Very excited - thanks for sharing the preliminary work with us!

73, John K5IT

[John Williams]

Can you propose a HP filter that is conservative on board space? My 30 MHz LP filter for RX is using some nice 1206 inductors. Can't do that at 1.7 mhz...

Thanks for the additional info. We are as we used to say in operating system development circles - "Peeling the Onion". Discovery is the fun part of the hobby.

I am worried about such an inexpensive relay. What is the MTBF and what are the RF characteristics of the contacts...

John

[Steve Haynal]

Hi John,

Thanks for the schematic. It does a lot to define a companion PA, is based on your past work, and is very much along the lines of what I'd like to see.

I like the price of the HK4100F-DC5V-SHG small signal relay, but it needs to be tested at RF and the current is a little high. I was hoping to test these out. Either the 12V or 5V is fine. The PA will supply 12V to v2.0 and v2.0 will make 5V and 3.3V available to the companion PA. I'm looking at the uln2803 which will handle the 7 band selects and PTT. I'd like to see it setup such that the companion PA specifies Vsupply of 5V or 12V depending on the relays. (OCREF on the Hermes schematic.)

I do like using 5V + transistor as 5V will be available to the companion PA. Also, 3.3V + transistor can work. 3.3V is need on the companion PA to power/switch the Peregrine device.

The 3dB pad is a nice option. A user can stuff a 0 Ohm resistor if they want no pad.

What is your thinking behind a separate filter for RX? I was assuming that one of the 7 TX filters would be LPF for 10M. The receive filter selection can always just use that filter. This would put the TR switch at TX out. Are you worried about the receive path going through 3 relays? If you use one of the TX LPF for RX, you could change your RX filter out for an optional broadcast HP filter.

It would be nice if the LPF filter layouts include surface mount footprints optional capacitors in parallel with the inductors so that an elliptical filter can be implemented as seen here if desired. 

A 4-layer board is probably best for the companion PA. The price is not that much more and you get a nice ground plane and easier placement and power routing.

I don't have a schematic yet, but the tap to measure final output power that I envision is based on the ASCII schematic in this thread. The "From Tee" side is a permanent connection to the final antenna connector on the PCB. The attenuation will have to be enough to protect the RX from the strongest signal. The "To Instrument" side will connect to RF1 of this device. RF2 remains unconnected. RFC connects to the RX input of the Hermes-Lite either through the HP filter if present or directly if not present. This is the same RX side of the TR switch. With this arrangement and another 3.3V signal from the FPGA, when in TX, the Peregrine device can connect the attenuated TX RF back to the RX for measurement. This is all done on the PCB, no external tap.

73,

Steve

KF7O

[Steve Haynal]

Hi John,

Thanks for your detailed and thoughtful feedback. It is appreciated.

I will keep the shift register idea in mind if I run out of FPGA pins, especially for the LED and DIP switch. There are only a few high-speed capable LVDS pins, so we may not be able to trade-off more LVDS for shift registers elsewhere. I did estimate FPGA pin usage and think we are okay, but it is good to have this option if pins become tight.

Maybe we shouldn't call this a J16 interface as it will only have the 7 band selects and PTT. 

A balun footprint that can support a SMT balun or or hand wound one is fine with me, but I need to have a specific part suggested. In v0.9 I used minicircuits but their baluns could not handle the TX current. Also, the reconstruction filter still uses hand wound toroids.

The bandwidth constraint for a LVDS pair will probably be due to the FPGA pin properties and not the header. I anticipate running the LVDS at twice the 73.728 MHz oscillator frequency.

Shielding the AD9866 is a nice idea. If we can find a shield with only four pins, one at each corner, we can add it to the layout. The routing density is high on the top, but four corner pins should be doable. I'd like help with people suggesting specific parts for this.

73,

Steve

KF7O

[ZL2APV]

My 30 MHz filter cannot be described as compact. I am using 2 coils wound on a 10 mm drill around 15 mm long and the caps are off the shelf polystyrene ones. It is a 5 pole with notches at x2 and x3 28 Mhz. It sits in the line from my antenna and is active on both Rx and Tx. The AM broadcast HPF is in the Rx leg and again external to the radio.

I suspect that the relay will handle the service very well. It would be marginal at 100 watts but I would trust it at up to say 30 Watts. You only need to look at the relays in radios like the TS-50 which go and go to see that small with RF can still work. The fact that the contacts are silver is good. MTBF is another issue and I am not sure how to get information on that from a practical installation.

Steve was looking at a power tap for getting some signal for pre-emphasis etc. and I guess it would need to be Hi Z at the Rx side as well to avoid loading the receiver input Z. This is also a good way to add another 20 dB loss to the device.

73, Graeme zl2apv

[John Williams]

Good observations -

thanks for pointing out the uln2803. I was trying to decide how to add the clamping diodes. This solves the issue. I will change my megabasic design to use this part and also will make the pullup selectable between 12v and 5v. That gives us a easy driver for 1.22 that can simulate the controls for pre-v2 testing.

on the 5v plus transistor - the amp needs a variable bias derived from 5V. For the Superband we used the regulator. If we bring 5V over from V2, we can forgo that part. I will take Graeme's suggestion to use a FET to switch it to insure no residual leakage. You will see this in the next revision of the schematic.

I was just using a prior approach for using a dedicated receive filter. Are you suggesting that both RX and TX go thru the LPFs and we switch between them on PTT?  I was trying to keep the filter paths short and separate to avoid excessive loss. Also to avoid excessive clicking and latency for quick RX/TX transitions. Think of the clicking going on during CW QSK breakin.

I am not opposed to making the 10M lpf inline on both RX and TX and using one of the relay positions for selecting a BC filter. Still need a design with easy to build inductors...  Let me see what I come up with...

I will investigate the tap...

John

[John Williams]

Steve,

On my test frontend revision, should we drive this direct from the bemicro fpga pins or should I buffer it with a 4050 part. It looks like the turn on voltage is 2.5 v min, and current is less than 1ma. I have been burned before by overdriving the fpga pins so want your thoughts...

On 8/15/2015 1:17 AM, Steve Haynal wrote:

[ZL2APV]

John, we need to be careful with the 4050. Both NXP and TI data sheets show Vih min as 3.5 volts for 5 V Vdd. Not sure which brand you are using but it would be important to get the same particular brand. It may be prudent to look at a purpose chip like the 74lvc4245.

73 Graeme

[Joe]

 As a daily user of the HL I would like to offer these comments.

The 30MHZ filter is required in the transmit path before any amplification

to minimize any aliases, in my case the filter is used in both the receive (roofing filter) and the transmit paths. I also find like others near AM broadcast stations a 1.7MHZ  HiPass is needed to prevent receiver overload, I copied the design from the original Hermes. 

 I'm concerned from what I've read that some current functionally is going to lost

with the new design I use  the J16 leads in a BCD reformat to switch final lowpass filters

and a 6 meter transverter using the SETUP options with HPSDR are these going to be lost?

What about remote SWR and Power output currently the FPGA has leads that support

the original Hermes circuit with board from John which uses a DAC to feed the FPGA  how is this going to work with the version 2.0? Also using some sort of feedback on board for Pure Signal would work with the PA board but what about people with outboard amplifiers?

   I'm concerned that trying to make a one size fits all HL version 2.0 may drop flexibility

 and features that we already have. I hope by going to a two board approach that what we have now is still accessible. 

    

Joe  wa9cgz

[John Williams]

Steve,

Will this LPF provide the pre-amplifier filtering that Joe is describing?

From the wiki => "This is a LPF for TX reconstruction and optional RX use. A LPF reconstruction filter is required for any digital to analog conversion, and the need is illustrated in this thread. The plan is for a simple 5 pole Chebyshev low pass filter as seen in the softrock SDR family with 3dB cutoff frequency in the low 30 MHz range. This will be a wind your own toroid filter."

Also, I have a design using AIML inductors for a 30MHz filter built on one of my boards. I can run it thru the VNA but it does not show much since the limit is 36.5 MHz. Jim Veatch suggested these inductors as having pretty good Q ... You may want to consider these instead of winding toroids for the low power TX path.  See attached datasheet.

John

--

[Alan Hopper]

Steve,

just a couple of thoughts:-

A simple serial io facility would allow flexible and virtually unlimited external interfacing.

A low power sleep mode with wake on lan would make my remote setup really slick.

73

Alan M6NNB 

On Monday, August 10, 2015 at 6:54:03 AM UTC+1, Steve Haynal wrote:

[Steve Haynal]

Direct from the FPGA. The ULN2803 will be on the v2.0 board, so no need for it on the companion PA. This is to protect the FPGA and support other PAs.

73,

Steve

KF7O

[John Williams]

ok on driving direct. My intention is to add it to a test frontend so I can simulate a v2 interface with a v1.22 config to test the pa controls... For v2 understood the chip will be on the main board.

We will need more control signals than 7 as is now. I need a ctrl (active high) signal for the PE4259.

[Steve Haynal]

Hi John,

Another possibility is to design the PA to interface to the new frontend card I am building to test v2 with v1.2x. This frontend card intends to match and test what v2 will have. From the architecture document, my goal is 4 extra control signals from v2 only for use by the PA companion board. These are unbuffered 3.3V. Given space will be tight, I prefer no extra logic and a very simple companion PA board. (7 filters only, RX shares TX filters, no relays except for the 7 TX filters, no BC HPF.) There will probably be several companion PAs so do what you like and don't feel constrained.

73,

Steve

KF7O

[John Williams]

Yes, I have 2 desires - one to test with, and a on-ramp for folks to still use their 1.22 boards with this new amp. This amp can replace my Megaband amp design for pre v2 users.

Will one of the 4 extra lines be used for controlling the 4259?

John

[John Williams]

Agreed. Can you define the connector pinouts so I have something to work towards?

My goal is to move the ball forward. I like the amp portion of what we have done. Filters have been a bust. Willing to take guidance on your thoughts and try out some of this as we iterate towards a final design. Don't care if I design the amp or someone else. Just want to keep things progressing.

John

On 8/17/2015 10:25 AM, Steve Haynal wrote:

[Steve Haynal]

Hi John,

Regarding a connector pin out, give me a week or two to design the v1.2x frontend that will resemble the v2.0. I think it would be great to test the amp with that as it will hopefully have the required connectors, although ribbon and coax will have to be used as the form factor won't be correct. People can then use this frontend with your PA even on v1.2x.

Yes, the extra lines can be used to control the 4259.

I don't think I can give the best guidance regarding RF and filter design as it is not my forte. Graeme appears to have expertise in this area and time to help. I tend to like design that is very simple and lean, even to the extent that others may think I am leaving off important stuff. A priority for me is that the companion PA fits the form factor, connects to the v2.0, and can be used as a driver for a larger PA or standalone QRP PA with 5-7 filters.

If you want to keep the ball rolling, you can create all the collateral like footprints, etc., that you will need. Sorry for the delay, but time is scarce for me and you too.

73,

Steve

KF7O

[John Williams]

I have ordered the relays so I can determine actual pinout and build some components and footprints. Datasheets are not specific on orientation to determine pin function.

John

On 8/17/2015 9:12 PM, Steve Haynal wrote:

Hi John,

If you want to keep the ball rolling, you can create all the collateral like footprints, etc., that you will need. Sorry for the delay, but time is scarce for me and you too.

73,

Steve

KF7O

On Monday, August 17, 2015 at 5:54:31 PM UTC-7, John Williams wrote:

Agreed. Can you define the connector pinouts so I have something to work towards?

My goal is to move the ball forward. I like the amp portion of what we have done. Filters have been a bust. Willing to take guidance on your thoughts and try out some of this as we iterate towards a final design. Don't care if I design the amp or someone else. Just want to keep things progressing.

John

[Steve Haynal]

Hi John,

Not sure if you've found the datasheet or how helpful it is but attached it to this post.

73,

Steve

KF7O

[John Williams]

I found that. I find it odd that they do not number pins nor show orientation on the diagrams. Guess the pin config controls orientation.

John

[Dave Miller]

Not unusual.  Just make sure your footprint and schematic symbol match.   It no doubt is a knock off of an existing relay.  This way they can't be accused of copying someone documentation. ;-)

Use normal pin 1 from top side.  Some people like to make it a 6 pin device.  Others would use a grid and make it a 10 or 14 with missing pins.   No numbers make is easy. :-)

Dave 

Sent from my iPhone

[John Williams]

latest schematic...

- simplified bypass per Graeme and Steve's suggestions. Required combining 60/40 filters.
- single fet bias switch
- pullup rpack and wiring for common signal from uln2803 to utilize clamping diodes.

Also have built footprints for the relays.

Graeme -

Can we do as good with a 5 pole filter as the current 7 pole filters? Would be lower parts count, esp the winding of toroids.

Here are the values for the 7 pole filters - http://www.qrp-labs.com/images/lpfkit/gqrplpf.pdf

Here are some 5 pole values - http://pa3csg.hoeplakee.nl/joomla25/index.php/rf-construction/84-5-pole-cauer-filters-for-the-hf-bands

John - W9JSW

On 8/18/2015 10:42 AM, Steve Haynal wrote:

[Steve Haynal]

Hi John,

I recommend keeping the 7-pole filters if layout allows. One can always leave off and jumper an inductor footprint to make it 5-pole. This goes for the capacitors in parallel with the inductors too. Since the 10M LPF is now used alone for 10M/12M TX, I think it should have the option to be 7-pole too as the intended PA is rich in second harmonics.

I like that the bypass option removes a relay. I wonder about the break in symmetry and layout challenges that may result.

73,

Steve

KF7O

[John Williams]

Steve,

I will do on the 7 pole filters. Will add some more places for the 12/10. I was thinking that a pair of relays would allow me to have them grounded when not in use, instead of letting both in and out terminating in one relay. If I can fit it I will add a second.

Can you outline how you organize the layers on a 4 layer board?

John

[ZL2APV]

Hi John,

Your new layout looks really good. You sure are putting a lot of effort into it. Picking up on Steve's recommendation of 7 pole filter layout, this is good advice.
More poles usually give better vswr numbers as well as sharper drop off.
5 pole Cauer filters often allow the notches to occur on 2nd and 3rd harmonic frequencies and can be better.
Using smd caps you could use only one footprint and if 2 caps in parallel to make up the value are necessary they can be piggy backed onto each other. With 1206 size caps this is easy to do and real estate gained.
If using smd caps be careful to get good high Q ones. Not all COG caps are necessarily good at RF.
Toroids will always give better filters than smd and will allow a small amount of adjustment but smd are OK.
Might be a good idea to invert the signal to the bypass relay so that it is "through" when released. It means that it will operate when any other band is selected resulting current for 3 relays but is another lever of fail safe.

I got a filter set from QRP Labs arrive in the post yesterday and will build them up and evaluate them today and post further information to you asap. Last night I started loading the filter values into Elsie and will post the files here when finished. I have 80, 40, 30 and 20 done so far.

73, Graeme ZL2APV

[Glenn P]

I must have missed a post here, but what is the logic in removing a relay (which is very cheap and bought in batches on ebay anyway) and using the Peregrine device?

glenn
vk3pe

[Steve Haynal]

Hi Glen,

The Peregrine device was also in the previous schematic. A relay was removed on the Band6 path as it is feed through now. The savings I hope for are in board area as we have a tight area budget.

The Peregrine device is for a tap to measure and inspect the output signal. Since we are using the highspeed ADC in full-duplex mode to do this, we can make complex measurements and do some antenna analysis, SWR, forward/reflected power, etc. measurements in the same style I did with pysdrvna. A relay is not used here as the cost for the Peregrine device is on the same order (60 cents), the higher power capability of the relay is not needed here, the Peregrine device is smaller and consumers less current, the Peregrine device is used on the v2.0 main board so a small quantity must already be purchased. The total cost for this optional tap is <$1 and very little real estate.

[John Williams]

Very clever, Graeme! I was trying to accomplish some sort of default
passthru if no lines were asserted. That is why I had the wide NAND
gate. This is much better!

John

[John Williams]

Here is the latest schematic, Graeme.

- annotated all filter values with components and toroid windings from
the QRP-Labs paper. Please verify against your elsie plots and suggest
changes if necessary. For instance, should we be using the T37-10 toroid
elsewhere rather than the T37-6? Forward final elsie plots (lct file) to
me so I can post them on github with the rest of the project files.

- made the relay change - thanks!

Also - please suggest brand/family on SMD caps to use. I usually use
Yageo caps like the ones posted in the super band BOM attached.
Additionally, for the superband I had to stack 805 caps to achieve some
values. Worked fine. I prefer to stay with 805 caps on this layout. It
is going to be very tight.

John - W9JSW

On 8/18/2015 5:09 PM, ZL2APV wrote:

[John Williams]

Steve,

I have a component and footprint for the 4259 on my github clone. Look there to bring over when you start your layout. The footprint package is SC70. I also have the HK4100 there.

Or I can email them to you.

[ZL2APV]

Hello John,

You are doing a great job and it looks like we are set to have a good reliably working radio suitable for QRP and WSPR etc. thanks to your efforts. I have attached the Elsie files so far and sorry but have not progressed with the rest. Its 2am and I am lying in the spare bed trying to avoid waking my XYL, can't sleep, alternating between shivering and freezing with some virus I have picked up.

Its a good idea to put the files on your github and that way they will be accessible to all. If I post them, the post will get forgotten in time.

I spent a fair bit of time yesterday testing and building filters and have posted the results in the filter thread The issue of signal feedthrough came up. With your setup using a relay at each end of the filter, you will have much superior isolation. Are you intending to use the filter boards from QRP Labs and solder or plug them into the PA board or will you lay out the filters on the PA board and simply use the components. The QRP Labs prices are so reasonable that it would be worth simply buying the filters required and at the end of the job there would be half a dozen nice experimental filter PC boards to play with. I will see if I can improve the filters with less components though and report back.

73, Graeme ZL2APV

[John Williams]

Pay special attention to the 60/40 filter. That is the only shared filter where the second harmonic is close to the passband. Rolloff will be critical.

John W9JSW

[John Greusel]

---

John,

I have completed building the QRP Labs LPF system with the relay board and filters for 40,20,17,15 and 10 meters.

I don't need it at the moment so I could send you the whole assembly if that would help you with your development efforts.

Being as we're both in IL it wouldn't take long to get there. :)

John

KC9OJV

[ZL2APV]

Here are the rest of the filters. On any that can be used on multi band, I have included markers for both bands.

ZZZZZZ
73 Graeme ZL2APV

[Glenn P]

Thanks John. Sorry to divert you from your good work.

glenn

[Steve Haynal]

Hi John,

Thanks! Github is perfect. I hope to start this weekend.

73,

Steve

KF7O

[Steve Haynal]

Hi John,

I would put RF/fast traces on the top/component layer, inner ground, inner power plane, not so critical traces on the bottom layer. This is a standard stackup. You can take a look at the Hermes-Lite v1.22 on github as it follows this stackup. There are a lot of online PCB trace/stripline calculators. You can control the impedance of the RF traces.

I am really excited about the BOM cost you posted in the $30-$40 range. I wanted to estimate the PCB costs but the elecrow site is down. I don't think going with 4-layer will be much more, and you will get better ground and easier layout.

73,

Steve

KF7O

[John Williams]

I think we will be fine on our current approach. Thanks for the offer. Will keep it in mind.

John

[John Williams]

Steve,

I have not yet posted a BOM prices for this design. Given that the
relays are dirt cheap, the cost should be quite good. Most of the
megaband cost was driven by the relays...

On another topic, I have a prototype push pull amplifier built up and
almost ready for power-up. It would be really nice to be able to
consider that for this project as well. Perhaps with the fall season
approaching, time will not be as precious... Adding another transistor
and a few components is not much more cost. Footprint and heat are
factors, however.

John

[Steve Haynal]

Hi John,

I sort of gleaned a price estimate from one of your posts that included a parts list. Elecrow is up and for 10cm by 8cm the cost in batches of 50 including shipping are roughly $4 for a 4 layer board and $1 for a 2 layer board. I think it is worth the extra $3 for the layout convenience, ground and density. I'd like to make these available on Tindie when ready too. I assume your design will be opensource?

Now that you have a preliminary filter layout, how much area do you estimate per band filter?

Good to hear that you are looking at push pull as well.

73,

Steve

KF7O

[Alan Hopper]

John,

this all looks great, I'm no rf engineer so can't be of much help but something just feels right to me about feeding the balanced output of the dac into a push pull amp.  There was a simple push pull design in the June issue of radcom that may be of interest.

73 Alan M6NNB

[John Williams]

Of course, open source. I have been pushing regularly to github...

Will get back to you on filter area. Graeme suggests I orient the middle toroid differently so I have to make that change and then measure.

John

[John Williams]

Steve,

The 6 filters consume approximately 41% of the 80 x 100 board space. They will occupy a 65mmx55mm space assuming a 1mm gap. If space allows I will go larger on the inter filter spacing to allow a metal shield option.

The push pull design is based on the PA-2008 amp, removing the first stage and using the balanced output of T1, the transmit balun, to drive a pair of RD16HHF1 mosfets. Will get back to that sometime later this fall, hopefully. If anyone wants to take that on, I am fine with that. I am including a snip of the circuit. I think it is close to this, but not guaranteed.

John - W9JSW

On 8/20/2015 11:24 PM, Steve Haynal wrote:

[Steve Haynal]

Hello List,

I've updated the Hermes-Lite version 2.0 architecture wiki page with more information on the power regulator plans. You can always go to the top of a wiki page and click on the "N revisions" link and compare past revisions to see exactly what has changed.

The update includes some part selection. Although we have not had noise issues with the prototype power supply, some of these new family members allow you to set the switching frequency. A switching frequency of 2.98 MHz will have no harmonic inside an amateur radio band. It would be nice to adjust the switching frequency from the FPGA...

73,

Steve

KF7O

  

On Sunday, August 9, 2015 at 10:54:03 PM UTC-7, Steve Haynal wrote:

Hello List,

I have created a wiki page describing the Hermes-Lite version 2.0 architecture. Please post discussion regarding v2.0 to this thread. This thread is sticky and will be displayed at the top of the GUI interface. 

73,

Steve

KF7O

[Steve Haynal]

Hello List,

You can find the initial schematic for the new frontend to test V2 features here. All of the files are on github as before. I hope to finish PCB footprints this week and start PCB layout new weekend. Some notes:

• There is an extra RF input switched by a PE4259. This is to remove the burden of switching the tap signal in and out from the PA/companion board. Any PA just needs to supply a properly attenuated signal to the specified RX input for measurements and the other RX input will be reserved for unaltered receive. No 3V or switching signal needs to be sent to the PA now, and this should work with other PA designs as well. This is optional an may be left off for a budget build.
• All RF connectors are listed SMA/BNC/HDR. I actually spent quite a bit of time this weekend working this all out. There is a triple use footprint that accommodates board edge SMA or BNC connectors as well as a single inline header for direct connection to a companion PA board in either the extension or sandwich configuration described in the V2 architecture. This will hopefully become clearer once the footprint is posted next weekend.
• 8 buffered signals are supplied to the PA companion board. The PA specifies what voltage level to use by driving COMR at the common reference. I had thought to make the ULN2803 bypassable, but those wanting to decode signals into more selects can just drive COMR to +5V and decode on a custom PA/filter card.
• 3.3V and FPGA signals will not be supplied to the PA. This is a great simplification and will enable more generic PAs to be used as no dependency on 3.3V or FPGA signals will sneak in. 
• The new frontend needs 4 FPGA signals that are not currently available on V1.22 hardware. RXCNT will connect to V1.2x CN3 or a CV to provide these signals. 
• C7 is optional given Jim Ahlstrom's experiments.
• Vcm can be 3.3 or 5V or something less if a resistor is stuffed in R2.
• The footprints for T1 and T2 support popular sizes from CoilCraft, minicircuits, Digikey and wind your own. More details will be in the BOM and footprints when released.

73,

Steve

KF7O

On Monday, August 31, 2015 at 9:04:59 PM UTC-7, Steve Haynal wrote:

Hello List,

[James Ahlstrom]

Hello Steve,

Thanks for posting the schematic.  I didn't realize you were going to produce a new front end board for v1,x boards before going to the v2 board.

I love the ULN2803.  Allowing direct connection to FPGA pins is a disaster waiting to happen.  And this part can switch 13.6 volts as well as 5 volts.

I think there is a problem with U1.  When transmitting, pin 3 will ground the receiver input.  This prevents full duplex operation, as would be required for using a reflection bridge.  For bridge use, we need generator output from the Tx and simultaneous Rx on the detector port.  Apparently you were trying to support Rx on P1.

I question the need for the tap at R7 leading to Rx through U2.  I think this will prove too inaccurate to be truly useful for reflection measurements, and too low power for power measurements.  An external resistive reflection bridge is no problem.  I realize the parts can be omitted.

I look forward to getting my board.  Sign me up!

Jim

N2ADR

[Steve Haynal]

Hi Jim,

There are several ideas I'd like to test with this new frontend board before committing to v2. If it works out well, others can benefit from some v2 features with v1.2x boards. I will send you a new frontend board or two. The MAX 10 parts won't be available until the end of the year so first v2 will probably be in December or January.

Thanks for seeing and reporting the problem with U1. This is certainly not what I intended. I've updated the schematic. There is another PE4259, but they take little space and cost 0.65. The KX3 uses 33 of the older PE4283 switch. I think it is important to have the ability to run RX through the LPF in standalone mode as we've had reports of images from FM stations. The AD9866 built-in LPF is active but doesn't quite cut it. I suspect it might be creating some mixing artifacts. I would like to eventually disable it.

I understand the need for full duplex to support VNA with a bridge. V1.22 Hermes-Lite like you have already works as a VNA with two different software programs. How hard would it be to enable Quisk VNA?

The tap at R7 is based on my previous work with pysdrvna and making VNA measurements with no bridge. I had decent results. At <$1.00, this tap is a cheaper option than adding a second ADC for power/SWR as done on the Hermes. With the LNA and tweaking of the tap, we can compensate for power levels. If builds are consistent to avoid calibration with OSL, I would like to have this tap provide reasonable power measurements, SWR, harmonic spectrum analysis with 2 receivers and antenna analysis. More serious measurements can be done with an external bridge. This is one of the ideas I want to test on this prototype frontend.

Here is a truth table of all the "care" configurations for the PE4259 switches:

TR TAP RX2 RX3

1100 - Transmit on P1 with receive used for measurement via tap

1010 - Transmit on P1 with receive on P2 used for measurement via tap on PA

1001 - Full-duplex TX/RX or external VNA use with TX on P1 and RX on P3

0000 - Receive through LPF and P1

0010 - Receive on alternate antenna P2 assuming no top on PA

0001 - Receive on antenna P3

Note that the PA tap can be placed on P2 or P3. If no PA tap is present, then two RX antenna connectors are available.

73,

Steve

KF7O

[Jim Smith]

Hi Steve, 

Apologies if I've missed something, but isn't the receive path through P1 always active even when the switches are set to receive through P2 or P3? 

73

Jim - G3ZQC

[James Ahlstrom]

Hi Steve,

Everything looks good now, and it will be fun to test out your ideas, especially the VNA with no bridge.  I just want to throw out one more idea for you to consider.

On Wednesday, September 9, 2015 at 1:42:25 AM UTC-4, Steve Haynal wrote:

 I think it is important to have the ability to run RX through the LPF in standalone mode as we've had reports of images from FM stations. The AD9866 built-in LPF is active but doesn't quite cut it. I suspect it might be creating 

The reconstruction filter is not ideal for Rx because the added zero compromises its attenuation.  The added switching is burdensome too.  The switches are cheap as you say, but the IO pins and traces to switch them are not.  A better plan might be to add a second 35 MHz low pass filter that is always in-line with the AD9866 receive pins.  It could be just a 5-pole Chebyshev made from 0806 parts.  Besides eliminating switching, the filter is useful for the other inputs as well as the Rx input.  Because the AD9866 has spurious responses at VHF, these could be excited by harmonics of test signals.

Chips like the AD9866 can filter by using digital filters at a multiple of the clock frequency.  This works well in the band of interest, say up through 70 MHz, but they can have responses above that.  The combination of an analog Rx LPF plus the sharp AD9866 digital filter at lower frequencies should be ideal.

Jim

N2ADR

[Alan Hopper]

Steve. Jim,

This all looks great, just thought I'd make a plea to keep an option for an unfiltered rx input, I've had reasonable results under sampling on 6m and fm broadcast  and it would be a shame to loose this facility, I realize the ad9866 prevents serious under sampling at higher frequencies but it is still a 'nice to have' if it doesn't cause other compromises.

73

Alan M6NNB

[Steve Haynal]

Yes, you are right. The last two lines should be the following just with no active TX signal. This reduces the care set to 4.

1010 - Receive on alternate antenna P2 assuming no tap on PA

1001 - Receive on antenna P3

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

Sharing the LPF is not my only reason for having a TR switch. I'd like V2 to be a standalone QRPp transceiver ready to go when connected to a single antenna as described in the V2 architecture wiki. Hermes-Lite users seem to be skewed to the WSPR side where 10-20 dBm is still interesting. Are you suggesting to move the TR switch to the other side of the TX filter and a new RX filter, or to do away with TX/RX on one antenna connector? How important or unimportant is it to others to have TX/RX on one antenna connector with V2.0?

I'm actually more worried about the space of an additional filter than the pins and traces for the TR switch. The switches add under 0.5 dB attenuation. But with this prototype frontend, I think there will be ample room so we can try several options before narrowing things down for V2.0. Here is what I propose:

• Provide a jumper from pin 1 to pin 5 of U5 to directly connect TX to the filter and test with the TR switch removed. This should be under 0.5 cm in length. U5 and U1 are left off.
• Since you indicate that the TX filter may not be the best for RX, add footprints to the TX filter to make it order 7. These can be bypassed to test your filter. I would like to try some other filters. I know there is a difference in opinion, but the spur at 73.728-Ftx goes up in frequency as Ftx goes down. Even without a filter it diminishes as it goes up in frequency. With a filter it should be even better. It is most problematic at Ftx=29.7 MHz, but I measured it at more than 60 dBc after the PA filters. This was with no filter between the v1.22 and PA and things should only improve even with a filter providing 10 dB at 73.728-29.7.
• If enough room remains, add an order 5 filter to one of the RX inputs for testing. The other RX input will be unfiltered to support under sampling and tap measurements.

 

The RX LPF in the AD9866 is before the ADC and not digital. It uses a switched capacitor array as part of the LNA to vary R-C time constants. Even when "off" it may just be on but forced to the highest frequency. If I had the right test equipment, I'd like to see exactly what it is doing.

73,

Steve

KF7O

[John Laur]

It is an interesting quandry where to "split the board" or even if it
should be done at all. If you stick with a 2 board design even for the
most basic 'barebones' QRPp/VNA operations, you only ever have to deal
with a single final assembly size for both that configuration and 5W
QRP configuration. You get plenty of room for components and full size
RF connectors and the ability to revise the design without having to
re-spin the "expensive" board. The downside is that any circuitry that
might be common to both configurations would need to be duplicated on
the companion board of each.

The more I think about it, why not simply move only the baluns,
clamping diodes, and other impedance matching circuitry to the
mainboard and leave all RF switching, routing, and filtering to a
second board? In the block diagram, the boards would split between the
two Baluns and the above circuitry in the RF Frontend section. This
circuitry has required very few revisions and has been commonly
required back to the very first Hermes-Lite. Even if you put the dac
reconstruction filter in the Tx path on the mainboard there would
still be use cases to bypass it (such as an undersampling transmitter)

73, John K5IT

[James Ahlstrom]

Hi Steve,

On Wednesday, September 9, 2015 at 11:45:56 PM UTC-4, Steve Haynal wrote:

 Are you suggesting to move the TR switch to the other side of the TX filter and a new RX filter, or to do away with TX/RX on one antenna connector?

I was suggesting keeping the TR switch and adding a second permanent LPF just prior to T2 to benefit both transceiver and VNA use.  Then all Rx inputs would be filtered.  But I missed your desire to have Hermes-Lite be a standalone transceiver.  And I missed Allen's point that this would eliminate Rx under sampling.  So let me think again.

My comment was directed more toward the V2 board, not so much the new front end board.  I always assumed that the final V2 board would have the Tx filter directly at the AD9866 Tx pins because it must handle VHF signals.  I am worried about VHF leaking around the filter if there is a lot of switching and trace length in the area.  This argues against trying to switch the single filter so it can be shared for Rx.  But maybe this is not really a problem because the parts and trace lengths are small and the V2 board will follow good VHF practice.

• Provide a jumper from pin 1 to pin 5 of U5 to directly connect TX to the filter and test with the TR switch removed. This should be under 0.5 cm in length. U5 and U1 are left off.

OK, but not necessary for me because I must solder pigtails to the filter to test it anyway.

• Since you indicate that the TX filter may not be the best for RX, add footprints to the TX filter to make it order 7.

It isn't that the Tx filter won't work for Rx; just that a separate 5-pole filter would be better.  The extra pads for 7-pole are OK and good insurance, but I hope we don't need them.

• If enough room remains, add an order 5 filter to one of the RX inputs for testing. The other RX input will be unfiltered to support under sampling and tap measurements

I think the VNA will need a filter, but I am not sure.  And and it can't use the Tx filter.  So this would be nice.  I think the tap measurements would benefit from a filter too.

 

The RX LPF in the AD9866 is before the ADC and not digital.

Thanks.  I didn't know that.

I think your latest design is good to go.  We can worry about leakage around the Tx filter when we get to the V2 board, not now.

Jim

N2ADR

[Alan Hopper]

Steve, 

 How important or unimportant is it to others to have TX/RX on one antenna connector with V2.0?

 I very much like the idea of being able to use the board standalone, it lowers the true cost of entry and provides a practical testable unit.  I can see myself with one battery powered and connected to my phone for walks up the large hill in front of my house.  Just to throw another option in, is it possible to cleanly amplify the output to say 0.5w on board, this way it will appeal to the less extreme qrp users and match the hermes and hiqsdr and work with a greater range of existing amplifiers.

73 

Alan M6NNB

[John Williams]

The pre-amp we just added to the V2 amp is all SMT and can put out that level of power with a transformer winding change, and appropriate heat sinking.

Of course, you still have the filter issue if running without a follow on amplifier.

John - W9JSW

[Alan Hopper]

John,

 this sounds good, I just like the idea of a raw hermes lite being useful on its own. If a filter is needed  it is quite simple to add one inline in the antenna feed as no power is needed. It can even be seen as a positive thing to interface from a delicate pcb to an antenna via a short flexible coax to the filter.  This is much simpler than adding a  ptt switch and filter, this may sound trivial to the technically capable but in reality it is very much  simpler.

73

Alan M6NNB 

[Steve Haynal]

Hi All,

I am very much on the fence regarding a low power TR switch on the main V2 board. I agree with Alan that it would be more interesting if it were 27 dBm output. The harmonics would have to be clean or inline coax filters used which is the biggest drawback to me. Do we have any data on harmonic output for the new driver? Once I got my PA, I've never gone back to QRPp at 20 dBm. A design without TR is simpler and cleaner. It would be good to hear opinions from more people.  

The AD9866 output is pretty clean and it could be argued that no filters below 30 MHz are required, even though it has an integrated amp after the DAC. I wonder if some of the inexpensive integrated amps for WLAN, CATV, etc., (1, 2, 3) could be repurposed as a clean 26-27 dBm out HF amp.

73,

Steve

KF7O

[John Williams]

I think you need to keep the v2 board as clean as possible. Nothing stopping someone from making a small 20mm add-on board with a 1W amp and a few filters. The more PA like function you add on the main board contributes to excessive bom cost for those that will have to manage ptt externally.

There is harmonic data on Claudio's web site. We used driver 6 for the first amp. 

John

[James Ahlstrom]

Hello Group,

We would need to add another power supply, since 3.3 volts will not power the amp.  And we will incur the added heat burden.  The reconstruction filter may be adequate for 20 meters assuming only the third harmonic needs filtering, but I would be surprised if we could avoid band switched filters for lower frequencies.

Jim

N2ADR

[Steve Haynal]

Hi All,

Thanks for the feedback. More will be appreciated. 

There are some WLAN amps that achieve near 500 mW output with only a 3.3V supply like this $0.93 part. This makes me wonder if it is possible at HF, especially if these parts could be made to work at HF.

If our goal is now not to worry about power output from the AD9866 for QRPp but have the cleanest unswitched signal, then why are we still using the AD9866's IAMP? On page 29 of the datasheet, "Applications demanding the highest spectral performance and/or lowest power consumption can use the TxDAC output directly." This might reduce the thermal drift we've seen. It might help with the 73.728-Ftx spur. It can be tested with the current frontend card as those inputs are run out, although it would require a small modification to the firmware that I'd be happy to make. Output can be a maximum 10 dBm, but we would probably stick to the 0 to 2 dBm range given recommended operating points for best linearity (page 30 of datasheet) and the filter. Can the driver design be easily modified to handle this lower input? Power output can be adjusted digitally by the AD9866 for a 7.5 dB range in 0.5 dB increments starting at the maximum of 0 to 2 dBm and going down.

73,

Steve

KF7O

[Alan Hopper]

Steve, 

thanks for considering the 0.5w option, I understand it may cause too may other compromises.  A 0.5w single board appeals to me as It could be made to easily fit in a pocket, combined with a phone and a roll of wire it would make a very neat trekking radio.  Whilst no filters at all would be great, I'm happy with  external filters especially if not needed for the higher bands. I suspect this route will still result in the lowest total cost of ownership of a working radio.  It is also a closer match to the very successful softrock.

On the other hand John,John and Jim all make very good points.

I'm sure I'll end up with a v2 and John's amp.

If I have it right the current new front end design has lower output than before and suspect this reduces the value and likely use of the on board rx/tx switch, without the extra power I'd forgo the switch.  

Reducing the drift would be good, I realise I've not looked at the drift in full duplex which may well have changed things. If I remember correctly the drift was worse at low drive, duplex may well have changed this.  I did some experiments correcting for drift in software that worked fairly well.

73

Alan M6NNB

[f6itu]

Hi Steve et all

This remark is probably irrelevant but ....

could it be possible to change the Si510 oscillator with a Si570 ?

Many reasons for this :

- it could easily be sourced in the US and in Europe

(http://www.sdr-kits.net/Webshop/products.php?43&cPath=7 for ex.)

- it don't need specific order details like the programming startup frequency or whatever

- it's spectral purity and phase noise is not particularly awfull

- it don't cost an arm and a leg in S&H fees for Europe like the 510 (many times the cost of the component, many times the price of a easy-to-find Si570)

- almost same footprint, same volume, exists in Cmos version

- adressing this I2C device don't need a major modification of the code as far as I can see

I love the "stack/in line" design. It offers a large variety of choices and extensions (well... a modular RX bpf based on John's lpf design for eg)

But I repeat, I may be totally wrong, I'm definitely not an engineer :- /

73'

Marc f6itu

[James Ahlstrom]

Hi Steve,

On Saturday, September 12, 2015 at 3:33:15 PM UTC-4, Steve Haynal wrote:

> If our goal is now not to worry about power output from the AD9866 for QRPp but have the cleanest unswitched signal, then why are we still using the AD9866's IAMP? On page 29 of the datasheet, "Applications demanding

The IAMP has a current gain of 4, and 20Log(4) is 12 dB.  If we lose 12 dB we need to get it back in another amp, and I doubt we can do better than the IAMP, especially since the IAMP is a balanced amp and reduces even harmonics.

   

>frontend card as those inputs are run out, although it would require a small modification to the firmware that I'd be happy to make. Output can be a maximum 10 dBm,

I don't know what that 10 dB number means, as I got nowhere near it in my calculations.  Maybe they mean PEP not RMS.

I think we should put the Tx filter right at the AD9866 Tx pins, and either (1) try to accomplish the switching to share it with Rx, or (2) use a separate Rx filter with an option to omit it or (3) have a second Rx input that is unfiltered.  Unfortunately the proposed updated 1.22 front end board will not answer these questions because they are related to the final layout.  Maybe we should just make a v2 board.  It will cost more than a front end board, but if we get lucky it will cost less in the end.  In any case it is quicker.

On another topic, I got my 50 ohm filter parts from Mouser, and I should have test results in a couple of days.

Jim

N2ADR

Jim

N2ADR

[Alan Hopper]

Jim, All,

  Claudio's 'wide band push pull low power amplifier' http://www.qsl.net/in3otd/ham_radio/PD85004_PP_PA_II/PD85004_PP_PA_II.html caught my eye.  My very naive reading of the graphs indicates filter-less operation at low power might just be possible,  Have I got this very wrong? I'm no rf engineer.  I'm not suggesting putting this on the v2 board but just interested if a filter-less solution is remotely possible.

73

Alan M6NNB

[Steve Haynal]

Hi Marc,

It looks very doable to support both the Si510 and the Si570+IC2. I'll try my best to do that. It will be convenient for testing to be able to easily vary the sampling frequency.

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

My understanding of the datasheet is that 10 dBm is peak power and available directly from the TxDAC (IOUT_P pins) with a 2:1 impedance ration transformer. You should see 1 dBm peak with a 1:1 transformer. You also have to disable the IAMP in the firmware. I thought the new driver on the PA board was underutilized and could provide the extra 12 dB.

When you refer to AD9866 Tx pins, I assume it is the pair we currently use from the IAMP, IOUT_N?

I think there is value in a new test frontend, especially for those with a v1.2x already. I'm not so concerned about the layout and have things pretty much floor planned. V2 to me is the full board with MAX10 and gigabit, both of which will be more work than the frontend. The MAX10 is also not available until late Q4.

For V2, I am seriously considering raising the sampling frequency to 79.872 MHz. The spur you have been battling at Fs-Ftx first evidenced itself when Fs=61.440 MHz. It was extremely bad and made TX on 10M essentially impossible. I changed Fs to 73.728 which puts the worst spur for amateur radio use at 73.728-29.7=44.03 MHz. My measurements convinced me that it was taken care of, as it lowers in strength as the frequency of the spur increases. But since you and possibly others are still concerned, I would rather raise Fs and not have to rely on a sensitive filter. If Fs is 79.872 MHz, then the worst spur for amateur radio use is at 79.872-29.7=50.172 MHz. It would essentially be the same strength as the spur at 47.099 MHz when you transmit at 26.628 MHz now with Fs=73.728 MHz. Is that low enough for you that you think a more standard reconstruction filter will work? This would also pretty much max out the capabilities of the AD9866, provide a bit more processing gain, allow us to share the filter with RX without the added attenuation, and perhaps add 1 or 2 MHz extra for the VNA. Since ~80 and ~160 MHz digital signals will not need to pass through connectors on V2, I am not concerned anymore about the frequency. I do need to update the firmware and verify that it will run at the speed on the lowest speed grade MAX10 device.

Fine business on the 50 Ohm parts. I look forward to seeing your results.

73,

Steve

KF7O

[Steve Haynal]

Hi Alan,

That is a very interesting amplifier. It does appear that for one variation the worst harmonic is at least 43 dBc. I too am very interested in a filter-less or single-filter solution. I'd love to see something in the 0.5-1.0W range (still switchable by the PE4259 or similar) that uses a single filter and extra RTL transmitters to add proper out of phase feedback at the harmonics to reduce the harmonics as you suggested. For V2, I am going to stick with no extra amplifier on the main board and go with the two board solution. It think we can eventually experiment with these ideas on a second board and if they work out, incorporate them in the main board for >V2.

73,

Steve

KF7O

[Alan Hopper]

Steve,

the no amp decision makes sense, I'm sure there is more than enough to do without the distraction of an amp.  One of the reasons I'm keen on a minimal filter solution is to use your multi tx feature to wspr on multiple bands at once, per band lpfs obviously prevent this.

73

Alan M6NNB

[Steve Haynal]

Hello Group,

I've updated the schematic for the test frontend board. I've also finished 95% of floorplanning and it looks like routing will be pretty easy. I plan to finish the BOM Monday or Tuesday so that I can order parts to arrive before next weekend and double-check all the footprints. Hopefully I will submit it for production next weekend. Here is a summary of what I plan to test with it:

• Relay driver with 5V and 12V relays
• Jim Ahlstrom's TX filter. This will be the default and you will have to cut 2 traces for other experiments.
• 5 or 7 element Chebyshev for TX with Fs at 79.872 MHz.
  
• Sharing of TX with RX
• Differences in chip inductors and wind your own
• Differences in baluns from DigiKey, CoilCraft and Minicircuits
• Differences between Vcm=3.3V and Vcm=5V
• 5 element RX filter
• TAP measurements

I am still hoping to get a clean 17 dBm out of this with switched RX/TX. This can still be useful for QRPp. This can possibly eliminate the driver on the PA. I hope this testboard will result in a useful configuration for other V1.2x users as well.

To do more frequent testing (I have to go to the OIT lab now and getting permission is quite the process), I am thinking about purchasing a Rigol DS1052E. The link shows how to convert it into a 100 MHz scope and a 500 MHz spectrum analyzer, plus you can program it in Python! Has anyone on this list done these hacks?

73,

Steve

KF7O