Hermes-Lite 2.0 Filters

[Graeme Jury]

Hi group,

I have started a new thread for this topic so that the filter discussion does not get mixed up with the discussion on the current filters for the CVA9 and SDK boards. I agree with Alan that the radio lends itself very well to skimming and other wide band activities so here is a proposal for a filter setup which could satisfy the single band operation and the wide band operation without requiring any more filters than are currently being designed for, just different types.

I have attached a sketch to show what I have in mind. The receive filters are a set of high pass filters designed for the start of each band and the transmit filters are a set of low pass filters as we are currently using.

The output of the Tx filters are always connected to the antenna and the input side is connected via the antenna changeover relay to either the transmitter on Tx or the receiver HP filters on Rx so a simple relocation of the aerial changeover on a standard transmit board is all that is needed there and the receive filters become HP filters instead of BP.

A little care is needed with the filter design to get low return loss over the full pass range of the filter but for the higher frequencies a 5 stage Cauer will do the job but as you go lower I find that 7 stage may become necessary. I have a bank of Tx filters from QRP Labs here which I will mate up with some High Pass filters as soon as I get a chance and let the group know how I get on. If you know of any pitfalls in the idea please let the group know.

73, Graeme zl2apv

[James Ahlstrom]

Hello Graeme,

I use a filter scheme very much like this with my HiQSDR-like hardware.  The only difference is that I route the low level Tx signal (before the power amp) through the high pass filters to eliminate low frequency spurs; although I am not sure this extra step is really necessary.  It works fine.  But the combination of high pass and low pass filters is not as sharp as a carefully tuned bandpass filter would be.

The only trouble I have had with my filters is that the Tx filters are reactive within the band and the power amp does not see an accurate 50 ohms even with a dummy load.  I fixed that by tuning my antenna tuner to match the combination of the Tx filters plus antenna, instead of just the antenna alone.

I see you are using Butterworth filters for Tx.  The advantage is insensitivity to exact component values as you have said.  But I think you will find that they are quite reactive in the band.  You may want to look at "Harmonic Filters Improved" by James L. Tonne, QEX September 1998 page 50.  These filters sacrifice pass band flatness below the band of interest for better return loss within the band.  There is an on-line calculator for these at http://www.tonnesoftware.com/optlowpass.html.  Of course, they are optimized within the band, and may be undesirable if you want broad band coverage including frequencies below the band.

Jim

N2ADR

[Graeme Jury]

I have tried the idea inspired by Jim's success with this system and have an acceptable result. Unfortunately I have a hole in my QEX library from early 90's to 2000 so can't read the article "Harmonic Filters Improved".
I have presented 4 screenshots from my Quisk VNA showing a fixed 2 MHz Cauer HP filter cascaded with firstly a 7 MHz LP Chebyshev and then a 14 MHz LP Chebyshev filter. These two filters are part of a switched bank on the QRP Labs LP filter referenced in my first post. There is a fair bit of blowby on the filters probably due to using a dual changeover relay rather than 2 single relays but the -40 dB floor is perfectly adequate for this test.

Method
Construct a 2 MHz Cauer HP filter from 3 toroids and junk box caps.
Connect in series with a QRP Labs LP filter set
Select the 40 Metre LP filter
Measure the through loss with my Quisk VNA and take screenshot
Select the 20 Metre LP filter
Measure the through loss with my Quisk VNA and take screenshot
Select the 40 Metre LP filter
Measure the return loss with my Quisk VNA and take screenshot
Select the 20 Metre LP filter
Measure the return loss with my Quisk VNA and take screenshot

Results
The combined passband loss of the 2 filters was around 0.7 dB on receive and the through loss on Tx was around 0.5 dB being the loss of the LP filter only
The return loss of the filters is better than -20 dB and in most areas much better than this. In SWR terms it is around 1.12 at worst and mostly 1.1 or less.
The out of band performance was better than 3 section Butterworth filters apart from the floor dictated by the LP filter performance.

Conclusions
This method is going to work well with toroids and relay switching. The next step is to see if Peregrine switches on the HP filters and relays or PIN diodes on the transmit filters will give good results too. I have passed some information on the GP10Y diodes to Andrew and I think he is going to look at switching filters with them. They hahttps://sites.google.com/site/rshfiqtransceiver/homeve less capacitance than the iN4007's as used by Jim Veitch in the RS-HFIQ here and previously referenced by Steve. The lower capacitance removes some of the requirement for a High voltage reverse bias and I am comfortably switching 100 watts with 12 volts reverse bias so filters for a 5 watt amplifier should be a switching possibility.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

This sound like a nice idea to me. If the consensus is to do this, I will move the TR switch to before the TX filters in HL2.

Do we have specifications for all TX filters using SMD inductors? The current board has specifications for wind your own, which is great, but I'd also like to have off-the-shelf alternates. What do people think of the TX filters used in the RS-HFIQ? There is only one set of filters past the PA on TX. All the SMD components are specified and it is open source.

Also, what are your comments on this previous post: I am happy but surprised that you saw no internal noise from the PE4259s. I would expect you would see noise around 900 kHz and odd harmonics similar to what I saw in this post by me on May 1 to the thread "PE425x switches noise at 900 kHz and harmonics (from the EMRFD group)."  The current RX BPF board uses 14 PE4259s, and I am curious and a bit worried as to how these 14 noise sources will add up. What noise do you see if you run an experiment similar to the one I did? Also, the PE4259 datasheet says performance degrades below 10MHz. Do you notice any differences or measure any distortion on 160M?

73,

Steve

KF7O

[Graeme Jury]

Hi Steve,

So far nobody has raised any objections and I hope they speak up if there is any as we don't want to be changing your new board after it is manufactured. I have only just become aware that Jim Veatch was using smd inductors on the transmit part of the filters and yes I agree that off the shelf inductors is a good thing if the total losses are not too great. His whole concept with the diode switching is a great idea and I am sure 160M can be accommodated with a more expensive PIN diode just for that band. There will be an issue in that the Tx output is rectified and used to derive the high voltage for biasing the Tx PIN diodes. Of course we would need to switch both the Tx and Rx PIN diodes on receive so would need a separate switch mode bias supply with an output of around +30 volts at a few milliamps. Of course if we use Tx relays the issue will go away.

I really can't answer you right now Steve on the Peregrine switches as I am doing the filter testing directly on an external test bed. I seem to be tied up with so many things that I just am not getting to the Rx filter board although I have all the semiconductors mounted ready for the filters. Glenn is the doer of the group and has his running and has made informal tests although I doubt he duplicated your tests. I am sure he will pop up with a careful analysis and we will know exactly what we are facing in terms of noise and distortion.

73, Graeme zl2apv

[Glenn P]

Still thinking about the LPF-HPF idea.....

Ref Peregrine switch 'noise', this morning I hooked up my BPF board which has all (12) switches fitted.

1st I did a run with the SA (HP8560A) connected to the output and no power applied to the BPF for a baseline check. SA set at 10Hz BW and with a noise floor around -120dBm. (sweep takes 3 minutes!)

2nd, I found a signal around 867KHz which seemed most prominant. Some other sigs present also) Same SA settings. Centre is 900KHz, span 300KHz ie +- 150KHz

3rd, i found what would appear to be 2nd harmonic of the 867KHz. Centre is 1.8MHz, span 300KHz [ I didn't check for other harmonics, it takes a long time and i have some errands to run now.]

It would appear to me there is no problem, band noise would mask these signals in most cases anyway.

glenn
glenn

On Tuesday, July 19, 2016 at 9:40:23 PM UTC+10, Graeme Jury wrote:

Hi Steve,

So far nobody has raised any objections and I hope they speak up if there is any as we don't want to be changing your new board after it is manufactured. I have only just become aware that Jim Veatch was using smd inductors on the transmit part of the filters and yes I agree that off the shelf inductors is a good thing if the total losses are not too great. His whole concept with the diode switching is a great idea and I am sure 160M can be accommodated with a more expensive PIN diode just for that band. There will be an issue in that the Tx output is rectified and used to derive the high voltage for biasing the Tx PIN diodes. Of course we would need to switch both the Tx and Rx PIN diodes on receive so would need a separate switch mode bias supply with an output of around +30 volts at a few milliamps. Of course if we use Tx relays the issue will go away.

I really can't answer you right now Steve on the Peregrine switches as I am doing the filter testing directly on an external test bed. I seem to be tied up with so many things that I just am not getting to the Rx filter board although I have all the semiconductors mounted ready for the filters. Glenn is the doer of the group and has his running and has made informal tests although I doubt he duplicated your tests. I am sure he will pop up with a careful analysis and we will know exactly what we are facing in terms of noise and distortion.

73, Graeme zl2apv

N2ADR

[Steve Haynal]

Hi Graeme,

The GP10Y diodes don't appear to be available as surface mount devices. Can you pick another device? A quick search with Digi-Key showed many surface mount diodes with lower capacitance than the 1N4007 and able to withstand >600V.

+30 V may be possible with an inexpensive boost converter, but I think we will need more to handle 10-15W PAs. It would be nice to keep the possibility for higher power in mind.

Regarding higher power, as the capacitors speced in the TX filters able to handle the expected voltages? The RS-HFIQ is using 100V capacitors. I think we are using 50V, which should be enough for 5W.

I was suggesting to use the RS-HFIQ TX filters without the PIN diode switching and biasing as a build option for surface mount devices. I still don't understand how he is able to use simpler 5-element filters and still meet emission requirements. Although the input to his PA mayh be a bit cleaner as it is passed through bandpass filters, the PA is pretty much the same as John's and is only filtered by these TX filters.

I'm not sold on pin diode switching, but it is an interesting direction to consider. The Russian design I linked to used pin diode switching. On the HobbyPCB 50W PA page, they warn that pin diode switching is good for QSK but expect power loss especially on 6M.

73,

Steve

KF7O

[Steve Haynal]

Hi Glenn,

Thanks for checking this. It is good to see that they are not all piling on top of one another such that a spur would be above the band noise. I bet WSPR could still see these spurs...

Are there any ways to measure distortion through these devices? I wonder about frequencies below 10 MHz. Also, if you look at the KX3 schematics, they double up some of the PE4283s on 12, 10 and 6M. I wonder if they experienced attenuation on these higher frequencies.

73,

Steve

KF7O

[Graeme Jury]

Hi Steve,

It is still early days on the PIN switching idea but I have a little data from my PIN antenna changeover switch. There are 4 illustrations to give some idea of the performance of the GP10Y. Unfortunately there is only a leaded version of this diode but I stayed with it because its reverse recovery time is 3 uSec for 5 pF junction capacitance at 4 volts rev bias which was the best I could find after trolling through data sheets for maybe a hundred possible candidates or more. Some did not give Trr like the 1N4007 although it is nearly as long so other unquoted ones may have been OK too.

Fig 2 is the circuit I used to test the diodes for Trr and used a 5 Watt signal with no bias on the diode.
Photo 1 is a 1N4148 showing how it is a good rectifier as its Trr is very short, I'm amazed it did not cook as I was terminated in 50 ohms..
Photo 2 is the signal through a GP10Y and there is just a trace of rectification on the -ve portion of the sine wave. Just a sniff of forward bias removed that.
Fig 4 is the through path of my antenna switch from the transmitter port to the antenna port, showing the loss going up to 0.4 dB at 50 MHz. Note that I used 2 series diodes for better isolation when on receive so the levels are the same as would be obtained in a filter switch.

The above results are indicative only of how it would perform and it would be necessary to build up a test circuit which I am not going to do at present as I am getting further behind with testing John's amplifier and filters but I will get to it soon.

I don't know much about what PCB manufacturers can do with a PCB as I have only ever had one made. I was hoping that it would not cost very much to have an oblong cutout the size of the diodes so the leads could be cut short and the diodes laid in the cutout with the leads soldered on lands on each end of the cutout. It would give a nice linear flow to the signal and no lead bending required.

I really can't be very specific Steve as I simply have not researched this top very well yet.

73, Graeme zl2apv

[James Ahlstrom]

Hello Graeme,

I have had some misadventures (smoke) trying to use 1N4007 diodes as switches, admittedly at higher powers.  The problem was not the diodes, it was the chokes.  The chokes must withstand high voltage and have high impedance at all frequencies.  Be careful of the chokes.

Jim

N2ADR

[Steve Haynal]

Hi Graeme,

A nonSMD package is a nonstarter for HL2. Any through hole assembly or deviation from a standard pick and place assembly costs more. Have you checked for SMD diodes with similar specs to the GP10Y that you might like?

73,

Steve

KF7O

[Graeme Jury]

I guess we don't need 1 amp diodes. I chose them to support 100 watt operation. The lower the current rating, generally the less the capacitance. Higher voltage is often an indicator of a longer Trr, so I will look in the 1/2 amp 1500 volt class and see what I can find. The problem is that suppliers don't include Trr in their parametric searches.

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

[Steve Haynal]

Hi Graeme,

Digikey includes Trr when searching "Diodes - Rectifiers - Single." Check these options out:

http://www.digikey.com/product-detail/en/fairchild-semiconductor/S1MFL/S1MFLCT-ND/5305090

http://www.digikey.com/product-detail/en/diodes-incorporated/RS1MSWFQ-7/RS1MSWFQ-7DICT-ND/5731846

http://www.digikey.com/product-detail/en/vishay-semiconductor-diodes-division/S07M-GS08/S07M-GS08CT-ND/3104489

http://www.digikey.com/product-detail/en/diodes-incorporated/RS1MSWF-7/RS1MSWF-7DICT-ND/5731845

http://www.digikey.com/product-detail/en/rohm-semiconductor/RFU02VSM8STR/RFU02VSM8SCT-ND/5113966

There are more options if you lower the current and voltage requirements.

73,

Steve

KF7O

On Wednesday, July 20, 2016 at 2:59:54 PM UTC-7, Graeme Jury wrote:

I guess we don't need 1 amp diodes. I chose them to support 100 watt operation. The lower the current rating, generally the less the capacitance. Higher voltage is often an indicator of a longer Trr, so I will look in the 1/2 amp 1500 volt class and see what I can find. The problem is that suppliers don't include Trr in their parametric searches.

73, Graeme zl2apv

[Steve Haynal]

Hi Jim,

The RS-HFIQ is using the Murata 82153C for the chokes. What do you think of it?

73,

Steve

KF7O

[Graeme Jury]

Hello Jim,

You made me smile about the smoking chokes. Yes I had the same experience with the 100 uH chokes I was using as they were series self resonant in one of the ham bands, I think it was 15M anyway I changed to 5 turns of kynar wire wrap wire through a BN2402 core which worked beautifully and was self resonant way above where I was working. The Murata 82153C cores quoted by Steve are self resonant at 31 MHz which may be a factor on 10 metres.

Cheers, Graeme

[James Ahlstrom]

Hi Steve,

I looked at the data sheet for these chokes, but there is not enough information to judge them.  We need a spec for maximum RF voltage, and the self resonant frequency.  These are directed at high current applications.

Jim

N2ADR

[Steve Haynal]

Hi All,

QRP Labs is now selling an interesting band pass filter kit. Has anyone tried these? For someone in ITU region 1 where there are many strong signals, it would be interesting to see if and how much a band pass filter such as one of these helps in practice. For example, are you able to spot more WSPR or JT65/JT9 signals with it in place? Ideally, one would also want to make proper lab measurements, but empirical data is also useful.

73,

Steve

KF7O 

[Alan Hopper]

Steve,

I've been trying to set up this sort of test with two hermes lites.

My purpose is purely to optimise my setup to maximise wspr and JT65/9 spots at my QTH.  

I realise that what I find may not be optimal anywhere else or even here next week, but hope to learn something in trying.

My basic setup is a 5 band 'coupled resonator' vertical 1/2 wave dipole resonant on 20,17,15,12 & 10m with no traps or tuner.

The antenna is fed into a mini circuits ZSC-2-1W splitter and then to both radios.

In an attempt to minimise noise the radios and a netgear GS105 gigabit switch are the only active devices in the shack and are all powered by battery.

The pc is in the house at the other end of a long network cable.

My measure of goodness is simply the number of wspr and JT65/9 spots, I have put some effort into making this measurement fair.

By using my radio software I avoid the vaguries of VAC and setting gains and I think I now have a fairly blameless digital path to the decoders.  I found small changes to audio agc could have a significant effect.  I now have reasonable control of frequency calibration using NTP ( I may try to improve on this using shared wspr spots).

As a baseline/sanity check I have been testing two basic frontends for the last few days, 

I realise they will never be exactly the same but I have been surprised by the level differences from the two the rx transformers, I wound new ones from identical wire and cores from the same order but still get a 2db level difference.  On 12hr tests on 20m the spot counts differences are less than 1%.  I did run the superband bandpass amp that John gave me against a basic front end for 6hrs and the basic front end was 2% ahead.  

I swap the frontends halfway through to cancel out radio/splitter errors.

I guess testing on 10m might be more revealing but spots are rare here at the moment.

I don't really know how my QTH IO91ue compares to others in terms of strong local signals (I'm not aware of any huge ones).

Before I go further I want to optimise rf agc setup as different frontends and filtering allow different levels of gain and to be fair this needs to be optimal.

All suggestions on experimental technique gratefully received. 

73 Alan 2E0NNB

[Graeme Jury]

Hi Alan,

Have you tried swapping the two transformers around on the two frontend boards? Bit of a pain to do this but might show some interesting results.

73, Graeme ZL2APV

[Steve Haynal]

Hi Alan,

Sounds like an interesting experiment. I've done similar experiments, but not as carefully as you are.

In the past, I've also seen variation of a couple of dB between different HL1.22 boards. I swapped out the frontends, but it appears dependent on the HL1.22 board, not the frontend. I measured the noise floor and SNR and it appears that the max SNR (dynamic range) is the same for all boards, but the noise floor and peak signal levels vary in lockstep by a few dB. I chalked this up to fabrication variations of the internal LNA and LPF and didn't investigate further. You can normalize the boards by setting the LNA gain differently for both. 

I think Claudio's graphs indicate the expected rise in the noise floor at various gain settings for the LNA. I think what matters is max SNR possible, or best dynamic range. It would be interesting to measure what LNA setting provides the best dynamic range. There is some data on this in the AD9866 datasheet.

I still hope to try your latest software later this weekend.

73,

Steve

KF7O

[Alan Hopper]

Graeme, Steve

I did swap my original transformers over and the difference swapped with them so I figured they were the significant difference. I then replaced them one after the other and on each change the difference changed.  In light of Steve's comments about variations in radios I'll double check.

73 Alan 2E0NNB

[in3otd]

Hello Alan,
I find rather strange that the transformers behave so differently; in the past measured a few identically constructed ferrite transformers (not like the ones used in the H-L, actually) to see if there were significant differences and found them to have almost identical performances (much less than 0.1 dB of difference), at least in the passband.
The RX transformer I have built for the H-L on a BN43-2402 core has about 0.1 dB losses, so I won't expect the possible variations in the core material to rise the losses up to 2 dB. For other core materials (75 or 77), which are more conductive, if the enamel is scratched while winding the transformer the wire could contact the core and this may lead to increased losses and distortion, but it should not be an issue with the 43 or 61 materials, AFAIU.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

reposted as I sent direct by mistake

Claudio,

thanks for that, I was surprised but didn't know what differences to expect. It was late when I did all the swapping so I might have done something silly.  I'll re investigate.

73 Alan 2E0NNB

Claudio,Steve,Graeme,

thanks for sowing the seed of doubt, a quick test with a signal generator confirmed it was the frontend not the radio, I then found a bad joint on the transformer and they are now as close to identical as I can measure, <1db for radio and frontend combined.  Not sure what was going on when I swapped them originally (I blame the wine).

73 Alan 2E0NNB

[Graeme Jury]

Glad to see it sorted. All transformers I have made have been very consistent. Go and have another wine and celebrate your success.

73, Graeme zl2apv

--

[Steve Haynal]

Is there no feedback on these these? I'm not sure about using pin diode switching, but would at least like to understand what are some good SMD choices.

73,

Steve

KF7O

[Graeme Jury]

Hi Steve,

I had a look at all of them and most would work OK. I went a bit further using the searches you suggested and the best I came up with was CMAD6001 which has a Trr of 3 uSec and a capacitance of 2 pF at 0 volts bias. The diode is rated at 75 volts and 250 mA which would be good for around 5 watts in Tx service. I was going to get some with my next Digikey order (nobody else seemed to have them) but I save it up until I get free postage so sometimes things take 6 months or so which is why I have not responded.

73, Graeme zl2apv

[in3otd]

Hello Graeme,
I see that the CMAD6001 is also available from Mouser; as I'm about to order some components there I can order some for you also, if you wish.

I wanted to include this diode also, do you think it could be suitable?

73 de Claudio, IN3OTD / DK1CG

[Graeme Jury]

Hello Claudio,

Firstly yes I would be very grateful for some CMAD6001's as I can start experimenting straight away with them. I will arrange payment via PM so will contact you soon.

The S1FLM diode is very interesting and has the potential to switch 100 watts. It has the lowest capacitance for a diode in its class that I have come across so far and would provide good isolation right up to 50 MHz I suspect. The dime a dozen 30 volt boost power supplies from China would provide enough reverse bias to bring the capacitance down to 2 pF. I prefer to do this rather than rectify the Tx output to get a reverse bias voltage.

What might be an issue is the Trr of 1.8 uSec which is quoted as a maximum value. If the diodes are near to this figure then they should work OK at 1.8 MHz but I have seen some types which are a lot faster than the max value. As is the usual case the real test is to build one up and measure how it goes. I would also point out that if the radio is built for an LF enthusiast then the option of having one or two TX filters relay switched could be provided on the board and the user can populate as they wish.

73, Graeme ZL2APV