HL2 users with Thetis PureSignal problems: please post here to see if we can solve this

[WU2O Radio]

I'm an Apache hardware user (NOT an Apache employee) who has worked closely with some of the dev's for a long time. I'm NOT an HL2 owner. Nevertheless, if we can solve this PureSignal issue that people are reporting on Thetis then it is hoped that any Thetis development that might occur in the HL2 community can carry over to the Apache community and vice versa.

Here's what I know (please correct any misperceptions):

- The problem exhibits itself by the generation of an aberrant, picket-fence (comb) like signal.

- HL2 users have observed this doesn't happen with PowerSDR , only with Thetis.

- HL2 users work-around the problem by using Single Cal.

- It does not happen on more modern Apache hardware, but it has been observed on older designs that are very similar to the HL2, in particular the ANAN-10E. Both of these designs provide minimal RF isolation between the TX and feedback receive paths.

- With Thetis people are seeing GetPk values of around 0.19 and using SetPk values of around 0.233. This is backwards, the SetPk value should be slightly less than the GetPk value.

Things that would be helpful to know in order to solve the problem:

- What version of PowerSDR are people using successfully? Please provide a link to that version. The intent is to compare the WDSP source code used there to the WDSP source code used in Thetis.

- How are people configuring their PureSignal options? At a minimum are you using PIN and STBL? Screen shots are preferred but even just a description would be helpful.

- Screen shots of the AmpView display and panadapter when a bad transmission is occurring would be very helpful.

Recommended settings for PureSignal: MOX wait 0.2, CAL wait 0.0, AMP delay default, Auto Attenuate, PIN and STBL ON, Quick Attenuate Response OFF (this last option is not in PowerSDR), TINT 0.5dB.

Instructions will follow on how to measure GetPK to develop a more optimal setting for SetPk (I don't have them yet).

73,

Scott

[WU2O Radio]

More on this for those with problems:

In the Linearity control panel ensure that the feedback path is such that Auto Attenuate drives feedback levels as follows:

Feedback Level:  >181 is too high; 128-181 is just right; 90-128 is weak; <90 is not a good idea.  PureSignal is quite tolerant, however.  IMPORTANT: Note that the value displayed won’t be correct if SetPk is incorrect.

Measuring GetPk in order to choose the right SetPk value:

- Set up the radio so that the feedback level with two tone is within the optimal range as discussed above.

- Auto attenuate should be enabled as part of this process.

- Assert two tone and then use Single Cal to obtain a GetPk measurement. Make 10 measurements. Ensure the 10 measurements do not vary by more than +/-0.0001. If they do then something is wrong with the quality of the feedback signal.

- Average the ten measurements, subtract 0.00015 and make that SetPk.

- Repeat the process since its accuracy depends on the value of SetPk so at least one iteration is probably wise.

Note theoretically it should be possible to calculate the value of SetPk mathematically. However this depends on knowing very precisely the gain cascade through the entire RF chain, that includes the effective gain within the DSP portion of the firmware. It is easier to just determine the value empirically. If we can get enough people to do this then the value for SetPk can be adjusted such that it will work reliably for every HL2 user running the same revision of firmware.

73,

Scott

[r webby]

I'll bite. Getting a screenie of the gross distortion is a tough ask, because it is apparently random.

But here is how mine looks when it's working as it should:

[WU2O Radio]

Thank you, Steve!

If you can catch a bad TX that would be wonderful. Maybe do a screen recording and bang at it hard enough to catch one, then screenshot a frame in the video.

That looks fine except for the SetPk value. If you have a chance please try to go through the process I outlined below and see if you can't develop a more optimized SetPk value. If we assume that your GetPk numbers hover right around 0.2415 then SetPk should wind up right around 0.2413. But the numbers need to come from a succession of Single Cal measurements and it's just a little iterative in that they will change when you change SetPk.

Historical perspective: Warren spent quite a bit of time to set the values for Apache/openHPSDR-firmware P1 and P2 variants. Again they are different because of the different effective gains in the P1 and P2 firmware DSP RF chains. The value for SetPk will be different again for the Saturn boards that will ship in the Apache G2 series, and are of course different for HL2.

Once we get a half dozen or dozen or so people optimizing their SetPk values and reporting out on them we should have a pretty good number for all to use. This will almost certainly remain the same unless some sort of gross bug is found in WDSP, or until a new HL2 firmware revision is released that changes the RF processing in the firmware.

73,

Scott

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/6c383585-ffb8-498c-a1ea-15ee2e3b71a9n%40googlegroups.com.

[Steve Haynal]

Hi All,

For those experimenting with PureSignal, please see the Hermes-Lite 2 wiki page regarding this:

https://github.com/softerhardware/Hermes-Lite2/wiki/PureSignal

Also, be aware that the HL2 gateware is not the same as 10E gateware. Because of the HL2 lower sampling rate and 12 bits, we are able to time multiplex some filters. Also, we use NCOs instead of Cordic for local oscillator as simulation shows this is just a clean and it requires less resources. The 10E is missing out and could adopt the NCOs. This is why we can fit 4 RXs,1 Tx or 10RX,0TX in the HL2. These changes also slightly alter the gain of the total DSP path, and is one reason why the SetPk and GetPk values differ for the HL2.

73,

Steve

kf7o

[WU2O Radio]

Steve,

I hope you don't think I was in any way denigrating the firmware.

There clearly needs to be a concerted effort to determine the correct SetPk value. Reading through the messages on the group people are all over the map on this.

If folks could go through the process I've laid out in this topic and post their results it would be most helpful to all.

73,

Scott

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/fd6334da-2825-4dab-85e4-e7313cc42366n%40googlegroups.com.

[Ezio Insinna]

Hello PS and HL2 aficionados

Here my values as shown. It works without any problem.

I found out that the best value from the sampler is 2 dBm (800 mV pp) with the 2.2 KOhm resistor (with a power Output of 200 W on 40 mtrs) from my "dirty" Zetagi amplifier (with LP filter).

I've not found the SetPk to be critical. I just set it to this value (0.24) at the beginning and the autocorrection does change it to other values......

But I might be wrong ?

73

Ezio

K3EK / F5MWA

 

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.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/CAPC_m5Gq1a1qQKc6_zNtf7TGmV81m8zCNNSr0i9ZtJXeD%3DFhzw%40mail.gmail.com.

[r webby]

Here''s the bug in all its majesty. I just kept hitting two tone until it (rarely) failed. I must have done it 50 or more times.

[r webby]

Note the feedback is a tad low on this test, but it will fail regardless of even a "perfect" feedback level. I normally run it in the green, or rather I _don't_ because I can't trust it not to do this mid-qso.

[WU2O Radio]

That's awesome! BUT...we need a screen cap like that that also shows the AmpView display. It's very important to get the AmpView display captured during one of these events. 

I know it's a PITA, but would you mind hitting two tone another 50 times, please? 

Thanks!

Scott

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/efa3fe69-5786-4256-aa68-87c8cf7e0aa7n%40googlegroups.com.

[WU2O Radio]

Ezio,

Automatic mode will not change your SetPk value. It is also stored and will not change if you close and re-open Thetis. At least this is the behavior with Steve's version 2.9.0.27.

It wasn't clear from your message below: have you or have you not experienced the bad transmissions on your HL2 using PureSignal with Thetis?

Would you please do the experiment whereby you hit Single Cal ten times and record the ten different values of GetPk, please?

Also, what version of firmware is loaded on your HL2, please?

Thanks,

Scott

[WU2O Radio]

Steve,

Hoping you can catch this again with AmpView visible. We do need to see what's up with AmpView when this occurs.

Also, would you please do the experiment where you hit Single Cal ten times and record the ten different values of GetPk?

Finally, what version of firmware are you running on your HL2?

Thanks,

Scott

On Mon, May 1, 2023 at 10:32 AM r webby <radiowe...@gmail.com> wrote:

[r webby]

I have tried many, probably thousands of times, but it likely won't put in another appearance until, well, who knows when!

Steve

[r webby]

0.239665895460918
0.233629326395074
0.233626960308559
0.233628154675623
0.233628568310198
0.23363040632719
0.2336283595137
0.233628500915057
0.240356887027181
0.240356887027181
0.233628311647693
0.233627870322566

[WU2O Radio]

That's perfect, thank you!

Now if we can get a few more folks to do that we can try to pick a value for SetPk that is more optimized.

Can you tell me what version of firmware (gateware) is installed in your HL2, please?

73,

Scott

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/6fa2921d-ff52-4eec-ab42-1721325b0fc0n%40googlegroups.com.

[r webby]

You can see it atop the Window title, though it did it just the same with older firmware.

[r webby]

Yeh, changing setPk  won't help. Been there, done that, many times :-(

[WU2O Radio]

Yes, but it's prudent to dot all i's and cross all t's when chasing something as elusive as this.

My bad on not catching the FW version on the title bar. Well done to all dev's for having it go up there in Thetis! :-)

73,

Scott

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/f09a56da-7a1a-4add-a276-5ba20dbc2567n%40googlegroups.com.

[Juergen Hussels]

Just my 2 cents on this topic,

I first discovered this on my HL2 . Suddenly the signal was much wider than without PureSignal. Next TX ok again. 

However, as I am building a new TRX with a Redpitaya 122.88 (14 Bit in TX) I observed the same issue. Even with 0dBm output from a naked Pitaya using a 3dB coupler for feedback and going into the Spectrumanalyser.

1. HL2 is running on 192KHz sampling rate. 

2. Redpitaya on 96 & 192KHZ - my  impression is that the misbehavior is more frequent with 192 KHz.

Software used are different Versions of Thetis. For the HL2 i used always the most recent Version from Raid MI0BOT.

While testing with Powersdr -I never observed wrong signals.

I believe the firmware is out of the match, as we have so many different hardware variants.

I also compiled a version of Thetis, where I nailed the SetPk to fix value, as sometimes this changed for some reason on a fresh start of the program. But this didn't help on the wide signal behavior.

As it seems from my point of view to relate a bit on sample rate, it might be a time critical thing in Thetis.

Only an observation- there are comments in Thetis which state that the PS-Bit was re-done at a certain point in time - was that the point the problems started ? As we don't see them in Powersdr?

73  de Juergen, DL4JY

[Pez]

Here are two screenshots I just grabbed. 

(FYI, my PS feedback levels are low as my amp was OFF, but the same thing happens occasionally with everything in the green... 

It might be unrelated, but also I want to mention that every time I release PTT in PS-A mode, the TX relays seem to switch OFF - ON - OFF, within a few hundred milliseconds. My amplifier doesn't like it much. It does not do this in Single Cal mode. 

Single Cal 2 Tone: 


PS-A SSB Voice: 


73

[Ezio Insinna]

Hello Scott,

Here the 10 values your requested

0.233627332057077
0.233627604048932
0.233626391749941
0.233626866060272
0.233627565250057
0.233626128547121
0.233630235124625
0.23362966813137
0.241083765163644
0.233629150610065

My HL2 has the 73.2 firmware (gateware) installed

The Thetis version is the last one v2.9.0.8 x 64 HL2 Beta 1

Here two screenshots

I hope this helps.

73

Ezio 

K3EK / F5MWA

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.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/a1934f94-255f-4c2a-98a6-6b915f4b5934n%40googlegroups.com.

[WU2O Radio]

Great input, everyone.

Pez--if you wouldn't mind getting us 10 measurements of GetPk using single cal that would be helpful.

What's most critical, and missing at the moment, is a shot of AmpView during a bad transmission. Perhaps some screen recording software and some patience might produce that?

Lastly, it's worth mentioning that the PureSignal algorithm in WDSP is optimized for a 192KHz sampling rate. If the DDCs (there should be two, one at the DAC output, the other receiving the feedback signal) used for PureSignal in the 7.3 gateware are running at other than 192KHz then the results will not be optimum.

73,

Scott

[Reid Campbell]

Hi Scott,

I have been following the thread and managed to get a video of Ampview while the fault was happening. You can get the video here

https://www.dropbox.com/s/akulnt60wvfh1i2/VID_20230502_205358713.mp4?dl=0

Let me know if there is anything else I can help with.

Cheers

Reid
Gi8TME/Mi0BOT

--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/b1a5d4e5-841c-43bf-b8d2-873ea5bde920n%40googlegroups.com.

[WU2O Radio]

Thank you, Reid!

Can you provide a screenshot or video of what it looks like when it is acting normally so that it can be compared?

73,

Scott

You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/44a474e1-1430-6b1a-f848-02fcdd4007ef%40gmail.com.

[Reid Campbell]

OK, the first one here is the normal condition.

https://www.dropbox.com/s/frgmh95erbzgh56/VID_20230502_215720935.mp4?dl=0

the next one is me switching the PS on and off. You will notice that some times the output starts to go flaky, almost like an oscillation which nearly hits bounds. If I do this enough, I get the full spectrum noise. Warning this file is big.

https://www.dropbox.com/s/qw5rh8sz7lqkgza/VID_20230502_215828864.mp4?dl=0

Cheers 

Reid
Gi8TME/Mi0BOT

[WU2O Radio]

Thank you, Reid!

73,

Scott

[Pez]

Here you go...

The HL2 was barefoot, using crosstalk but still with -67.5dBm on the RF3 port as well. (External amplifier was connected, but turned OFF). PS Level is 110 if I test in PS-A mode (low-ish, but corrects ok).

Here are multiple readings, 2 of which of which are with the problem to occuring:

0.233626453836743
0.233620528435096
0.233627882286252
0.233628183914257
0.233630215101233 - this was with the issue occurring
0.233624562218041
0.233624250765239
0.233624201793431
0.23363064946133
0.233624325038427

0.233628186069398 -  this was with the issue occurring


Single cal - Normal operation:

Single cal - Problem 0.233628186069398: 

The other thing I see, is an occasional "pulsing" of the TX spectrum during the 2 tone tests. This is in both PS-A and Single Cal modes. It looks almost like it's about to go into crazy mode, but it pulls it back, just in time... That's how it looks anyway. 

I hope that helps. 

73

[Pez]

Hello Scott,

Thanks for this information: 

- Lastly, it's worth mentioning that the PureSignal algorithm in WDSP is optimized for a 192KHz sampling rate. If the DDCs (there should be two, one at the DAC output, the other receiving the feedback signal) used for PureSignal in the 7.3 gateware are running at other than 192KHz then the results will not be optimum.

FYI, in my tests above, I was in 384000 mode. Although, I have been having the same issues occur when set at 192000. 

73

[Reid Campbell]

Scott,

I was playing a bit more with this and I have seen the wide spectrum with PS turned off. I'm not sure if it is maybe some logic fault were PS is still on but displayed as off or the wide spectrum can happen without PS?

Cheers 

Reid
Gi8TME/Mi0BOT

[Pez]

Hi Reid and Scott,

I have seen a similar thing at one point in my testing. It was confusing, but it looked like PS was ON when I had it set to OFF. 

73

[Reid Campbell]

Just got it to do it this morning, screen shot attached. Now every time I transmit 2 tone, I get the wide spectrum, without PS.

Cheers 

Reid
Gi8TME/Mi0BOT

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/b249cdcd-9867-42d7-a8b1-d5d3891d992bn%40googlegroups.com.

[WU2O Radio]

Thank you Pez and Reid, this is very helpful!

73!

Scott

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/abe1c199-4d9f-25ea-6115-684080f89523%40gmail.com.

[Reid Campbell]

Scott, at this moment in time I have my system in state where I get the wide signal for single tone or dual tone with out PS on. I can change band and the wide signal is still there.

Is there anything I can look at which would help you?

Cheers 

Reid
Gi8TME/Mi0BOT

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/CAPC_m5GESeR%3D78VrDB9LZOoGW4z9hDz66X2U4_G2ht0cCgPCGg%40mail.gmail.com.

[Pez]

I'm no expert on this, however is it possible to check if the filter board relays are actually behaving and switching correctly - to make sure the hardware filter switching is being told to do the correct things. (As opposed to the PS algorithm being the fault... For example, what if the filters are switching incorrectly and the output (and the PS) is acting on the resulting harmonics?). It's just an idea I had - I could be a long way off the track. 

73

[WU2O Radio]

Pez,

I'm noticing that your feedback level is far too low and your S-ATT far too high. It's somewhat surprising that it is working at all.

What's going on with your feedback levels, please?

73,

Scott

[WU2O Radio]

Not sure on either of your questions. However, I do have an experiment I'd like you to run, which I will post in a separate message in about an hour or so, because it's time for me and my dog to go for our morning hike! :-)

So standby, and please look for something in a little bit.

And, Pez, what's up with the feedback level and S-ATT on your setup?

Thanks!

Scott

[Pez]

Hi Scott, 

In my screenshots, the setup was as follows:

HL2 (5w) > Amplifier (TURNED OFF) > 60dB RF sample port  > High impedance converter box with an additional -7.5dB) = -67.5dB on a 5 watt signal. 

The Amplifier is usually ON, so it's usually fine with that attenuation,]. But it was OFF when I did the tests because I did it remotely (via remote desktop). 

Does that all add up? Or is there something else I need to look at? 

73

[WU2O Radio]

Hi Pez,

Unfortunately that disqualifies it as a valid data set. Not sure what sort of amp you have, but -67.5dB is about 10dB too much attenuation even at 1500W. With hardware that has poor RF isolation performance it is important to keep feedback-to-interference levels as good as possible. Thus you want a feedback level that puts you in the neighborhood of about 25dB S-ATT (not really attenuation in the HL2, it's a programmable gain amp, but relatively speaking).

If you could repeat the data collection exercise with the amp turned on such that you achieve the best possible feedback levels it would be most appreciated. Being able to see S-ATT, the linearity control panel, and the AmpView display is super important. A video is better than a screenshot.

Also see my other experiment which I will post momentarily.

Thanks!

Scott

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/d1f120ac-f36b-4282-9bba-3c8dcb8efcd3n%40googlegroups.com.

[WU2O Radio]

All,

To the HL2 users who have posted their results: a big thank you!

The dotted nature of the line is a natural consequence of the two tone waveform when using the standard 700 & 1900 Hz tones.

In this example from my radio we see a single blue line of dots:

Changing the tone pair to 70 and 190 Hz on my radio we can see a nice, solid, single line of data:

Just to get a better look at the data, if people would post their results with a 70 & 190 Hz tone pair it would be most appreciated. Being able to see S-ATT, the linearity control panel, and the AmpView display is highly desirable. Video is better than screenshots, if possible.

BTW, leaving aside the PS difficulties we are looking into, when you use 70 & 190 PureSignal works just as well, it can help with certain amp's that are picky about 700 & 1900 (e.g. ACOMs) and, most of all, it doesn't sound like you are "singing the song of our people" 😄

Thanks & 73!

Scott

[Reid Campbell]

Here is the control panels

https://www.dropbox.com/s/7hzmsw0afvgzbw6/VID_20230503_144818336.mp4?dl=0

and what is on the spectrum display

https://www.dropbox.com/s/kma5ymanhppf6go/VID_20230503_144835271.mp4?dl=0

Cheers 

Reid
Gi8TME/Mi0BOT

--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/CAPC_m5EAt%2Bs24bcGRypeP7%3DwkubHivjV7PFuWG1f%3D3uP4XSRbg%40mail.gmail.com.

[Pez]

Hi Scott, 

I need some guidance here. Firstly, it is actually working and correcting well. I am seeing -50dBc at times. With and without the external AMP.  But I understand, you want to see a stronger signal at RF3... However, would doing this test in barefoot crosstalk mode be valid too? Isn't that what my results look like? What am I missing? 

I was doing Single Cal tests as asked, and as far as I can see by the settings above, - this is going to mean -31 regardless of the feedback... Is this setting incorrect for me? Can you please let me know if that is correct or not, and also exactly what I should be seeing in the S-ATT window, and in what PS modes? 

FYI - The amp is run at 400W when it is ON. It is capable of 2000W. It has a -60db 50ohm port built in. That's why I'm using it with a high impedance add-on (-7.5dB)  - and it works! 

I have an external RF sampler I can use here, that will be closer to -47dB.  I can try this if you like, but please let me know what you think I should expect to see in the S-ATT.. 

Thanks in advance for your help.

73

[WU2O Radio]

Hi Pez,

Working, and working optimally, are two different things. And, given the difficulties that some HL2 users are going through with this, there is some reason to pay attention to detail, if only to rule out any possible issues with adjustments that ordinarily might not matter.

Good thinking to post that screenshot. The "Force ATT ... to 31 ..." option is likely to screw things up when using Single Cal. If you perform a two tone transmission with PS-A ON in order to obtain an automatic attenuator setting, then click Single Cal, after that transmission the next two-tone will see the S-ATT go to 31 and subsequent Single Cal's won't be as optimum. I recommend turning "Force ATT..." OFF for testing, and even turning it off all the time unless you are getting ADC overload indications during normal operations.

Understanding that I don't have an HL2 to play with, and that I have not had the opportunity to read the firmware (gateware) source code, what I *have* been able to read about the HL2 hardware and firmware suggests that:

At an RF3 input level of +12dBm the AD9866 programmable gain amp will be set to -12dB, thereby achieving 0dBm internally on the AD9866, and that this is mapped to an S-ATT value in both Thetis and PowerSDR of 31dB. Conversely, while the maximum programmable amp gain in the AD9866 is +32dB, this is limited in the firmware to +20dB in order to match the dynamic range of the S-ATT setting in software (0 to 31dB), hence a feedback level of -20dBm will send the programmable gain amp to +20dB and the S-ATT value in the software will be shown as 0dB.

If anyone who *really* knows what is going on can chime in here with additions and corrections that would be great!

Assuming those numbers are all correct (and I did write "assuming"), then it is easy to infer what your feedback levels are and what they should be from looking at the S-ATT value.

At 400W, or +56dBm, less 67.5dB coupling and attenuation, that's -11.5dBm at RF3, the AD9866 programmable gain amp will be set to 11.5dB, and thus the S-ATT indication should be driven to about 8dB, give or take given how accurate or not the exact gains and losses are at a given frequency.

If you did this with your external, -47dB sampler, and assuming you still want to use the high impedance stage, then that would be a total path loss of 54.5dB and under the same conditions S-ATT would, of course, merely change by the difference between 67.5 and 54.5, i.e. instead of 8 you'd see somewhere around 21dB of step attenuation. This would present a much better feedback-to-interference level and PS performance might be improved under those conditions.

Finally, noise performance might also be improved if, rather than use a high impedance transformation at the input to RF3, instead place a relay there similar to K2 and drive it with the INTTR signal that drives K2. Obviously wire the relay such that during TX it presents the feedback signal to RF3 and during RX it is open. This would be equivalent to what is done on the Apache hardware.

Mind you that this is all "merely" optimization and, while worthwhile in its own right in order to achieve maximum PS performance, there is still a good chance there really is a bug in Thetis that is causing the problem of bad PS solutions.

73,

Scott

On Wed, May 3, 2023 at 10:13 AM Pez <perrin...@gmail.com> wrote:

Hi Scott, 

["Christoph v. Wüllen"]

Hi all,

since Thetis and piHPSDR use exactly the same engine (WDSP) under the hood,
and even exactly the same version of WDSP, it might be instructive to
see what piHPSDR makes out of this.

Setup: RX running at 48k (!) sample rate. This has been done since it has
been speculated that PureSignal does not work to well there.
The results with 192k are essentially the same.
piHPDSR was running on a Pi4 (but the same results are obtained
on a Macintosh).

Two pictures (PDF) show the feedback signal. "ON" means PS is running
and I see IMD3 about -48dBc which is what I usually get with PS. The HL2 is
transmitting barefoot into a dummy load and feedback comes from the
crosstalk of the T/R relay. When I hit the "OFF" button to switch
off PureSignal (see the "OFF" picture) then IMD3 goes to -35 dBc which
is not bad at all and is the "native" IMD of the HL2 PA.

BTW the SetPK value is hard-wired to 0.230 for the HL2 in the program, and when
running, I see GetPk values slightly larger, as reported here recently.
I use automatic attenuation to adjust the feedback level, here it is close
to optimal (feedbk=154).

This is meant to give some additional info for the developers of "other" SDR
software, hope this helps.

[WU2O Radio]

Thank you, Christoph.

Question: are you seeing the occasionally aberrant PS results with piHPSDR that Thetis users are seeing? I would guess not?

73,

Scott

[Graeme Jury]

Hello Group,

This is very close to what I get with piHPSDR running on my Linux Mint
desktop. My attenuator is at 19 and everything is about the same except
that I don't get quite as good a result as Christoph with my corrected
signal about -45 to -50 dBc although I am using a not so linear 100 watt
amplifier. PS just works for me and very occasionally it seems to auto
re-sync and away it goes for a couple of hours again. My pickup is 30t
on a 25mm toroid in a little PCB box with the coax with its shield
broken passed through it. Its about -32 dB and flat to 60 MHz. The
attenuator is 470 ohm series and 47 ohm terminator on the toroid side.

73, Graeme ZL2APV

[Pez]

Hi Scott,

Thanks for the excellent reply - I appreciate it. I will spend some time obtaining some better PS values and will repeat the tests. I will say that the PS issue has appeared the same for me (the same issue as everyone else is seeing here), regardless of when I tested barefoot with crosstalk, or a using the 400w amp with the -67.5dB sample, or even a with a different 100w amp that I tried with my the -47.5db sampler. (Its a fixed -40db inline sampler with the -7.5db impedance changer box = -47.5dB). 

In all cases, my PS corrections have been very effective (but unreliable, as per the observations in this thread from us all). It's funny that general consensus is that PS won't work the way mine is currently configured? But I can reliably hit -48dBc to -50dBc results with my setup. It's actually correcting OK. It's almost like the PS levels are far less critical than we think? But I won't commit to that, I'm just noting my experience with this so far. 

73

[r webby]

Hello everyone,

Important note here: PS _never_ works properly on an HL2 with samplerate > 192000

I am not sure what version of Thetis you are using, but I considered not making anything higher available for the HL2 because of this. Then I relented, thinking that it precludes wide(r) band use for RX only.

Therefore:

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/fa4737d1-37ae-4bb4-832b-552072b37002n%40googlegroups.com.

[Pez]

Hi, I have been running at 384000 in Single Cal mode with a result of -48dBc to -50dBc. It mostly works fine for me - only the occasional issue with that. It is the PS-A mode that is causing the real trouble. 

I'm not sure I understand the changes you mention totally, but please don't take my PS Single Cal away at 384000!!! I like my -48dBc at this rate. Even if it's not perfect! 

Best 73

["Christoph v. Wüllen"]

> Am 03.05.2023 um 19:15 schrieb WU2O Radio <wu2o....@gmail.com>:
>
> Thank you, Christoph.
>
> Question: are you seeing the occasionally aberrant PS results with piHPSDR that Thetis users are seeing? I would guess not?
>

Hi Scott and Reid (and others),

do you think it is possible that the WDSP PureSignal algorithm sometimes produces
IQ samples that exceed the range (-1.0, 1.0)? This would perfectly explain the
"crazy" spectrum (integer overflow in double ==> sint16_t conversion).

The reason for this speculation is

- in piHPSDR, I had to switch off/on PS upon "TUNE", a typical max-amplitude
situation. The comment says that I have done so to avoid PS artifacts.

- in piHPSDR, I have reduced the amplidude of the two components of a
TwoTone signal from 0.5 to 0.49. I do not remember why I did this put
perhaps had a good reason.

@Reid:

It is easy to do one or all of the following tests

- detect and flag "unitiy interval" violations of the floating point
TX IQ samples

- slightly reduce the amplitude of the TwoTone generator

- multiply the TX IQ samples with a constant slightly less than
32767.0 before converting to int

and see whether this has an effect.

Yours,

Christoph.

[r webby]

Don't worry, its still there. It will stay, too but as far as I concerned it is for RX only.

Even if it turns out that 384000 is not in the Metis (P1) spec, I will leave it there, but do please note the warning. We will definitively find out whether or not PS _should_ work on 384000 hopefully from someone that knows for sure in this thread.

:-)

Regards,

Steve, KLJ

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/644a4d2d-45ba-4ea6-aaa9-1e1d3fb21a30n%40googlegroups.com.

[r webby]

Hello Chris!

It is possible. How about setting a conditional breakpoint to detect overflow? I can certainly do this in the Windows wdsp, perhaps if you have the source for piHPSDR. you could do the same?

Even an assert() would probably do.

Steve

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/8B6195EF-B688-4F8C-8D5B-94207DF7DA84%40darc.de.

[r webby]

Caught it! Finally.

SR 192k. You'll see the actual real issue here. Then I re-key, and it clears it, and everything works properly once more.

audioenhancedps.co.uk/GOTIT.mp4

HTH

Steve

["Christoph v. Wüllen"]

I located the following code in the Thetis repo, at the end of file
ChannelMaster/networkproto1.c. Since I do not have a Windows
development machine, I would love to see if somebody could
change the two constants 32767.0 below to, say, 32500.0, and
see whether this has an impact on the "crazy PS spectrum"
after recompiling Thetis. You simply need a recompiled exe file
and can replace the existing one (after backing up the original).


DWORD WINAPI sendProtocol1Samples(LPVOID n)
{
DWORD taskIndex = 0;
HANDLE hTask = AvSetMmThreadCharacteristics(TEXT("Pro Audio"), &taskIndex);
if (hTask != 0) AvSetMmThreadPriority(hTask, 2);
else SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_HIGHEST);

int i, j, k;
short temp;
double swap;
double *pbuffs [2];
pbuffs[0] = prn->outLRbufp;
pbuffs[1] = prn->outIQbufp;

while (io_keep_running != 0)
{
WaitForMultipleObjects(2, prn->hsendEventHandles, TRUE, INFINITE);
// if ((nddc == 2) || (nddc == 4))
if (pcm->xmtr[0].peer->run && XmitBit)
{
// if eer/etr mode and transmitting, overwrite LR data with EER data
memcpy(prn->outLRbufp, prn->outIQbufp + 256, sizeof(complex) * 126);
}
if (!XmitBit) memset(prn->outIQbufp, 0, sizeof(complex) * 126);
// WriteAudio (30.0, 48000, 126, prn->outIQbufp, 3);
// WriteAudio (60.0, 48000, 126, prn->outLRbufp, 3);
for (i = 0; i < 4 * 63; i += 2) // swap L & R audio; firmware bug fix
{
swap = pbuffs[0][i + 0];
pbuffs[0][i + 0] = pbuffs[0][i + 1];
pbuffs[0][i + 1] = swap;
}
for (i = 0; i < 2 * 63; i++) // for each sample from both sets, 8 bytes per
for (j = 0; j < 2; j++) // for a sample from each set, 4 bytes per
for (k = 0; k < 2; k++) // for each component of the sample, 2 per
{
temp = pbuffs[j][i * 2 + k] >= 0.0 ? (short)floor(pbuffs[j][i * 2 + k] * 32767.0 + 0.5) :
(short)ceil(pbuffs[j][i * 2 + k] * 32767.0 - 0.5);
if (prn->cw.cw_enable && j == 1)
temp = (prn->tx[0].cwx_ptt << 3 |
prn->tx[0].dot << 2 |
prn->tx[0].dash << 1 |
prn->tx[0].cwx) & 0b00001111;
prn->OutBufp[8 * i + 4 * j + 2 * k + 0] = (char)((temp >> 8) & 0xff);
prn->OutBufp[8 * i + 4 * j + 2 * k + 1] = (char)(temp & 0xff);
}
WriteMainLoop(prn->OutBufp);
}
return 0;
}


> Am 04.05.2023 um 14:43 schrieb Toonces Cat <toonces...@gmail.com>:
>
> Group,
>
> A couple of clues... The wide TX on my 10E with Thetis ALWAYS happens on the initial KEY UP and NEVER part way through a transmission. Also when I use PS-A the pure signal never locks completely for the first few seconds.
>
> 73's
>
> Mike WA3O

> 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.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/9cd8d6ca-3f1c-43dc-a361-64693cd5df92n%40googlegroups.com.

[WU2O Radio]

Mike,

Can you please do that again but with the AmpView display visible also? We need to be able to see the AmpView display clearly.

Is this the barefoot output of the 10E, or through an amp?

Finally, would you please confirm this is a US spec. 10E, i.e. a 14 bit Hermes?

Thanks,

Scott

On Thu, May 4, 2023 at 9:00 AM Toonces Cat <toonces...@gmail.com> wrote:

Quick video with a voice transmission. I will try to get a better one later.

https://photos.app.goo.gl/n614zWnZ49UnMvzT6

73's

WA3O

On Thursday, May 4, 2023 at 8:43:21 AM UTC-4 Toonces Cat wrote:

Group,

 A couple of clues... The wide TX on my 10E with Thetis ALWAYS happens on the initial KEY UP and NEVER part way through a transmission. Also when I use PS-A the pure signal never locks completely for the first few seconds.

73's

Mike WA3O

On Thursday, May 4, 2023 at 6:04:56 AM UTC-4 radiowe...@gmail.com wrote:

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/dc7e1cfb-bba3-439e-b2c5-89eaca928500n%40googlegroups.com.

[Reid Campbell]

Hi Christoph,

I tried that and I'm still getting the wide signal. What I have noticed
is that it might depend on the loading of your computer. I was just
running Thetis and couldn't get the fault but when I put my usual load
on (Firefox with 20 pages, Chrome with 20 pages, Thunderbird, MSHV, Grid
tracker, What's App and Visual studio) it was a lot easier to reproduce
the fault.

Cheers

Reid
Gi8TME/Mi0BOT

[WU2O Radio]

Thank you, Mike.

73,

Scott

P.S. It looks like wdsp has been ruled out. Now other parts of the code need to be looked at.

On Thu, May 4, 2023 at 12:19 PM Toonces Cat <toonces...@gmail.com> wrote:

OK here it is...The best i can do without a tripod.

https://photos.app.goo.gl/cB7VUAjSeYu4iuEt9

1KW out. LDMOS amp . on a Optibeam 12-6 on 20 meters. The radio is a 10E with the firmware switch on the bottom cover and a blur front panel.

73's

Mike WA3O

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/6779baf8-d399-46e6-8b65-5a7981d4aca3n%40googlegroups.com.

[WU2O Radio]

All,

A small status report for everyone follows. Steve G7KLJ: please correct me if I get anything wrong!

At this point people can probably relax a little bit on the testing. Everyone's inputs so far have been very appreciated and very helpful. Steve and Warren have worked VERY hard to rule out the WDSP code and that has been ruled out with very high confidence. So has the firmware/gateware. The problem is associated with Protocol 1 firmware support in Thetis and seems to affect any hardware running P1. With the vast majority of Apache hardware users running Protocol 2 this leaves a few Apache users and all HL2 users running P1. Maybe some Red Pitaya folks, too. Focus is now shifting to other parts of the code. Additional tests may be devised to assist with this and, if so, then volunteers will again be needed.

A few other things have come to light that would assist in achieving optimal PureSignal performance and that may not be given as strict attention in the HL2 community as they are in the Apache community. None of these things has anything to do with the bug, but with or without the bug they are worth paying some mind to.

First is the understanding that in Thetis, with P1, only the 192KHz sample rate is supported for PureSignal.

Second is the value of SetPk. It appears that 0.2330 is a very good choice, and if it is not the default in your version of Thetis then certainly consider changing that yourself.

Third is the attention to detail in the adjustment of RF feedback signal level as measured at the RF3 connector. Feedback to TX interference isolation in the HL2 design is similar to that of the original Hermes, of course. That is to say it is not very high. Heck, some folks are using the onboard T/R switch as a coupler! Adjusting the feedback level to provide a relatively high feedback-to-interference ratio can't hurt. To that end the RF coupling factor and any additional attenuation or impedance transformation losses should be arranged to result in a level of at least 0dBm or maybe even a little more, but don't exceed +10dBm. If the auto-attenuate function arrives at an S-ATT of between 20 and 25 at your maximum power level you are doing well. Obviously this will degrade at lower than maximum power levels, but as the power level drops the need for superior IMD drops with it.

Another thing that was interesting is the idea of presenting a high impedance to the RF3 port to decouple it from the normal RX path. In the Apache universe people simply arranged an RF switch that closed during TX on the feedback port and opened during receive. It's an option that might improve both noise performance and isolation.

More to follow certainly as the debugging continues.

73!

Scott

[WU2O Radio]

.233 will not work for the Hermes. It is only for the HL2. Remember you are posting here in the HL2 Google Group so discussions and direction will be largely HL2 specific.

Use the Thetis default for your Hermes.

Apache discussions are, of course, the norm over at the Apache forum where I know you are also a member.

That TAPR board is an excellent example of what I was describing to the HL2 folks regarding switching on the feedback input.

73, 

Scott

On Thu, May 4, 2023, 18:07 Toonces Cat <toonces...@gmail.com> wrote:

Scott,

I am running 192khz and Protocol 1 and I tried Setpk of 0.2330 and pure signal doesn't lock. I want to also mention that I am using the Pure signal relay box from TAPR seen here 

https://tapr.org/product/tr-plus-transmit-receive-switch-for-pure-signal/

Currently "OUT OF STOCK" : (

A week ago I tried several Protocol 2 firmware's with my 10E and Pure signal did not work at all.

I know the 7000dle is a different firmware/ radio but, I used Protocol 1 exclusively with never a fail with PS with the same exact coupler and amp I used on the 10E now.

73's

Mike WA3O

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/af233d02-f1ba-4494-8cc8-a9d3c42ec2c4n%40googlegroups.com.

[WU2O Radio]

Mike,

I'm not sure what points you are trying to make, but the tone of your post seems quite negative.

First, when someone receives "ABSOLUTELY NO" help on a forum, group, mailing list, whatever, that's because nobody has a way to help. People want to help. But sometimes there is no help to be had. And at that time it was certainly true. So trying to blame someone or some group for something like this is pointless. Others that may have had the problem also had nothing to offer, they even posted in your thread six months ago. But at that point six months ago nobody had anything to contribute. It was not because the forum is a bad place or people were ignoring you.

Second, emailing developers in the casual, and very tiny, open source environment that we exist in is generally unlikely to get a positive response. Some developers are OK with it. Others want you to use a public forum post. Others want you to use their Git repo issues tool. And even if you do what they want in that respect they are going to prioritize both their life and hobby activities in the manner with which they please. Getting ignored by dev's is almost the rule rather than the exception. Nobody's paying them, indeed they don't want any money because then they often feel beholden. And this issue doesn't really get any better when you actually pay for your software. The complaints about Flex software are legion.

Yes it sucks to have a problem with the software, and it sucks worse when that problem is not shared by a large percentage of the users of that software, and it sucks even worse than that when there is a work-around identified because then dev's will tend to prioritize the issue even lower. And if people get shrill about it, they may walk away entirely. At the end of the day all you can do is hope to get lucky, i.e. hope a skilled dev. decides the problem personally bothers them sufficiently to do something about it.

73,

Scott

On Thu, May 4, 2023 at 8:23 PM Toonces Cat <toonces...@gmail.com> wrote:

Scott,

 I am fully aware where I am posting. I posted this very similar problem on the Apache labs site over 6 months ago with ABSOLUTLY NO help from anyone then I posted a posible way around the problem with Steve's program and now there is some movement.

https://community.apache-labs.com/viewtopic.php?f=20&t=4396

I even tried to PM Ramdor about this and was ignored.

Mike WA3O

[dsolbe...@gmail.com]

Has this bug in Thetis been corrected?  I am experiencing the same problem.  I can get PS working with Single Cal.

Don K9AQ

[WU2O Radio]

I can't speak for any HL2-specific builds, but this was fixed in the Apache-focused version 2.9.0.8 in June 2023. The problem affected any Hermes-based system running Protocol 1 firmware, not just HL2.

Also check your setpk value. I'm not an HL2 owner, but I believe for HL2 folks it should be set to 0.233 or thereabouts.

73,

Scott

[Don Solberg]

Thanks Scott,

I am running an older version of Thetis, 2.9.06.  There is a newer version, V.2.10.3.5 Beta 1 available.  I will try upgrading to that.

Thanks for pointing that out.

73,

Don K9AQ

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/S39CWXGFmyE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/3bf330d8-d7fb-4201-af2e-1ac35a1def32n%40googlegroups.com.