Frequency Drift

[Nathan Long]

I'm seeing some pretty bad frequency stability issues with my new HL2 board.  FT8 decodes are 60%-70% fewer than they should be while the frequency drifts around between transmit cycles, and even for several minutes after ceasing transmit.  See screenshots below.

Semi-stabilized (no transmitting)

During FT8 Transmit cycle @ 2W (rfi is from hvac) 60% - 70% fewer decodes during rx cycle.

After stopping transmit cycle.  60% - 70% less decodes than there should be.  Frequency stability seems to go back and forth for several minutes before settling. 

[Steve Haynal]

Hi Nathan,

This is an uncommon problem and hasn't been reported in years. I operate pretty much 100% FT8 and do not see this problem on multiple HL2s. With the Hermes-Lite 1, we had drift issues like this, but the Hermes-Lite 2 updated the oscillator to eliminate this drift. Can you please provide some more information?

** Is this a Hermes-Lite 2.0 purchased from Makerfabs? If yes, is it a recent purchase within the last one or two months?

** How are you dissipating heat from your HL2 during TX? Via the standard enclosure?

** What peak temperature does software report during TX?

** Please post high resolution photos of the front and back of your HL2 board. Modern phone photos work fine. In particular, I'd like to see U6, X2 and the components near U6.

73,

Steve

kf7o

[Nathan Long]

Hi Steve, 

Yes, this is a HL2 from Makerfabs.  I received it about 10 days ago.

I am dissipating the heat using the standard enclosure, shim, and thermal paste. 

The peak temp while transmitting at 5W is approx 37C @ 1.57A.

Since going back to get these measurements and pics just now, and operating with the top of the enclosure removed, the drift seemed to be much better, and then out of nowhere it began drifting again.   I've attached the pics of the top side of the board.  Pics of the bottom to follow once I've taken it out of the enclosure.

Thanks and 73, 

Nate - W9CO

[Nathan Long]

The markings on U6 are 5923B000 048UFN.   X2 seems to say R3844, but it's very hard to make out.

[Brock Nanson]

Just another data point for you... My board was likely in the same production run as Nathan's.  When I got it going with WSJT-X the first time, I also saw the waterfall slowly angle across the display.  After a few decodes it straightened up to be rock steady and I haven't noticed it since.

When I'm in front of it again, I'll try to connect right after it powers up and see if the drift happens when completely cold.

Brock

[Kristofer Danner]

37C seems off to me unless ambient is quite low. I see higher temps than that when doing FT8 with both of my HL2 units and they have fans. Could something be going on here where it is actually a bunch hotter than indicated?

KD

[Nathan Long]

This is curious.  I wish I had a thermal camera.  Nothing feels particularly hot to the touch.   The ambient temperature in the shack (basement) is probably 19C/66F or lower.  

The bad behavior seems to be independent of temperature, or at least not caused by high temp readings.   I've watched it be rock steady and then go to unstable without any significant temp change. 

[Steve Haynal]

Hi Nathan,

Thanks for the pictures. The X2 markings are hard to read, but could you check if it is R38A4 or R3844? It should be R38A4. 

A high of 37C seems okay to me. My HL2 in a similar basement environment never goes above that.

Did you install the heat shim?

Can you try some finger pressure on X2 and U6 and let us know if that changes the drift you see?

You have an extra pigtail on the RX line. I don't think that makes a difference, but out of an abundance of caution, can you try with this desoldered to see if there is a difference?

If you are handy with a soldering iron and SMT parts, there is an exposed ground pad on the back side of U6. Can you make that is well soldered?

Since you are in the US, it is easy for me to repair or exchange your board if needed. My address on www.qrz.com is correct. You can send just the HL2 board. I like to use the small prepaid priority mail boxes, ~$7.50. I will reimburse you from the repair pool.

73,

Steve

kf7o

[Brock Nanson]

Hi Steve,

In case you're interested, I powered up my HL2 from cold, which in my shack is around 20 Celsius.  I immediately started WSJT-X to see if I could replicate what I had seen previously.  And, yes, I was able to.  See the attached photo... and note how the direction of drift changes(!).  I watched this happen for a minute or two (back and forth), then hit Tune and watched the temperature climb into the low 30's (Thetis) before toggling back to receive.  At that point, the drift was gone.  And I mean, gone.  I watched a few stations over several FT8 cycles and their decoded data sat at the same frequency, right to the hertz.  A small few varied by a hertz, up or down, but that is more likely to be at the transmit end, IMHO.

After 10 minutes of receive, the temperature is just under 26.  Still no drift.  I've been running it for hours since it was assembled and have seen no drift issues once it's going.

This may not be perfection, but it's not going to worry me any.  It seems like HL2 was trying to track to the right frequency as it warmed up, based on how it went back and forth...but I don't know if that's even possible.  It would be interesting to find a really strong WSPR signal to compare over a full two minute cycle, but I'd wager that the waterfall would wander back and forth, with less variation as time goes on.

Brock

[Nathan Long]

Steve, 

R38A4 is correct.  I took all the suggested steps and still have the drifting.  The heat shim is installed with some good quality thermal paste.  I sanded the paint off the enclosure with increasing degrees of fine sandpaper, along with the shim, so they're both very smooth.  I imagine they're getting a pretty good thermal connection there.  I'll get the board shipped out asap.  Thanks again. 

73, 

Nate - W9CO

[SM7EGM]

Hello
 I only see that the frequency operation is shown in the pictures with wsjt-x.

Is there the same frequency deviation if you listen to a frequency normal station?

 Tried running wsjt-x on another computer?

 // Sm7egm, Ulf

[Nathan Long]

Hi Ulf, 

Thanks so much for that suggestion. I never suspected software at all.  Now to try and get my shipping label money back from USPS, hihi. Here is the waterfall from wsjt-x for 10MHz WWV (using USB).  Thetis is at the bottom and PowerSDR is at the top.   Any ideas of a setting in Thetis that would cause this problem?  

[DL1YCF]

> Am 29.03.2021 um 19:53 schrieb Nathan Long <natha...@gmail.com>:
>
> Hi Ulf,
>
> Thanks so much for that suggestion. I never suspected software at all. Now to try and get my shipping label money back from USPS, hihi. Here is the waterfall from wsjt-x for 10MHz WWV (using USB). Thetis is at the bottom and PowerSDR is at the top. Any ideas of a setting in Thetis that would cause this problem?
> On Monday, March 29, 2021 at 1:22:52 PM UTC-4 SM7EGM wrote:
> Hello
> I only see that the frequency operation is shown in the pictures with wsjt-x.
> Is there the same frequency deviation if you listen to a frequency normal station?
> Tried running wsjt-x on another computer?
>
>

To me this looks like the artifact of a resampler, most likely in the audio sample chain.

I have no intimate knowledge of the SDR softwares mentioned, but they do have
an audio resampler (to bridge the possible sample rate mismatch of the
„produced“ and „consumed“ audio samples). Try if you can switch it off.

Technical background: a resampler will try to get the „input buffer“
at half-filling, but upon startup it is empty. If ill-designed, it
will have oscillations before it settles down. This could be
an explanation why the problem vanishes after the first RX/TX/RX
sequence.

Perhaps the ThetisMaster can comment on that.

Yours, Christoph DL1YCF.

[Alan Hopper]

Hi,

that is interesting, I wonder if it is the audio resampler in Thetis.

73 Alan M0NNB

[Nathan Long]

For the screenshot I posted above, I have the audio device and audio settings set identically (sample rate,  buffer size, vac sample rate, rx1 sample rate, etc)  I've tried this on two different computers and seen the same behavior. I don't see the drift using SmartSDR or SparkSDR either.

[Nathan Long]

One other thing to note that might help narrow this down...  I monitored the signal from another radio and there doesn't appear to any discernable drift when using the tune button in Thetis.  However, using the tune button in WSJT-X does create a signal that changes frequency.

[Wayne Maxwell]

I had this same issue, using multiple Thetis and PowerSDR versions. Testing by monitoring a stable frequency indicated the RF appeared stable but WSHT-X was showing the same drift as the in the pictures. I cured this on my system by checking the "Force" boxes under Setup/Audio/VAC1 VAC1 Monitor and confirming all of my virtual devices were set to the same sample rates (the VB cables defaulted to 44.1K on my system so I changed them to 48K and if you are using VB Banana or similar check those settings). I have been updating Thetis to Beta1 and then Beta2 with no drifting noticed on WSJT-X. I am running my system on an i5 and Thetis typically shows 40-45% processor loading, the problem would get worse if I opened the browser as Chrome really likes to push the CPU to 100% when preforming some activities.

73 Wayne W1WNS

[Nathan Long]

Wayne,  

Thank you!   Checking the "Force" boxes to "Disable the resampler matching loop" completely fixed the issue!  I made sure that my settings were identical between PowerSDR and Thetis, along with VAC settings (48K for all), but the "Force" option doesn't exist in PowerSDR unless the Resampler box is ticked in the Audio -> VAC settings.  I hadn't discovered this setting previously, as it was working fine with an ANAN-10e.  As soon as I enabled that, the problem appeared again.  It's interesting that this setting by default is Off in PowerSDR and On in Thetis.

73!

Nate  - W9CO

[Steve Haynal]

Hi Nate and Wayne,

Many thanks to Wayne! This should be added to the #documentation  I would not have suspected resampling but can see how it may cause this issue. This issue never shows up on Linux with SparkSDR or Quisk with the automatic and free PulseAudio virtual audio connections. As Brock pointed out, you may see some drift as temperatures stabilize from cold in the first one or two minutes of use, but there should be no drift during normal operation.

73,

Steve

kf7o

[Ulf Hansson]

Picture taken with my CAT S64 Mobile after 5 min two tone test

--
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/vD_hmoyAa9A/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/458d612b-5368-4ea2-9e14-968a07b96284n%40googlegroups.com.

["Christoph v. Wüllen"]

Well, if one had „zoomed“ the waterfall in the SDR program one would have
noticed that the drift is only in WSJTX, not in Thetis/PowerSDR.

So in the first place, when reporting „frequency drifts“, it should be
based on the first-place IQ signal and not on what happens downstream
in the audio chain.

> During FT8 Transmit cycle @ 2W (rfi is from hvac) 60% - 70% fewer decodes during rx cycle.
>
>

> After stopping transmit cycle. 60% - 70% less decodes than there should be. Frequency stability seems to go back and forth for several minutes before settling.
>
>

> --
> 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.

[Brock Nanson]

This seems to have sorted it out for me as well, Wayne - thanks for your answer!

I'm running an older i5 and see similar CPU use, although the live graphics mode does cause variation.  Chrome... as much as I like it as a browser... seems to be a huge memory and CPU hog... LOL

73

Brock

VA7AV