Limesdr mini loopback failed solved (?) - tweak ICT_VCO or return it?

Projects LimeSDR
#1

Hello everyone,

I bought an limesdr mini. LimeQuickTest gave me the classic loopback error in an intermittent way. It would sometimes fails when cold ~30°C , but always when hot.

random-test-fail
random-test-fail983×1360 116 KB

I have a month to return it. During that time, I tried to investigate what might me happening.

Using the gui, I turned on only TX, and checked what it outputtted an clean signal - it was OK.

However, RX VCO have a weird behavior: Even if sucessfully tuned, it was not in fact stable it seems.

By cranking the VCO (0x0120[7:0] ICT_VCO) bias current from 128 to 200 the VCO responded adequately, and I could receive signals without ‘holes’ while sweeping.

I also noticed that increasing PW_DIV2_LOCK also helps a bit.

I reflected that change in the library, and test is now always passing, event when hot.

My Question: Is there any drawbacks in increasing ICT_VCO ?

#2

I think telling people that it is fine that LimeQuickTest fails is not right. My case looks very similar to many people it seems, and I when the test fails, it might means that RX VCO is not really locked, automatic tuning may fail for some frequencies.

It seems I can indeed manually get the VCO to properly lock at a bias of 128 by tweaking CSW_VCO, but the spot is quite narrow. When it get a bit more hot, I can tweak it at least with a bias of 132. With a bias above 200, it can automatically tweak at 70°C and even 80°C around 1000MHz, and then limequicktest can pass at 70°C.

The automatic tuning with low bias, sometimes stops at an non ideal csw: tx loopback is ~40db less than expected with an increased bias.

I think the automatic tuning is too simple, or VCO lock detection is somewhat defective. The automatic tune sometimes select the middle of an array of csw value that just does not works, when the first and last value works ok.

Should I return my board for being too delicate ?

#4

Apologies, thought I’d replied to this post already. With LimeSDR Mini the VCO bias current was being set lower than with other boards and this was, as you noted, causing issues. After discussing with the team it was decided that this must have been done in error, as there was no recollection of it being set lower for any particular reason. This has now been fixed in Lime Suite master and a new release is planned for around the end of July, when updated Ubuntu packages etc. will be available.

#5

Hey! A bit late to the party, but was this fix applied? I’m getting this intermittent test failure problem. It seems to oscillate consistently, the first run succeeds, the next fails and so on. It may have a different cause but the behavior is very similar.

Edit: probably a different problem. I’ve only reproduced this on Windows 10 (limesuitegui build date 2021-07-25, version 20.10.0). Have tested the same device on linux (arch, the community/limesuite package, same version, build 2021-06-22) and it works reliably.