LimeSDR-mini LTE constellation

I measured LTE QAM256 constellation on OAI LTE downlink. Compared to USRP b210 LimeSDR-mini has bigger EVM, though almost the same downlink speedtest. Is it as expected or something wrong with calibration?

USRP

LimeSDR-mini

Not sure if there are many people using OAI with LimeSDR hardware and most seem to use srsLTE or Amarisoft, so I’m not sure what the current status and what you should expect. Though it’s quite possible that there is scope for optimisation. May be worth asking on the OAI list/forum.

I can retest it with srsENB, but the result will be the same, I think. Looking at the CW there is a quite high shelf of noise at 10…100 kHz offset (-70…-80 dBc)

it looks, it was a gain issue.

In SRSLTE gain 65 has the same constellation. GAin 60 an less has worse constellation than USRP but acceptable.

Gain 60

Gain 55

Gain 65

Hi,

I don’t have anything to add but may I know,

  1. what the test equipment You use ?
  2. what the PSU You use ?
  3. What Clock You use.
    Thank You
    Anton
  1. R&S FSV
  2. USB powered
  3. Internal

Sorry, but isn’t it QAM64?

My mistake. Of course QAM64

Would have been nice to see the allocation and EVM summary for the second picture similarly as for the first one.

If you can do vector analysis, I would recommend you to check the frequency error. My Mini was off by 12kHz on 1820Mhz (band3 LTE), which was a lot worse than the standard specifies, and it caused newer phones not to find the network.

To be fair, the Mini is not sold as it can be as precise as modern telecom standards require it to be (mostly 0.5ppm or 50ppb), but in most cases it is a lot worse than the advertised 1-2ppm. It can be calibrated to meet the 1-2ppm precision, which is more than enough to do LTE emulation with good performance, however it has to be calibrated first.

The DAC offset can be changed with LimeSuite, but you have to go back and forth between SrsLTE and Limesuit and constantly check the frequency error on the FSV to find the best settings. With the FSV you can do this relatively easily, I would highly recommend you to do so.

Frequency error is not an issue for me. The problem persisted on high modulation schemes only.

Hi Leased,

I assume Your FSV have K40 capabilities so please can You do phase noise measurement on the freq You working on and if You have a resource for low noise freq reference it will wonder if you can compare the onboard ref to ext low noise reference.

Regards,
Anton

No, FSV has no phase noise options installed

in that case, Could You use a better reference for limeSDR.
Maybe the onboard TCXO is not good enough in terms of phase noise.

Unfortunately no. I have 10MHz reference only, but LimeSDR mini has 40MHz input, that is not common

You can use a 10MHz reference. See:

And notes at:

https://wiki.myriadrf.org/LimeSDR-Mini_v1.2_hardware_description#Clock_Distribution

I understand that the phase noise will not be great, due to limited space and cost optimised nature of LimeSDR Mini. If you need higher performance, would suggest LimeSDR-USB plus external GPSDO, or LimeNET Micro, which includes this and can be used as a USB peripheral.

From my understanding the phase noise of fractional synthesizer can not be improved by using of external clean oscillator. Only frequency error and maybe phase drift at low offsets can be improved.
Anyway, with gain reduced as I mentioned above the EVM is good enough to have clean LTE QAM64 transmission.