LMX2571: EVM Module, higher than expected distortion when using I2S Analog FM

Dean,

Thank you for the response!

1) Yep, already tried that one!

2) Lower the frequency to 1 KHz, and also boosting the I2S transmission frequency increased the number of samples. Doesn't really make it "better" so to speak.

3) The loop bandwidth is set to 204.8 KHz (default EVM settings), which I figured would be enough? I also read that "twice the modulation rate" was recommended somewhere, I forgot what documentation exactly.... But I figured that meant modulation frequency (in my case 100 KHz analog FM which is perfect), perhaps it means FSK step frequency? Or perhaps I am completely wrong?

4) I am not apply gaussian shaping to the pure tone Analog FM waveforms... I don't think my particular application would benefit from this actually.

One thing that immediately stuck out like a sore thumb to me was the use of the shielded coaxial jumper cables from the picture inside the PowerPoint slides you uploaded. I have never seen jumper cables like that, and in fact the ones I am using are unshielded (I2S clock/data/fs probably coupling all over my board and ground plane!). Do you have any insight on a manufacturer or source for those types of board to board connectors?

Thank you!

Brandon

Hi Brandon,

I have couples of question.
What is the reference clock (OSCin) frequency?
fpd frequency?
Output frequency = 450MHz only?
Delta sigma order?
Fractional denominator (DEN)?
Frequency deviation?

-OSCin is the onboard 20 MHz LVCMOS device that comes with the EVM module. The frequency accuracy is quite poor, more like 19.95 MHz, this shows up in the Carrier Frequency output at 450 MHz, where is is actually 449.8 MHz or so. I don't think this really matters in the long run, the frequency will just be wrong! (Unless I am wrong about that :) )

-fPD set to 100M, tried 80M but same results

-Currently only using 450 MHz for testing, planning on using other frequencies later once I have a good baseline

-Delta Sigma Order was not configured in my case for Analog FM. Can I benefit from enabling dithering here?

-DEN set to 2^24 = 16,777,215

-Tried frequency deviation at 30 KHz and 15 KHz with very similar performance, in fact there was about 1% more distortion at 15 KHz.

Also a side note, I am using the onboard 3.3V LDO regulator, taking 5V directly from the development kit I am attached to for programming. I did not see any noticeable difference when using a Linear lab power supply (surprisingly).

Let me know if you have any other questions!

Thanks.

Hi Brandon,

Your configuration looks OK to me. I will give it a trial when my test equipment is back (sorry, I don't have an exact date for this, something I cannot control).
Since your sampling rate is very high, please try to increase the loop bandwidth. If you don't want to modify the loop filter in the EVM, try increase the charge pump current, disable the 3rd and 4th pole filter. Dithering is not necessary.

Noel,

Just gave it a shot. I doubled the charge pump current, and disabled the 3rd/4th pole internal resistors. The distortion measured was basically the same (around 0.2% less), just moving the board around changes the distortion more than this change. I'm not sure what the cause of this issue is...

The distortion wave riding inside of my demodulated sinewave is almost exactly 250 KHz as measured by the oscilloscope. I don't think anything in the system is running at 250 KHz, which is particularly odd.

Please let me know if you have any other ideas.

Thanks!

Brandon

I just upped the CP current to max * 2.5 gain, and I was able to reduce the distortion further (1.6%)! Seems like that is the issue right now. Will modify the loop filter tomorrow with physical components and report back.

Thanks!