TxTSG output

#1

Hello @Zack,
We connected the output of our board with TxTSG configured to generate /8 tone.
The center frequency is set to 2.25GHz and the clock freq is 200MHz so /8tone is @2.275GHz.
There is a 18dB amplifier at the output of the chip.
The concern is the spurious or noise near the tone.
Question: is this correct? (most likely not). How to fix it?
IMG_20190513_182317%20(Small) IMG_20190513_182257%20(Small)

Looking forward to your inputs and thank you for your time again.

#2

Hello @Zack,
Another measurement that might be useful.
The first one is on our board and second one from the LimeMini board. Our Ref clock is 50MHz and are using the api functions to set the frequency.
IMG_20190514_093000%20(Small) IMG_20190514_093411%20(Small)
Regards,

#3

A lot of variables to consider. I would check ref clk and power supply performance first.

#4

Thanks @Zack. May I ask specifically what you would like us to check for the ref clk? The spurs here seem to be PFD spurs (I could be wrong). The spurs are at harmonics of 136KHz.
We are using the same ICs used in the LimeMini design for the power supply and have taken care of the routing to the best of what we could.
Please advise,
Regards,

#5

Hello @Zack,
We think that the spurs are related to the settings (may not be 100% because of the clock and power supplies) because of the following reasons:

  1. By changing CGEN, the spurs move a bit (and this is repeatable)
  2. The spurs were present even when the Refclk was turned OFF.
    Looking forward to your inputs,
    Regards,
#6

Hi @EnthuMan,

Power regulators from LimeSDR-Mini works at 2MHz, hence I have no idea where 136kHz comes from.
Do you have provision to supply your 50MHz clock to external devices? If yes, you could supply it to LimeSDR-Mini and see if the same spurs appears. Check this for more information.

#7

Hello @Zack,
We will try feeding the LimeSDR-Mini with an external clock but I suppose that will involve changing quite a few things.
Another update since the previous post:
We changed the CLKGEN frequency to different numbers and observed the following:
spurs2
Notes:
Delta freq1 and freq2 are first and second harmonics that change with the freq being set with CLKGEN (blue column in MHz). The 254KHz spur remains constant (so probably related to the 50MHz reference).
We are trying to go through the GUI code to understand what is the difference in the way the frequency is set in the GUI and the driver.

#8

Hi @EnthuMan,

Few more thoughts:

  1. Try to supply LMS7002M from lab power supply if you have a chance.
  2. Check CGEN and SXT VCO Comparators state to be sure that these are locked.
#9

Hello @Zack,
Checked the CGEN and SXT VCO comparators. Both are locked (and read the same register values as the LimeMini using GUI).
Also, fed 1.25V externally and the result is the same.
The way to configure SXT seems to be a little elaborate in the GUI code. We are trying to understand the differences but may I ask if you could please highlight.
Also, the output of the LimeMini is remarkably clean, almost as if it is because of some additional filtering.
Regards,

#10

Hello @Zack,
Apologies for writing again. We really need to narrow down on the problem (whether it is a software (the way the registers are being written) or hardware issue).
Any help will be greatly appreciated.
Thanks,

#11

Hi @EnthuMan,

Can you share schematic?

#12

Hello @Zack,
Please send me your email address.
Regards,

#13

Hello @Zack,
We noticed something and seek your help for that.
CGEN needs to be configured before SXT else SXT output is not stable.
The reference clock for SXT is xclk_tx so why would CLKGEN matter here?
Further the spurs seen at the output change with CGEN frequency setting.
In short, CGEN configuration and SXT are related but it is not clear.
Regards,

#14

Hi @EnthuMan,

It should not depend.

#15

Hello @Zack,
I have checked multiple times now (even on the LimeMini board). If I go directly to the SXT tab in the LimeSuite GUI, enter the frequency and press calculate I can see signal show up at the entered frequency but it is very unstable with a lot of noise.
However, if CGEN is set first and then SXT is configured, the output is normal.
Regards,

#16

Well, if CGEN is not configured, it still generates some not stable frequency. Hence TxTSG provides digital noise to DAC which is visible in RF. What else to expect from not configured part?

#17

Hello @Zack,
I agree that if the DAC is generating noise then it will translate to a noisy spectrum. (the first one is the clean spectrum and the second one is the noisy).
IMG_20190516_151324%20(Small) IMG_20190516_151239%20(Small)
May I ask you to please confirm the following:

  1. If CGEN is not configured, then it is generating some frequency (or is it not generating any frequency at all?)
  2. For TxTSG to produce “significant noise”, doesn’t it have to receive DIQ data, FCLK and IQSEL. Is the LimeMini doing that?
  3. When we enable TxTSG mode from the API, we still see the spurs and again these spurs are related to the CGEN frequency because they move if the frequency is changed.

Sorry for repeatedly bothering you with this but we are in a fix at the moment.
Please guide on the exact sequence to follow for configure TxTSG? (maybe the portable driver API function call sequence).