search

sinara-hw / Fastino

Fast 32-channel, 3MS/s per channel, 16bit DAC EEM card compatible with Zotino
11 stars 2 forks source link

5 MHz Broadband Output Noise #58

Closed pathfinder49 closed 4 years ago

pathfinder49 commented 4 years ago

I'm having a look at the performance with the Fastino DACs set to various output values.

pathfinder49 commented 4 years ago

I've configured my system using the spur-free configuration with common mode chokes from #56.

With a 1kHz bandwidth, I observing broad noise peaks around ~4.6 MHz and ~9.2 MHz. These are present with a range of DAC voltages. In the following plots channel 31 is shown. This is appears to be representative of other channels.

Up to 5 MHz scan image

4-10 MHz scan image

Edit: These measurements were taken in a different (hopefully lower EMI) building to previous measurements presented in #56)

Edit 2: Fixed axes.

hartytp commented 4 years ago

A few questions here...

  1. These plots look rather different to the data you posted previously. The noise floor seems quite a bit higher in the new data (for the same measurement BW). Also, comparing to the orange trace in the old data there are fewer spurs, but I guess that's because the old narrow-band spurs are swamped by a broad noise peak at around 4.75MHz? From memory, it the spectra you posted above feel cleaner than the ones we were looking at last week (lots of spurs around 6-7MHz). Is that because you moved the equipment to a new (quieter) lab?
  2. Are the DACs updating continuously in this? I'm surprised not to see a big peak at the DAC clock frequency.
pathfinder49 commented 4 years ago

Unlike my measurements in #56, Kasli and Fastino were placed in a floated rack. I observe the following:

By eye, 5-10 MHz noise performance is better with Fastino in contact with the rack and an additional grounding strap.

pathfinder49 commented 4 years ago

Playing around with this, a ground strap to Kasli-ground also appears to resolve this. However, I now have trouble reproducing the broad previous spectrum.

pathfinder49 commented 4 years ago

A few questions here...

  1. These plots look rather different to the data you posted previously. The noise floor seems quite a bit higher in the new data (for the same measurement BW). Also, comparing to the orange trace in the old data there are fewer spurs, but I guess that's because the old narrow-band spurs are swamped by a broad noise peak at around 4.75MHz? From memory, it the spectra you posted above feel cleaner than the ones we were looking at last week (lots of spurs around 6-7MHz). Is that because you moved the equipment to a new (quieter) lab?
  2. Are the DACs updating continuously in this? I'm surprised not to see a big peak at the DAC clock frequency.

  1. The previous spurs may have been pickup in the previous building. I have not observed these in the new location. Whenever the broad peaks are eliminated, I observe no high frequency spurs. As for the noise floor, the analyser is not configured for the same sensitivity here.

  2. The LDAC (active low) is low and CLR (active low) is high. So the DACs should be updating.

pathfinder49 commented 4 years ago

With Kasli and Fastino ground tied to the analyser ground I get this noise spectrum.

image Edit: fixed axis

hartytp commented 4 years ago

Is the clock line toggling?

hartytp commented 4 years ago

Also I’m somewhat suspicious of you y-scale here. Is that really dBV/kHz? If so the analyser noise floor would be 300nV/rtHz which seems very high. Unless I’m missing something the units are either off or the analyser is configured to give a very high noise floor (or its much noisier than I expect).

hartytp commented 4 years ago

For the close-in noise floor Robert measured -106dBV in 7kHz BW which is inline with expectations. You’re posting -85dBV in a 1kHz bw which doesn’t seem right. I suspect you’ve got the analzyer config wrong/aren’t using the units you think you are...

hartytp commented 4 years ago

In some of your other posts I see similar numbers quoted as being in dBmV/Hz. -85dBm/Hz would be -105dBV/kHz which Is more in line with expectations...I don’t think this changes any conclusions but it’s better to get the units right

pathfinder49 commented 4 years ago

In some of your other posts I see similar numbers quoted as being in dBmV/Hz. -85dBm/Hz would be -105dBV/kHz which Is more in line with expectations...I don’t think this changes any conclusions but it’s better to get the units right

The analyser gives spectra in units of dBmV over the band-width. I supect this is what happend, but will check tomorrow.

jordens commented 4 years ago

-85dBm/Hz would be -105dBV/kHz

In a 50 Ohm system -105 dBV is -92 dBm and -105 dBV in 1 kHz is -122 dBm in 1 Hz.

pathfinder49 commented 4 years ago

I messed up my dBmV to dBV conversion. I subtracted 30 instead of the correct 60 dB difference.

pathfinder49 commented 4 years ago

Is the clock line toggling?

The DAC SCLK is not toggling.

hartytp commented 4 years ago

In a 50 Ohm system -105 dBV is -92 dBm and -105 dBV in 1 kHz is -122 dBm in 1 Hz.

Oops, adding to the confusion, that dBm was meant to be dBmV.

hartytp commented 4 years ago

I messed up my dBmV to dBV conversion. I subtracted 30 instead of the correct 60 dB difference.

I thought it was something like that. Well...oops...anyway at least with that factor all the numbers look consistent.

hartytp commented 4 years ago

The DAC SCLK is not toggling.

Well in that case the DACs aren't updating. Explains why we don't see any spurs ;)