Official bladeRF forum

Hello everyone,

I'm using the BladeRF as a neutral hydrogen spectrophometer for radio astronomy at ~1420 MHz and have run into a problem with subtle noise within the bladeRF itself that causes broadband noise pulses of about 0.1 to 0.2 dB that come and go (which wouldn't be evident to most users who aren't integrating samples), even if I'm hooked up to a 50 Ohm termination as a load with the bladeRF within a shielded box. So far, gain settings for the LNA, rxvga1 and rxvga2 don't resolve the problem.

My suspicion is that the local oscillator in the BladeRF could be the issue. When I look at the magnitude of the DC spike in the center of my spectrum, it generally rises and falls in unison with the rest of the signal in my spectrograms and this phenomenon is even more evident when I tune to some frequencies, especially at 0.3 GHz where the DC spike goes up and down like a yo-yo with a magnitude of ~1 dB.

I've attached a spectrogram that shows the problem that I'm having (see below). Any thoughts or tips on how I might control/tame the LO within the bladeRF?

Thanks,
David


https://4e5efa02-a-62cb3a1a-s-sites.goo ... edirects=0

Have you tried using a 5v power supply instead of the USB supplied power?

I'd be more suspicious of the supply than the oscillator. The oscillator is pretty good.

I included a cap and inductor to filter the 5V source that is channeled into the USB3 cable while still within the computer, but perhaps that is not enough to clean it up sufficiently or it is picking up RF from the data wires. I'll try out a separate 5V supply.

I'll try a similar experiment on my setup as well with a 50 ohm dummy load, tuned to the same center frequency.

Do you have a script you run to compare results?

I'm using this Python script to process IQ data files that were generated from the BladeRF and Gnuradio (16 bit intervealved IQ binary files). It is doing a lot of averaging on 400 Gb data files saved using a SSD to capture data arriving at 30 MSPS, but it could work on files captured at any sample rate. Vmin and Vmax will likely need to be adjusted to visualize the data in a satisfying way.
--------------------------------------------------------------------------------------------------------------------------------
#Open a Gnuradio IQ complex data file and convert it into a waterfall
#spectrogram then save it as a FITs file

import numpy as np
import math, struct
#import pyfits - use if you want to generate a FITs file
import matplotlib.pyplot as plt

scount = 0
sciavg = np.zeros((4000,1024),dtype=float)
f = "/media/david/SSD/30mtest.bin"
fin = open(f,"rb")


for yy in range (0,3999):
scount = scount + 1
print "block #",yy

sciraw = fin.read(80000000)
sa = np.array(struct.unpack('20000000f',sciraw))
scidata = sa.astype(np.float32).view(np.complex64)
spectra = plt.specgram(scidata,NFFT=1024,noverlap=64)
specf = spectra[0] * 1E7
plt.clf()
sciavg[scount,:] = np.median(specf,axis=1)

#use the next three lines for saving as a FITs file if desired
#sciavg = sciavg.astype('int16')
#hdu = pyfits.PrimaryHDU(sciavg)
#hdu.writeto('/home/david/gnuflow/30mtestf.fits')

plt.imshow(sciavg,vmax=10000,vmin=9000)
plt.show()

I found out what my problem was with the noise problem. In my Gnuradio companion flowgraph Osmocom source block following an example I had seen in the forum, I had "Ch0 DC Offset Mode" and "Ch0 IQ Balance Mode" set to "automatic". When set to "off", everything is o.k., my IQ balance issue is also fixed as well.

Thanks for letting us know!

Brian