Official bladeRF forum

In another thread we were discussing the large (25-30 dB) spike on center frequency in SDR. Today, I spent a few hours exploring the spectrum from .3 to 3.8 GHz just to get re-familiarized with some of the features of SDR and how to best interact with the bladeRF 115. This caused me to have some questions...

1. I noted, as I spun the dial, that I found "anomalies" in monitored spectrum. For example, I looked at the 430-450 MHz spectrum for a few hours (good thing it was a slow day at the office, eh?!) and I would see interesting wide spectral 'bumps' of 10 dB around .25 MHz wide that would come and go at seemingly random times, but in consistent frequency ranges;

2. Looking at the 2.4 GHz spectrum, I noted a significant increase in the noise floor (expected) with sharp spikes indicative of the frequencies associated with "channels" of the 2.4 GHz band. What caught me off guard, however, was the 30 dB or so of wide-band noise just below 2.5 GHz. Wide band meaning 20+ MHz of space with 30 dB of increased noise floor;

For fun, I pulled out the Moto 2660 to make some relative comparisons to the 430-450 MHz observations and didn't see similar behavior as noted in #1. Given that the 2660 was just recently calibrated, I'm fairly comfortable with its accuracy, although its response time to spectral changes, etc., leave me

Referring back to the conversation about SDR and center-peak noise, etc., I'm wondering how to isolate the source of these anomalies. Terminating the RX side into a dummy doesn't appear to be the answer, and I wonder about a methodology to calibrate the 115 to better understand its limitations, and what role the computer, USB interface, processor speed, connectors on the board, etc., play in observed anomalies, accuracy, etc.

On the subject of the on-board RF connectors: Has anybody toyed with building a shield to cover the center pin, maybe a 3-sided small enclosure soldered to the ground side of the connector to completely enclose the signal line - and whether this contributed any significant improvement to observed noise?

The party's just starting...

Hi,

Why not post screenshots? This always helps understand what you're describing. Others may see things you have missed...

This, sir, is why you're "the man"... I have added a PNG of the situation. Please note the two regions highlighted in yellow - although the screen capture doesn't indicate the amplitude at it's peak, approx -100dBm normally, it is definitely a "pulse" without a pattern of time that would allow me to say it occurs every 5 seconds. Sometimes the pulses are less than a second in duration - other times 2-3 seconds. I see a similar 'bump' 446.7500 with similar behavior.

Again, pulling out the 2660 I do not see the same signals.

I'm wondering if this is aperiodic noise caused by something environmental - USB Switches, Ethernet Switch, high-frequency ballasts, who knows - and trying to come up with a way to determine - of the 'signals' I see in an observed spectrum, what is real and what is a result of the board's shortcomings, if any. I question whether comparing the observations in SDR to the 2660 is valid - I note odd little birdies in the 2660 that don't show up on the SDR display -- and the same applies in the opposite direction.

So, I return to my original premise: Is there a way to terminate the Rx side of the board so that a true zeroing of the data can be observed - that would seem to give me the board/computer/cpu and environmental impacts without the receiver in-play.

Thanks!

Hi,

To answer your question - I don't know. Why not take a screenshot with a small dummy load on the RX antenna and the bladeRF in a shielded enclosure (or as close as you can get)?

I don't understand - what's the 2660?

Show me what you have - I don't use the bladeRF at the moment and am not involved in the development, I'm just a very simple programmer.

Hola, Simon -

The 2660 is a Motorola R2660 service monitor with spectrum analyzer.

For "fun" I have terminated the Rx input with a 50 Ohm terminator, placed it in a shielded cage and taken some snapshots for comparison. I think my next bet is to sort out some additional shielding so I can provide more accurate answers. In short, I have not seen the 'bumps' using termination and shielding, or just termination. That's a good sign... I guess.

However, we still see spikes at various intervals on the spectrum that I would think would have disappeared as well. As previously discussed, on the center frequency - but other frequencies as well.

Here's the capture from terminated + shielded...

Hi,

The center spike I can understand. The others must be coming from the board, time to shout at a hardware guy .

My next step is to completely enclose it and see what the story is from there. Thanks for the info and thoughts.

Simple programmer... bah!

LOL

drbob wrote:My next step is to completely enclose it and see what the story is from there. Thanks for the info and thoughts.

Simple programmer... bah!

LOL

Stupid programmer

In case it was lost in translation, I was making fun of your statement about yourself, "I'm just a simple programmer..." You're definitely not. Don't sell yourself short. Thus, "Bah!"

An update, of sorts...

I had a custom built case made that entirely encloses the unit in a milled and routed aluminum enclosure with an RF seal between the upper and lower portions of the enclosure, and used SMA <-> N connectors through the enclosure. RF choke on the USB cable. Re-measured and I note that my terminated receive tests did not result in any significant decrease in spikes across the observed spectrum.

For fun, I'm now moving to GNUradio and will see what the outcome is there, along with DC-blocking to determine if that is the source of the peaks seen in the pictures provided.

Hi everyone,

I have doubts too in this issue. I'm trying to observe gsm and umts signals in the air by GNU Radio Companion with my bladeRFx115. Although I am able to observe some specific signals, I want to remove the noises by filtering. So, I'll be able to detect center frequency of the signal. I've tried the filters in Gnu Radio Companion, however I couldn't remove the noise. Also, I've tried to use peak detector in flowgraph but there are lots of type conversion issues and I didn't observe even if type conversion is done. I don't want to lose the exact signals by using peak detector with type conversions. I've added some screenshots about the spectrum.