Issues with spurs and noise floor on the LimeSDR

I received my LimeSDR board earlier this year and have just started getting some time to experiment with things. I seem to be experiencing some serious challenges with this board. Specifically, there appears to be some sort of serious instabilty/oscillation happening with the device. The unit is mounted in a metal enclosure with 1/4 inch metal standoffs and UFL to SMA pigtails on each of the connections (each mounted securely to the case.) As one moves a hand above the board (or attempts to get the case cover near) the sprogs change amplitude on the order of 20+ dB. I have provided two images to illustrate this issue. In addition, in order to get the noise floor to a “more reasonable” level, I have found that running the antenna cables through ferrite bar helps some with the noise floor through the placement of the bars (1"x 3" by 0.25"). The noise floor moves about upward of 30 dB depending on the placement (or absence) of these ferrite EMI products.

The screen shots are made with 50 ohm loads on each of the antenna terminals.

The LimeUtil --update and LimeUtil --cal routines have been run on the unit though no observable difference has been noted.
I have also used nylon standoffs as an experiment and, as would be expected, it was actually a bit worse.
The problem certainly increases with increasing frequency–especially above 400 MHz.

Has anyone else experienced, and more importantly, been able to resolve these issues?

In short, this unit acts as if something is oscillating in a manner in which it should not be…somewhere in it’s spectrum.) Any advise would be greatly appreciated.

What were the LNA, TIA and PGA gain settings?

This is mine (50 ohm terminator is on BNC so not good above 1 GHz)

Hi n5

The USB cable was a source of a lot of interference in my setup originally.

I added a shield vertically between the digital and rf sections of the sdr board and also fashioned a shield around the usb connector which helped.

I power the sdr from a separate DC supply which also runs a fan.

After all the shielding the spectrum looks a bit better and you can actually see a cw signal down to -125 dBm using SDR Console - if you shift the generator signal a bit off the centre frequency.

A 75 MHz BW is pretty wide - try a little narrower. What does 5 MHz look like ?


Thank you for your question Andrew. The gain settings do affect the level of the spurs above the noise floor as one would expect rolling back the gain. However, no matter what the gain LNA (max to -9), TIA (0,-3,12), PGA(-10,0,+10) there does not appear to be any change in the frequency or relative levels (one spur to another) evident. Adjusting the ferrites external to the board does, however, alter the relative levels of the spurs somewhat as if the spurs are being picked up from multiple sources. It’s all very odd. Changing the sampling bandwidth appears to have no effect on the spur frequency either. Whether 2.5 Mhz or 10 or 50 Mhz is set as the sampling rate, the spurs remain at the same frequencies.

In fact, if I simply remove the UFL connector from the board, the spurs remain–just at lower amplitude (and some go under the noise floor.)


Here’s the equivalent on mine with an attempt to duplicate your settings. At this particular frequency range, the noise floor is about 5dB higher than yours. While the amplitude of the spurs is less at this frequency than others, they are quite evident. I am, however using an SMA terminator.

Thank you,

Using LNA W

Switching to LNA H

I have not yet attempted to power the board from an external supply, but it’s something that I should try. I’ll just have to acquire that particular plug to do so. The lower bandwidth (2.5, 5, 10) as I mentioned in another reply this morning does not change the amplitude or frequency of the spurs appreciably outside of the expected differences in the “binning” of power in the FFT. The proper term for spreading the power across multiple FFT bins escapes my memory for the moment.

I may have to also experiment with shielding. Thank you Tom,

Just to mention that my RX1_W is modified for HF and have die cast
sealed (RF and air) case with metal stand off for lime.
No internal shields.
Power through USB.
Internal and external fans powered on separate DC supply.
Connectors are UFL pigtail to N, home made.

Just a bit more info after a few hours of investigation.
I disconnected the UFL connections to the external clock (they were not in use anyhow) and experimented with various attempts at shielding, etc.

While shielding efforts thus far didn’t seem to have any substantive effect, I did notice that when placing my finger in the area of the yellow circle, I could get the spurs/noise floor down about 17-20 dB while not losing any RF sensitivity. The positioning is very sensitive.

Today’s findings. I re-arranged all the coax such that it remains to the side of the board (Not above or below the plane of the board.) This has largely eliminated the spurs. I am still struggling with getting the noise floor down as placement of the ferrite EMC material does affect it on the order of 10-20 dB on the upper bands.

Did you run the calibration — and if so, more inportant to my question … which one?

Over the course of the past couple of weeks, I have run three different programs for calibration. The first one attempted was the LimeSuiteGui. The second attemped was LimeUtil across the full spectrum. Most recently, I attemped LimeSuiteCalibrate after getting all the python dependencies going on my Win machine. However, when this was run, it ended in an error condition and I have not given it a re-run yet. I must say that it is not at all clear which cal routine should be used. I would have hoped that a more concise set of guidance would be in place by now, but there seems to be a lot of info scattered in a incoherent way. This has to be causing confusion among the user base. At least this forum is helpful in its own way.

I made an attempt earlier today to switch to an external reference clock with the LimeSuiteGui to rule out any issues with the on board oscillator. I followed the guidance for modifying the Si5351C, but it did not switch things to external clocking. The learning curve continues. This board has a lot of potential–it’s just quite a challenge, even with some knowledge of SDR, to get things going. Hopefully, I can get the noise issues solved. I guess I’m at a loss at why things so fidgety and sensitive to external factors.


I do want to clarify that the spurs are only gone when I place a bar of ferrite material (insulated) across the board.

Today’s experiment was to completely remove the board from the metal enclosure and observe behavior. Any moving any ferrous object within about 15cm of the board causes fluctuations in the noise floor.

So, the next step I’ve taken is to see if there is anything measurable coming from the board…perhaps an oscillation. Well, I didn’t find that, but I did measure this with an AirSpy at a distance of about 20 cm.

If I depress the reset button on the LimeSDR and/or remove power, the broadband noise disappears leaving only some spurs from the AirSpy unit. This is evident in the darker blue band toward the bottom of the waterfall.

This was also observed, when setting the antenna connector to “None” within SDR Console. (This is independent of tuned frequency.)

Some sort of broadbanded floater signals.

To me, it would seem as there is some sort of problem with this particular board. Has anyone else reported such issues?


I see the same things but it is intermittent as frequency changes, as if certain software or firmware decisions about IF and clock frequencies are making a difference. I can even hear the LED blinking in a demodulated signal. I ran the cal routine with USB2 because of all the noise. I may try to make a can around the RF section using PCB material on the ground traces on the board. A divider I made has already helped a little.

Ok … then perhapse changing the .db file to something else in /home/.limesuite will not help

My thought being that your supressing vhf noise ( that lime has more gain for) and have a situation where limeutil --calibrate has killed any hf reception but the vhf noise is the dominant signal.

Try the rx tests with the .db file inaccessible just for fun on the 30M band … i use WWVH durring the day because its almost always able to receive it.