Page 1 of 6

Testing RSPduo receiver sensitivity with noise generator

Posted: Wed Jun 20, 2018 4:39 pm
by glovisol
After reading the excellent articles of this thread:

viewtopic.php?f=7&t=1562

I purchased one specimen of the wideband noise generator described therein in order to test the RSPduo receiver sensitivity at the various inputs and to set up reference data which would allow me to check receiver performance from time to time. This activity was also stimulated by the realization that receiver sensitivity on the Hi Z port appeared to be significantly better than on the 50 Ohm inputs, so relative measurements among inputs would be very useful.

This post concerns the HF range from 1.9 to 14 MHz. The noise generator in this frequency, with a 12.6 V supply, outputs a level of roughly - 50 dBm. To test sensitivity NF style it is necessary to feed the inputs with a very weak signal which can barely increase the S/N ratio.

To bring the noise level to the needed weak signal of nominal -130 dBm, I built two -40 dB attenuators. The receiver is set up as shown in TABLE 1:

- IF gain in manual @ 30 dB. Note that in the table it is wringly mentioned as "RF gain".
- RF att. minimum value.
- LSB.
- Filter 1800 Hz.
- AGC OFF. Note that somentimes the receiver audio mutes in these conditions. Just quickly turn the AGC ON and then OFF.

The test procedure is simple. Just connect the text fixture to the desired input with the NG (Noise Generator) OFF and note the noise floor on the receiver's (RSP1) dial & the field noise. Then turn on the NG and note the MINIMAL difference: you should barely hear the difference in noise. If the difference is more than 10 dB, it means you have to increase the attenuation, to come near the receiver's noise floor. The DIFFERENCE between the NG OFF and ON gives you a fair idea of the receiver's sensitivity, as well as telling you if (God forbid!) one input is not operative. Test results can also be recorded as a reference for future tests.

Looking at TABLE 1 it is evident that the HI Z input is the most sensitive up to 7MHz. I wa also aware that at 7 MHz and below the receiver was significantly more alive using the HI Z port with a 9:1 balun (and a long wire antenna) as compared to the 50 Ohm ports. This is confirmed in TABLE 1, where the advantage is between 15 and 20 dB up to 3.8 MHz and goes down to 6 dB at 7 MHz. This advantage is much lower at 14 MHz, probably due to the fact that a different, lower inductance balun must be used here.

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Wed Jun 20, 2018 5:03 pm
by glovisol
Sorry: the system did not let me post additional files. Here they are. To conclude the tests have demonstrated excellent sensitivity in all HF bands, considering that the minimum detectable signal is always at or below 0.14microvolts.

ERRATA CORRIGE: Please note that the final single attenuator value that was used in the tests is 46 dB, not 40 dB, so the total attenuation to use in the test behind the NG is 92 dB, i.e. the output of the NG in this frequency range is in excess of -40 dBm.

WARNING AS OF 05/07/18 :I used the noise generator with a supply voltage of 12.6V. The last amplifier stage of the generator failed after 15 days, probably because this supply voltage is NOT suitable. Better be safe and use no more than 10.0 V

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Fri Jun 22, 2018 4:38 am
by glovisol
I post here a clearer version of TABLE 1.
Sensitivity Table (2).jpg
Sensitivity Table (2).jpg (94.41 KiB) Viewed 33177 times

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Fri Jun 22, 2018 9:37 pm
by someYguy
Thanks for posting this! Very good way to test receiver sensitivity and one which is new to me. I'm going to use this technique to measure the receiver of my Yaesu FT-817nd also :).

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Mon Jun 25, 2018 5:49 am
by glovisol
Hi someYguy,

Please post your results (the FT-817 as well) as it would be very interesting to compare all measurements!

glovisol

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Mon Jun 25, 2018 7:24 am
by vk7jj
Hi glovisol, thanks for your post, most informative. Just for comparison here are a couple of examples of real world low noise performance on 40m from an RSP1A.

The only noise on the waterfall apart from that generated by the multitude of ham SSB stations appears to be at 7.042MHz and that appears to be generated within the RSP1A as it's not present in the RSP2.

Only one of the SSB stations was within Australia, all the CW and the remaining SSB stations are around 13,000 - 17,000km away from my Tasmanian location being mainly in Europe and the US.

As you say the RSP2 appears to offer generally slightly better low-noise performance from it's HiZ input but it was active on WSPR at the time of these pics. Looking at the traces makes one wonder how WSPR manages! Excuse the fuzziness of the low quality jpgs.

The slight slope on the noise floor follows the slope of the antenna's gain.

The RSPs truly are excellent receivers.

Regards, Phil VK7JJ

Overview of 40M
RSP1A_40mSSB.jpg
RSP1A_40mSSB.jpg (276.88 KiB) Viewed 33336 times
CW segment
RSP1A_40mCW.jpg
RSP1A_40mCW.jpg (245.06 KiB) Viewed 33336 times

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Mon Jun 25, 2018 11:13 am
by glovisol
Hi Phil,

Thanks for your message and for the PC screens, which show the really clean spectrum devoid of self generated interference offered by these RSP receivers. What is amazing to an old timer like me (I was active in the fifties of last century as I1LOV) is that with the RSP 1A or the RSPduo you can receive a perfectly readable SSB signal at S1/2, with another S9++ signal a few KHz away: sometimes you read the signal and you cannot see it in the display, because it is buried in noise! In the fifties the very best seemed to be the Collins 75A4, you had a 1 KHz resolution, adiacent channel selectivity was good, but weak signal reception was a problem. The Racal RA17 was also a good (and horribly expensive one!) but strong signals even a long way off would block or otherwise render reception very difficult. HRO's and Hallicrafters were a different compromise: better sensitivity and strong signal performance, but frequency accuracy and re-settability were another matter altogether.

I am evaluating the RSPduo with a long wire (135 m.) antenna and I am working on balun optimisation. I look forward to posting results in the near future.

Kind regards,

Gianfranco/glovisol

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Tue Jun 26, 2018 12:08 am
by vk7jj
Hi Gianfranco, good morning again from VK7.
What is amazing to an old timer like me (I was active in the fifties of last century as I1LOV) is that with the RSP 1A or the RSPduo you can receive a perfectly readable SSB signal at S1/2, with another S9++ signal a few KHz away
My obvious enthusiasm and delight springs from the same beginnings, from actually using Lecher lines and light bulbs, Nuvistor powered GDOs and germanium diode detectors on microamp meters walking around Yagi V's Quad antennas trying to visualise their fields. Now with SDRs and antenna modelling and high performance antenna analysers we can see everything ... well nearly!

Two things I wondered about with your tests, the first being the apparent lack of signal leakage, did you notice any?

The second was whether you tested or allowed for the effects of decimation on your noise figures, I've never done any measurements, have you?

eg. This excellent and to the point article:

http://play.fallows.ca/wp/radio/softwar ... rning-dsp/

Regards, Phil

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Tue Jun 26, 2018 5:55 am
by glovisol
Hi Phil,

You made me recall my first transistor: a Rayteon CK 722 in 1956.....blown up because of polarity inversion after a few minutes' experimentation....

- Please explain what you mean by signal leakage.

- Interesting article on decimation effects. To tell you the truth I started checking for decimation effects on sensitivity at the beginning of May, as I received my first RSP2pro. Unfortunately I had to return it, as Tuner 1 input was not operative, so I purchased the RSPduo. My first Beverage antenna was only 30 m. long and received only weak signals, therefore my attention was concentrated on methods to improve the quality of the S2 / S5 signals I was receiving. Unfortunately never noticed any improvement by manipulation of decimation. Since last week I installed a new Beverage, 135 m long and now I am really in business, 160 to 15 m. bands and have been able to get far away signals with fields up to S9+10.

In any case it is difficult to test for decimation improvement with on the air signals due to fading and other effects. Some kind of lab test fixture should be needed.

Best regards,

Gianfranco

Re: Testing RSPduo receiver sensitivity with noise generator

Posted: Tue Jun 26, 2018 7:27 am
by Tech_Support
Hello Folks,
Decimation will have no discernible affect on sensitivity or noise figure. Decimation increases the ADC resolution and decreases quantisation noise. This will only be noticeable at very low gain settings where the ADC quantisation noise starts to be come significant in the overall system noise figure. Oversampling and decimation only effects quantisation noise. It does not affect Johnson noise. At maximum gain the ADC quantisation noise is at least 20 dB below the Johnson noise level of the RF front end and will therefore be undetectable. People often confuse the fact that weak signals become easier to see on the FFT display when decimation is applied with an improvement in sensitivity. This is simply not the case. Decimation reduces the final sample rate and therefore for a fixed size FFT reduces the bin size (resolution bandwidth) of the FFT. As the FFT measures the total power in each FFT bin, the noise floor appears to drop. However the integrated noise power across the total channel bandwidth will remain unchanged.

Sincerely

SDRplay Tech_Support