Cheap "MSI SDR" (SDRPlay RSP1) clones, anybody have one?

[A.Z.]

I would be surprised if the SDRplay had only 42 bits of dynamic range. If that was the case I think I would have heard people bitching about it!

And the things going even worse, because SdrPlay also use MSi001 tuner chip (the same as MSi.SDR) which dynamic range is limited to about 45-92 dB, depends on the gain settings. It means that you will get just about 80-90 dB dynamic range for receiver. Almost the same as RTLSDR (80 dB).

Ok, I'm off from this; someone please, buy one of these so that they could be compared to the RTLSDR, I suspect they'll come out just a bit better, but the RTLSDR will still be the winner; oh well 

[radiolistener]

someone please, buy one of these so that they could be compared to the RTLSDR, I suspect they'll come out just a bit better, but the RTLSDR will still be the winner; oh well 

What you're want to achieve from receiver? What is your goal?

With RTLSDR and proper LNA and bandpass filter you can get almost the same result as with ETTUS or any other top receiver... 

It doesn't means that RTLSDR is better. This is very like smartphone photo camera with "just press the button" functionality vs professional photo camera. Professional photo camera allows to do much more quality photo. But it also requires much more knowledge to control it.

ADC bits here is the same as megapixels. You can have 10-20 megapixels with cheap optics on smartphone and 5 megapixels on professional photo camera with high-end optics and large matrix. And your 20 megapixels cannot beat 5 megapixels of professional camera with good optics...

[cdev]

I've been doing that. (building filters and LNAs and various antennas)

And its true, for many applications the RTLSDR is "good enough".  A few years ago, it would have taken a rack full of hardware to receive 24-1700 MHz in any fashion. Also, the direct sampling, although a bit of work to get going, is an amazing extension of the functionality to lower frequencies the hardware was never supposed to be able to receive.

The RTL's  low price makes it very compelling. And like perhaps even millions of other people, all around the world, (and that was definitely because they were so affordable) I've learned a lot using them.

I'm not challenging their usefulness. I'm just saying that they have their limitations.


The analogy you draw between a small image sensor and a larger one is a very good one, and there is no way around it, when you have a 3.1 megapizel camera's most high resolution image, it wont enlarge to be printable over a certain size. If you want to print out a large image and have it not be blurry, you need to invest in better cameras and better lenses.

Also, even if you have a great lens, the sensor matters a lot.

Similarly with the RTL, They have a lot of birdies (which vary from unit to unit quite a bit) and they can't receive really weak signals.  They do inspire a lot of experimentation - to work around all these problems.

Which is probably my goal, learning. So in that respect they have been great.

[radiolistener]

The difference between top end receivers, such as ETTUS and cheap one such as RTLSDR is not ability to listen weak stations. The difference is reception quality - signal distortions, spurious performance. And if we compare MSi.SDR and RTLSDR from this point of view, we will found that MSi.SDR is more cheaper receiver than RTLSDR. But actually MSi.SDR cost twice more than RTLSDR.

The only interesting feature of MSi.SDR is just wideband spectrum view, with almost 8 MHz at once on the screen. This is really interesting feature. But this is not receiver feature. This is spectrum view feature. For double price you will get wideband spectrum view, but worse reception quality and a lot of issues. It's up to you to decide whether to double pay for it or not.

By the way MSi.SDR has text "Panadapter" instead of receiver. So, Chinese guys don't lie and said truth. MSi.SDR is really good and cool as a panadapter with 8 MHz bandwidth on the screen, but not as receiver.

I really like to see all 8 MHz of spectrum on the screen in real time. This is really amazing!
But when I want to listen for the radio, I prefer to use my RTLSDRv3

[radiolistener]

With FPGA and ADC module you can build your own setup and get even better features, for example see attachment for screenshot with entire 24 MHz bandwidth at once in realtime on the screen. This is really amazing

I can see almost full short wave bandwidth from 0 Hz till 24 MHz in realtime... 

On the second screenshot you can see spurious performance of my setup from 0 Hz to 24 MHz. The scale is dBFS (0 dB is the maximum allowed level with no overload = 1.697 Vrms). There is some little DC offset on the ADC input, so it shows about -28 dB on DC.

As you can see, noise floor is about -150 dB/Hz or -75 dB for 24 MHz bandwidth.
This is correspond with ADC datasheet specification SNR=75 dB.

For CW (500 Hz) it means dynamic range SNR = 123 dB.
With take into account reference voltage 1.697 Vrms, it correspond to CW MDS sensitivity about -105 dBm or 1.2 uV.

As you can see, sensitivity is not so good, but good dynamic range

But RTLSDRv3 is a small and don't eat much CPU, so this is where RTLSDR shine. It is very small cheap and allows you to listen almost any frequency. Also RTLSDR has a lot of different software. This is why use of RTLSDR is a piece of cake.

[radiolistener]

You need images that contain more strong signals, next to weak ones.  You'll notice that the dynamic range is less than 50 db.

Here we are 

RTLSDRv3 connected to NWT with 10 MHz 0 dBm output through 30 dB attenuator.
There is -30 dBm on the RTLSDRv3 input. Vertical scale is reset (0 dB equals to max ADC input).

CW bandwidth is 500 Hz. RTLSDR shows noise floor about -89.5 dB within 500 Hz bandwidth on the background of -4 dB carrier.

So, the dynamic range for CW mode with 500 Hz bandwidth is

SNR = 89.5 - 4 = 85.5 dB

Also we can calculate sensitivity for CW mode 500 Hz:

noise floor = -85.5 - 30 = -115.5 dBm = 0.375 uV

Now we know, that RTLSDRv3 in CW mode at 10 MHz has:
- sensitivity about MDS = -115.5 dBm = 0.37 uV
- dynamic range about SNR = 85.5 dB

[fourfathom]

With the RTLSDR-V3 I saw 70dB SNR (1 KHz demodulated CW tone, 25uV input to preamp, 5KHz audio bandwidth).  I hadn't quite hit the SNR limit, so I can easily believe you can get 80db+ with a 500Hz bandwidth.

[radiolistener]

With the RTLSDR-V3 I saw 70dB SNR (1 KHz demodulated CW tone, 25uV input to preamp, 5KHz audio bandwidth).  I hadn't quite hit the SNR limit, so I can easily believe you can get 80db+ with a 500Hz bandwidth.

This is very predictable. RTLSDR has 8 bit ADC working at 28.8 MHz.

ADC SNR = 8 * 6.02 + 1.76 = 49.92 dB (for 14.4 MHz bandwidth)

Processing gain for 14.4 MHz to 500 Hz is 10*log(14400/0.5) = 44.59 dB

So, the theoretical maximum of dynamic range for ideal ADC is:

SNR = 49.92 + 44.59 = 94.51 dB

Since ADC is not ideal, the dynamic range is a little smaller. I think it's about 89.5 dB, the same as measured noise floor. I cannot put more powerful signal on the input, because my NWT output is a little noisy, so there is needs a little margin to avoid ADC overflow.

SNR = 89.5 dB means that the effective ADC resolution of RTLSDR is:

ENOB = (89.5 - 44.59 - 1.76)/6.02 = 7.17 bit

[cdev]

If the rtl-sdr mailing list still exists, this would be a question to ask them. A very long time ago there was a big debate about dynamic range there. In theory I think the maximum dynamic range possible with an 8-bit SDR is around 50 db. That said, its my understanding that some people, notably Leif, SM5BSZ, an expert in such things,  managed to eke out a substantial bit more dynamic range using custom modifications of the librtlsdr library (with several different possible modes) which he published. However I dont think his suggestions were ever incorporated into the librtlsdr trunk.


I think this debate basically died without ever having been resolved. Similarly, there were other modificatiojns that improved the selectivity which never seemed to be incorportated into the librtlsdr.

Please fill in the blanks here with some facts if anybody knoews them. I'll be honest, I lost interest in the issue when it became obvious to me that with the exception I think of the GPIO for antenna power the various interesting mods people were making were not going to get adopted and nobody set up an alternative center for librtlsdr development I could trust. the convenience of using a package manager and easier upgrades and my resistance to running sftware whose origins I was not so certain of unless it was vetted by somebody I trust won out for me.

Most poeople seemed to move on to other hardware. Given that so much more of it became available. Can you blame them?

That said, I would love to see somebody squeeze out more dynamic range from of an 8 bit SDR..

With regard to Mirics, this looks interesting..

http://www.sm5bsz.com/linuxdsp/hware/mirisdr/mirisdr.htm


The plot thickens..

Also see

http://www.sm5bsz.com/linuxdsp/hware/rtlsdr/rtlsdr.htm 

Perhaps this is what you mean?

"The data delivered from the rtlsdr is only 8 bit. An ideal 8-bit A/D converter should have a SNR of 50 dB The bandwidth delivered by the rtlsdr is 2 MHz while we measure MDS in a bandwidth of 500 Hz. The bandwidth factor is 4000 and that provides a gain of 36 dB. Note that the dongle reaches the theoritical limit in sensitivity mode at gain settings of 0 to -15 dB. Measurements have to be made with a third signal present because otherwise the A/D converter might give the same value most of the time. With a GAAS FET LNA having the gain required to lift the noise floor by 3 dB in sensitivity mode with 0 dB gain setting and with adequate filtering one can get very good performance with the E4000 dongle. With a 1 MHz wide filter at i.e. 432 MHz one can get a dynamic range near 86 dB with a sensitivity only about 3 dB worse than optimum. By increasing the gain to 15 dB one would then have a sensitivity very close to the theoretical limit at a dynamic range of about 68 dB. Note that LO spurs and reciprocal mixing can be the limiting factors. "


Note that both of these web pages are around 8 years old a very long time in technology)

What I am saying is that additional tricks may be possible, but hooks would have to be provided and agreed upon to utilize them.

[fourfathom]

This is very predictable. RTLSDR has 8 bit ADC working at 28.8 MHz.
[...]

Yes.  I'm just saying that my measurements, your measurements, and the theory all support each other.

As for sensitivity, unless you have a very low-gain (lossy) antenna, at HF frequencies the atmospheric and other noise is going to be the limiting factor.  My measured 0.005 uV noise floor (using a preamp) is far below the atmospheric noise.  I am far more concerned about IMD, overload, and aliasing.

[cdev]

The noise level increases a lot in the summer it seems.

This is very predictable. RTLSDR has 8 bit ADC working at 28.8 MHz.
[...]

Yes.  I'm just saying that my measurements, your measurements, and the theory all support each other.

As for sensitivity, unless you have a very low-gain (lossy) antenna, at HF frequencies the atmospheric and other noise is going to be the limiting factor.  My measured 0.005 uV noise floor (using a preamp) is far below the atmospheric noise.  I am far more concerned about IMD, overload, and aliasing.

If one can live with their other various quirks, (more spurs than the Rafael Micro ones)  the elonics
e4000 dongles have better UHF sensitivity than the others.

[radiolistener]

In theory I think the maximum dynamic range possible with an 8-bit SDR is around 50 db.

Yes, 8 bit ADC has dynamic range:

SNR = 8 * 6.02 + 1.76 = 49.92 dB

But this is for entire bandwidth. RTLSDR ADC running at 28.8 MHz, it means that it has 14.4 MHz bandwidth. So, these 50 dB is for 14.4 MHz bandwidth.

But there is no need 14.4 MHz bandwidth to listen CW, AM or SSB. For example SSB needs just 2.7 kHz bandwidth. Due to processing from 14.4 MHz to 2.7 kHz we get additional 10*log(14400/2.7) = 37.2 dB of dynamic range.

It means that for SSB mode RTLSDR 8 bit ADC has theoretical maximum of dynamic range:

SNR = 49.92 + 37.2 =  87.12 dB

[fourfathom]

Due to processing from 14.4 MHz to 2.7 kHz we get additional 10*log(14400/2.7) = 37.2 dB of dynamic range.

Yes.  [To some extent] You can see this effect in the one-bit delta-sigma D/A and A/D devices used in audio work. [But since this uses feedback it isn't quite the same thing.] 

And since the SDR sample rate is not coherent to the signal being detected we get an effective wideband dither which improves the narrow-band SNR.  This dither is probably assumed in the bandwidth-effect calculations, I'm not brushed-up on the math.  But I have designed quantization logic (in ASICs) that used high-frequency dither to good effect, greatly increasing the effective narrow-band resolution -- so I know it works.

[radiolistener]

Yes.  [To some extent] You can see this effect in the one-bit delta-sigma D/A and A/D devices used in audio work. [But since this uses feedback it isn't quite the same thing.] 

It works like average. When you average several samples you will get sample with more bit resolution. The same thing happens in the filter which is applied in order to reduce bandwidth. Before do decimation, there is needs to apply filter to remove high frequency components (to prevent aliasing). So, after low pass filter the result samples will have more bits resolution.

Reduce bandwidth for 4 times equals to add 1 bit to the sample resolution. This is why 1 bit = 10*log(4) = 6.02 dB

[fourfathom]

It works like average. When you average several samples you will get sample with more bit resolution.

Yes, but you still need dither (noise or deliberate) or equivalent to make this work.  Example: with a 1-bit A/D (a fixed-threshold comparator) digitizing a DC input you will only get one-bit resolution no matter how fast you oversample or how much you post-filter.  Add the proper noise or dither to that DC input and then you can obtain *much* better quantization resolution after filtering. 

The asynchronous sub-Nyquist-rate sampling (or bandwidth-limited undersampling) also acts as a sort of dithering in the time-domain.  I assume that this means that with input signals at or very close to integer fractions of the sample clock where the effective resolution is closer to that raw 49dB resolution of the 8-bit converter.  I will try to play with this when I get a chance.

[radiolistener]

Yes, but you still need dither

There is no need dither for receiver, because there is atmospheric noise from antenna. Dither is needed when you measure some signal from device with no noise. But when you connect antenna, it already have enough noise for dithering

[fourfathom]

Yes, but you still need dither  (noise or deliberate)

There is no need dither for receiver, because there is atmospheric noise from antenna. Dither is needed when you measure some signal from device with no noise. But when you connect antenna, it already have enough noise for dithering

Yes, as I said.  Just pointing out that something else is required for increased resolution.  In this case, atmospheric or thermal noise is that thing.

[cdev]

So, it seems that this dongle comes with the generic USB vendor and product IDs from Mirics which make it very similar to the previous Mirics dongles. (except that more of its antenna inputs are connected)

http://blog.palosaari.fi/2013/10/naked-hardware-13-logitec-ldt-1s310uj.html

If it works with the free drivers it may be a lot like the sigrok situation. 

Does anybody know?

[cdev]

If you are willing to use a translator or speak Chinese there is a bit of stuff about efforts to make DIY MSI SDRs on a chinese BBS, kehuang.org. Aparently the two main chips that are used to make an SDR cost so little its inspired a lot of people to try to make their own.

Here are two examples:
kechuang.org/t/83757

kechuang.org/t/85213?page=0

Nobody seemed so overjoyed with the results. however its kind of interesting.

I donnt know if whomever made the blue dongle is a user there or not or where it stands in comparison to the other efforts.

A German ham,  recently put a lot of work into reviewing it, using the sdrplay software, not the free libraries.

"Measurements on an MSI SDR stick from Gerd DC6HL"
https://saure.org/cq-nrw/2020/06/10/messungen-an-einem-msi-sdr-stick-von-gerd-dc6hl/

[Gods69]

then I own one, and two friends of mine another pair (we ordered them together); the unit is just like the one depicted above, although it hasn't that electrolytic capacitor soldered on the board

Hi,
I also bought this receiver and it also does not have this electrolytic capacitor mounted on the PCB. I think he was soldered after.
Worth it to put this decoupling capacitor?  Thank you.

[cdev]

What kind of software supports the AD DAC as an SDR under Linux?

I am not a windows user. Windows makes me ill, the stuff it does. I cant handle it and remain sane.



So better not even try..

However, I'd love to try building this into an SDR (assuming all thats needed is filters being it direct samples.. right?)

it if it plays well with Linux.

As for comparing the units, I thought we were discussing about cheap commercial SDR units

yes, we're talking about cheap SDR with price < 50 USD.

I believe that comparing those with highly priced ones would be like comparing apples and oranges, otherwise we may just compare an RTL or MSi unit with an ETTUS USRP and conclude that the RTL/MSi are crap

MSi.SDR cost about 40-45 USD. This is a lot. By spending just a little more money (about 70 USD) you can build fair 12 bit SDR setup with much better performance.

For example: https://www.aliexpress.com/item/4000796488236.html

[charliedelta]

Stop playing with toys. Just buy a Perseus which gives you a great receiver, spectrum analyzer and measuring receiver all for the one price.

[cdev]

Too expensive for me, last time I looked. 
Also, I still enjoy playing with toys.

That said, there are fewer high dynamic range SDRs now that I would wish. Since then Ive looked into the actual commercial SDRplay and it doesnt look as good on closer inspection than I would hope either.

If I had more in the way of programming skills, I'd definitely be interested in building a receiver or more likely, a ham transceiver kit.

A big thing is software. Finding something  that meets all of the various requirements isn't easy.

[vinlove]

I have a SDRPlay RSP1A, and the overload problem is not serious issue.  When it overloads, you just scroll down the RF Gain to the level under the overloading, and all is fine.  I find RSP1A a great SDR coverage from 1 kHz -2000 MHz all modes.   I run it under the old Windows 10 loaded ACE laptop, kindly donated by my daughter.

[A.Z.]

Too expensive for me, last time I looked. 
Also, I still enjoy playing with toys.

That said, there are fewer high dynamic range SDRs now that I would wish. Since then Ive looked into the actual commercial SDRplay and it doesnt look as good on closer inspection than I would hope either.

If I had more in the way of programming skills, I'd definitely be interested in building a receiver or more likely, a ham transceiver kit.

A big thing is software. Finding something  that meets all of the various requirements isn't easy.

SDRplay units aren't bad at all, same goes for AirSpy ones; sure, they aren't "ettus" or the like, but they offer quite a lot of bangs for the buck, plus there's quite a bunch of sw around (check github) to get one started, and then (but that will work for most/any SDR) there's the redhawk sdr framework which allows to use (or write) logical blocks and play quite a lot of cool experiments