SDR receiver for low frequencies from 50/60 kHz and up

[petert]

Hello,

I am working on a receiver for the dcf77 time signal radio station in Mainflingen/Germany, which broadcasts on 77,5 kHz.

SDR receivers usually start at much higher frequencies. I had some success with using a sound card, but ideally I would like something that goes up a bit higher in frequency, up to around 120 MHz to also see FM radio stations. The purpose is testing without requiring a spectrum analyzer, which is currently out of my budget.

Are there reasonably cheap priced SDRs that allow for that? Are they also suited for filter development/testing, e.g., when developing a superheterodyne receiver?

[DaJMasta]

SDRPlay and a few others make some options that go down to 1kHz, but there are upconverters available for others (or as separate units) to put your low frequency signal into a usable band.  I think with the cheapest SDRs, this is usually the approach, and most software has an option to do the math for you on the spectrum display and show the input frequency even when capturing the upconverted one.

[petert]

Nice option. Is the support for low frequencies (1kHz an up) built-in into SDRPlay, or do you need an additional upconverter?

Is it suitable for analog filter development/testing?

[awallin]

We have looked at LF and HF with an Ettus N210. Samples at 100MS/s and about DC-30 MHz can be streamed to GNU-radio over 1 Gbit ethernet. It might be a bit expensive as new but around 7-800eur on ebay.

[ogden]

For DIY'er direct sampling (2Msps) mod of RTL-SDR is an option: https://www.rtl-sdr.com/rtl-sdr-direct-sampling-mode/

[OwO]

It's very trivial to make a simple board with Spartan 6 FPGA, ADC, and a USB PHY. I can put together something like this in a day or two. The trickier part will be the frontend. You need a highly linear LNA and lots of filtering, which is where almost all commercial SDRs fall short. There is no way for a SDR manufacturer to know what band you want to use ahead of time, so none will provide any real filtering.

[petert]

It is ok if the filtering is not perfect, I can do it in software well enough. (Like in the soundcard example.)

But the question is if I can attach *external* filters I made myself in hardware, and test their performance reasonably well with an SDR. Obviously it wont be as good as with a spectrum analyzer, but is it possible? If yes, any suggestions?

[eliocor]

a simple and economical way to receive DCF-77 is to use ARM Radio by I2PHD: 
http://www.weaksignals.com/
source code and schematics included
here a sample of DCF77 reception (from North Italy): http://i2phd.org/code/dcf-77.mp3 
the same SW, can be ported on different DISCOVERY boards, eg: STM32F746G-DISCO

[DaJMasta]

On the frontend side, one of the nice things about the low frequency component is that a lot of the tolerances are more forgiving since the small parasitics don't have big effects in the kHz band, and reasonable linearity in that region in achievable with standard opamps or whatnot.  In any case, direct sampling could be a good option, and the SDRPlay SDRs support 1kHz signal on the primary antenna input, no special external stuff required.


With direct sampling in mind.... you could probably just get a high end audio ADC evaluation board without the audio band low pass filtering (very often these things go to a few hundred kSample a second with very high resolution) and put that directly into your SDR application.  A 50 ohm termination and a low pass filter on an opamp with some gain would probably be quite servicable as a fixed-gain frontend.

[coppercone2]

a 500KSPS+ SAR 18bit/16 ADC would be a good choice for this IMO

then you can sample 10 times over the highest frequency 50KHz.

[radiolistener]

It is ok if the filtering is not perfect, I can do it in software well enough. (Like in the soundcard example.)

No, it is not ok. You cannot do it in software, because filtering should be done before ADC.

There are three main issues which require filtering before ADC:

1) All frequencies higher than Fs/2 (where Fs is ADC sampling clock rate) will be folded into first Nyquist zone:


2) Strength out-of-band signals can overload your ADC, if it cannot handle enough dynamic range. This is why more bits resolution is better, because SNR = 6.02 * N + 1.76 [dB], where N is ADC bits resolution.

3) High ADC clock jitter leads to degrade dynamic range. Also it leads to degrade filter parameters for digitally implemented filter, due to the high jitter in the input signal.

Even if you will use ultra low phase noise oscillator for ADC clock and high dynamic range ADC, you will still need to cut off all frequencies above Fs/2. In order to avoid aliases in the first Nyquist zone. This cannot be done in digital domain. So, there is need at least anti-alias LPF before ADC.

You can build your own SDR with high speed ADC and FPGA. For short wave, components will cost you for about 100..200 USD. For example, you can use these ADC:
- 14 bit, 105 MHz ADC: https://www.aliexpress.com/item/1PC-14-105M-high-speed-ADC-module-data-acquisition-module/32730197994.html
- 12 bit, 65 MHz ADC: https://www.aliexpress.com/item/High-speed-AD-module-AD9226-module-parallel-12-bit-AD-65M-data-acquisition-FPGA-development-board/32832584280.html

Both modules are widely used for SDR receivers and suitable for VLF receiver. Both use operational amplifier on the input, this is very good to receive frequencies below 2 MHz. It also works for 2..30 MHz, but if you want to get even better noises and distortions, you can replace op.amp with RF transformer, such as ADT4-1WT.

I tried it and it works good to receive VLF stations. You can even listen for CW at 50 Hz carrier.

Of course, you will need some FPGA borad for DDC processing, because computer cannot handle streams at speed 100-300 megabytes per second. You can use DE0-NANO education board or some cheap Chinese FPGA board.

If you want to cover FM band it will be more complicated, because it will needs ADC with sample rate above 220 MHz. So,it will be more expensive and more complicated to process the data flow.

But you can receive FM band with 65 or 105 MHz ADC through aliases. All what you need is just to put FM band-pass filter (instead of LPF) before ADC to cut off all unwanted frequencies. You also probably will need some amplifier, but actually FM stations usually so strong, so it can be received with no amplifier. I can listen it even without filter on the ADC input.

If you want more cheap solution, then you can use sound card with some mixer and filters. For example this one:
https://www.aliexpress.com/item/1PC-New-Arrival-AD831-High-Frequency-RF-Mixer-Inverter-0-1-500MHz-Board-Module/32823467841.html

Just put BPF before mixer and LPF after mixer. Also connect LOin to some RF generator and you will be able to listen for 0...500 MHz

[9aplus]

My advice is to buy second hand SDRPlay RSP1 which is perfect for requested task.
That units now go for 50 to 80 euro.
One ready application for spectral analyze, here -> https://www.sdrplay.com/community/viewforum.php?f=12&sid=df5d613fb3340793302b46f17b35ef7e

I am regular user of RSP1 for NDB beacons RX in 300-450 kHz range.

[NiHaoMike]

Get a cheap RTLSDR and add the direct sampling mod. Then add a preamp and some filters.

[coppercone2]

keep in mind high end DDS and DAQ systems will have different AA filters you can switch in for different wave shapes (i.e. stanford research DDS does this). There is no one size fits all realty.

You need to think about impulse response and your filter bode plot with regards to each other, along with the domain you are interested in.

[radiolistener]

For filter testing and tune you can use NWT.
This is a sweep gen + logarithmic/linear detector. It is available in different mods.
The most popular mods are:
- NWT70 (50 kHz - 80 MHz):
https://www.aliexpress.com/item/NWT70-NWT7-Sweep-with-0-50DB-attenuation-network-points-50k-85M/32763448183.html

- NWT500 (0.1 - 500 MHz):
https://www.aliexpress.com/item/1PC-NWT500-0-1MHz-550MHz-USB-Sweep-analyzer-attenuator-SWR-bridge-SMA-Cable/32808842510.html

For SDR receiver, you can use also RTLSDR but it has a lot of spours and too small dynamic range. But still usable.
https://www.aliexpress.com/item/RTL-SDR-Blog-RTL-SDR-V3-R820T2-RTL2832U-1PPM-TCXO-SMA-RTLSDR-Software-Defined-Radio-Dongle/32939551915.html

note: be careful, there are a lot of fake RTLSDR, they will not fit to your needs. I provided link to official seller. Don't try to buy from other sellers, it will be 99.99% fake.

RSP1 is a little better than RTLSDR and has some filters on the input.

[DaJMasta]

I ended up getting the RSP1A when I was looking for an SDR with low frequency capability (why I mentioned SDRPlay in particular), not the cheapest option, but it uses a better ADC than most budget SDRs (14 bit maximum), so you get some more dynamic range, pretty good specifications regarding the frontend available, and all of it is handled for you.

There are definitely cheaper options for the band of interest, and there are definitely good ways to do it getting your hands dirty, so to speak, but something like the RSP1A is a convenient, well documented option.

[radiolistener]

RSP1A uses a better ADC than most budget SDRs (14 bit maximum)

actually RSP1A uses 10 bit ADC: MSi2500

"14 bit" is just marketing, these 4 marketing bits is just SNR gain taken with DDC (due to decimation).

Real 14 bit ADC will get much better SNR with DDC, but nobody says that this is "18 bit" ADC

For example, MSi2500 works at 24 MHz and 10 bits, it equals to SNR=62 dB.
By reducing bandwidth from 24 MHz to 48 kHz, we get SNRbw=86 dB (equals to 14 bit).

If we get real 14 bit ADC AD6645, it works at 105 MHz and 14 bits, it equals to SNR=86 dB.
By reducing bandwidth from 105 MHz to 48 kHz, we get SNRbw=116.4 dB (equals to 19 bit).

If we get RTLSDR stick, it works at 28.8 MHz and 8 bit, it equals to SNR=50 dB
By reducing bandwidth from 28.8 MHz to 48 kHz, we get SNRbw=74.7 dB (equals to 12 bit).

Now you know what these marketing "14 bit" means

[DaJMasta]

According to a relatively in depth review of the RSP1A where SDRPlay commented on the chip used and the number of bits, it's not a 10 bit converter.  The MSi2500 is a 12 bit native (10.4 ENOB) that has been firmware 'upgraded' to actually output 14 bits worth of data, which do increase dynamic range.  Yes, it requires the lower maximum bandwidth and the extra bits maybe be derived from oversampling and decimation.... but they do result in higher dynamic range, and since the ADC is part of the whole integrated package, it seems like sort of an arbitrary distinction to say a "14 bit" converter which outputs 14 bits and has notably better dynamic range than the "12 bit" converter the chip normally runs (they mention 10dB dynamic range increase over the same chip in 12 bit mode when the frontend isn't really noisy from high gain settings) somehow is just marketing bull over a 12 (or 10) bit converter.


Perhaps more notably when I was looking around at options...... the vast majority of SDRs just use 8 bit converters.  Anything beyond that is an improvement worth noting, and I haven't really been in the practice of calling out manufacturers using their ADC's specified number of output bits in their datasheet vs. the effective number of bits (24 bit audio, for example)... it's just understood that dynamic range and noise figures are the limiting component of ADC output precision in a lot of cases and it's important to pay attention to those figures.


In any case, I am a fan of >8 bit converters, and they are available.

[radiolistener]

According to a relatively in depth review of the RSP1A where SDRPlay commented on the chip used and the number of bits, it's not a 10 bit converter.  The MSi2500 is a 12 bit native (10.4 ENOB)

according to the datasheet, MSi2500 has 10 bits resolution. And there is no mention about 12 bits.
But unfortunately there is no mention if these 10 bits is ENOB or real native resolution.
So, may be you're right and it is 12 bit.

but they do result in higher dynamic range, and since the ADC is part of the whole integrated package, it seems like sort of an arbitrary distinction to say a "14 bit" converter which outputs 14 bits and has notably better dynamic range than the "12 bit" converter the chip normally runs (they mention 10dB dynamic range increase over the same chip in 12 bit mode when the frontend isn't really noisy from high gain settings) somehow is just marketing bull over a 12 (or 10) bit converter.

This is incorrect to say 14 bit, because any SDR DDC receiver has SNR improvement due to reduced bandwidth.
This is not unique feature of RSP1A.
So, if we say that it has 14 bit, then we need to say that RTLSDR also has 10-12 bits.
Otherwise we will not be able to compare these bits, because they has different meaning.

Even more, we can go further and get any desired bandwidth and write for example that RTLSDR stick has "24 bits ADC"... But you know, that these "24 bits ADC" is worse than native 12 bits ADC. 
Since there is no mention for which bandwidth these "24 bits" were specified exactly, the buyer will think that this is native 24 bits at ADC clock frequency. And will be deceived.

[DaJMasta]

They do say in their documentation that it is "native 14 bit", as in, the data is output as 14 bit.  Sounds like there is some additional processing, but if that's the granularity that shows on the output, I don't know why calling it that would be incorrect.  Yes, you're going to end up with some LSBs worth of noise and the ENOB/dynamic range/ADC noise floor figures are more valuable, but 14 bit is still accurate, and they do specify that the 14 bit output is only at 6MHz or below input bandwidth, and that the 12 bits/10.4 ENOB is available for the full receiver bandwidth so..... I just don't think it's meaningless marketing fluff, it's a description of how the architecture works.

If your 12 bit scope has <11 ENOB, I don't have any problem with being labeled as a 12 bit scope as long as there's ENOB/dynamic range/ADC noise figures.  Likewise, if the internal ADC architecture is 8 bit and it oversamples to output 10 bits, I don't have any issue with it being advertised as a 10 bit ADC so long as you're also advertising the effective sample rate of the 10 bit output.  If you advertise your SDR receiver as 16 bit and 50MHz bandwidth, but your 100MS/s 10 bit converter can only achieve 16 bits of output at a few hundred Hertz bandwidth, then it's false advertising/marketing wank.  If you describe the behavior of the system in the specification.... that's just showing off your strengths.

   We're splitting hairs about wording, and I don't see how their advertisements are inaccurate even if a better understanding of the system gives you a better picture of real-world performance.  If I had said 14 bit at 6MHz bandwidth in my initial post, would this whole conversation have been avoided? 

[radiolistener]

If I had said 14 bit at 6MHz bandwidth in my initial post, would this whole conversation have been avoided? 

yes, technically it will be more clean.
But I don't think that RSP1A has 14 bit at 6 MHz.

In order to get 14 bit SNR at 6 MHz bandwidth, 12 bit ADC should work at 200 MHz.
I'm not familiar with MSi2500 chipset, but as I understand it's ADC works at 24 MHz.

With 12 bit @ 24 MHz you can get just 12.5 bits @ 6 MHz.

I think they specified 14 bit for bandwidth 750 kHz.
If you say 14 bit @ 750 kHz it may be true.

But we also didn't take into account that SNR improvement is possible when ADC has ideal Gaussian noise. Which is not the case in the real world. There is need dithering. Also such SNR improvement will be real for zero jitter ADC clock. Which also is not possible it the real world. With Ultra Low Phase Noise oscillator for ADC clock and very good quality ADC we can get some close value. But this ADC has ENOB just 10 bits. I think, in better case it will be 10.5-11 bits @ 6 MHz or even worse.

This is why talking about native ADC resolution at native ADC bandwidth is more correct.

May be there is some mistake, you can check my calculations:


processGain = 10 * log10( 24 / (2*6) ) = 3.01 dB
1 bit = 6.02 dB

So, processGain is just 0.5 bit.

I don't see how to get 14 bit @ 6 MHz bandwidth with 12 bit @ 24 MHz ADC. It's just impossible.

[David Hess]

2) Strength out-of-band signals can overload your ADC, if it cannot handle enough dynamic range. This is why more bits resolution is better, because SNR = 6.02 * N + 1.76 [dB], where N is ADC bits resolution.

This is a killer.  Even high performance wideband receivers (like common scanners) have problems with this because the first mixer stage gets overloaded creating cross modulation products.  Single band receivers with a wide tuning range alleviate this somewhat by using tunable preselectors (often using varactors) or switchable input filters.  A DC through broadcast FM band receiver is going to have is performance limited by this.

Incidentally, in the HF band by itself this is not quite as much of a problem in many areas because the background noise level can be very high but it still becomes an issue when an in-band transmitter is in close proximity.  HF receiver features like pass band tuning, notch filters, and antenna nulls do not exist without reason.

I would use a dedicated receiver for DCF-77 and probably still include a preselector filter.  But this could take the form of a separate receiver front end for a more general purpose receiver.  Another advantage of this is that many RF amplifiers and mixers have rather high flicker noise which will degrade sensitivity at such a low frequency anyway.

[petert]

Thank you everyone for the great input! I really appreciate it.

I guess I will start with SDRPlay as a spectrum analyzer and go from there. Dave announced a video about this very topic, so I'll wait a little longer to see if there is more insight there as well.

Apparently there are fake SDRPlays around to watch out for:
https://www.rtl-sdr.com/cloned-sdrplay-and-airspy-units-now-appearing-on-aliexpress-ebay/

Another nice spectrum analyzer software (for SDRs): https://github.com/xmikos/qspectrumanalyzer

A related eevblog video:

In the last video ( ) people mentioned useful pre-amplifiers:

Check out the SDR Play receivers for making a Poor Man's Spectrum Analyzer. 10 Mhz visible bandwidth, front end filters, built-in preamp, etc. and the software is very well done. Well worth the extra price which is very reasonable considering how well they work compared to a cheap "SDR Dongle".

I've got a small question: how good are those <10$ eBay "low noise" amplifiers? Are they actually low noise?
In my experience these things from minicircuits cost 10 times more, if those low noise amplifiers from eBay work fine that would be great for the cheapo student radio interferometer telescope project I'm trying to build!

Answer:

Most of the ones I've tried actually are. They use amplifiers from Mini-Circuits and other high-quality providers.

[petert]

It is ok if the filtering is not perfect, I can do it in software well enough. (Like in the soundcard example.)

No, it is not ok. You cannot do it in software, because filtering should be done before ADC.

The soundcard has a built-in filter to limit the signal bandwidth.

I meant to test an analog filter's performance I made in hardware, by using SDRPlay as a spectrum analyzer that gets fed by the analog filter.

[radiolistener]

I meant to test an analog filter's performance I made in hardware, by using SDRPlay as a spectrum analyzer that gets fed by the analog filter.

SDRPlay uses bandpass filter on the input, so it's frequency response will not be flat.

You can try to use some ADC with cheap Chinese logic analyzer, like that:
https://www.aliexpress.com/item/EZ-USB-FX2LP-CY7C68013A-USB-logic-analyzer-I2C-serial-and-SPI-core-board-Source-Code-development/32851287266.html

Here is example on how to implement it: http://oh2eat.dy.fi/widebandsdr.html