Cheap and Small Dynamic Signal Analyzer?

[electrolust]

NanoDSA!

[maxwell3e10]

AD4003 ADC and OPA828 input amplifier. Anyone has better suggestions?

[Berni]

Sounds like a pretty good fit alright. Perhaps also add a DAC to also let it do frequency response or distortion measurement.

Pair it up with a ST MCU with hardware floating point to do the FFT. The more gruntry ones are fast enough to do live real time FFT with no gaps at a few MSPS. Slap a LCD on the front and you got a NanoDSA.

[electrolust]

always best to start from requirements. THEN decide the components. The hard part is going to be the software.

[Berni]

Well if you are making something similar to the NanoVNA then the number one requirement is cost.

Hence you just pick the best components that the BOM budget will allow and squeeze the best specs possible out of those. But yeah most of the work here is indeed the software. Its not difficult to do but a lot of work to give it all the useful features of a DSA.

[gigavolt]

NanoDSA!

  my thoughts exactly.  I'm surprised nobody has tried to do it yet, but I suppose that the combination of low frequency and low noise makes it difficult.

[Berni]

NanoDSA!

my thoughts exactly.  I'm surprised nobody has tried to do it yet, but I suppose that the combination of low frequency and low noise makes it difficult.

Guess more of a problem is that very few people actually want one. They are perfectly happy with there existing Rigol scope.

The low frequency and noise is not that much of a problem, you just pick a really nice opamp for the AFE. More of a difficulty is getting a spurious free response over the whole range. The noise floor is so low that even the tinyest of signals show up as a spike. This means a great deal of care must be taken in the shielding and PCB design to avoid any power supply noise or digital signals from getting in there and showing up. For example in my diy sound card design i had a LDO become just slightly unstable, not enough to cause obvious problems or a huge sine wave on the output when when poked with a scope, but it was singing enough to produce a small spike in the FFT, was quite a hunt to find where it was coming from.

[KE5FX]

For example in my diy sound card design i had a LDO become just slightly unstable, not enough to cause obvious problems or a huge sine wave on the output when when poked with a scope, but it was singing enough to produce a small spike in the FFT, was quite a hunt to find where it was coming from.

Let me guess, the part number contained the digits "1117"?

[Berni]

Let me guess, the part number contained the digits "1117"?

Forgot what it was exactly but it was some fancy pants high PSRR LDO whos job was to prevent crap from getting into the sensitive front end. So yeah... you had ONE job.

As for 1117.. yeah i don't even need a scope anymore to tell when one of those damn things becomes a oscillator. Last time i started wondering why my board is making a whistling sound even when there are no switching regulators on it. Turns out it was oscillating so wildly that the ceramic capacitors around it started signing loud enough to be noticed even in a not so quiet room.

[2N3055]

Let me guess, the part number contained the digits "1117"?

Forgot what it was exactly but it was some fancy pants high PSRR LDO whos job was to prevent crap from getting into the sensitive front end. So yeah... you had ONE job.

As for 1117.. yeah i don't even need a scope anymore to tell when one of those damn things becomes a oscillator. Last time i started wondering why my board is making a whistling sound even when there are no switching regulators on it. Turns out it was oscillating so wildly that the ceramic capacitors around it started signing loud enough to be noticed even in a not so quiet room.

"The singing regulator", new Broadway show..... LOL...
Yep, I avoid them too...

[gigavolt]

Guess more of a problem is that very few people actually want one. They are perfectly happy with there existing Rigol scope.

That too, because ultimately the reason I would want it is for better noise performance/higher dynamic range, but in a much smaller bandwidth (<100 kHz) than my scope.  I'm guessing most hobbyists can "get by" with 8 bits.

[Frex]

Hello,

Just for information, I currently design such type a device.
It's a full open source 2 channels analyzer that use the LT2380-24,
a  24 bits 2MSPS high performance ultra low THD ADC.
It's not really a very low cost device and the full project is not yet finished.
(I'm writing the FPGA firmware for now).
Anyway, maybe that can interest some people.
I publish the project advancement on a DIYaudio thread HERE

The design has been firstly checked with tweaked EVM boards few years ago,
all details and history can be read on this other thread HERE .

Regards.

FRex

 

[maxwell3e10]

This is a cool project! I guess for a general purpose device one might make a more versatile input stage, with several different gain settings and a good single-ended noise level.

[KE5FX]

Very nice.  I was about to say that any effort along these lines needs to be multichannel.  I'm short on free time at the moment but will bookmark the diyaudio thread(s) for later perusal. 

[trampas]

I was looking for something like the DSA that can measure amplifier noise floor and was wondering what options are out there?

[TopQuark]

I was looking for something like the DSA that can measure amplifier noise floor and was wondering what options are out there?

I was thinking the same. Couldn't find anything I like in the market, so I decided to roll my own. I ordered a AD4630-24 eval board (back ordered till late Nov unfortunately) to pair it with a beefy Zynq ultrascale dev board I already have, so the hardware should be plug n play.

It will take the AD4630-24 32.768ms to collect 65536 samples at 2 MSPS. Vivado estimates the zynq chip can do 2 channel, 30 bit input, 65536 bin FFT in roughly 1.5ms, so it should be no problem at all to have 32768 lines FFT resolution and 1MHz real time bandwidth. The AD4630-24 has 105.7 dB typical dynamic range as stated in the datesheet.

For context, the SRS785 cost 14k USD, can only do 800 line FFT, 102.4 kHz real time bandwidth, 90 dB typical FFT dynamic range. The zynq dev board I have cost around 1k USD (massively overkill for the job), and the AD4630-24 eval board is 200 USD. All in all, the cobbled together solution should beat the pants off the SRS785.

Of course, talk is cheap, and I haven't written a single line of code for the project yet, so don't have your hopes set high for now 

[Berni]

I was looking for something like the DSA that can measure amplifier noise floor and was wondering what options are out there?

For just noise measurement you can get away with a preamplifier going into an oscilloscope. Mainly because you don't need a lot of dynamic range for this, just high enugh sensitivity.

There are 1000x gain preamplifiers out there from Stanford, but you can just as easily build your own using low noise opamps.

As for measuring the noise of an opamp itself you just build a high gain amplifier circuit out of the opamp you want to test and it will amplify up its own noise to a point where it is easily measurable.

I was thinking the same. Couldn't find anything I like in the market, so I decided to roll my own. I ordered a AD4630-24 eval board (back ordered till late Nov unfortunately) to pair it with a beefy Zynq ultrascale dev board I already have, so the hardware should be plug n play.

That ADC does look like it could make a nice "NanoDSA" when paired up with a fast MCU with hardware floating point. Don't think you quite need the power of an FPGA to do FFTs that fast.

Tho the chip is out of stock everywhere as per chip shortage apocalypse.

[TopQuark]

That ADC does look like it could make a nice "NanoDSA" when paired up with a fast MCU with hardware floating point. Don't think you quite need the power of an FPGA to do FFTs that fast.

Yea you can probably get away with a fast MCU like a STM32H7 or something similar. Though once you add in windowing, digital anti aliasing filtering, network analysis, averaging, digital mixing etc. I think it will be a pain to maintain real time analysis capability with a MCU. Perhaps a happy medium would be a cheap Chinese FPGA (e.g. the ones in sipeed tang boards) paired with a decent MCU.

To be honest, the main motivation for wanting to use a overkill zynq ultrascale dev board is because the ADC eval board has a pain in the butt FMC connector, and the FPGA dev board just so happens to support that connector  .

[Messtechniker]

ESI Juli@—www.esi-audio.com
Sound Blaster X-Fi—www.creative.com
EMU 1212, 0202, 0404—www.emu.com
Lynx L2, L22—www.lynxstudio.com
Asus Xonar Essence—www.asus.com

M-Audio Audiophile 192—www.m-audio.com
(The one I use, no longer available. Note: is PCI,not PCIe)

These have no reconstruction filters which stop an 20kHz regardless of sample rate.
Some may only be available used....

[TopQuark]

AD4003 ADC and OPA828 input amplifier. Anyone has better suggestions?

Another option would be AD7768-1, seems purpose made for the job, in stock and not that expensive. I might get a few to play around with.

[TopQuark]

NanoDSA!

my thoughts exactly.  I'm surprised nobody has tried to do it yet, but I suppose that the combination of low frequency and low noise makes it difficult.

Guess more of a problem is that very few people actually want one. They are perfectly happy with there existing Rigol scope.

The low frequency and noise is not that much of a problem, you just pick a really nice opamp for the AFE. More of a difficulty is getting a spurious free response over the whole range. The noise floor is so low that even the tinyest of signals show up as a spike. This means a great deal of care must be taken in the shielding and PCB design to avoid any power supply noise or digital signals from getting in there and showing up. For example in my diy sound card design i had a LDO become just slightly unstable, not enough to cause obvious problems or a huge sine wave on the output when when poked with a scope, but it was singing enough to produce a small spike in the FFT, was quite a hunt to find where it was coming from.

I do have a need for a DSA. The question is more about if I will do it as a one off or am I trying to build a "NanoDSA".

If I am just building one for myself, I'd just wait for the AD4630-24 eval board to ship, plug it into my fpga dev board, write just enough code to have the project do what I want to do, and be done with it.

On the other hand, if I am building a "NanoDSA", I'd go with a couple of AD7768-1, or even a stereo audio ADC, pair it with a STM32 (or clone, even), and build something down to a cost.

I think I'll build some form of a DSA, because I genuinely have a need for one, just not sure if I have the will and motivation to do a "NanoDSA" project.

[maxwell3e10]

If you think of making it as a project, I would go for at least 1-2 MHz sample rate. This would put it clearly above the sampling rate of audio products, which otherwise are hard to compete with in terms of hardware performance.  Software is another question. It would be useful to take input from an audio interface (ADC+DAC), which often have a USB connection and standard drivers, and implement all software functionality of an SR785.

[David Hess]

Guess more of a problem is that very few people actually want one. They are perfectly happy with there existing Rigol scope.

That too, because ultimately the reason I would want it is for better noise performance/higher dynamic range, but in a much smaller bandwidth (<100 kHz) than my scope.  I'm guessing most hobbyists can "get by" with 8 bits.

Exactly, and it is too specialized of an instrument.  For most network analysis, 8 bits is plenty and a DSO with signal generator only needs appropriate firmware.  There are some relatively low cost products intended for the audio market that operate up to 192 or 384 kS/s like the QA403.  At higher cost, Cleverscope operates at much higher frequencies, but with lower resolution and an oscilloscope input stage will compromise performance.

I have thought about designing a more general bidirectional 2-port low frequency VNA for low frequency network analysis, but the bidirectional design would compromise performance as a dynamic signal analyzer because of limited common mode rejection, and I am not sure that is a problem worth solving.

[TopQuark]

Guess more of a problem is that very few people actually want one. They are perfectly happy with there existing Rigol scope.

That too, because ultimately the reason I would want it is for better noise performance/higher dynamic range, but in a much smaller bandwidth (<100 kHz) than my scope.  I'm guessing most hobbyists can "get by" with 8 bits.

Exactly, and it is too specialized of an instrument.  For most network analysis, 8 bits is plenty and a DSO with signal generator only needs appropriate firmware.  There are some relatively low cost products intended for the audio market that operate up to 192 or 384 kS/s like the QA403.  At higher cost, Cleverscope operates at much higher frequencies, but with lower resolution and an oscilloscope input stage will compromise performance.

I have thought about designing a more general bidirectional 2-port low frequency VNA for low frequency network analysis, but the bidirectional design would compromise performance as a dynamic signal analyzer because of limited common mode rejection, and I am not sure that is a problem worth solving.

https://www.testunlimited.com/pdf/an/5988-6774EN.pdf
According to page 20, a 2 channel DSA (one ch for ref, one ch for DUT output) with a noise source should make for a blazingly fast if not "real time" network analysis / bode plotting tool. I think if I want to include this function, I will indeed want to have at least 2Msps for the ADC so that the bode plot can cover lets say 100Hz to 1MHz in a single sweep.

DC2390A eval board from ADI (https://www.analog.com/media/en/technical-documentation/user-guides/DC2390AF.PDF) would be the perfect solution paired with a FPGA board, with two LTC2500-32 and two fast DAC. Shame the connector on the board is a HSMC which doesn't plug into anything I own.

[David Hess]

https://www.testunlimited.com/pdf/an/5988-6774EN.pdf
According to page 20, a 2 channel DSA (one ch for ref, one ch for DUT output) with a noise source should make for a blazingly fast if not "real time" network analysis / bode plotting tool. I think if I want to include this function, I will indeed want to have at least 2Msps for the ADC so that the bode plot can cover lets say 100Hz to 1MHz in a single sweep.

An impulse source is even faster.  Or an edge can be used as the source and then the signal differentiated before the FFT.  Limited only by the FFT execution time and the time to capture the acquisition, the display of the network response can be real time.  I am actually disappointed that the recent oscilloscope bode plot generators cannot use this method.

DC2390A eval board from ADI (https://www.analog.com/media/en/technical-documentation/user-guides/DC2390AF.PDF) would be the perfect solution paired with a FPGA board, with two LTC2500-32 and two fast DAC. Shame the connector on the board is a HSMC which doesn't plug into anything I own.

There are a bunch of products which almost do it, but at least for me the software is not trivial either.

I have been thinking more in terms of a VNA because it would be so easy to use as a super LCR meter and I have never seen one that operates in the way that I am thinking.  Of course maybe there is a good reason for that.  HP published an application note on the subject which discusses the advantages and disadvantages of the various designs, and I should reread it.

I am not really interested in distortion measurement.  There are too many products which already handle that well so making another distortion analyzer is less interesting.