Realtime FFT for 0-30kHz. Best bang for the buck?

[superUnknown]

I want a good realtime FFT for mechanical vibe analysis. Can you recommend a good value instrument?
Accuracy isn't critical, as vibe analysis is very interpretive.  It's important to have a simple display, non-glitchy, non-fiddly, and simple to set up. I have a Rigol 1054z but it doesn't cut it. How's the Digilent analog discovery 2?  Thanks!

[daveshah]

I'm not sure what the signal levels you expect are like but have you considered a USB sound card - ones with sample rates >= 96kHz are still cheap and cover your range of interest - with some free or paid software? There seem to be plenty of options out there.

[LaserSteve]

Sound card and Daqarta?

http://www.daqarta.com/

Steve

[DaJMasta]

If you can find a sound card that doesn't filter out the high frequencies, that's a good option.  You can get very low noise and very high dynamic range with standard drivers and a selection of programs.  My current audio interface though, for example, has a pretty sharp rolloff at about 21.5kHz, so it wouldn't be suitable for your application, and cheaper options probably don't specify whether they do this or not - so the best value would probably be had in finding user frequency response or noise floor plots that go that high.


They don't make them really anymore (I think the soundcard based solutions have taken over), but if you're willing to consider used gear, look at Dynamic Signal Analyzers.  Dave's mentioned them in a couple videos and there are a few options available, they are basically spectrum analyzers for the low end of the frequency spectrum, and were designed for vibration or acoustic analysis.  There are newer ones, but not from many manufacturers and not at nearly the same value proposition as the older units.

If you find one with the frequency response, I do think the soundcard route is the best bet.  Something like the QA401, which has a generator and is designed for audio work, may be ideal because they even use BNCs for a differential input and you wouldn't have to go swapping cables everywhere, but I don't know whether or not they filter the high end.

[rstofer]

I have the Analog Discovery 1 and it does a very nice job.  I sent a 10 kHz 2V P-P square wave into the scope channel so I could see all the odd harmonics.  No real reason, it just seemed like the way to demonstrate the unit.

[wraper]

If using a scope, cheap GDS-1054B will do the job very well.

[wraper]

I have the Analog Discovery 1 and it does a very nice job.  I sent a 10 kHz 2V P-P square wave into the scope channel so I could see all the odd harmonics.  No real reason, it just seemed like the way to demonstrate the unit.

Actually poor result. Should be very low number of points used for FFT.

[awallin]

QA401 looks quite interesting, also for voltage-regulator noise investigations and similar low-noise work (QA401 could be fed from an LNA)
https://www.quantasylum.com/content/Home/tabid/40/Post/268/QA405-and-QA401-Status

any other sound-cards out there with 192 kSps and similar specs?

[jpb]

QA401 looks quite interesting, also for voltage-regulator noise investigations and similar low-noise work (QA401 could be fed from an LNA)
https://www.quantasylum.com/content/Home/tabid/40/Post/268/QA405-and-QA401-Status

any other sound-cards out there with 192 kSps and similar specs?

The Creative Labs X7 has some nice specs and it has (apparently - I don't have one myself yet but have e-mailed Creative about it) has a mode by which it can record directly without using the sound processor which distorts things. In the UK at least it is much cheaper than it used to be.
http://uk.creative.com/p/sound-blaster/sound-blaster-x7

Here is a quote from their support department who are quite helpful though not used to answering questions from electronics engineers:

Regarding your last email, please be informed that the max. Bit-depth/sampling frequency for recording is up to 24-bit/192KHz and there are two Line In RCA-jacks for left and right channel, hence the input is stereo. The size limit of the recording is not related to the X7 itself but depends on the sufficient hard drive space as well as on the recording software itself: some recording programs only allow recordings up to a specific file size.

I was interested in it as it allows unlimited recording to get very large data sets (though in your case I think you want real time) whilst the Q401 is limited by its own memory. (I was looking to record data for doing oscillator stability measurements over long periods of time.)

Another idea I had was to use a 32 bit DAC developer board (Linear Technology DC2222A-B)  coupled either to a logic analyser or using Linear Technology's own demo scope board (but this is pricey):
http://www.digikey.co.uk/products/en/development-boards-kits-programmers/evaluation-boards-analog-to-digital-converters-adcs/791?k=LTC2508

[orolo]

If you go the soundcard way, and since you mention 0-30kHz, make sure the card is DC coupled. Mine isn't, and when presented with a low frequency squarewave, the droop in the on state is quite noticeable.

[rstofer]

I have the Analog Discovery 1 and it does a very nice job.  I sent a 10 kHz 2V P-P square wave into the scope channel so I could see all the odd harmonics.  No real reason, it just seemed like the way to demonstrate the unit.

Actually poor result. Should be very low number of points used for FFT.

I spread the time base to eliminate higher harmonics.   A more representative FFT would look something like the attached.  The question is:  Are the higher harmonics of any interest in a 30 kHz vibration analysis?

[snarkysparky]

Wouldn't all sound cards be DC coupled.  After all what purpose would the DC level serve.

[rstofer]

I got to thinking that my 10 kHz examples were pretty useless when the frequencies of interest are very low.  So, I set the square wave down to 10 Hz.  It's kind of weird waiting for the buffer to fill but the AD actually produces something useful.  See attached...

ETA:  Which brings up a question of how "real time" the FFT really is when the buffer is 10 seconds long.  I suppose it is possible to come up with an arbitrary waveform or even play a recording of real samples but those are outside my range of interest.

[awallin]

QA401 looks quite interesting, also for voltage-regulator noise investigations and similar low-noise work (QA401 could be fed from an LNA)
https://www.quantasylum.com/content/Home/tabid/40/Post/268/QA405-and-QA401-Status

any other sound-cards out there with 192 kSps and similar specs?

The Creative Labs X7 has some nice specs and it has (apparently - I don't have one myself yet but have e-mailed Creative about it) has a mode by which it can record directly without using the sound processor which distorts things. In the UK at least it is much cheaper than it used to be.

The X7 looks interesting! Do you know of anyone using it for electronics/noise analysis?
Google finds this test report - but I think it shows only tests of the DAC, not ADC..
http://reference-audio-analyzer.pro/en/report/dac/creative-x7-le-pf.php
some graphs over here also, they might show loopback, DAC->ADC, not sure
http://www.head-fi.org/t/756102/creative-sound-blaster-x7-detailed-review-impressions
that page mentions the ADC PCM4220. page 9-11 of the datasheet have noise floor plots:
http://www.ti.com/lit/ds/symlink/pcm4220.pdf
for 192 kS/s there's a rising noise-floor of the FTT towards 100kHz... (similar to what the QA401 page talks about..)

[jpb]

QA401 looks quite interesting, also for voltage-regulator noise investigations and similar low-noise work (QA401 could be fed from an LNA)
https://www.quantasylum.com/content/Home/tabid/40/Post/268/QA405-and-QA401-Status

any other sound-cards out there with 192 kSps and similar specs?

The Creative Labs X7 has some nice specs and it has (apparently - I don't have one myself yet but have e-mailed Creative about it) has a mode by which it can record directly without using the sound processor which distorts things. In the UK at least it is much cheaper than it used to be.

The X7 looks interesting! Do you know of anyone using it for electronics/noise analysis?
Google finds this test report - but I think it shows only tests of the DAC, not ADC..
http://reference-audio-analyzer.pro/en/report/dac/creative-x7-le-pf.php
some graphs over here also, they might show loopback, DAC->ADC, not sure
http://www.head-fi.org/t/756102/creative-sound-blaster-x7-detailed-review-impressions
that page mentions the ADC PCM4220. page 9-11 of the datasheet have noise floor plots:
http://www.ti.com/lit/ds/symlink/pcm4220.pdf
for 192 kS/s there's a rising noise-floor of the FTT towards 100kHz... (similar to what the QA401 page talks about..)

No I don't know anyone using it for electronics/noise analysis but  the specs looked good and the second of your links (which was one of the reviews I read) had a loop-back test which looked good to me.

I was thinking of getting it and if it didn't work out I would at least have a good external sound card/amp for my computer

The reason why I'm hesitating is that for my application (measuring the beat frequency from oscillators for Alan Deviation measurements) I really wanted to be able to use an external clock locked to my reference oscillator. But you have to spend ten times as much to get a ADC that has this (i.e. a pro-audio one).

This is when I looked into using something like the Linear Technology ADC demo board. The board itself is not too expensive (around £150 + VAT) but the problem then is getting the data.  The test board that Linear Technology sell which comes with FFT software is expensive (several hundred pounds) and also it has limited memory (you can't stream directly to a pc). But, it says in the datasheet for the demo board that you can connect it to a logic analyser so I thought perhaps using a cheap USB analyser but none of these are 32 bit.

So my latest thoughts are to perhaps use the demo-board plus a high end Arduino and write my own software but its starting to become a bit of a major project!