instrument for audio signal analysis

[Rax]

Hi all,
First post here though I seem to be coming here more and more often when researching something having to do with my bench...

I'm looking for something suitable for audio signal spectrum visualization (some call them "spectrum analyzers," some "dynamic signal analyzers...") I'd like to be able to cover 10Hz to 100kHz at a minimum, have a decent noise floor. It seems that most digital scopes come with some FFT options, so they may be an option. Especially given my budget for this - I'd keep this well under $1,000, though under $500 would be really awesome if useful - they may be all I got. I am based in the US, if that helps with the suggestions.

I've been looking at some Instek models (for instance, GDS-1000A-U), Tektronix (TBS1000, though turned off by the backlash on this forum following their release), Rigol DS1054Z. Agilent may have some affordable models. In the way of specialized spectrum analyzers, I've also been looking at some vintage HP models, such as 3582A etc. I admit to being a fan of these old HP/Tek instruments, of which I have some (a Tek 453A and an HP334A, for instance) and they are built like works of art. Having said that, I will listen to the voice of reason and look for something that's the best for the intended use in the given budget.

It may be said that a true solution to what I'm looking for may be an Audio Precision box and such niceties, but those are far out of my hobbyist reach...

Thanks a lot for your input! Looking forward to your thoughts.

[Messtechniker]

Why spend so much money, if a good sound card and the program
called "Audio Tester" for a few bucks might serve your needs?
See enclosed example. Bandwith can go up to just under 100 kHz
with the sound card I use (M-Audio - Audiophile 192) .
Just noticed. The Audiophile 192 is no longer available.
So you might have to look at an ESI Juli@ or something similar.

But make sure that the frequency response extends to around 100 kHz.
Some high sampling rate sound cards have a fixed built-in filter,
cutting in already at 30 kHz regardless.

Yours Messtechniker

[Messtechniker]

Addendum: The graph shown is that for a small power amp just
before clipping (add +20 dB to the levels shown) .
The soundcard itself has a much lower distortion.

Yours Messtechniker

[Rax]

Why spend so much money, if a good sound card and the program
called "Audio Tester" for a few bucks might serve your needs?

Messtechniker - thanks for pitching in. Doing this is an interesting idea, though I should have probably make clear I am looking for an instrument to put on the bench/shelf. I built a buffer to place between the computer and DUT a while ago (which I never really ended up using), but I remain very weary of risking the computer assigned for this use with a direct route to the testing end of the probes. This is also because a lot of my work is with tubes.

Also, I'm not sure what happens to the calibration when this solution is adopted.

Besides, I learned in time to appreciate dedicated instruments for given tasks - they just tend to perform better, take much less time to perform, present much better logistics around them. A touchpad/mouse and a large screen (with a computer attached to it) near or on the bench are pretty inconvenient. If what I need ends up being rendered to be a scope, it may also replace my old 453 for most tasks... That much more an advantage, as the space I have is at a dear premium.

I am willing to do this as a temporary solution, though.

[nctnico]

A Tektronix TDS500/TDS700 series with the FFT option could be a good one for little money. These scopes have high-res acquisition and the FFT also has averaging so it isn't jumping allover the place. The FFT isn't very fast though especially with a lot of averaging. I have used these scopes for some audio analysis work in the past to measure the bandwidth of a system. I'm quite sure the FFT on the low end Chinese scopes are not up to this task.

Another option would be a spectrum analyser which goes down to 10Hz. IIRC some radio communication analysers have audio analysis capabilities as well but I don't know if they work up to 100kHz. You'll need to adapt your signal to 50 Ohm inputs though.

[deadlylover]

I have a Rigol DS1054Z and it is bloody hopeless for audio FFT measurements (for measuring distortion anyway). 

If you're more interested in measuring THD and such, how about taking a sort of half measure and getting an audio analyser? You can then perform FFT on the analyser's post notch monitor outs with a sound card, with no risk of blowing up your PC.

IIRC, the Shibasoku 725 with a few simple mods can dance with the Audio Precision flagship at the -120dB THD+N mark for 1kHz@20kHz BW. They are still very much state of the art, and if you look hard enough you can easily get one for <$1000 delivered. And for when you don't want to perform FFT, you still have a nice instrument that can measure levels on individual harmonics up to the 5th separately.

Not quite what you were looking for, but it's another wallet destroying option perhaps. 

[amirm]

Scopes are designed to have very high speed but pretty low resolution.  For audio, you want it the other way around.  An 8-bit Analog to Digital converter is not going to be at all useful for audio.  As said, if you want cheap, just get a sound card and use the free software out there.  You would have to be careful with input levels and such (don't try to measure an amp!) but should be good for line level outputs.

[vk6zgo]

You could go "retro",& use a Noise & Distortion meter.(N&D Set),which was the standard way of testing in Broadcasting for many years.

OK,in a straightforward Distortion test,they really show you THD+Noise.
Noise is normally much lower in level & has very little effect upon the result,but if the reading appears abnormally high you can then do a Noise Test to see if you have a Noise,rather than a Distortion problem.

You can chase up various harmonics by nulling them out with the filters,if you really want to know how much each contributes.

Another possibility is to approach it from the other direction,& use a Selective Millivoltmeter,which is what they used in testing Hearing Test equipment.
With these,you can tune in the various Harmonics & so determine their contribution to THD.

You can look for Intermods with a Selective Millivoltmeter,too---that's how we used to check our ISB transmitters back in the day!

On eBay,I have often seen HP Audio Spectrum Analyser plugins for the various Spec An mainframes.
Not many people want them,so they languish for ages,only to come up again in a new listing.

[Keicar]

If you're on a budget, there's a lot you can do with the demo version of ARTA - although you've got to deal with the fact that your simultaneously testing your audio interface in addition to the DUT - which can be frustrating. I ended up getting a HP 3562a Dynamic Signal Analyzer - which does a fine job, although the limitations imposed by the 13-bit ADC mean that while the dynamic range is acceptable, you can't necessarily see down to microscopic levels of distortion - but it's certainly more trustworthy than using a PC audio interface that was never intended for test & measurement duty. Another product that has piqued my interest is the QuantAsylum QA-400 https://www.quantasylum.com/content/Products/QA400.aspx which might be worth a look.

[Cubdriver]

With some patience and judicious sniping you should be able to snag a functional 3582A, 3561A or 3562A from the 'bay for well under $1k; likely even cheaper if you're willing and able to repair it.

-Pat

[kc8apf]

After playing with a HP331A for a while, I jumped on the 3562a bandwagon.  As others have mentioned, it doesn't have the impressive noise floor of an AP but I find it to be a pleasure to use.  I got mine for $500 on eBay.  They usually run around $1k but occasionally you find them lower.

[radhaz]

Check this repair:
http://amp8.com/vv-amp/onlife/u-22_3.htm

[singapol]

A HP8903B perhaps although this is not a SA? You can find one on ebay at hobbyist prices.

http://www.keysight.com/en/pd-1000002339%3Aepsg%3Apro-pn-8903B/20-hz-to-100-khz-audio-analyzer?cc=SG&lc=eng

[SteveLy]

The way I do audio signal analysis is to use a digital recorder to capture the signal, convert it to floating point and then do the maths on the computer. It's cheap, easy and very effective if you don't need real-time. (I code what I need in C/C++; but you could also use higher level tools such as Matlab, Octave, Scilab, etc - even Audacity will do basic FFT stuff with a few clicks.)

You could of course also use a sound card for capture but stand-alone audio recorders tend to give better signal-to-noise per buck than sound cards. Most of them will also give you live digital audio via USB so real-time processing should also be doable, but I have not explored that. If you don't need real time, then the convenience of having all the data in a file and being able to analyse the heck out of it is very nice.

[Marco]

What did you use for a pure oscillator? (Although I guess you could do a pass through first and subtract the oscillator's harmonics in the complex frequency domain.)

[Marco]

How do you cancel the harmonics from the DAC?

[GK]

Depends on requirement. I am working on a high end audio DAC project that goes all the way down below the latest AP's limit, specifically -126dB THD, 140+dB DR. In this case, some improvisation needs to be done.

Dynamic range (DR) is essentially the same as SNR. Most high-performance 24-bit audio DACs only manage 16-bit performance with the best 32-bit DACs reaching 20 or 21 ENOB. 21 ENOB returns a SNR of 126db - well short of 140dB, which requires 23-bit performance.

For instance this one only does -123dB SNR and its THD performance is well short of bettering -126dB.

http://www.ti.com/lit/ds/symlink/pcm1795.pdf

I'd be interested to know the part # of the DAC you are using.

[GK]

I built my own analyzer. Feeding the distortion residual into a PC sound card for FFT gives the best of both worlds.

http://www.users.on.net/~glenk/thd/thd.htm

[Rax]

Wow. Such great feedback - thanks all! Let me itemize a little.

A Tektronix TDS500/TDS700 series with the FFT option could be a good one for little money. These scopes have high-res acquisition and the FFT also has averaging so it isn't jumping allover the place. The FFT isn't very fast though especially with a lot of averaging. I have used these scopes for some audio analysis work in the past to measure the bandwidth of a system. I'm quite sure the FFT on the low end Chinese scopes are not up to this task.

Another option would be a spectrum analyser which goes down to 10Hz. IIRC some radio communication analysers have audio analysis capabilities as well but I don't know if they work up to 100kHz. You'll need to adapt your signal to 50 Ohm inputs though.

Thanks a lot for the Tek TDS recommendation. I am closely examining the series. I wonder how this would stack up against a dedicated HP DSA/SA, such as the 3582/3561/3562 types. In other words, if getting a dedicated such instrument is really needed, or assigning two majors tasks (scope & DSA/DA) to one instrument is enough for most intents and purposes.

On the (RF) spectrum analyzer option - it just seems too much of a complication and I am uncomfortable with the different standards that would interface. A little like I would replace the compromise of using a sound card with another compromise.

[Rax]

I have a Rigol DS1054Z and it is bloody hopeless for audio FFT measurements (for measuring distortion anyway). 

If you're more interested in measuring THD and such, how about taking a sort of half measure and getting an audio analyser? You can then perform FFT on the analyser's post notch monitor outs with a sound card, with no risk of blowing up your PC.

Not a bad idea, though I'm not sure which ones exactly allow for such pass-through use. If sticking with my current guns, I could do this with my HP334A. I'm not sure if the piece of gear I'm considering right now - HP 8903B - allows for this. In any case, I may be contemplating getting an audio analyzer (such as HP 8903B) and a DSA/SA....

IIRC, the Shibasoku 725 with a few simple mods can dance with the Audio Precision flagship at the -120dB THD+N mark for 1kHz@20kHz BW. They are still very much state of the art, and if you look hard enough you can easily get one for <$1000 delivered. And for when you don't want to perform FFT, you still have a nice instrument that can measure levels on individual harmonics up to the 5th separately.

Not quite what you were looking for, but it's another wallet destroying option perhaps. 

I was completely unaware of the brand, even - sorry. I guess they're not that easy to come by in the US (or I should change my hangouts). I will take a closer look.

[Rax]

With some patience and judicious sniping you should be able to snag a functional 3582A, 3561A or 3562A from the 'bay for well under $1k; likely even cheaper if you're willing and able to repair it.

-Pat

After playing with a HP331A for a while, I jumped on the 3562a bandwagon.  As others have mentioned, it doesn't have the impressive noise floor of an AP but I find it to be a pleasure to use.  I got mine for $500 on eBay.  They usually run around $1k but occasionally you find them lower.

The 3562a sound like a very interesting piece. My best pick this far for an DSA/SA. Ideally, I think I'd like one of these and an 8903a on my shelf. $500 for a fully working one is a fantastic deal.

[kc8apf]

With some patience and judicious sniping you should be able to snag a functional 3582A, 3561A or 3562A from the 'bay for well under $1k; likely even cheaper if you're willing and able to repair it.

-Pat

After playing with a HP331A for a while, I jumped on the 3562a bandwagon.  As others have mentioned, it doesn't have the impressive noise floor of an AP but I find it to be a pleasure to use.  I got mine for $500 on eBay.  They usually run around $1k but occasionally you find them lower.

The 3562a sound like a very interesting piece. My best pick this far for an DSA/SA. Ideally, I think I'd like one of these and an 8903a on my shelf. $500 for a fully working one is a fantastic deal.

What does the 8903 do that the 3562 does not?

[Rax]

If you're on a budget, there's a lot you can do with the demo version of ARTA - although you've got to deal with the fact that your simultaneously testing your audio interface in addition to the DUT - which can be frustrating. I ended up getting a HP 3562a Dynamic Signal Analyzer - which does a fine job, although the limitations imposed by the 13-bit ADC mean that while the dynamic range is acceptable, you can't necessarily see down to microscopic levels of distortion - but it's certainly more trustworthy than using a PC audio interface that was never intended for test & measurement duty. Another product that has piqued my interest is the QuantAsylum QA-400 https://www.quantasylum.com/content/Products/QA400.aspx which might be worth a look.

Thanks a lot for the details on the limitations of the 3562A (may you be talking about the 3563A, though?...). I'd also be interested to hear if anybody knows the QA-400 hands-on. I've seen it before and pondered on it for a while.

[Rax]

With some patience and judicious sniping you should be able to snag a functional 3582A, 3561A or 3562A from the 'bay for well under $1k; likely even cheaper if you're willing and able to repair it.

-Pat

After playing with a HP331A for a while, I jumped on the 3562a bandwagon.  As others have mentioned, it doesn't have the impressive noise floor of an AP but I find it to be a pleasure to use.  I got mine for $500 on eBay.  They usually run around $1k but occasionally you find them lower.

The 3562a sound like a very interesting piece. My best pick this far for an DSA/SA. Ideally, I think I'd like one of these and an 8903a on my shelf. $500 for a fully working one is a fantastic deal.

What does the 8903 do that the 3562 does not?

They are different animals, to my mind (never had either, so I go by spec sheets etc.). I'd use the 3562A to visualize the content of the audio spectrum (invaluable, surely), while with the 8903A you can obtain similar information, but all numerical: distortion, S/N, RMS AC measurements, even DCV if one needs it, etc. They may look at the same content, but their output I see useful in fundamentally different ways. Also, it's true that the 3562A has a signal source, but if I go for another DSA/SA, it may not have it. Besides, I've learnt to appreciate having multiple instruments with overlapping duties, which is never truly superfluous. But seriously, never mention this to my wife.

I'm looking forward to feedback on this though; I am still figuring out my needs, including (instrumentally) through this conversation.

[kc8apf]

With some patience and judicious sniping you should be able to snag a functional 3582A, 3561A or 3562A from the 'bay for well under $1k; likely even cheaper if you're willing and able to repair it.

-Pat

After playing with a HP331A for a while, I jumped on the 3562a bandwagon.  As others have mentioned, it doesn't have the impressive noise floor of an AP but I find it to be a pleasure to use.  I got mine for $500 on eBay.  They usually run around $1k but occasionally you find them lower.

The 3562a sound like a very interesting piece. My best pick this far for an DSA/SA. Ideally, I think I'd like one of these and an 8903a on my shelf. $500 for a fully working one is a fantastic deal.

What does the 8903 do that the 3562 does not?

They are different animals, to my mind (never had either, so I go by spec sheets etc.). I'd use the 3562A to visualize the content of the audio spectrum (invaluable, surely), while with the 8903A you can obtain similar information, but all numerical: distortion, S/N, RMS AC measurements, even DCV if one needs it, etc. They may look at the same content, but their output I see useful in fundamentally different ways. Also, it's true that the 3562A has a signal source, but if I go for another DSA/SA, it may not have it. Besides, I've learnt to appreciate having multiple instruments with overlapping duties, which is never superfluous. But seriously, never mention this to my wife.

I'm looking forward to feedback on this though; I am still figuring out my needs, including (instrumentally) through this conversation.

I highly suggest reading the 3562 manual and Fundamentals of signal analysis