Looking for an audio analyzer

[DaJMasta]

I know there are a bunch of audio types around here and while I think I know what I want and what sort of specs to be looking for, I don't have any immediate experience using audio analyzers so I'd like some input to help me figure things out.

I'm a musician with a bit of a fascination with audio gear, and it's been a fairly longstanding goal of mine to do some measurements to characterize the sound of different instruments, techniques, spaces, etc..  I'm also interested in being able to characterize the performance of my gear - frequency response, THD+N, etc - both as a reference for further experiments and so I can see the effect of different configurations or modifications.

To those ends, I want an analyzer that has:
A high resolution FFT analysis mode that can update at least several times a second (minimum) and which all data can be saved
A built in audio signal generator for characterizing equipment or spaces
Software to compute distortion, snr, mark fft peaks, visualize the input in real time (or near to it), etc.
Two input and two output channels
Some degree of portability - I don't care if it's heavy and takes a bit to set up, but it can't be half a dozen bits of gear that require a dolly to move around
Digital audio signal measurement does not concern me, I'm working entirely with analog signals

I don't get the impression that a DSA will give me the signal generation and software options I want.  I don't think the HP8903 is an option either because best I can tell, it doesn't offer FFT capability.  While something like a very nice sound card and some software could potentially do the job, I don't know what software that would be and I do want it to be just one piece of software - I want the input and output to be run through the same stuff so it can handle the math in the output as I don't want to be programming my own stuff for the standard measurements.  I also know a lot of test equipment related software can be quite pricey.  I've seen the QA401 suggested a few times and it looks promising, but I don't know if the software is on the level of other brands - ease of use is important and built in features are important, but I also know nothing about the update or capture rate for the visualizations or data.

If the QA401 is on the inexpensive side, of what I've seen, the Rohde & Schwarz UPL and the Audio Precision ATS-2 are in the same price ballpark (at least buying used) and have both excellent performance figures and the reputation of their respective brands behind the quality and usability of their software (all I really have to go on without personal experience).  On the very high end of prices I'd consider, there's the R&S UPV and the Stanford Research SR1.

Having the complete system integrated without a computer has its benefits in terms of usability and portability, but I get the impression that a computer interface will make it easier to store data and will increase visualization performance (for example, the UPL uses a pre-USB era computer, so while it's performance is adequate, it's probably not fast and it's probably a bit of a pain to get the data out and on to a modern computer).


So does anyone have experience with the software on any of these devices?  Know of something else I should consider?  I'm not in a rush to buy, I'm comfortable taking at least a few months to decide and watch for deals, but I'd appreciate some input.

[Assafl]

Have you looked at Audiomatica? http://www.audiomatica.com
or Linearx? http://www.audiomatica.com

The back of the Loudspeaker Design Cookbook has lots of ads for stuff like this...

[Loboscope]

As I did some time ago in an other thread concerning a similar question I will recommend again "hpw-works" [http://hpw-works.com/].
This is a very powerful software for exactly all the measurements you want to do. It also generates the clean test-signals you must have to achieve reliable results and it is able to work with two channels (and also with two independent interfaces). It is a really professional solution and so it is not really cheap but worth its price.
You need to have a good audio-interface to gain the maximum profit from hpw-works like RME or similar high-quality interfaces. But if you will make recordings and other audio-stuff at a high level, you will have such an interface anyway.

[amirm]

For sure no DSA or Scope provides sufficient dynamic range for audio analysis.

Since you are into music production, if you have a capture card/ADC, you can start with that and free software from Rightmark: http://audio.rightmark.org/index_new.shtml.  PC is quite powerful so you can show you real-time FFT depending on length and sample rate.

If you have a few thousand dollars, then dSound from Prism Sound is a good option.  They are small and pretty portable to carry and the software interface is very good.

Step up from that is Audio Precision and there we are talking $15K to $25K with better analog front-end and specs.

Rhode and Schwartz is also up there with an embedded display and high prices to go with it.

[eeFearless]

What would people think about something like the PicoScope 4262.  I've never used it.  But, it might be appealing as a more general piece of test equipment for audio circuit use (instead of just directly plugging in mics)?  16bit scope.  AWG.  I've heard the PicoScope software permits very long FFT record lengths for finer freq resolution.  Thoughts?

[DaJMasta]

I think even "high res" scopes are out simply because that high resolution or high resolution mode is still actually somewhat low for audio sampling.  Sure they can take many more measurements in the same time frame, but it seems that the enhanced resolution of the ADCs in audio circuits gives them better dynamic range and lower noise floors by a good margin.


Anyways, I've been taking a look at the suggested stuff and there seems to be some real promise in there, the CLIO seems to be about right in terms of specs and while there isn't a ton of info on the software that I've found, it does seem like it's built to do the kind of things I'm looking it, it's just that the interface is firewire    Certainly not insurmountable, but of all the interfaces to choose... I know it's better for low latency audio than USB 2, but even thunderbolt would be preferable to me 

The LinearX LX500 seems capable and the software suite seems very comprehensive, I wouldn't be going for it but the 3d analysis stuff is pretty impressive too.  Price is a bit higher, but it's worth further looking into I think.

The Prisim Sound dScope seems like it's also hardware that will do the job admirably, and the vote of confidence in the software is a good sign, but the price could be high... no immediate indicators on the site :p

For clarification's sake, yes I know a new UPV is probably still significantly more than that, but for these pricier bits of gear I'm looking for second hand stuff and the going rate seems to be a stretch, but potentially manageable.


I do have a question about the software-only options though, and that would be what sort of interface would actually qualify as one good enough to do the sort of characterization I'm looking for?  I do have a decent ADC built into my mixer (that's at least comparable with some of the dedicated interfaces I've used in the past), a pretty good USB DAC for my normal listening, and I always have the option to get something better for better measurements, but while I don't doubt that studio quality gear will be able to match the fidelity requirements, are they really suitable for the measurements?  I think my main areas of concern would be absolute voltage references (if you put a 1V sinewave out of your generator, run it through a follower amp and into an input, is studio gear going to be able to guarantee that it spat out a 1V sine and is reading a 1V sine or is it just going to be able to give you accurate measurements relative to wherever the reference point is at?) and clock jitter.  I haven't taken apart high end interfaces before, so maybe I'm crazy, but I would think jitter plays a pretty decent part in getting low-noise FFT measurements without lots of averaging, and I don't know of any studio gear that would use, for example, an ovenized oscillator, or accept a reference frequency input.  Basically, I'm concerned that gear just designed to sound really good won't actually be precise enough to be measuring and characterizing gear that also sounds good, but I don't know enough about what interfaces are available to be able to say for sure.

If there are audio interfaces that aren't specialized analyzer gear, then I think some of the recommended software is definitely up to the task, but I don't know if they exist.... and if they do and they're just real expensive, I sort of would prefer a unit designed for measurement specifically though that's really only based on a 'gut feeling'.

For others' info, I also heard from a member with a UPL and they said that usability and update rate were no problem even on the older hardware, which is great to hear.



Anyways, I'll keep the research going, thanks for the input so far!

[amirm]

Your studio gear is fine.  With FFT and fast PC you can use long number of samples which gives you very nice "process gain."  In simple English, you can hugely lower the noise level in FFTs using even average ADCs.

The above allows you then to see jitter as sidebands of your signal.

As for voltage measurement, you can do that too by playing files with some voltage and seeing what your capture shows.  Then compensate for the level difference.

You can see a sample of software analysis in blogger Archmiago's write ups.  Here is an example: http://archimago.blogspot.com/2016/10/measurements-raspberry-pi-3-hifiberry.html

You can certainly start at this level with no money out of pocket to learn the topic itself. 

That said, I am not a fan of software solutions as it is not a system where results can be easily duplicated and compared. 

[Vgkid]

What about the QuantAsylum qa401.

[Assafl]

Clio has a USB "pocket" version (pocket - as in very big pockets). Not sure it does everything you need. My guess is with these types of systems is to try a few. Depends on their slant (loudspeaker, acoustic analysis, vibration analysis...) - sure - some are so task specific they are clearly out (like red rock acoustics SpeaD and reverse SpeaD) - but others may or may not satisfy your requirements.

Liberty Instruments (Praxis, Liberty audiosuite, etc) now has a $299 package with Parts Express: https://www.parts-express.com/dayton-audio-omnimic-v2-computer-based-precision-room-measurement-system--390-792 From the picture it looks like the old Behringer measurement microphone (remember that? used to be supplied with Velodyne subwoofers).

Even for iPhones there are pretty capable packages (like AudioTools) - that with a calibrated microphone can be pretty cool - not perfect but can replace most uses for "inexpensive" SPL meters (but probably not for work environment certification / noise pollution). Pretty expensive, but they used to have 70% off discounts once in a while.

ETF was a favorite 15 years ago (seems like they have a new product but the site states 2013 so don't know if they went the way of the Dodo): http://etfacoustic.com/

[_Wim_]

For acoustical jobs, I would always choose for a software only option, even if cost was not an issue. A half decent sound card will have at least 10x less distortions than any acoustic environment or speaker you will ever measure. But the advantage is you can use more software tools (no dedicated hardware) and much more types of testsignals and much more ways to visualize and report the measurements (very important to make sense of the measurements)
TIP: always use constant latency ASIO drivers for your sound card to avoid timing issues
For measuring audio equipment (electronics), it is different (because you need much better specs, you want to avoid ground loops, you’re working with low and high voltage signals, you want very low noise...) so here a decent hardware solution (AP, R&S, Prismsound...) can make things a lot easier, but it is not absolutely necessary.
In my opinion, recent audio equipment (electronics) are nowadays more than good enough not to invest your time in that (avoid so called “high end” gear, as these sometimes deliberately distort to sound “different” than the competition).
Speakers and acoustics however still have lots and lots of issues (due to the laws of physics), so most benefit is to be had here. So if I were you, I would buy a calibrated mic, a sound card with recent and stable ASIO drivers, and start learning to work with software tools like ARTA, Holmimpulse, REW, Abec3 AND read lots and lots of books on acoustics and speaker design…

[jackenhack]

I'm currently finishing up a headphone amplifier build and I got the QuantAsylum QA401 when I started developing the amp. With todays extremely low noise and distortion figures on op amps, You'll hit the bottom of what the analyser can measure very quickly. I had to build a twin-T notch filter and get hold of a extremely low THD signal generator to be able to measure below -108dB THD. Getting a industry standard Audio Precision is the dream, but they are way to expensive for me...

[_Wim_]

I'm currently finishing up a headphone amplifier build and I got the QuantAsylum QA401 when I started developing the amp. With todays extremely low noise and distortion figures on op amps, You'll hit the bottom of what the analyser can measure very quickly. I had to build a twin-T notch filter and get hold of a extremely low THD signal generator to be able to measure below -108dB THD. Getting a industry standard Audio Precision is the dream, but they are way to expensive for me...

But the question is, why would we need to see lower than that? It can be fun as a technical exercise, but it will not improve the sound quality we hear any more…

[DaJMasta]

I mentioned the QA401 in the original post and I just don't have any experience with the software, which would probably be the big component of it.  A 32 bit DAC and ADC sound fancy, but it doesn't look like the specs are any better than the 24 bit competitors, and I think it just comes down to physical restrictions of the design and parasitics at that point.  Since that is very software dependent too, I'm not sure how much performance would actually differ from just using my existing equipment.

The CLIO pocket seems like a similar little device, but the big drawback is that it's only a single channel.  Sure you could get two, but it seems like there's several options for this sort of USB box with some converters so it may be easier just to pick another if I go that route.  I think the measurement mic pack is probably not worth it for me - just the mic itself isn't that expensive and I've got interfaces that are probably better than a built in USB thing, so then it comes down to the software and it does seem like there are affordable options.  I've also got an older dedicated SPL meter, a CEL-231 with a calibrator box.  Sure, the cal seal on the inside is from 1992, but the spec is down to +-1dB , which isn't really common with most inexpensive meters


I think in the short term, getting rightmark and experimenting with my current interfaces is basically a no-brainer, but I agree with the liking a hardware solution sentiment.  It also seems to afford you some more input protection, better filtering, more EMI rejection, and a cleaner power supply without having to put everything on an external power conditioner.  Maybe a good interface can give you performance good enough that you can't really hear a difference, higher performance will still get you more fidelity in the visualization and the same amount of accuracy with a shorter sample time, and since I do want close to real time FFT visualization, the extra dynamic range will help keep the visualization from getting too noisy when turning up the responsiveness.  Another concern is that I'm trying to be as scientific as possible with my main use - the measurement of instrument sounds - mostly because I don't know what results I'll find.  Having better equipment will let you see more detail in the minor differences and reduce the chances of running into hardware artifacts or limitations in the measurements that are difficult to identify.  I'm looking for an instrument that can beat the specs on my current hardware because my current hardware seems to edge out my own ear across the board (I've been honing my listening skills for a while now as part of my trade), and I don't want to run into a situation where my measurements can't match that level or my ears can match it with another couple decades of experience.  I want measurements that are good enough that no one can hear beyond them, because some of the people who may eventually be interested in my data may literally be the most discerning listeners in the world.


Couple new questions from reading through datasheets and such:
Are there audio analyzers that provide phantom power?  Are commercial phantom injectors also noise sources?  I assume it's a concern with ultra low noise gear, but is that a common issue?
Do you think it's valuable for the generator side of the instrument to produce more than sine waves?  It seems like some of the lowest distortion generators are pure sine only, but I've already been using some AM and added harmonics in signals on my existing function generator... is that something that's likely to be useful?  I assume all the normal audio rating specs are done with sine only, but it's definitely not a "realistic" test signal.

[Loboscope]

Although human ears and human hearing are fantastic and amazing in respect of its construction and function, there will be no doubt that any reasonably decent technical solution will be far better in terms of precision, reliability and reproducibility of its results.
The frequency response of any decent high-quality (omni-)microphone will be easily outplay any adult ear and there are for example some new Sennheiser Studio-Microphones which will reach up to 50 KHz. Also the dynamic range of a really good microphone/amplifier will exceed the total amount of our ears because there are microphones which will deliver a low distortion output up to 140 dB and a decent amplifying circuit will also let pass the signal unspoiled.
And the great amount of the dynamic range of our ears from 0 to nearly 120 dB (normally the maximum pain threshold) will only be achieved by an time-consuming adaptive process. If your ear has been exposed some time to a strepitous environement you will afterwards not be able for some time to hear really quite sound for some time. This is due to the internal protective mechanism of our ear.
And all the functionality will only be completely there, if our ears are sane and undamaged.

But the most erratic part will be our brain itself which must interpret all the signals coming from our ears. And the brain will interpret continuous - because this is the task of our brain. And there will be expectations, mistakes, misconceptions and so on.
Mostly the expectations will colour our perceptions.

So I will underline that any (half-)decent microphone, amplifier, interface and software will give you the all the reliable results you want to have measuring audio sources like musical instruments, voices and so on. And today it is not necessary to pay a fortune for having professional results.
Even a simple and dirt cheap app like "Speedy spectrum analyser" will show astonishing results.

[amirm]

Couple new questions from reading through datasheets and such:
Are there audio analyzers that provide phantom power?  Are commercial phantom injectors also noise sources?  I assume it's a concern with ultra low noise gear, but is that a common issue?

XLR inputs are for line level capture and not microphone.  So they neither provide phantom power no mic amplification.

Do you think it's valuable for the generator side of the instrument to produce more than sine waves?  It seems like some of the lowest distortion generators are pure sine only, but I've already been using some AM and added harmonics in signals on my existing function generator... is that something that's likely to be useful?  I assume all the normal audio rating specs are done with sine only, but it's definitely not a "realistic" test signal.

The common way to do this testing is to use a synthetically created test file and using that through a DAC for the source.  These are mathematically produced and with dither can go down to any depth you want.  No need to use any analog generator.  And for testing things like DACs, you just feed them the digital stream eliminating.

You can even create the files using your DAW.  Just make sure you add dither (TPDF) so that you don't see the distortions caused there in your analysis.  YOu can create single frequency and sweeps as needed.

[DaJMasta]

Ah ok, it seems like some of the built in generators are specified as sine only, but not by any means all of them.  Having the capability to do effectly "true arb" patterns by just playing an audio file seems like a good way to do it.

If it's line level inputs, would you also want an external preamp for connecting a mic directly, or is the gain high enough and noise floor low enough on the inputs that they can interface directly with the mic (given phantom and AC coupled as needed)?

[FrankBuss]

I used an Audio Precision SYS 2722 for years (not mine unfortunately, at a client). Don't know what you want to pay, you can get it used on eBay for $6000. It is worth the price.

The software is easy to use. See the manual here for all the details. Can do FFT, too. It is heavy, but portable in the way you described it, only one big device, plus the required PC or laptop. You can enhance the software with own Basic scripts. Once I implemented a complete calibrate and test script for production for another high priced audio gear. It really has all you ever need, like balanced and unbalanced analog inputs / outputs (with phantom power) and digital in/outs (AES/EBU, up to 192 kHz sampling rate). In the software you can configure the outputs with signal generators and configure which inputs you want to measure and you can save setup configurations. Can output all kind of useful test signals, like sin, noise, arbitrary waveform (IIRC the length was a bit limited) or many different kind of pulses and bursts (useful for loudness measurement tests, or verifying peak meters).

And very useful if you develop hardware or want to do in-depth tests: the analog features of the digital inputs, like jitter measurement, see e.g. this video:



I don't like their software of the newer devices, like for the APx500. The GUI looks more polished, but in my opinion it is less flexible and less useful, because you can't organize all the measurements and generators in one easy to access configuration and I don't know if it is scriptable, too.

[amirm]

If it's line level inputs, would you also want an external preamp for connecting a mic directly, or is the gain high enough and noise floor low enough on the inputs that they can interface directly with the mic (given phantom and AC coupled as needed)?

I am not a recording guy but I am pretty sure the mic will need a pre-amp to get up to the line level inputs.  You can feed it lower level signal but then you lose a ton of your dynamic range.

What will you be doing with the mic anyway?  If it is for room measurements then that is an entirely different topic.  There, I 100% recommend computer software and any microphone you can hook up to your computer.  The weapon of choice is free software called Room EQ Wizard.  It is a very sophisticated piece of software, outperforming many commercial ones.  The down side is that despite what it calls itself, it is hard to learn to use.  At least initially.  To that end, I have written two tutorials to get people started:

Part 1: http://www.audiosciencereview.com/forum/index.php?threads/room-measurement-tutorial-for-dummies-part-1.4/

Part 2: http://www.audiosciencereview.com/forum/index.php?threads/room-measurement-tutorial-for-dummies-part-2.5/

I plan to finish the series by showing what to measure and what to do to improve room acoustics.  It will include much of the research into psychoacoustics as opposed to simple cookbooks out there.

Anyway, tell us more and we can give more relevant advice. 

[ruairi]

I'm racing out the door to the studio here but I would advise the O.P. to slow way down and walk before you can run.  You are talking about some very high end test equipment but making very basic mistakes in your understanding of audio, electronics and measurements.

I totally understand the idea of doing things right and getting the best but understanding the fundamentals and how things are done will help you make better choices when you buy higher end gear later.

I use REW (free) and a Prism Lyra 1 interface for acoustics work and on the road electronic testing, on the bench I use a Prism dScope.  You can go a long way with something like REW or audio tester and an interface for a couple of hundred dollars.  Start learning the fundamentals - for example why we test with tones and not music, then buy gear as you know more about what you really need.

Can you tell us more about your application?

[DaJMasta]

Maybe the issue is that I have two applications:  There's the testing application that involves the generator and the low noise floor and distortion, and the more primary real time FFT visualization with recording (of the FFT frames, I can get the audio however if needed).

The idea was to try to get them both on the same instrument, since most of the testers which had enough visualization power to generate a Bode plot (not a Keithley 2015 or an HP 8903) also had a good FFT mode and the frontend to keep it nice and clean.  I'm not trying to use music to test the performance of gear, but I've used AM and tone + harmonics modes on my current function generator for some audio stuff before and I'd like to have the capability (my current gen is only .075% THD in the audio band, so it wouldn't be suitable for use with these instruments).

So I'm looking for, specifically:
Good FFT mode (reasonably quick updates, lots of bins available for detailed analysis, wide dynamic range, real time visualization, recording of each frame)
Ability to test studio grade gear for distortion, noise, and frequency response

The rest is much less important, but if it's fascinating and it doesn't cost a ton more, I'd certainly like it.  I don't plan on characterizing the acoustics of any halls, but it would be great if I could pick up an omnidirectional measurement mic and dabble.  Measuring timing differences or harmonic differences between input signals would also be really cool to play with, but it would be just for that, I don't have a use for those features I know of.  Something like setting up a mic and a speaker in a room and evaluating how performance changes with different furniture or mic/speaker placement is definitely an extra but would be fascinating.

[_Wim_]

I'm racing out the door to the studio here but I would advise the O.P. to slow way down and walk before you can run.  You are talking about some very high end test equipment but making very basic mistakes in your understanding of audio, electronics and measurements.

I totally understand the idea of doing things right and getting the best but understanding the fundamentals and how things are done will help you make better choices when you buy higher end gear later.

I have to agree with this. Good audio measurements are much more about understanding WHAT and HOW you are measuring, then about having very good measurement gear. I would recommend to read a few good books on the subject. I can recommend Floyd Toole "Sound Reproduction", Joseph d'appolito "Testing loudspeakers" (a bit dated, but explains the basics really wel)and Bob Cordell "Audio Amplifiers" (contains a good section about measuring amplifier and electronics).

To understand "how" we hear, I can recommend JanSchnupp "auditory neuroscience making sense of sound". This book will give you a better idea what parameter a important, and which need to be only good enough.

[montemcguire]

I'm currently finishing up a headphone amplifier build and I got the QuantAsylum QA401 when I started developing the amp. With todays extremely low noise and distortion figures on op amps, You'll hit the bottom of what the analyser can measure very quickly. I had to build a twin-T notch filter and get hold of a extremely low THD signal generator to be able to measure below -108dB THD. Getting a industry standard Audio Precision is the dream, but they are way to expensive for me...

But the question is, why would we need to see lower than that? It can be fun as a technical exercise, but it will not improve the sound quality we hear any more…

The numbers one gets from an analyzer are completely artificial - nobody listens to sine waves. So, stating that such and such distortion levels are audible or not is hard to defend 'a priori'. It's reasonable to expect that, given a complex musical signal, the distortion products are greater than with a simple sine wave.

Despite this uncertainty, I have found that these simple sine wave tests can be used to expose the behavior of a circuit very reliably, and that knowledge can help a designer to avoid specific distortion mechanisms, or optimize the circuit to minimize these distortions. Something as simple as changing the frequency compensation of an amplifier might have a small effect or no effect at all, and relying on a listening test to say whether the change improved performance, worsened it, or had no effect might be extremely difficult. This is the beauty of an analyzer: if one's goal is a clean circuit, an analyzer will provide a tireless, repeatable, and honest assessment of a circuit. With an analyzer, some circuit "tweak" that proves to have no effect can be understood as "irrelevant" right away, without the vagaries of a series of tedious listening tests.

It is still up to the operator to choose the test and understand the meaning of the results, but given a high resolution distortion measurement, it's my opinion that one can more accurately and rapidly evaluate a circuit's behavior using a high resolution distortion analyzer than is possible with listening tests.

[FrankBuss]

The numbers one gets from an analyzer are completely artificial - nobody listens to sine waves. So, stating that such and such distortion levels are audible or not is hard to defend 'a priori'. It's reasonable to expect that, given a complex musical signal, the distortion products are greater than with a simple sine wave.

You are right, because just measuring the THD (or THD+N) doesn't say anything about the interactions between different signals. This is one reason Audio Precision has the multitone test:

https://www.ap.com/technical-library/using-multitones-in-audio-test/

Additionally, any amplifier exhibits some filtering. This means that the phase shift of signals at different frequencies can be different (depending of the kind of filter, even if not intentionally). For this it is useful if you can do a bode-plot:

https://www.ap.com/technical-library/using-bode-plots-in-apx/

Disclaimer: I'm not associated with Audio Precision, but it is just the best  for audio measurement

[montemcguire]

The numbers one gets from an analyzer are completely artificial - nobody listens to sine waves. So, stating that such and such distortion levels are audible or not is hard to defend 'a priori'. It's reasonable to expect that, given a complex musical signal, the distortion products are greater than with a simple sine wave.

You are right, because just measuring the THD (or THD+N) doesn't say anything about the interactions between different signals. This is one reason Audio Precision has the multitone test:

https://www.ap.com/technical-library/using-multitones-in-audio-test/

Additionally, any amplifier exhibits some filtering. This means that the phase shift of signals at different frequencies can be different (depending of the kind of filter, even if not intentionally). For this it is useful if you can do a bode-plot:

https://www.ap.com/technical-library/using-bode-plots-in-apx/

Disclaimer: I'm not associated with Audio Precision, but it is just the best  for audio measurement

I'm a big fan of AP gear as well :-)

Still, while multitone and IM tests seem useful, I have found that, for my work, the APx-555 high precision analog generator has a lower residual for simple 1-2kHz sine wave tests, and thus seems to more honestly 'sort' and differentiate different devices under test, especially when the devices are so clean that they approach the AP residual.

For gear that is, for example, worse than 1 ppm distortion, the multitone and other DAC based stimulus signals can be OK, and can do some useful, quick characterization. But, if you're like me and you're scraping the residual with 32-64 FFT averages, the IM and DAC based signals won't match the low residual of the high resolution generator / analyzer, and they fail to properly 'sort' DUTs according to their real errors.

It'd be neat to have an IM test set that could match this performance, but keep in mind that a (basically linear, time invariant) device that fails an IM test also has to fail a simple sine wave test. One thing that an IM test can do is torture the device with high frequencies, but that can also be done with a sine wave test, measured at several frequencies. This can show an indication of the distortion rise with frequency, which can point to the mechanism for the circuit's behavior - e.g. 6dB/octave rise or 12dB/octave can point to different causes etc.

Well, that's my world, and the APx-555 helps me out. But still, I could use another 20dB of distortion floor, just to be able to measure silly things accurately. I suspect that this will be possible in a decade or so, and I'll queue up for that box when I can!

[FrankBuss]

Well, that's my world, and the APx-555 helps me out. But still, I could use another 20dB of distortion floor, just to be able to measure silly things accurately. I suspect that this will be possible in a decade or so, and I'll queue up for that box when I can!

Well, personally I think anything better than -120 dB THD+N (like some DACs and ADCs can do, meaning 20 bits effective resolution) doesn't make a difference and it gets really expensive. Be careful with additional 20 dB, because if you move while measure something, the induced current caused by your body-electricity, or if a truck is moving outside and then the piezoelectric effect of the vibrating capacitors might cause measureable distortions