Spectrum analyser or a scope with good FFT?

[JSE009]

I can surely need some help/advise 

I am working a lot on audio stuff and especially pre- and power amplifiers (tube and transistor).
And for measuring distortion etc. I use two Keithley 2015 and they do the job perfectly 
But what I miss is a closer look on the spectrum of a (test)signal, yes I can use my Rigol scope (DS2072A) but as Dave showed nicely in his review episode #845,
the Rigol is not really suitable for this part of the job.
I also use a pico scope (2205A) but the noicefloor of this unit is to hight to measure audio signals so I use this unitI only when I am in the field so to say and want to perform
some quick measurements.
I can buy a standard spectrum analyser but they almost all started at 9kHz and I need less than that.

I did like the Oscilloscope FFT Comparison in Dave’s review  and now I am wondering, is it smart to look for a scope with a “good” FFT function or do I have to look for
a spectrum analyser that start at a low frequency? Sometimes they can be found on eBay.
I do not want to spent more then $2.000,=  for a unit.

Thank in advance for feedback.

[Wuerstchenhund]

I am working a lot on audio stuff and especially pre- and power amplifiers (tube and transistor).
And for measuring distortion etc. I use two Keithley 2015 and they do the job perfectly 
But what I miss is a closer look on the spectrum of a (test)signal, yes I can use my Rigol scope (DS2072A) but as Dave showed nicely in his review episode #845,
the Rigol is not really suitable for this part of the job.
I also use a pico scope (2205A) but the noicefloor of this unit is to hight to measure audio signals so I use this unitI only when I am in the field so to say and want to perform
some quick measurements.
I can buy a standard spectrum analyser but they almost all started at 9kHz and I need less than that.

I did like the Oscilloscope FFT Comparison in Dave’s review  and now I am wondering, is it smart to look for a scope with a “good” FFT function or do I have to look for a spectrum analyser that start at a low frequency? Sometimes they can be found on eBay.
I do not want to spent more then $2.000,=  for a unit.

As you already found out, spectrum analyzers start at comparably high frequencies (usually 9kHz, some even higher i.e. 10MHz) and are not much use for audio. There are a few audio SAs but they are somewhat rare and pretty pricey (even for some pretty old ones with mono CRT).

A scope with good FFT is in my opinion a better alternative. However, FFT on pretty much all Rigol scopes (up to the ridiculously priced DS6000) is nothing more than pathetic as it only uses a few thousand (4k, in some cases even just 2k) points for FFT. That means Rigol is is out.

You also want a scope with a somewhat decent processing subsystem so that FFT calculations are done reasonably fast, and that pretty much rules out most B-brand scopes.

With a $2k budget, you could get a new Keysight DSOX2000A Series scope which offers decent FFT capabilities. However, I'd consider the 2nd hand market, which has several scope alternatives that give you fast FFT, plus you get a much more capable scope on top (so you could sell your DS2000A and recoupe some of the money).

Having said that, the cheapest alternative is a good(!) PC sound card and a software SA for audio.

[nowlan]

Would the quantasylum gear be better than your 2015?

https://www.quantasylum.com/content/Products/QA400.aspx

[ADT123]

The more points you can have in the FFT the lower the noise floor - going from 1000 points to 1,000,000 spreads the noise floor over 1000 times as many bins so will drop the noise floor by 30dB.

With your PicoScope 2205 make sure you have the max no of points enabled in the spectrum (probably 8000 as it has 16K memory) then enable averaging to drop the noise floor more.  If thats not enough you need more resolution and more memory - I did a test with 1 million points and 16 bits resolution on a PicoScope 5444B and got the noise floor 120dB down.  See reply 25 and 26 here https://www.eevblog.com/forum/blog/eevblog-845-oscilloscope-fft-comparison/25/

Noise floor is of course not the only issue - a high res scope with a low distortion front end is also important for audio.

[markone]

With a $2k budget, you could get a new Keysight DSOX2000A Series scope which offers decent FFT capabilities. However, I'd consider the 2nd hand market, which has several scope alternatives that give you fast FFT, plus you get a much more capable scope on top (so you could sell your DS2000A and recoupe some of the money).

8 bit DSO for High End audio analysis ?

[_Wim_]

Scopes never have a good enough AD to make usefull audio measurements (well, maybe for valve amplifiers they can be of some use due to their typical high distorsion figures).You have scopes with 16 bit AD (accomplisched though oversamping), but even those get only a THD+N of around -80db (-96db is the theoretical max for 16 bit). The quantasylum is much better then this (THD+N > -98 db, but also most good pc audio cards with free software (for example ARTA, Holmimpulse, Rightmark Audio Analyzer, Room EQ Wizard...) can reach level up to -100dB

Personnaly I use ARTA with an Echo Audiofire 2 sound card (balanced in and out) and this has been very usefull to me. I also have a rigol ds1054z for general electronisc work.

[markone]

Would the quantasylum gear be better than your 2015?

https://www.quantasylum.com/content/Products/QA400.aspx

I was going to suggest the E-MU 1212M PCIe :

http://www.creative.com/emu/products/product.aspx?pid=19169

but the QA400, that honestly i did not know, even if with lower specs, @200 USD seems a good deal due to the fact that has all the metrics ready out of the box.

Have you direct experience with that ?

[AF6LJ]

I am thinking that maybe going old school would give better results...
Get a selective voltmeter. They are made for this kind of stuff.

[_Wim_]

Would the quantasylum gear be better than your 2015?

https://www.quantasylum.com/content/Products/QA400.aspx

I was going to suggest the E-MU 1212M PCIe :

http://www.creative.com/emu/products/product.aspx?pid=19169

but the QA400, that honestly i did not know, even if with lower specs, @200 USD seems a good deal due to the fact that has all the metrics ready out of the box.

Have you direct experience with that ?

I have the E-MU 1212M PCIe also, but the drivers are really unstable and cause BSOD very often. It is a very good card (see spectrum plot), but the instability makes it very annoying to use...

[Wuerstchenhund]

With a $2k budget, you could get a new Keysight DSOX2000A Series scope which offers decent FFT capabilities. However, I'd consider the 2nd hand market, which has several scope alternatives that give you fast FFT, plus you get a much more capable scope on top (so you could sell your DS2000A and recoupe some of the money).

8 bit DSO for High End audio analysis ?

Most better scopes offer high resolution modes (i.e. 11bits on older LeCroy scopes, 12bit on older Agilents like the DSO5k, 16bits on the R&S RTM) which work pretty well, and the reduction in analog bandwidth is not a problem for audio applications.

It does require a decent processing backend but on a good scope FFT performance will still be good even with high res/ERES modes enabled.

[G0HZU]

I can buy a standard spectrum analyser but they almost all started at 9kHz and I need less than that.

If you avoid the mid range models you will find that a decent RF spectrum analyser will usually work down to 100Hz or less and it will maintain a decent noise floor down to audio frequencies but it isn't really the right tool for HiFi audio analysis. Too slow and not enough SFDR for audiophools This class of analyser also tends to be very big and heavy. I've got several decent RF spectrum analysers here and they all work very well down to 100Hz or less. 

I normally use a soundcard for stuff like this but care is needed not to overload or damage the soundcard. The same would apply to a decent RF spectrum analyser in terms of ease of damage. An RF spectrum analyser will usually have a 50R input impedance which often isn't suitable either.

I would have thought that a Picoscope would work well here although I haven't used a Picoscope for many years. The last one I used was over 12 years ago.

Like you I have a Keithley 2015 THD and this can be configured over GPIB to show the levels of harmonics with a reasonably fast update rate on a PC display. So it looks like a spectrum analyser even though it is really just measuring and displaying the individual harmonics. But it can do this very accurately in my experience. See the image below.
The Keithley is also very rugged in terms of damage level. But it isn't going to deliver the ultimate THD (or THD and noise) performance that some die hard audiophools crave.

[markone]

With a $2k budget, you could get a new Keysight DSOX2000A Series scope which offers decent FFT capabilities. However, I'd consider the 2nd hand market, which has several scope alternatives that give you fast FFT, plus you get a much more capable scope on top (so you could sell your DS2000A and recoupe some of the money).

8 bit DSO for High End audio analysis ?

Most better scopes offer high resolution modes (i.e. 11.5bits on older LeCroy scopes, 16bits on the R&S RTM) which work pretty well, and on a good scope FFT is still pretty fast even with the resulting larger datasets.

Everything below a real 100dB dynamic range and proper noise floor is useless for the purpose, your suggestion was for Keysight DSOX2000A, that is utterly out of specs for audio meas.

[_Wim_]

With a $2k budget, you could get a new Keysight DSOX2000A Series scope which offers decent FFT capabilities. However, I'd consider the 2nd hand market, which has several scope alternatives that give you fast FFT, plus you get a much more capable scope on top (so you could sell your DS2000A and recoupe some of the money).

8 bit DSO for High End audio analysis ?

Most better scopes offer high resolution modes (i.e. 11.5bits on older LeCroy scopes, 16bits on the R&S RTM) which work pretty well, and on a good scope FFT is still pretty fast even with the resulting larger datasets.

11.5 bits wont cut is, and the R&S RTM have a native 8-bit dac, and only reaches 16 bit through oversampling. They do not state the THD+N figure in their datasheet, but the picoscope 5000 series (https://www.picotech.com/oscilloscope/5000/picoscope-5000-specifications) do something simular but only reach a THD of -70db which is insufficient for audio work. If you really want to use a scope for something like this, you should buy a scope the a native 16 bit ad like a picoscope 4262 (https://www.picotech.com/oscilloscope/4262/picoscope-4262-specifications). But due to their use of a 16-bit AD, these sort of scopes have a very low bandwith, making them less usefull as a general purpose scope. So in my opinion (and as other have posted), you much better off with a general purpose scope, and a dedicated audio measurement system. Due to the power of modern pcs, a pc based solution clearly offers the best bang for buck

[markone]

Would the quantasylum gear be better than your 2015?

https://www.quantasylum.com/content/Products/QA400.aspx

I was going to suggest the E-MU 1212M PCIe :

http://www.creative.com/emu/products/product.aspx?pid=19169

but the QA400, that honestly i did not know, even if with lower specs, @200 USD seems a good deal due to the fact that has all the metrics ready out of the box.

Have you direct experience with that ?

I have the E-MU 1212M PCIe also, but the drivers are really unstable and cause BSOD very often. It is a very good card (see spectrum plot), but the instability makes it very annoying to use...

I have the old PCI version, i do not have stability problem but i have to say that its mixer control panel is a huge PITA at first glance, performance wise i think its the best bang for the bucks, much better than even pricer USB versions.

[_Wim_]

I can buy a standard spectrum analyser but they almost all started at 9kHz and I need less than that.

If you avoid the mid range models you will find that a decent RF spectrum analyser will usually work down to 100Hz or less and it will maintain a decent noise floor down to audio frequencies but it isn't really the right tool for HiFi audio analysis. Too slow and not enough SFDR for audiophools This class of analyser also tends to be very big and heavy. I've got several decent RF spectrum analysers here and they all work very well down to 100Hz or less. 

I normally use a soundcard for stuff like this but care is needed not to overload or damage the soundcard. The same would apply to a decent RF spectrum analyser in terms of ease of damage. An RF spectrum analyser will usually have a 50R input impedance which often isn't suitable either.

I would have thought that a Picoscope would work well here although I haven't used a Picoscope for many years. The last one I used was over 12 years ago.

Like you I have a Keithley 2015 THD and this can be configured over GPIB to show the levels of harmonics with a reasonably fast update rate on a PC display. So it looks like a spectrum analyser even though it is really just measuring and displaying the individual harmonics. But it can do this very accurately in my experience. See the image below.
The Keithley is also very rugged in terms of damage level. But it isn't going to deliver the ultimate THD (or THD and noise) performance that some die hard audiophools crave.

I can agree the Keitley is also a nice solution. I can also agree that the final device does not need to have THD levels of -100dB or more (audiophool territory), but around -80dB is necesarry for high qualitiy listening, especially so if you not playing at the optimal sound level for the lowest distorsion. Also, when devellong audio gear, it very handy that you measurement system has at least a 10db lower distionsion and noise floor, so you know for sure that what you are seeing is the DUT and not the measurent system itself.

This makes using a sound card sometimes difficult, as ground loops cause distorsion in the measument system that can sent you on the wrong track. To avoid this I have been thinking of buying a AD-card en DA-card (evalution boards) and powering these with an isolated power supply individually. Both cards would then be connected to a pc audio card using an optical toslink connection. With some additional input protection this creates a high-end audio measurement system that is very robust...

[_Wim_]

Would the quantasylum gear be better than your 2015?

https://www.quantasylum.com/content/Products/QA400.aspx

I was going to suggest the E-MU 1212M PCIe :

http://www.creative.com/emu/products/product.aspx?pid=19169

but the QA400, that honestly i did not know, even if with lower specs, @200 USD seems a good deal due to the fact that has all the metrics ready out of the box.

Have you direct experience with that ?

I have the E-MU 1212M PCIe also, but the drivers are really unstable and cause BSOD very often. It is a very good card (see spectrum plot), but the instability makes it very annoying to use...

I have the old PCI version, i do not have stability problem but i have to say that its mixer control panel is a huge PITA at first glance, performance wise i think its the best bang for the bucks, much better than even pricer USB versions.

Yes, the mixer control panel is a pain, but once you have your usual setups configured and saved however, you are in the clear. It was (it is not sold any longer) indeed the best bang for buck performance wise.

[Wuerstchenhund]

11.5 bits wont cut is, and the R&S RTM have a native 8-bit dac, and only reaches 16 bit through oversampling. They do not state the THD+N figure in their datasheet, but the picoscope 5000 series (https://www.picotech.com/oscilloscope/5000/picoscope-5000-specifications) do something simular but only reach a THD of -70db which is insufficient for audio work. If you really want to use a scope for something like this, you should buy a scope the a native 16 bit ad like a picoscope 4262 (https://www.picotech.com/oscilloscope/4262/picoscope-4262-specifications). But due to their use of a 16-bit AD, these sort of scopes have a very low bandwith, making them less usefull as a general purpose scope. So in my opinion (and as other have posted), you much better off with a general purpose scope, and a dedicated audio measurement system. Due to the power of modern pcs, a pc based solution clearly offers the best bang for buck

You're probably right that the THD+N figures of most scopes will be insufficient for high end audio work, but then OP did only broadly specify what he's looking for so a scope with high res mode (which is what a friend of mine who's dabbling with HiFi gear uses, amongst other gear like an audio analyzer) might (or not) have satisfied what he wants. I never worked much with audio stuff so I can't speak of experience.

Which is exactly why I also wrote that a PC sound card is probably a better alternative.

[deadlylover]

I'll echo the suggestions for a sound card + PC for FFT, that QuantAsylum 400 looks pretty good.

I have a spectrum analyser (Advantest TR4171) that starts from 10Hz, but the problem is the spectrum analyser's distortion is in the range of -70dBc or so, way too high for audio work, it is basically useless to have a look at individual harmonics.

You can use a notch filter to reject the fundamental, but now the problem becomes the veeeery long sweep times, expect anywhere between 3-15 minutes just for a 1kHz or 10kHz span at 3Hz or 10Hz RBW. Not very practical for most uses in audio.

Another half measure would be to get a dedicated audio analyser with a monitor output, they're pretty standard so you should find them on basically everything. You can then perform FFT on the monitor output without any risk of blowing up your PC or sound card (they can usually take up to 150V input too, good for power amps), and depending on the audio analyser you can get very decent performance.

Sometimes all I really need to look at are the first 5 harmonics, and any good audio analyser can display them individually, no need to set up math functions on a scope or fire up a PC, 1 button and you're golden.

Attached is an FFT on the monitor out of a ShibaSoku AM51B audio analyser loopback using the crappy built in sound card from a cheap laptop, that second harmonic is at -139dB from the fundamental.

ShibaSoku make great audio analysers, one to look out for is the old 725B or 725C models, with a few simple mods (upgrading 25 year old op amps  ) they will bring any entry level audio analyser today to their knees in terms of residual noise+distortion anyway. They'll measure down to -120dB THD+N @ 1kHz, 30kHz BW no problem, and for reference the new flagship Audio Precision APx555 goes to like -124ish I think, but they start at 27 grand. 

[markone]

snip, but they start at 27 grand. 

 

[rosbuitre]

Hi
I am considering purchasing this audio interface, MOTU Ultralite MK3 that comes with software to measure FFT, someone use?

http://www.motu.com/products/motuaudio/ultralite-mk3

http://www.motu.com/products/cuemix-fx/tuner-analysis.html

Regards

[_Wim_]

snip, but they start at 27 grand. 

 

That's "only" the price of a pair of "good" speaker cables: http://enjoythemusic.com/superioraudio/equipment/0113/siltech_royal_signature_series.htm

Euhm correction: one the price for the short version of 1m. The 2.5m version sets you back 62000$. Bargain if you are a real audiophool...

[Muxr]

There is also the Signal Hound, it goes down to the audio frequencies and it's a pretty capable PC based SA. For my audio work I just use a keithley 2015 for THD measurements and a scope FFT if I need to look at the frequency domain.

[joeqsmith]

Years ago when I was working with audio, the company I worked for owned an HP3582A.   At the time, there was no Labview and I remember using a Polaroid camera to get data off of it, besides collecting it by hand.    We also had some ADC board that we used with the VAX at that time.    Once the PCs started picking up steam, we had some software / hardware package that could plot waterfall curves.  It was pretty impressive as I was doing this by hand up till  then. 

I would bet the worlds worse modern PC sound hardware is better than what we were doing.   

[Muxr]

I think the problem with your standard audio PC hardware is jitter. Nwaguy who kind of set the PC audio community on fire (interesting read, many on here will enjoy his story) has an article on it. dScope is what audio industry uses nowadays but it's out of reach of hobbyists.

http://nwavguy.blogspot.com/2011/02/testing-methods.html (it was updated on 4/15)

Anyways those who haven't heard of this guy. He's some anonymous EE, who popped up out of nowhere and started challenging these common myths in the Audiophool circles we all like to make fun of. Like a $2 op-amp being superior to these nude virgin op-amps, people were after. But he backed it all up with measurements and facts. He claimed he could make a headphone amplifier that was more objectively accurate than these multi thousand dollar alternatives, and then he did just that and open sourced it. It's called O2. It's kind of the ultimate, I told you so story.

[joeqsmith]

I think the problem with your standard audio PC hardware is jitter.

Care to expand on this?