Audio Analyzers (but not a toy)

[trobbins]

I am attaching a diagram with the output from 4i4 for 20kHz sine somewhere at 1/2 FS with the specification that it is not in the loop. Everything is satisfactory, but higher in frequency we have artifacts. I am attaching two more diagrams, one up to 250kHz and the next up to 2.6MHz. The noise generated by the switching sources can be clearly observed, although it is possible that some artifacts have another origin. The noises in the last diagram do not disappear if the signal is off, but are proportional in any situation to the level of the master potentiometer. These disturbances are best seen with the oscilloscope on a rectangular signal because they have about 12-15% of the useful value. The spectral components used in the composition of the rectangle are stable over time, but the noise of the switching sources wanders continuously. I use an LC filter with fc =136Khz, N=7 at the output of Focusrite to clean all the radio frequency when I need a sine wave 20Hz-100kHz. After all, it's a 200 Euro sound card and for the hobby it's just fine.

Does '1/2 FS' mean you use a sampling frequency of 96kHz with a nominal 48kHz bandwidth?
Can you post a spectrum from say 10kHz to 2.6MHz, rather than with separate spectrum widths, to give confidence that the artifacts are shown along with the generated signal and its harmonics.
The 240kHz artifact is pretty much at the noise floor level. 
The 1.1MHz to 2MHz artifacts may be just a result of your measurement system and ground/parasitic loops, given they don't disappear when the signal is off. 
Can you elaborate on what you mean when talking about a 'rectangular signal'.

For use in audio measurements, I can't see what influence any soundcard artifacts in the low MHz range (or even at 250kHz if that is an actual artifact) would have on any measurement accuracy or outcome or

If the soundcard was in use at the same time, and in the same measurement setup along with an instrument that has wider bandwidth, then yes the other instrumentation may well be affected (eg. a DVM with 300kHz bandwidth for rms voltage, or an alternate instrument for Bode plotting like a Picoscope), but that seems to be getting into some exotic test setups where one would have to be aware of lots of potential issues.

[macboy]

I have a Panasonic VP7722A which is a very nice analyser. Being older it does not have a big graphical display that the more modern ones have, but it is simple to use and has a very low THD floor (around 1ppm or 0.00010%). Unlike some pure analog analysers like the HP 8903B (which I owned in the past) this one uses DSP to individually measure the fist 10 harmonics. The front panel can only display the first 5 but you can grab the others over GPIB. The DSP can also average out noise so that you can measure only THD in addition to THD+N. It also does IMD, SNR, and other things.

I have also used a sound card plus software (REW, Rightmark audio analyser, others), but I prefer a standalone instrument for basic measurements. There is something about cabling up a DUT to the front panel jacks, pushing a few buttons, and seeing a measurement appear on an LED display. The main benefit of the software solution is the FFT display, plus easily saving the results. The drawbacks are plenty: booting PC and software, using keyboard and mouse instead of buttons (touch screen would be nice but I don't have), and the need to roll your own interface between amplifier output and sound card input so that you don't blow up the latter.

[nenea dani]

I am attaching a diagram with the output from 4i4 for 20kHz sine somewhere at 1/2 FS with the specification that it is not in the loop. Everything is satisfactory, but higher in frequency we have artifacts. I am attaching two more diagrams, one up to 250kHz and the next up to 2.6MHz. The noise generated by the switching sources can be clearly observed, although it is possible that some artifacts have another origin. The noises in the last diagram do not disappear if the signal is off, but are proportional in any situation to the level of the master potentiometer. These disturbances are best seen with the oscilloscope on a rectangular signal because they have about 12-15% of the useful value. The spectral components used in the composition of the rectangle are stable over time, but the noise of the switching sources wanders continuously. I use an LC filter with fc =136Khz, N=7 at the output of Focusrite to clean all the radio frequency when I need a sine wave 20Hz-100kHz. After all, it's a 200 Euro sound card and for the hobby it's just fine.

Does '1/2 FS' mean you use a sampling frequency of 96kHz with a nominal 48kHz bandwidth?
Can you post a spectrum from say 10kHz to 2.6MHz, rather than with separate spectrum widths, to give confidence that the artifacts are shown along with the generated signal and its harmonics.
The 240kHz artifact is pretty much at the noise floor level. 
The 1.1MHz to 2MHz artifacts may be just a result of your measurement system and ground/parasitic loops, given they don't disappear when the signal is off. 
Can you elaborate on what you mean when talking about a 'rectangular signal'.

For use in audio measurements, I can't see what influence any soundcard artifacts in the low MHz range (or even at 250kHz if that is an actual artifact) would have on any measurement accuracy or outcome or

If the soundcard was in use at the same time, and in the same measurement setup along with an instrument that has wider bandwidth, then yes the other instrumentation may well be affected (eg. a DVM with 300kHz bandwidth for rms voltage, or an alternate instrument for Bode plotting like a Picoscope), but that seems to be getting into some exotic test setups where one would have to be aware of lots of potential issues.

Sorry for rectangular and FS. I had in mind square wave and Full Scale respectively. Now I have calibrated the board, I have disconnected all the unnecessary devices nearby and the LED lights. I use RG316 cables and everything is well shielded. I am attaching two diagrams with a 20kHz signal / around 55mVeff measured with the oscilloscope at the input of the analyzer and then removed. I immediately connect with the generator off and it can be seen that the switching sources are still there or somethings. You are right that devices connected together can influence each other and that radio frequency artifacts do not matter in audio. However, some high-performance amplifiers can exceed 500Khz and I heard that ten times the band limit is needed for  tracking tests phase change. In short, to work correctly between 20Hz and 20kHz, it must have minimum attenuation between 2Hz and 200kHz.

[trobbins]

Sorry for rectangular and FS. I had in mind square wave and Full Scale respectively. Now I have calibrated the board, I have disconnected all the unnecessary devices nearby and the LED lights. I use RG316 cables and everything is well shielded. I am attaching two diagrams with a 20kHz signal / around 55mVeff measured with the oscilloscope at the input of the analyzer and then removed. I immediately connect with the generator off and it can be seen that the switching sources are still there or somethings. You are right that devices connected together can influence each other and that radio frequency artifacts do not matter in audio. However, some high-performance amplifiers can exceed 500Khz and I heard that ten times the band limit is needed for  tracking tests phase change. In short, to work correctly between 20Hz and 20kHz, it must have minimum attenuation between 2Hz and 200kHz.

Was the last plot made with the same test connections as the first two plots, but just with the soundcard not being asked to generate the 20kHz sinewave tone?
Did you do a reference plot with the same equipment and test connections in place, but the soundcard itself unpowered ?
Can you elaborate on what "ten times the band limit is needed for  tracking tests phase change.", as I'm not sure what 'tracking tests phase change' means.
Do you make tests that use the 4i4 to generate signals and apply them to equipment, and have the Sigilent scope connected at the same time for measuring performance of the equipment in the frequency range beyond that of the 4i4 test signals, and is that what you are concerned about?

[pdenisowski]

Discontinued, but definitely not a toy:

https://www.rohde-schwarz.com/us/products/test-and-measurement/audio-analyzers/rs-upv-audio-analyzer_63493-7558.html

Edit:  looks like you can still pick up a used one from Ebay for 15-20K USD

[KungFuJosh]

Edit:  looks like you can still pick up a used one from Ebay for 15-20K USD

I offered them $200. For some reason they didn't accept. 🤷🤣

[nenea dani]

Sorry for rectangular and FS. I had in mind square wave and Full Scale respectively. Now I have calibrated the board, I have disconnected all the unnecessary devices nearby and the LED lights. I use RG316 cables and everything is well shielded. I am attaching two diagrams with a 20kHz signal / around 55mVeff measured with the oscilloscope at the input of the analyzer and then removed. I immediately connect with the generator off and it can be seen that the switching sources are still there or somethings. You are right that devices connected together can influence each other and that radio frequency artifacts do not matter in audio. However, some high-performance amplifiers can exceed 500Khz and I heard that ten times the band limit is needed for  tracking tests phase change. In short, to work correctly between 20Hz and 20kHz, it must have minimum attenuation between 2Hz and 200kHz.

Was the last plot made with the same test connections as the first two plots, but just with the soundcard not being asked to generate the 20kHz sinewave tone?
Did you do a reference plot with the same equipment and test connections in place, but the soundcard itself unpowered ?
Can you elaborate on what "ten times the band limit is needed for  tracking tests phase change.", as I'm not sure what 'tracking tests phase change' means.
Do you make tests that use the 4i4 to generate signals and apply them to equipment, and have the Sigilent scope connected at the same time for measuring performance of the equipment in the frequency range beyond that of the 4i4 test signals, and is that what you are concerned about?

     In fact, I did the last test with the generator connected as in the first two tests, but switched from on to off. I also did tests with the 4i4 board disconnected from the USB and indeed there are artifacts at about 100Mhz where it probably receives local FM stations and something else around 1.17Mhz but no more. I want to start a real analyzer to properly investigate. As soon as I find out, I'll let you know the result. I understand that this Siglent that I am currently using, is a teaching device and is not 100% reliable. This SA does not behave very well between 9 and 75Khz and therefore in TG mode it has the possibility of normalization for "correction".  For now, I have to take out of the warehouse a 25Kg Systron Donner ultra-old mastodon, but rebuilt and tested. I don't want to use the 4i4 board at more than 20Khz because the manufacturer only guarantees up to this point, but I'm intrigued by the noise that appears at the same time as the useful signal and which is dependent on the master level and that's why I think it comes from inside.
     I abandoned Focusrite and now I'm testing a white noise generator that seems to be fairly constant between 20Hz and almost 100kHz.
Regarding the 2Hz-200kHz or even 500Khz spectrum, this is the range on which the amplifier must be checked. "Such frequency extension gives the possibility to maintain the phase errors to a very low level". There are many articles on the net and the first one I found quickly is http://www.audioanalogue.com/en/maestro-monoaural/maestro-monoaural-se-power-amplifier
  Surely this issue is known to those who build hi-fi, but my expression was not correct because I am just an audio hobbyist.

[Doctorandus_P]

My preference would also be to use a (high-end) sound card.
If you want a "stand alone" device, you can also combine it with some tablet PC. There are plenty of second hand of these,

[trobbins]

nenea dani, this is all quite interesting, and I commend your efforts.

The various dc/dc on the soundcard I use (EMU 0404 USB) causes a known corruption within the 96kHz bandwidth of its measurement spectrum, and users identified a way to suppress that by re-routing a signal line away from a dc/dc internal to the soundcard.  For the audio work I do, which certainly extends to 96kHz, and beyond, I know there can be some spurious signals down in the weeds on one channel of the soundcard - which can either be overlooked, or the other channel used (some users find that noise aesthetically displeasing, even though it typically doesn't detract from the measurements they are making).

When I need to extend measurements beyond 96kHz, which I certainly do for Williamson amps and typically have to go out past 500kHz to assess feedback gain and phase margins, I use a separate signal generator and a scope and some voltmeters that have enhanced bandwidth, and so there is no opportunity for the soundcard to pollute such measurements.  So I'm intrigued as to what type of measurement situation could be corrupted by a soundcard emitting noise well outside its nominal measurement bandwidth.  Are you envisaging that the soundcard would be in use at the same time as when measurements were being made by other equipment (which could be corrupted)?

Ciao, Tim

[nenea dani]

nenea dani, this is all quite interesting, and I commend your efforts.

The various dc/dc on the soundcard I use (EMU 0404 USB) causes a known corruption within the 96kHz bandwidth of its measurement spectrum, and users identified a way to suppress that by re-routing a signal line away from a dc/dc internal to the soundcard.  For the audio work I do, which certainly extends to 96kHz, and beyond, I know there can be some spurious signals down in the weeds on one channel of the soundcard - which can either be overlooked, or the other channel used (some users find that noise aesthetically displeasing, even though it typically doesn't detract from the measurements they are making).

When I need to extend measurements beyond 96kHz, which I certainly do for Williamson amps and typically have to go out past 500kHz to assess feedback gain and phase margins, I use a separate signal generator and a scope and some voltmeters that have enhanced bandwidth, and so there is no opportunity for the soundcard to pollute such measurements.  So I'm intrigued as to what type of measurement situation could be corrupted by a soundcard emitting noise well outside its nominal measurement bandwidth.  Are you envisaging that the soundcard would be in use at the same time as when measurements were being made by other equipment (which could be corrupted)?

Ciao, Tim

   It seems that we don't have to worry about what happens above 100kHz if we use the Focusrite board only up to a few tens of kHz. Indeed, the sine signal appears polluted and this makes us think, but now I understand that it does not affect the measurements. I'm trying to reach 100kHz in analog mode and that's why I built an up-mixer that moves the whole 20Hz-100kHz spectrum to 1,000,020MHz-1,100.000Mhz and thus there is no longer a problem that SA starts with 9kHz. In this way, I see everything that happens in almost 80dB. I still have to build a pure sine generator at least for 25KHz-100kHz or lower(1kHz-100KHz) and synchronize it with SA. For now I'm struggling with a white noise generator that doesn't allow a precise measurement for fc(-3dB). I still have subtle problems that don't allow me to go below 15-16Hz because I have some phase noise from the local 1MHz oscillator, but I really don't do the performance, just a hobby. I am very careful with the noise and that is why I use multiple decouplings including tantalum and only linear sources. Even if my work is in the area of switching sources here in audio, I don't even want to see it. Sometimes I use an ultra low noise LDO, but here it was not necessary. I cannot connect the SA Siglent input in parallel with the oscilloscope input because the first disturbs the second in the sweep rhythm. However, I think that the Focusrite board is not disturbed by SA Siglent.

[GordonB]

I appreciate that I'm a bit late to this thread having just joined the Forum, I'm a retired noise and vibration consultant and I have one sound analysing system left that I'd be happy to sell, but I'm not prepared to ship overseas, only in the UK. The system is a Norsonic 121 that will do environmental noise and building acoustics in 1/3 octaves, I'll put a listing on the For Sale section but I'm looking for £350 for it complete. There's an intermittant issue with battery charging but it works perfectly from an external 12V source.

[metebalci]

I also searched for AA options recently. I think the options are roughly:

- cheapish sound card + a free or not expensive software
- qa403 which is quite unique in this space
- expensive sound card/dac + a free or not expensive software
- quite an expensive dac (rme) + virtins mi
- virtins hardware rtx + software
- proper AAs, dscope, ap etc.

Which one to choose depends on various factors and what you want from the software and the hardware. Qa403 looks great for the price, but I never find it in stock. So I went for rme dac and virtins mi (and not only for this purpose, I use the dac for other things as well, I dont need an AA 24/7).

[pope]

I also searched for AA options recently. I think the options are roughly:

- cheapish sound card + a free or not expensive software
- qa403 which is quite unique in this space
- expensive sound card/dac + a free or not expensive software
- quite an expensive dac (rme) + virtins mi
- virtins hardware rtx + software
- proper AAs, dscope, ap etc.

Which one to choose depends on various factors and what you want from the software and the hardware. Qa403 looks great for the price, but I never find it in stock. So I went for rme dac and virtins mi (and not only for this purpose, I use the dac for other things as well, I dont need an AA 24/7).

AFAIK, virtins hardware rtx has been discontinued long time ago.

[metebalci]

I also searched for AA options recently. I think the options are roughly:

- cheapish sound card + a free or not expensive software
- qa403 which is quite unique in this space
- expensive sound card/dac + a free or not expensive software
- quite an expensive dac (rme) + virtins mi
- virtins hardware rtx + software
- proper AAs, dscope, ap etc.

Which one to choose depends on various factors and what you want from the software and the hardware. Qa403 looks great for the price, but I never find it in stock. So I went for rme dac and virtins mi (and not only for this purpose, I use the dac for other things as well, I dont need an AA 24/7).

AFAIK, virtins hardware rtx has been discontinued long time ago.

According to website, rtx6001 is still being sold.

[pope]

I also searched for AA options recently. I think the options are roughly:

- cheapish sound card + a free or not expensive software
- qa403 which is quite unique in this space
- expensive sound card/dac + a free or not expensive software
- quite an expensive dac (rme) + virtins mi
- virtins hardware rtx + software
- proper AAs, dscope, ap etc.

Which one to choose depends on various factors and what you want from the software and the hardware. Qa403 looks great for the price, but I never find it in stock. So I went for rme dac and virtins mi (and not only for this purpose, I use the dac for other things as well, I dont need an AA 24/7).

AFAIK, virtins hardware rtx has been discontinued long time ago.

According to website, rtx6001 is still being sold.

I really don't think so. There is some possibility it might get produced again but very little as far as I'm concerned. It is on the the website but have you tried to purchase it. I don't think there's an option.

[metebalci]

I also searched for AA options recently. I think the options are roughly:

- cheapish sound card + a free or not expensive software
- qa403 which is quite unique in this space
- expensive sound card/dac + a free or not expensive software
- quite an expensive dac (rme) + virtins mi
- virtins hardware rtx + software
- proper AAs, dscope, ap etc.

Which one to choose depends on various factors and what you want from the software and the hardware. Qa403 looks great for the price, but I never find it in stock. So I went for rme dac and virtins mi (and not only for this purpose, I use the dac for other things as well, I dont need an AA 24/7).

AFAIK, virtins hardware rtx has been discontinued long time ago.

According to website, rtx6001 is still being sold.

I really don't think so. There is some possibility it might get produced again but very little as far as I'm concerned. It is on the the website but have you tried to purchase it. I don't think there's an option.

You have right, I asked to virtins, they said (company) RTX (I didnt know there is another company actually producing it) discontinued it. So an option between DAC+software and proper AA is gone. So there is no option (other than a 2nd hand proper AA) from ~1.5K$ to ~5K$ or so.

[arvidb]

Maybe this thread from Audio Science Review could be of interest.

Topping D10s + USB isolator as signal source, E1DA Cosmos ADC for measurement, and REW for processing. Total cost ~€400, and a THD+N better than -110 dB (16 Hz to 22 kHz, 1 kHz fundamental, highest harmonic 3rd at -128 dBc in this particular measurement) seems pretty good to me! A downside is the low input impedance of the Cosmos ADC I guess.

If I'm not mistaken, both the DAC and the ADC have maximum sample rate 384 kHz so a pretty wide bandwidth for audio equipment.

[arvidb]

Here's a measurement of my (very recently arrived) Topping D10s + Cosmos ADC (grade B), with a Hifime High-Speed USB isolator "v2" on the ADC.

With better matched levels I could apparently gain another 5-7 dB of THD+N, but the current level of performance will be plenty for my uses anyway so I might not bother to optimize things further.

[nenea dani]

Finally, what is the upper frequency limit in this configuration?

[harha]

Hi
I use the same setup-but with E1da 9038D6K as generator---just amazing performance..
Hardy

[bozidarms]

RTX 6001            THD:  0,000017%  =  -135,391dB

[tshackelton]

Do you have a link to that python notebook? I'm toying around with python/jupyter controlling my chinese AWG and scope, and was looking to add some of this functionality. I'd love to peek over your shoulder....