So either you are and consider us too stupid for the benefit of your experience, or you were 2 decades ago and have missed the whole development of bit perfect drivers for windows.
You are on a DIY forum, you asked for something for home use, then have shot down every sensible suggestion from people who are measuring this stuff on a daily basis. Methinks you protest too much for the level of trolling shown.
+1, I'm also totally lost on the purpose of discussion now
Well why? You can claim everything you like, but every claim requires an independent verification. I do not see any. So i am in the position of the skeptic.
For example you have meter stick, and only this tool for measurement of distance, you can never know whether your measurements are correct unless you verify the calibration of your one and only meter stick with an accepted standard. There is only one accepted standard , and it is not your meter stick.
I am grateful for the responses, but I see only two which make sense to me, the rest is just claims.
soundcards can "self calibrate" in loopback mode - if the results are good enough then there's little point in some sort of independent verification
and added diff amp, filters, indirect IMD product techniques can get around some limitations, or identify them as being specific limitations of the soundcard's ADC, DAC or analog electronics
for speaker measurement even cheap motherboard soundcards are way beyond actual needs in resolution
for electronics better soundcards with 192k, 80 kHz BW exceed actual audible measurement need - but you need a 100 MHz 'scope too today to look for stability, parasitic oscillation
and added diff amp, filters, indirect IMD product techniques can get around some limitations, or identify them as being specific limitations of the soundcard's ADC, DAC or analog electronics
for speaker measurement even cheap motherboard soundcards are way beyond actual needs in resolution
for electronics better soundcards with 192k, 80 kHz BW exceed actual audible measurement need - but you need a 100 MHz 'scope too today to look for stability, parasitic oscillation
I think you're missing something in the way it works.
When measuring things like THD, you never need precise absolute values. All the measurements are relative. You've got some value for the center frequency, and then you've got -80db for the 2-nd harmonic. This is enough for precise calculation.
The software is just a high-speed digital calculator.
jcx is absolutely right - resolution and accuracy very much depend on the hardware - ADCs, DACs, filters, linear amplifiers, etc.
The software sends the digital values to DAC, it generates corresponding voltages - you get the waveform. The more linear the hardware is - the better (less distortion) waveform you get at analog output.
Same with the input - software just reads the values from ADC.
Now, external analog loop-back allows checking the whole measurement system. The software sends the values sequence for sine wave to DAC that converts it to analog waveform and sends it to analog output, which is connected directly to analog input, forwarding the waveform to ADC, from where the software receives the values sequence again.
Now, it's time to run FFT algorithm, see harmonic levels (relative to central frequency) and calculate THD.
Done.
Software is nothing to do with the measurement process, which is done by the card's hardware. The more linear and fast the hardware is, the more accurate results in wider bandwidth you get.
I have a couple of hardware signal analyzers - both of them are wider bandwidth, but less resolution, than the Lynx card.
Try to run an external loop-back FFT and get the level of resolution I published earlier. If you reach the level of this order, I'll be the first one to ask about the details on the hardware you have used
Cheers,
Valery
P.S. I'm not a Lynx sales agent
Last edited:
You can easily verify that the audio signal path in your operating system is "bit perfect". For example, you can connect 2 computers together with a SPDIF cable, play on one machine and record on the other. Then you can compare the recorded file to the original.
If the recorded file is identical to the original then that proves it is working as well as digital audio ever can. There is no need to worry about the operating system architecture.
Performance of ADCs and DACs is a separate issue, that has nothing to do with the operating system. I also own an older Tektronix AA501 distortion analyser. It only reads down to about 0.002% THD+N, but nevertheless I've found it to be somewhat more sensitive and more useful than the various semi-pro soundcards I've used. I bought it because I got frustrated trying to make high frequency THD and IMD measurements with soundcards.
Disclaimer: I only wrote the firmware for about a dozen scientific instruments, so I don't really know anything about computers.
If the recorded file is identical to the original then that proves it is working as well as digital audio ever can. There is no need to worry about the operating system architecture.
Performance of ADCs and DACs is a separate issue, that has nothing to do with the operating system. I also own an older Tektronix AA501 distortion analyser. It only reads down to about 0.002% THD+N, but nevertheless I've found it to be somewhat more sensitive and more useful than the various semi-pro soundcards I've used. I bought it because I got frustrated trying to make high frequency THD and IMD measurements with soundcards.
Disclaimer: I only wrote the firmware for about a dozen scientific instruments, so I don't really know anything about computers.
Last edited:
Disclaimer: I only wrote the firmware for about a dozen scientific instruments, so I don't really know anything about computers.
Oh, yeah - same thing, just an Intel GPU-based high-performance video encoders manufacturer. Don't have a clue about computers
Funny .. ok. I am the ( almost sole) author of a complete real time operating system. Not for Intel architecture, mind you. Inclusive of app interface.... but ok. I have my conclusion and you have yours. if you trust your measurements its ok. for you. If I don't, its ok. for me. So soon we will see numerous emergency sales of Audiolab analyzers at a bargain at ebay. Take it easy, folks.
Take it easy, folks.I guess it is somewhat valid to doubt the precision of soundcards as measuring devices. As other people already pointed out:
The gain is not calibrated. You don't know how many millivolts per bit unless you measure it yourself using some other calibrated audio frequency voltmeter. If you do this, most software will let you enter the calibration constant and give you plots scaled in volts. A "real" audio analyser will already be calibrated.
The input voltage range of a soundcard is somewhat limited. An audio analyser will usually have a high impedance balanced input with switchable gain ranges, so you can measure anything from a tiny line level signal to the output of a big power amp. Pete Millett designed an interface that gives a soundcard this sort of input.
The frequency response may also be limited, coupling capacitors (probably cheap electrolytics) will roll it off at LF, and some makers cheat and don't give the full Nyquist bandwidth when running at higher sample rates.
Last but not least, there is the matter of sigma-delta converters. All modern audio converters work like this, they only have about 4 bits and these are dithered at high frequency to give the impression of up to 24. This means that a soundcard-generated sine wave contains a certain amount of HF crud that looks dreadful on an analog THD analyser.
There is no evidence that this crud is audible and it doesn't seem to bother a sigma-delta ADC either, but you have to wonder if it is getting in the way of doing precision high frequency measurements somehow.
The gain is not calibrated. You don't know how many millivolts per bit unless you measure it yourself using some other calibrated audio frequency voltmeter. If you do this, most software will let you enter the calibration constant and give you plots scaled in volts. A "real" audio analyser will already be calibrated.
The input voltage range of a soundcard is somewhat limited. An audio analyser will usually have a high impedance balanced input with switchable gain ranges, so you can measure anything from a tiny line level signal to the output of a big power amp. Pete Millett designed an interface that gives a soundcard this sort of input.
The frequency response may also be limited, coupling capacitors (probably cheap electrolytics) will roll it off at LF, and some makers cheat and don't give the full Nyquist bandwidth when running at higher sample rates.
Last but not least, there is the matter of sigma-delta converters. All modern audio converters work like this, they only have about 4 bits and these are dithered at high frequency to give the impression of up to 24. This means that a soundcard-generated sine wave contains a certain amount of HF crud that looks dreadful on an analog THD analyser.
There is no evidence that this crud is audible and it doesn't seem to bother a sigma-delta ADC either, but you have to wonder if it is getting in the way of doing precision high frequency measurements somehow.
At the high end of "sound-card-ish" technology, but below the highest of the high end test equipment, sits this AP model...
AP High Performance Audio Analyzer & Audio Test Instruments : apx515
AP High Performance Audio Analyzer & Audio Test Instruments : apx515
Calibration takes about 2 minutes max and is very easy to do. I have found it to be quite stable using the relatively moderately priced AP192 card. It's something I recheck every now and then, just like I would with any instrument.
Apparently showing actual experimental data is insufficient, despite it actually answering the original question.
10ppm THD would be exceptionally good for a soundcard-based measurement. If I saw a figure better than that on a loopback test, I would suspect someone had put absinthe in my Buckfast.
Another thing to ponder: The guys at AKM, Wolfson etc. who design the audio converter chips. How do they measure the performance? They probably stick the new chip on an eval board and hook it up to an Audio Precision system. So the graphs in the ADC datasheet really show how the chip performs compared to AP's latest gizmo.
Another thing to ponder: The guys at AKM, Wolfson etc. who design the audio converter chips. How do they measure the performance? They probably stick the new chip on an eval board and hook it up to an Audio Precision system. So the graphs in the ADC datasheet really show how the chip performs compared to AP's latest gizmo.
I forgot to insert a smiley. An audio expert would not come up with such a claim of 10 ppm thd obtained with sound card measurement.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.