I think a look at history will help rationalize this IM discussion. https://www.rfcafe.com/references/e...istortion-electronics-world-february-1960.htm
Look at how audio IM was measured before digital systems were possible. Modern digital systems are trying to emulate the analog solutions originally for compatibility. The analog systems used filters and demodulators. RMS sensing was not practical so RMS indicating average reading instruments were used.
The digital emulations become difficult when there isn't clarity on how to measure amplitude in a narrow band the way the analog system did.
Look at how audio IM was measured before digital systems were possible. Modern digital systems are trying to emulate the analog solutions originally for compatibility. The analog systems used filters and demodulators. RMS sensing was not practical so RMS indicating average reading instruments were used.
The digital emulations become difficult when there isn't clarity on how to measure amplitude in a narrow band the way the analog system did.
The flaw in IMD SMPTE measurement (12 dB too good), that I described has been removed a couple of hours ago with QA's software update v. 1.197, available on their web page
https://github.com/QuantAsylum/QA40x/releases
https://github.com/QuantAsylum/QA40x/releases
In the meantime clarification on the method (for deriving SMPTE IMD from FFT) was obtained.
While there seems to be no agreement here in this forum on the question, if the IM-signals are generally uncorrelated or maybe correlated within each linear combination pair of the same order, the answer is clearly given by the methods prescribed by SMPTE and DIN and TGL (and others).
In a nutshell:
Within a (symmetrical) pair of the same order both IM-signals are correlated due to their origin from amplitude modulation. Hence, f1+ nf2 and f1 - nf2 are both correlated for the same order n, but are not correlated for different orders n.
This explanation is eplicitely given in a remark contained in the TGL Intermodulation measurement standard 17175-04 (TGL was the GDR equivalent to DIN).
That reason lead to the SMPTE and DIN formula for calculation from FFT, which is applied i.e. in ARTA (well known audio measurement software usable with soundcards) and described in DIN 45403:
and is consistent with the same method applied by Audio Precision in its measurement equipment, as described by Richard C. Cabot of that company:
"The modulation components of the upper signal appear as sidebands spaced at multiples of the lower frequency tone. The amplitudes of the sidebands are added in pairs, root square summed, and expressed as a percentage of the upper frequency level."
"added in pairs" = linear addition first, then the sums of the linear addition for each order are rms-summed.
As a matter of fact, the theoretical background for the formula is "nice to know" but actually not essential. It is sufficient to follow the method prescribed by SMPTE or DIN (the same in this case) according to the formula contained therein and shown here.
So far, QA403 v. 1.197and former QA40x software versions do not follow the SMPTE procedure but apply throughout rms-summing for all IM-signals up to the chosen order, regardless if the belong to (symmetrical, linear combination) pairs of same order or not. This results in IMD-distortion values that are too good (too low) by typically 3 dB.
I assume, that QA may want to address this also in one of the future software versions to make Quant Asylum tools readings compatible with the ordinary SMPTE standard. Up to then, with software versions prior to v. 1.197 one has to add hefty +15 dB (explained before, +12 dB due to wrong carrier reference and +3 dB because of non-SMPTE/DIN summing formula) and from v. 1.197 until this final issue is corrected, one has to add +3 dB (typically, if only one order dominates the IM) to the SMPTE IMD readings from the QA40x equipment.
Best Regards,
Reinhard
While there seems to be no agreement here in this forum on the question, if the IM-signals are generally uncorrelated or maybe correlated within each linear combination pair of the same order, the answer is clearly given by the methods prescribed by SMPTE and DIN and TGL (and others).
In a nutshell:
Within a (symmetrical) pair of the same order both IM-signals are correlated due to their origin from amplitude modulation. Hence, f1+ nf2 and f1 - nf2 are both correlated for the same order n, but are not correlated for different orders n.
This explanation is eplicitely given in a remark contained in the TGL Intermodulation measurement standard 17175-04 (TGL was the GDR equivalent to DIN).
That reason lead to the SMPTE and DIN formula for calculation from FFT, which is applied i.e. in ARTA (well known audio measurement software usable with soundcards) and described in DIN 45403:
and is consistent with the same method applied by Audio Precision in its measurement equipment, as described by Richard C. Cabot of that company:
"The modulation components of the upper signal appear as sidebands spaced at multiples of the lower frequency tone. The amplitudes of the sidebands are added in pairs, root square summed, and expressed as a percentage of the upper frequency level."
"added in pairs" = linear addition first, then the sums of the linear addition for each order are rms-summed.
As a matter of fact, the theoretical background for the formula is "nice to know" but actually not essential. It is sufficient to follow the method prescribed by SMPTE or DIN (the same in this case) according to the formula contained therein and shown here.
So far, QA403 v. 1.197and former QA40x software versions do not follow the SMPTE procedure but apply throughout rms-summing for all IM-signals up to the chosen order, regardless if the belong to (symmetrical, linear combination) pairs of same order or not. This results in IMD-distortion values that are too good (too low) by typically 3 dB.
I assume, that QA may want to address this also in one of the future software versions to make Quant Asylum tools readings compatible with the ordinary SMPTE standard. Up to then, with software versions prior to v. 1.197 one has to add hefty +15 dB (explained before, +12 dB due to wrong carrier reference and +3 dB because of non-SMPTE/DIN summing formula) and from v. 1.197 until this final issue is corrected, one has to add +3 dB (typically, if only one order dominates the IM) to the SMPTE IMD readings from the QA40x equipment.
Best Regards,
Reinhard
Last edited:
A couple of days ago QA have issued software v. 1.198 which is addressing the remaining 3 dB mismatch of v. 1.197 when compared to SMPTE IMD measurement with other FFT-based gear.
Finally, from v. 1.198 the QA40x (QA403) yields SMPTE (MOD) IMD measurement results which are in line with those from other common instruments, like i.e. soundcard / audio interface based REW (Room EQ Wizard), ARTA and the latest (Feb. 2024) version of Multi Instrument (Virtins Technology), hence complies with the standards
A comparative measurement for a particular test amplifier gave identical SMPTE IMD results (2nd + 3rd order IM) within the boundaries of repeatability (approx. +/- 0.5 dB):
1. QA403 with software v. 1.198 IMD (up to 3rd order) = - 46.9 dB
2. REW IMD SMPTE = - 46.8 dB
3. ARTA SMPTE IMD 60:7000 Hz, 4:1 = - 46.6 dB
RMAA:
It shall be noted, that RMAA (Right Mark Audio Analyzer) IMD measurements apply the so called power method with the 60:7000 Hz 4:1 signal pair. Different from the SMPTE (MOD) method, the power method relates the rms power sum of the IM-products to the total power (= essentially the power of the 60 Hz signal). That results in an IMD (power) value, that is always 9 dB (= factor 2.82) smaller than the corresponding SMPTE IMD value. That's why RMAA IMD looks so much "better" than SMPTE IMD.
Finally, from v. 1.198 the QA40x (QA403) yields SMPTE (MOD) IMD measurement results which are in line with those from other common instruments, like i.e. soundcard / audio interface based REW (Room EQ Wizard), ARTA and the latest (Feb. 2024) version of Multi Instrument (Virtins Technology), hence complies with the standards
DIN IEC 60268-3:2019-02
IEC 60268-3:2018
SMPTE RP 120 (2005)
A comparative measurement for a particular test amplifier gave identical SMPTE IMD results (2nd + 3rd order IM) within the boundaries of repeatability (approx. +/- 0.5 dB):
1. QA403 with software v. 1.198 IMD (up to 3rd order) = - 46.9 dB
2. REW IMD SMPTE = - 46.8 dB
3. ARTA SMPTE IMD 60:7000 Hz, 4:1 = - 46.6 dB
RMAA:
It shall be noted, that RMAA (Right Mark Audio Analyzer) IMD measurements apply the so called power method with the 60:7000 Hz 4:1 signal pair. Different from the SMPTE (MOD) method, the power method relates the rms power sum of the IM-products to the total power (= essentially the power of the 60 Hz signal). That results in an IMD (power) value, that is always 9 dB (= factor 2.82) smaller than the corresponding SMPTE IMD value. That's why RMAA IMD looks so much "better" than SMPTE IMD.
Last edited:
Has anyone out there bothered to characterize the high frequency performance of this thing? What is the THD+N at 20kHz when sampling at 192kHz?
I have yet to see a published FFT plot that goes higher than 20kHz or anyone even test the thing with an input signal higher than 2kHz. Can the software for this thing go higher and look at the harmonics of 20kHz to at least 80kHz?
I have some very high-Q audio-frequency filters for a non-audio application to characterize. These require a very slow frequency sweep rate when generating bode plots - a user-programmable sweep rate speed being a rather basic requirement not accommodated by all of the "audio" analyzer software I've had the misfortune to evaluate so far. Is the QA403 software any different?
Thanks.
I have yet to see a published FFT plot that goes higher than 20kHz or anyone even test the thing with an input signal higher than 2kHz. Can the software for this thing go higher and look at the harmonics of 20kHz to at least 80kHz?
I have some very high-Q audio-frequency filters for a non-audio application to characterize. These require a very slow frequency sweep rate when generating bode plots - a user-programmable sweep rate speed being a rather basic requirement not accommodated by all of the "audio" analyzer software I've had the misfortune to evaluate so far. Is the QA403 software any different?
Thanks.
Have you tried REW stepped-sine measurement https://www.roomeqwizard.com/help/help_en-GB/html/spectrum.html#steppedsine ?
Tried the QuantAsylum forums?
BTW I've posted several here with 80kHz bandwidth: https://www.diyaudio.com/community/threads/new-version-of-my-tonearm-phono-preamp.411809/
https://www.diyaudio.com/community/threads/modular-phono-preamp.338801/#post-7631321
Not sure I've used a signal much above 20kHz, as its audio, not RF
I shouldn't have to say this, but if THD at 20kHz is to be measured, the measurement bandwidth should extend to at least the 3rd or 4th harmonic (60 or 80 kHz) with minimal attenuation. This is about the best anyone can hope for with something like the QA403 or a generic sound card at a sampling rate of 192kHz. Still distinctly inferior to my analog analyzer with a switchable 2-pole Butterworth measurement bandwidth up to 500kHz,
The QA403 generator distortion out to 20kHz would be interesting to see. I don't know how many independent review videos by incurious copycat twits just running a loop-back test at 1kHz the world needs.
REW looks like it might be the ticket for my filters. Thanks.
The QA403 generator distortion out to 20kHz would be interesting to see. I don't know how many independent review videos by incurious copycat twits just running a loop-back test at 1kHz the world needs.
REW looks like it might be the ticket for my filters. Thanks.
QA403 is available again in the EU.
https://www.elektor.de/products/quantasylum-qa403-24-bit-audio-analyzer
https://www.elektor.de/products/quantasylum-qa403-24-bit-audio-analyzer
I can get the distortion of mine better than that by juggling the input attenuation and generator level settings.
- Making some circuit optimizations to a low-noise pre-amplifer I designed a while ago pending a revamp and improvement of the design. The 50Hz pick-up isn't surprising but am scratching my head as to why the 3rd and 5th harmonic of the mains frequency is so strong. The shorted-input pre-amplifier is battery powered and floating and everything in my lab is turned off, except for the PC running the QA403. I can't imaging what could be generating this other than the PC PSU.
How good is the galvanic isolation on the USB port of this thing? I am wondering if there is a common-mode component polluting things. Only the expected 50Hz component is present at the pre-amplifier output when not connected to the QA403 and examined otherwise.
Will get this software installed onto my laptop next.
- Making some circuit optimizations to a low-noise pre-amplifer I designed a while ago pending a revamp and improvement of the design. The 50Hz pick-up isn't surprising but am scratching my head as to why the 3rd and 5th harmonic of the mains frequency is so strong. The shorted-input pre-amplifier is battery powered and floating and everything in my lab is turned off, except for the PC running the QA403. I can't imaging what could be generating this other than the PC PSU.
How good is the galvanic isolation on the USB port of this thing? I am wondering if there is a common-mode component polluting things. Only the expected 50Hz component is present at the pre-amplifier output when not connected to the QA403 and examined otherwise.
Will get this software installed onto my laptop next.
Last edited:
Powered from the PC. Got this thing connected to my laptop now - no significant difference, even after cutting all power to the lab at the breakers.
Must be the neighbors HVAC or something enveloping me inside some kind of local force field or something, lol.
Will have to get this gear packed into the ute and driven to the beach.
Must be the neighbors HVAC or something enveloping me inside some kind of local force field or something, lol.
Will have to get this gear packed into the ute and driven to the beach.
Here is my setup, battery pack lasts me at least 5 hours:
And yes my laptop is running off of a brick too.
https://1010music.com/product/usbbsplitter
Since this pic, my data cable from my laptop is short as is the cable from the power bank. The voltage I am detecting via the software for USB is right at 5.0 volts. I’ve been fairly successful with this setup but I’ll admit, I haven’t used it for continuous measurements for more than 5 hours or so. Charging the Anker power bank takes about 1 hour if completely discharged.
Works for the hobbyist!
Best,
Anand.
And yes my laptop is running off of a brick too.
https://1010music.com/product/usbbsplitter
Since this pic, my data cable from my laptop is short as is the cable from the power bank. The voltage I am detecting via the software for USB is right at 5.0 volts. I’ve been fairly successful with this setup but I’ll admit, I haven’t used it for continuous measurements for more than 5 hours or so. Charging the Anker power bank takes about 1 hour if completely discharged.
Works for the hobbyist!
Best,
Anand.
Last edited:
Now that I've had time outside of other preoccupations to actually think about this I'm guessing those harmonics are ADC artifacts. As far as the displayed peak amplitude can be trusted, that 50Hz mains hum fundamental at 1.2nV is only 1.2uV as presented to the QA403 input, factoring in the 60dB pre-amplifer gain: about -118dBV.
Probably asking a bit much of the ADC to sample a signal that small without distortion. I need to jury rig an intermediate amplifier to give at least another 40dB of gain after the pre-amplifier under test; tomorrow - too late now.
Last edited: