If you are referring to the THD, that is why CCIF IM is a much more important distortion measurement for class D amplifiers, since the lower-sideband odd-order products are all in-band, and should not be affected by the output filter. For 19+20kHz CCIF IM, 3rd order picket is at 18kHz, 5th order picket is at 17kHz, etc. Even order distortions are low in the band, such as 2nd order being at 1kHz. 4th order at 2kHz, etc.
IM is what we are usually hearing when HF distortion degrades sound quality. HF IM is what puts the spit on a cymbol, for example.
Cheers,
Bob
Self-declaration to me means automatic disqualification
A simple test for a DAC is to reduce a properly dithered sine wave or music recording to -20dB below the properly dithered noise floor and play this back with loadsa amplification (careful in case a FS signal comes next
But this only tests the bottom couple of LSBs, not likely to induce much noise modulation.
Piano is one of my best DAC torture tests. Along with voice. Analog tape softens piano (by adding noise modulation) enough to be fairly easily discernable.
No, there wasn't a project running for a classD amp when I was there, but I did play with an eval board or two (HIP4081 comes to mind). We listened to a few OEM designs - didn't like ICEpower but a pre-production UcD held more promise.
I believe that B&W use Hypex modules in the Zeppelin and the PV1D sub-woofer.
Someone suggested to Jan that i take a look at the THD of the regulators tested in the bakeoff we had last year with the New Jersey Audio Society volunteers. I used the same load as in WJ's TAA tests (100R+100uF) and injected 50mA onto the positive rail and looked at the resulting FFT. This isn't the audio output of the line amp, just the rail modulation. (The PSRR of the Borbely All-FET is about 40dB so regulator issues manifest themselves on the output.)
The regulator with all the harmonics is extremely quiet, but messes up the sound, at least in this application for which the PSRR of the discrete line amp is much worse than an OPA637 etc., etc.
must be more contex - as stated its wrong - any nonlinearity causes IMD with multitone stimulus
and we can have feedback gain phase rotating many degrees in small frequency range if loop gain has peaks - without a trace of distortion in the closed loop response as long as the system is linear
and we can have feedback gain phase rotating many degrees in small frequency range if loop gain has peaks - without a trace of distortion in the closed loop response as long as the system is linear
Last edited:
May I respectfully suggest this would be clearly seen in simple THD of the amplifiers with these 2 different supplies.
Self-declaration to me means automatic disqualification
Don't forget those who hear 'chalk & cheese' differences too.The test at -20dB LSB was cos in da old days, some crude PCM DACs had the signal go crackly as you neared 1 LSB and even disappear below 1 LSB. The old Philips 14b oversampling ones actually did pass this test. I think there was a Denon Test CD with such a signal.
You can also do this test at eg +10dB LSB ie -83dB FS and get distinct differences between da old DACs.
The reason for testing at very low level is that at high level, its difficult to distinguish the very small differences. The very best ears (and I've tested some of the best) will report a sense of unease which they can't quite articulate on the poor DACs. I was flabbergasted to find some of my Panel could reliably distinguish between correct & partial dithering on 70's & 80's Pop at near FS. The low level tests allowed mere mortals to distinguish DACs reliably.
Of course there was program where it didn't make any difference (even to my true golden pinnae) if it was clipped 50% of the time. I suspect much 'modern' music now falls in this category.
On piano, noise modulation has a distinct 'shush' sound accompany each note. On a bad case, the 'shush' sounds crackly.
BTW, with 21st century processing, you can measure THD for signals well below 1 LSB if you average long enough. Or you can listen to sine waves at that level and see which ones sound like clean but noisy sines and which ones sound crackly.
Last edited:
I believe the reason that they could is not that they actually heard the dither variance, rather that the DAC and following circuitry behaved differently in the sense that distortion at a much, much higher level was introduced which varied per dither pattern.The reason for testing at very low level is that at high level, its difficult to distinguish the very small differences. The very best ears (and I've tested some of the best) will report a sense of unease which they can't quite articulate on the poor DACs. I was flabbergasted to find some of my Panel could reliably distinguish between correct & partial dithering on 70's & 80's Pop at near FS. The low level tests allowed mere mortals to distinguish DACs reliably.
Distortion this obvious is way, way up in volume, around the -60dB level. As soon as the test signal is simplified the distortion levels drop right back down where they "should be" by standard measurement, the -90dB or so. Digital distortion is a very will-o-the-wisp, dynamic, behaviour, difficult to pin down ...
Yes of course - we continue to measure THD because its easy to do. But its close to irrelevant, so we need new measurements. Noise modulation measurement is harder to do with FFT. It also used to be easiest to generate sine-shaped test stimulii - that excuse no longer exists with digital generation being cheap and plentiful.
I just noticed peufeu posted up the link to these plots : DIY Test Equipment for Audio and Ham Radio Enthusiasts
What I'm curious to know is - some of these show interesting shaped noise floors. Are those noise floors the same as when unloaded or is the 50mA stimulus inducing some noise modulation in the regulators?
I'm rather pleased at the erudite nounce of this forum. If I'd posted this only a decade ago, I would have been inundated with posts claiming evil digital can't possibly resolve anything below 1 LSB
Perhaps people are actually reading Lipsh*tz & Vanderkooy on dither.
Hi Harry,
I believe that several of the posts above have done a pretty good job of describing noise modulation in the context I intended. In general, any way in which signal influences the noise amplitude or the noise statistics is noise modulation.
Its likely that there are noise modulation mechanisms at work in many class D amplifiers. Whenever there is quantization, there is an opportunity for noise modulation, I believe. I don't know of any class D amplification scheme that does not have some kind of quantization, be it PWM, a variant of PWM, or PDM, or something else. However, some class D schemes mitigate the effects of the quantization quite well. For example, closing the feedback loop around the class D amplifier in a linear way like Hypex does mitigates such effects a lot. Sigma-Delta class D amplifiers mitigate noise and noise modulation through noise shaping.
Compression schemes like MP3 are God-awful examples of noise modulation.
Cheers,
Bob
"Compression schemes like MP3 are God-awful examples of noise modulation."
MP3 should be outlawed.
Not before it could be objectively proved that 320kbps MP3 is audibly distinguishable from any other lossless codec.
Hint: it's not, and no golden ear could ever identify such in a double blind test.
Of course, 48kbps or 96kbps MP3 are a different story.
probably one of those stuck in time audio memes - a decade ago mp3 128kb was pretty easily heard - the coding side has been improved
and at higher bit rates, especially using other good lossy codecs that scale better with higher bit rates hydrogenaudio has dropped public trials above 128kb rates becasue there is no statistical significance in the blind listening results
and at higher bit rates, especially using other good lossy codecs that scale better with higher bit rates hydrogenaudio has dropped public trials above 128kb rates becasue there is no statistical significance in the blind listening results
All I hear with any MP3 is dynamic compression. I don't see how bit rate can make up for that. I'm not saying they can't sound pleasing but there is no dynamics. Anyone old enough to remember what dynamic recordings sound like I think will agree.