Hi everybody, If somebody out there has this actual issue of 'Stereophile' April 1992, could you please put the review article up on the HCA2200. I have seen the review, but I have lost track of the actual review. I am pretty sure that there was little or no prejudice against IC's regarding this amp, but I could be wrong.
Well, I found the issue and article. More later, but I was forced to fix the unit, before it could be successful in future. Bob Harley found great measurements, (a little noisy) but lousy overall sound. Why? I took a chance and removed the IC, and voila !
Well, I found the issue and article. More later, but I was forced to fix the unit, before it could be successful in future. Bob Harley found great measurements, (a little noisy) but lousy overall sound. Why? I took a chance and removed the IC, and voila !
Attachments
Last edited:
Sure, agreed. But some people may not have heard the news. And since we can hear that, it might not be too surprising if it were to turn out we could hear other types of distortion at more or less similar levels. It would be interesting to find out how many people can hear things a little harder to discern.
It comes down to our being able to 'listen through' the noise, and is to be expected imho.
When 'listening through' the noise, the nature of the system noise and consequential signal noise artifacts are of critical importance.
By reducing and tailoring system noise, very low level information becomes available.
This is nothing new, indeed very good design achieves most of this, but being able to do it at will (instant retrofit) is very much so.
Dan.
Mark, that's fine. At the same time, I feel you keep moving the goal posts. And again, I'd have to ask what you mean by undithered 16/44 -- I mean if you took an adc that was linear throughout its range and sampled a vinyl record at 16 bits, you'd have natural dither that would mask any LSB issues. All of which is well covered by Vanderkooy and Lipshitz in 1984.
I'm not taking an extreme position that I completely reject all understanding because we cannot say with 100% confidence, but in the audio world it does appear my stance is rather hard lined, because, well, we're pretty bereft quality research. And the breadth of fantastical claims is rather wide. Likewise, a lot of claims ride the line of plausibility, which demands serious effort to tease apart.
Take some of the more outrageous cases made here in this very thread. Or the plenty of rationalizations that straight up violate core tenets of physics. I don't even have to jump to a different page on this thread to bring up an example. And that includes getting a bunch of folk haphazardly in a room with a bunch of amplifiers and zero controls and saying "this one is best". Even if there's consensus, it, at best provides a hypothesis (as Demian rightly corrects me). Most inquiries will turn null under closer examination, that is just the nature of science.
With regards to heresay, I could make up my listening impressions of XYZ whole cloth, as could you. I don't want you to trust my listening impressions, especially if I don't give you my methodology. We've got a number of rather amusing cases here on DIYAudio where someone *clearly* heard the difference between two bitwise identical files with ever-so-slightly different metadata (date and filename IIRC). This is the world we unfortunately live in. If you give a scope trace or some sort of harder data, at least I have a fighting chance of reproducing it, and vice versa.
I'm not assuming you, or anyone else is making stuff up -- I do truly assume people are acting in good faith. I also do truly believe human subjects are very unreliable. Most fields are ripe with irreproducible studies, and that's under the umbrella of academia where data is supposed to be better attended. No one in my lab group could reproduce the results I got in my most cited paper, and that is a purely semiconductor-based topic with clear end points. So whatever magic I was doing in my process left with me even though I left meticulous notes and followed up with my colleagues to troubleshoot. That's a field that's a far cry from human performance or human preference, where the natural noise in the tests is high and the ability to inadvertently "flavor" a test is quite easy. I realize it's fashionable to flog on how many of the landmark psychological experiments have been disproven or at least shown irreproducible, but it shows clearly the difficulty of the work, and sets the bar for test conditions quite high.
This is an overarching scientific skepticism that permeates my work as much as it does my play. Again, I don't understand why this is controversial if you're remotely interested in the truth versus preserving the world view you live in. I've cited Ioannidis enough times to hopefully give a clue where I'm coming from.
Well NP is a top designer all right, but he fails miserably to succeed in his repeated suicide attempts
Well, you (and Dan) are aware that I can’t hear undithered 16/44, at least with my audio equipment (and I wouldn’t blame the equipment for that).
http://www.diyaudio.com/forums/everything-else/169484-what-wrong-op-amps-475.html#post4954936
http://www.diyaudio.com/forums/everything-else/169484-what-wrong-op-amps-476.html#post4955946
Thank you for the kind invitation. But you are living a bit faraway from where I am, probability for doing it is really low
George
The keyword here is 'tenet'.
Physics is borne of observations, and the maths describes these observations, at least in over view and can enable predictions.
However, not all is described and there are still holes in the knowledge.
Gravity, the origin of the universe etc are not explained yet.
You have it correct.
Six people have reliably heard, identified and described these difference on a range of playback systems.
I do not yet have explanation but I do have suspicion that it is down to error correction processing inbuilt in Flash memory.
The other side of the coin is to start with observations, and then derive testing methodologies that measure and prove such subjective observations.
Yeah, I'm dragging the chain on coming up with hard data, but regardless the subjective evidence is unequivocal on a wide range of applications.
I have my GF moving in today, and this will free me and push me to get on with the objective testing that I have so far not rigorously and peer repeatably pursued.
If my theories/rationalisations are in error so be it, that does not change the fact of the repeatable observations and the usefulness of my device.
If my theories are correct and proven then it's time to roll these understandings out for the benefit of all...call it majorly disruptive in some ways.
Dan.
Well NP is a top designer all right, but he fails miserably to succeed in his repeated suicide attempts
George
....so have others.... my key word to focus on is ALL. NP included.
THx-RNMarsh
To me, I wouldn't say it's a matter of moving goalposts, rather it's just an ongoing, back and forth conversation, IMHO. For example, I couldn't ask you every possible question I might have about your position in my first reply to you because I don't know what you are going to say next, or how you may further describe or clarify your position. New questions may arise in the normal course of conversation and discussion.
However, for the most part I think we do agree on most of what you just said. We seem to agree that sometimes people are mistaken about what they think they hear. In particular, when people talk about subtle differences in vague terms like "sound stage" or "depth," I tend to share your concerns, especially since perception of those things may be strongly affected by small changes in level.
On the other hand, perception of some sounds is much less affected by any level differences, and instead primarily involves recognizing the presence or absence of something very distinct. Undithered bit-depth reduction would be something more of that nature. As an arbitrary example anyone could probably understand, it would be unlikely for for someone to mistake an orchestral triangle ping for a tympani hit with a felt beater, or maybe to mistake the triangle sound for that of a sustained string section chord, due to some result of imagination, mood, psychological priming, small level changes etc. The sounds are just too different in nature to be easily mistaken, at least in most normal situations, including at pretty low level playback.
Therefore, I would be very reluctant to lump all perceived differences in sound to be equally at risk for mistaken identity. Most likely, I think you would probably agree more if you someday had a chance to hear for yourself some things that might not be distinct coming from your laptop sound system, but would be very distinct if you had the opportunity to listen in a well-equipped mastering room.
Anyway, getting back to the question at hand in reference to dither or the lack thereof, perhaps we could say we are talking about bit-depth reduction from a high quality, low distortion 24/44 source to 16/44, in order to produce an audio CD?
Last edited:
Maybe there could be reason to blame your equipment. Some people do their listening on systems that do not perform as well as the specifications might suggest.
For one example, some computer built-in sound cards claim to be hi-rez or hi-def quality, but if you look specs for the headphone output, it is rated at maybe -68 dB THD+N. It might in fact be possible to play hi-def content, but not with what I would consider to be fully hi-def resolution.
For another example, people usually don't know much about the exact distortion and time domain performance of their speakers. Many speakers have rather poor transient response, and long decaying waterfall plots. Some of those types of things may make speakers with very nice looking frequency responses still not suitable for hearing some kinds of issues in source material.
Last edited:
You're addressing the wrong person, I never thought THD was much of a factor in much of anything. I was simply repeating others stated preferences however they arrived at them.
To be clear my comment was in reference to certain early CD's where A/D flaws would have been cured with properly applied dither, certainly not vinyl rips which in most cases are noise dithered.
And that brings another subject, the nature of the vinyl noise and it's dithering effects....noise shaping ?.
That said, I have plenty of vinyl rips that sound pretty good, BUT there is variation.
Dan.
No demons in the grooves doing maths just random noise and yes there is often some peaking somewhere in the passband but as George demonstrated it is also often that the noise floor of the mic mix is considerably above the raw floor of the vinyl.
I plotted a histogram of the sample to sample difference on a silent groove and found a minimum of 6 bits of dither at the 24 bit level (why would anyone record original material at 16 bits anymore)? The histogram very closely matched a random Gaussian.
Attachments are from a synthesized signal in 44.1K/16bit format (worst possible case), generated and recorded in software (best possible case in terms of abysmal low noise floor and non existent added distortion).
Data (8190 freq points) imported into Excel for viewing.
The question is, under real life acoustic recording conditions what a good ear can pick up some x dB beneath the noise floor in a commercial recording ?
The quietest HR recordings I have analyzed sofar, have a noise floor (mics on, silent section before music starts) of –130dB FS @20KHz, -125dB FS @10KHz, -115dB FS@1KHz, -90dB FS @100Hz, and anything btn –90dB FS to –50dB FS @ 20Hz depending on the recording space
And some related literature
http://www.ece.rochester.edu/courses/ECE472/resources/Papers/Lipshitz_1992.pdf
http://drewdaniels.com/dither.pdf
AES E-Library Noise: Methods for Estimating Detectability and Threshold
George
Attachments
Scott, that's a very good method for checking noise.
Now, 6 bits is the minimum you found.
Can you look how many bits was the mean of the sample to sample difference please?
Noise occupying the 6 lowest bits in a 24bit recording, is approximately –105dBFS noise level, phenomenally low noise for vinyl.
http://www.diyaudio.com/forums/atta...09d1488491256-digitizing-vinyl-level-bits.png
George
George, the best test is the decay of a pure piano note into silence, the problem is un-dithered at few bits full scale. You can take a chance and cheat a little by turning up the volume at the right moment. It becomes painfully obvious.
- Status
- Not open for further replies.