Daniel, I understand and I would have agreed completely some years ago.
However, now I find there are some nagging issues. It is uncontroversial that some people can hear undithered 16-bit audio. In addition to replacing quantizing noise with random noise, dither exposes a few bits of information buried in noise below the 16th bit. Some of that is below -100 dB relative to full scale. Of course not all music is constantly at full scale either. But however one cuts it, for these and other reasons I can no longer agree with the point of view you so ably expressed, at least under all listening conditions.
Having said all the above, for normal listening in many situations most people are probably unlikely to notice very small imperfections in sound reproduction. For those types of situations, such as in a noisy night club, or when electronics have been engineered to mask distortion with noise from circuit resistances, I would expect it to be about like you say.
However, now I find there are some nagging issues. It is uncontroversial that some people can hear undithered 16-bit audio. In addition to replacing quantizing noise with random noise, dither exposes a few bits of information buried in noise below the 16th bit. Some of that is below -100 dB relative to full scale. Of course not all music is constantly at full scale either. But however one cuts it, for these and other reasons I can no longer agree with the point of view you so ably expressed, at least under all listening conditions.
Having said all the above, for normal listening in many situations most people are probably unlikely to notice very small imperfections in sound reproduction. For those types of situations, such as in a noisy night club, or when electronics have been engineered to mask distortion with noise from circuit resistances, I would expect it to be about like you say.
Last edited:
As far as I'm concerned this is analogous to the low order distortion vs crossover at the same THD level issue. At low levels 16 bit undithered has gross THD and a transfer function with multiple abrupt discontinuities especially at zero (i.e. horrible crossover).
Problem identified, problem fixed, end of story. Do you know any serious producer or recording engineer that works at 16 bit undithered?
Any idea how those undithered 16 bit tests were run? It very much changes the landscape. It's dangerous to generalize that to mean we can hear in the range of -96 dB in any way, shape, or form. I mean, yes, it point to the lack of overarching utility of THD, but I think we can easily find ourselves in a situation where someone is doing the equivalent of my 1000W vs 1W amplifier example (essentially 10 bits!). We can contrive situations to bring it out, but when digging down a like for like, in more realistic circumstances, these differences fall apart. E.g. flipping between two decently specced opamps where the difference is in the range of discussion.
There was a lot of research done in this area in the late 80's - early 90's. Lots of papers published in the JAES (Journal of the Audio Engineering Society) and in app notes from Crystal Semiconductor, Analog Devices, etc. Rigorous double-blind studies, optimization parameterizations...you can Google it, or in my copious free time 🙄 I can look for some titles in my research library.
Yeah, unfortunately my access to JAES expired with my previous .edu email account. And the Google(s) outside of scholar (see: JAES) has been less than fruitful for getting to the rigorous stuff.
If you have stuff on quick draw, that's cool, if it requires digging, not worth the effort. More curious how effectively all the 16 bits were being utilized (as to elucidate the -96 dB figure of merit).
If you have stuff on quick draw, that's cool, if it requires digging, not worth the effort. More curious how effectively all the 16 bits were being utilized (as to elucidate the -96 dB figure of merit).
Being third in command, he was far enough down the totem pole that to offer corrections to his superiors' spotty English would just NOT be done. Since the two I dealt with were father and son, things we less...um...'formal'. I had no problem, on the other hand, letting my boss (the president of SRS) know, before the meeting, that he let me handle it so he didn't blow it by accidentally doing something stupidly rude.
As for the PMI takeover - there was actually one before that, when PMI acquired SSM (Solid-State Microtechnology for Music). I always wondered how those went. I left NexGen when AMD bought us because I was a systems-level designer, not a chip designer, and did not want to be relegated to uninteresting work (and I guess the fact that I told the new Prez of NexGen that he could f**k himself - as a Moslem, he really didn't like that
- didn't help much). Besides, my stock options were 96% vested - so I took a year off, just riding my MotoGuzzi around and hang-gliding in Santa Cruz. People were just starting to surf Mavericks, and I was going to jump into that when I got tapped to join the startup of SRS. But how did the PMI takeover go down? and the SSM buyout/merger?
Last edited:
Vishnu -- JAES stuff would need to be .pdf's, but whatever is easiest, really.
Feel free to PM and we can make an honest take of it all. And thanks!
Feel free to PM and we can make an honest take of it all. And thanks!
SSM was very short money, I never got past that they were a bunch of Derek Bower's buddies or some such (before the PMI thing so I have little or no knowledge). The PMI thing was classic the new boss boots the old boss and East coast conservatism IMO. Hard feelings on both sides still linger (and rumors) after all these years and I can't share some of the personal stories in a public forum. One thing though anything that could be construed as royalties is off the table at ADI. I have some stories about ex-AMD folks that would land me in court if I shared them.
BTW your praise of Bedini and Otala baffle me (for different reasons).
Last edited:
The NexGen thing was the same - new boss boots out old boss, and I can't share some of those stories for the same reasons. But Atiq Raza - the new boss - has touted himself in multiple public arenas as one of the founders of NexGen.
Funny - he and I were hired the same day, employees number 38 and 39.
The one regret I had about not transitioning over to AMD was Jerry Sanders' epic million-dollar Christmas parties - but I had enough friends at AMD, even before the buyout, that tickets were no problem. But it got frosty indeed when I ran into my ex-colleagues - loud music, alcohol, and having been pigeonholed into cubicles did not improve their mood at all. And all the ones who were my friends had done what I did - took their stock and pulled the ripcord. I just went cuz my GF loved to watch nerds party like rockstars.
As for the Otala / Bedini "praise" thing. I could explain that, but maybe this isn't the place. And as far as praise goes, Derek is up there for me, as are Robert Adams and Malcolm Hawksford, Edward M. Cherry, Gunther Theile..
Where can I find a fairly comprehensive comparison of OPS configurations and their distortions?
THx-RNMarsh
THx-RNMarsh
Hi Richard,
Probably the most honest comparisons would be what you can see and poke at on your own bench. I'm not trying to be smart about this. You could always obtain various pieces of equipment that represent each configuration and simply test them. Minimal work needed on your part, and the test subjects can be sold afterwards, or given back if they were loans.
The Nakamichi 620 amplifier is a good representative of a power diamond buffer. I'm suggesting it because there isn't a lot of changes or modifications over the basic design, not because I might have an affinity for it or not. You probably have a very good idea of which amplifiers represent the purest form of whatever output configuration they represent.
-Chris
Probably the most honest comparisons would be what you can see and poke at on your own bench. I'm not trying to be smart about this. You could always obtain various pieces of equipment that represent each configuration and simply test them. Minimal work needed on your part, and the test subjects can be sold afterwards, or given back if they were loans.
The Nakamichi 620 amplifier is a good representative of a power diamond buffer. I'm suggesting it because there isn't a lot of changes or modifications over the basic design, not because I might have an affinity for it or not. You probably have a very good idea of which amplifiers represent the purest form of whatever output configuration they represent.
-Chris
Chris ---- yes... and I have started to do this.... but then thought someone surely has done this by now.
Can you put up that diamond OPS for me --- so I can see values and xistor numbers.
THx-Richard
Can you put up that diamond OPS for me --- so I can see values and xistor numbers.
THx-Richard
I'm the guy who put the 100u x 100u hammer and sickle logo on the NexGen/AMD "K6" microprocessor rev.A. Atiq Raza thought it was kinda funny, until the Real Muricans in the Texas fab screeched and hollered with righteous indignation. Then he gave me a stern talking-to. There's my name right on the JSSC paper.
_
_
Attachments
We had a guns and ammo layout guy that put and American flag on a chip and one of the stars was over the wrong layer and shorted the rails. All decorations of layouts was banned after that.
Big uP projects lavish megabuxx on CAD tools, and even in '96 there were super extensive DRCs and LVSs that worked very hard to discover chip-killing layout errors. Still, fearing for my life, I put the logo on Metal-4 only, and made sure every one of its polygons were connected to VSS.
It was a fun program to write actually: TrueType font file (suffix .ttf) + ascii text file ---> Cadence Virtuoso polygons. SKIL code. Ah the memories.
I see Ravi's name on that paper. We used to go to lunch together at this all-you-can-eat buffet-style restaurant. He would PILE his plate on the first trip through...I told him "You can make a 2nd or third trip..." and he explained that growing up, there were like 14 kids in the family, and if you didn't grab it fast at chow time, it was gone.
As far as Atiq...the less I say about him, the better...
- Status
- Not open for further replies.