Hi Waly,
Maybe I am missing something here, but the only spec. for distortion I can find is with supplies of 5 V and a 10K load. I do not see anything that gives me a higher load impedance in that data sheet. A bipolar supply of 8 VDC and higher load impedance will develop lower distortion numbers. I have measured distortion lower than 0.4% in equipment that used a 4066 for switching, and the 4966 (higher supply voltages) worked similarly as well. I can't remember what the exact distortion numbers were now, but drive it with a low impedance source and a high impedance load (100K is fine) and you will measure lower distortion.
One day I'll have to set this up on a breadboard and test it.
-Chris
Edit: I see, you're equating a transfer curve to distortion performance. Until I see a distortion vs load Z plot, I'm not convinced. Not even a little bit.
Maybe I am missing something here, but the only spec. for distortion I can find is with supplies of 5 V and a 10K load. I do not see anything that gives me a higher load impedance in that data sheet. A bipolar supply of 8 VDC and higher load impedance will develop lower distortion numbers. I have measured distortion lower than 0.4% in equipment that used a 4066 for switching, and the 4966 (higher supply voltages) worked similarly as well. I can't remember what the exact distortion numbers were now, but drive it with a low impedance source and a high impedance load (100K is fine) and you will measure lower distortion.
One day I'll have to set this up on a breadboard and test it.
-Chris
Edit: I see, you're equating a transfer curve to distortion performance. Until I see a distortion vs load Z plot, I'm not convinced. Not even a little bit.
And with what should I equate the distortions, if not the transfer curve, with the shoe sizes of the design team?
Although you are indeed completely missing my point, rest assured I don’t need to convince anybody in particular, you are safe in your beliefs. Those who can read a data sheet and have an idea how a cmos transfer switch works will know the truth, anyway.
BTW, the Nak Dragon distortions are spec’d at 0.8% on metal tape and 1% on other tape types. Good design decision to use the el cheapo 4066 analog switches in the signal path.
Nakamichi Dragon - Manual - Three Head Auto Reverse Cassette Deck - HiFi Engine
Last edited:
Didn't Earl Geddes say something about THD not being related to how we hear? He said he could provide an example of 0.1% THD nobody could hear, and one anyone could hear.
So, what kind is it for magnetic tape and CMOS analog switches, or to put it another way, where does each lay along that audibility continuum?
So, what kind is it for magnetic tape and CMOS analog switches, or to put it another way, where does each lay along that audibility continuum?
Last edited:
Almost anyone average competent can provide such examples. I was hoping we no longer have considered distortion a single number. This is going in circles, as everything here, that's why the thread had become boring with not too many relevant inputs.
With all due respect Mr. Geddes speaks the obvious in this case. 0.1% crossover buzz (for instance) vs 0.1% pure 2nd harmonic is just too simplistic. It's basicly a discussion for the 70's, today an amplifier with noticeable crossover should simply be considered broken.
... or Ultra Highend aspiring for 'AAA' rating
darTZeel NHB-458 monoblock amplifier Measurements | Stereophile.com
darTZeel NHB-108 Model One power amplifier Measurements | Stereophile.com
Hi Scott,
Hi Waly,
I'll have to put something together to actually measure the distortion.
What you are missing is that the scale of the transfer curves is linear, and a tiny shift, unseen at that scale, is enough to reduce THD from 0.4% to a far lower number. In order for that graph of transfer characteristics to be useful to you for this purpose, it would have to be an expanded scale to separate your 0.4% THD from the lower numbers. I think most of us could agree on that.
Discussing this in theoretical terms will go round and round forever. At some point someone (maybe me) will have to actually measure the distortion. It will indeed be a worst case measurement as it will happen on a breadboard with loads of stray capacitance and iffy connection resistance. But then, you would probably argue that as well, won't you?
-Chris
Yes, absolutely!
Hi Waly,
I'll have to put something together to actually measure the distortion.
What you are missing is that the scale of the transfer curves is linear, and a tiny shift, unseen at that scale, is enough to reduce THD from 0.4% to a far lower number. In order for that graph of transfer characteristics to be useful to you for this purpose, it would have to be an expanded scale to separate your 0.4% THD from the lower numbers. I think most of us could agree on that.
Discussing this in theoretical terms will go round and round forever. At some point someone (maybe me) will have to actually measure the distortion. It will indeed be a worst case measurement as it will happen on a breadboard with loads of stray capacitance and iffy connection resistance. But then, you would probably argue that as well, won't you?
-Chris
Millennia Media across the canyon from us. John has been at it probably thirty years. Chances are we have all heard something through one of his mic amps.
Sure, easy to agree on that.
However, I just mentioned that to set some context for discussion about the fact that the metric in question, THD, has been known to correlate poorly with hearing for a long time, yet here at DIYAudio and out there in the world of commercial manufacturing it's still the de facto standard.
People continue to use it as though its more useful than it is for predicting audible and/or objectionable distortion. Including, of course, for tape machines.
It's interesting to me that some people can find that really good tape sounds great, but really good digital sounds merely good. Obviously, they must be referring to some subjective experience, not measured accuracy. Yet, enjoyment of music is in the end what it's all about.
Since the 70's was so long ago somehow it seems we should be making more progress than we have in finding some metric that will let us move away from THD to something that's a better fit for matching engineering specifications to how people hear the end results of our engineering efforts.
Last edited:
No People use it because they are expected to publish it. I don't think anyone on here who is has any clue takes THD seriously, not least because there are a handful of amplifiers published that are nudging single digit ppm distortion. These wonderfully pointless beasts are designed because people can for the sheer fun of it.
I'm curious which "analog design guru" from NAK you met in '88. When I was working at SRS Labs in 92-93, two "guys" from NAK came by, the president and his son, who was VP of Engineering. The son spoke flawless English; dad, not a word. Nak senior went nuts when he saw I was using a NAK preamp on my bench - he had designed it when he was an up-and-coming design engineer in the company. It was at that time an entirely family-owned affair, and HIS father, the founder, made him work his way up, just like he was doing with his son. With his son translating, he asked why I was using that preamp, and I replied "Because it's a perfect preamp." Very close - multiple low-noise bipolars paralleled right at the input, to drive the noise floor into oblivion, just like Analog Devices (PMI) did with the MAT-02 and MAT-04. What amazed me was that he could tell just by listening that
i'd modded it, and insisted on seeing the schematic for the mod. It actually did improve the performance, and that intrigued him. He asked (again, through his son) why I modded it, and (careful of the Japanese faux pas of being seen as criticizing) I said "To make it MORE perfect." That pleased him to no end, and SRS got the licensing deal then and there. And after I left SRS, they asked for my expertise on a consulting basis, up until I moved to Canada to work for QSound. Where I bumped into Mark Alexander (again) when he came by (working for Crystal Semiconductor) to look at licensing QSound tech.
Due to the nature of a voltage bias in the majority of modern class AB amps, dynamic distortions near the zero crossing are all over the map, and their spectrum is modulated by envelope.
Do you have any measurements and quantifications of this phenomenon, performed on a well designed SOTA class AB amplifier? This must be easy to measure. If you only verbally declare your thoughts on solid state class AB flaws, the same way you may declare anything.
The mid-1980's Nak pieces were high performing for the time, for sure, we had been Nak dealers since the early 1980s and had at least one of everything (wish I still had those Stasis amps!), and tens of 1000ZXLs, Dragons and MR-1s.
My interactions with Nak were largely through Dolby, where I spent a bunch of time working on improvements to high-speed duplicated cassette performance and implementing HX for high-speed replicators. I remember working with a young Japanese engineer, cannot remember his name who had perfect English. With Nak we were finalizing mods to the MR-1 intended for high-speed dupe QA use, and ended up being the shop that did the mods to several hundred MR-1s.
I'll bet Nelson Pass has some good Nak stories considering he licensed tech to them in this period...? hint hint
In some ways, Mark, I'm really surprised you keep flogging this without looking more carefully at the subtleties of the position.
It's easy to make the argument that THD is pretty meaningless when we're talking -60 dB THD+N line-level. The harmonic components are, in essence, huge, and where they sit spectrum-wise is going to have a fighting chance of making an audible difference. Dr. Geddes makes a very safe argument here by claiming he can, by switching around the harmonic content of the distortions from one extrema to another, make something audible. He doesn't make that argument for lower levels of distortion, where you might well have the same levels of higher order harmonics as your "pure HD2 -60 dB THD" previous test case did.
The funny thing is the -60 dB THD mark is roughly where the literature stops. While certainly the threshold on hearing is blurry and human-dependent, it's falling off fast somewhere around there, which makes complete sense especially when you look at how performance across a species is conserved (healthy vs healthy, you really don't get 10x "better" humans at anything) Your own trials with various opamps in PMA's test bears this out, where it took you an extremely contrived test to unreliably/questionably differentiate between errors sitting around the -75 dB mark or so in diffmaker. (that's not a sign of disparagement on your behalf, but highlights the difficulty and meaning of your one-off test)
That argument looks a lot less solid when you start talking about -80 THD+N and really goes off the rails when it gets down towards the -100 THD+N level or lower. There just isn't any energy *left* in the harmonics worth discussing. Even if THD is itself a poor metric, it's safe to say that a very low THD is going to correlate with very low distortion by any possible test. Sure, we can take a distraction by saying a -100 dB THD near-full power 1000w amp might sound different (within their common output range and below clipping) to a -100 dB THD near-full power 1w amp, but under more equal comparisons, these ultra-low distortion electronics (especially when we're talking line-level like-for-like) have all converged to the same output.
One can be entirely comfortable with Dr. Geddes's assertions and argue that audible differences at -80 dB or smaller are specious at best. Which is why you see a huge amount of push-back when people claim to hear huge (or really any) difference between two generally equal pieces of electronics that are both sitting at -90 dB THD or less. What differences are there legitimately left to hear?
As to whether a cmos switch that looks like it has ~1/10 the overall media's distortion is sufficiently masked, I don't know. I can see where Waly's going with the argument, even if I'm willing to give wiggle room. If it were a -80 dB switch in a -60 dB error media, then I'd be pretty firmly in agreement with him.
I have long forgotten names and exact titles but highly probable it was the son. They were in a fairly remote location, a group of school children made quite a fuss as we walked by embarrassing their teacher.
Funny thing the only other engineer with perfect English that I met only talked after the meeting. Apparently he was like third in command at the table and was not allowed to speak even though the meeting had been painfully difficult due the the poor translations.
BTW I met Mark but knew him briefly, the PMI takeover was not that smooth.
Last edited:
- Status
- Not open for further replies.