Question for Jan
Didn't you publish a multipart article (or post it online) a couple of years back with a DIY amp design on multiple PCBs that had vanishingly low distortion?
Sorry I can't pull it out of memory right now - brain tired from running math models all night and writing the above opus during compile times....
Didn't you publish a multipart article (or post it online) a couple of years back with a DIY amp design on multiple PCBs that had vanishingly low distortion?
Sorry I can't pull it out of memory right now - brain tired from running math models all night and writing the above opus during compile times....
Vishnu,
It is always a pleasure to read such stories from a junior
but one comment:
It is not that TIM, IIM etc are ignored. It is just that 'we' know it, know how it is created, and know how to avoid it. It isn't an issue, and pretending it is, does grave unjustice to any competent designer worth his salt.
I too respect JC for his accomplishments, but his incessant hammering on IIM, IMD etc to justify his aversion for anything with 4 legs or more is simply misleading and disrespectful.
Do we really have to believe that people who design opamps which are essentially distortion-free up to several MHz have no clue about these things?? Gimme a break.
Jan
It is always a pleasure to read such stories from a junior
but one comment:It is not that TIM, IIM etc are ignored. It is just that 'we' know it, know how it is created, and know how to avoid it. It isn't an issue, and pretending it is, does grave unjustice to any competent designer worth his salt.
I too respect JC for his accomplishments, but his incessant hammering on IIM, IMD etc to justify his aversion for anything with 4 legs or more is simply misleading and disrespectful.
Do we really have to believe that people who design opamps which are essentially distortion-free up to several MHz have no clue about these things?? Gimme a break.
Jan
Last edited:
Yes indeed, it was similar to the TA-N7 but the bootsrap devices were also BJT's (2N3773 iirc). The output devices were MJE200 + complement.
It in this book: Audio Amateur Power Amp Projects - KCK Media Corp.
Jan
Ed Simon has said that he can hear differences between op-amps operating at G=10. He has a project going with 10 sealed line amplifiers and said that so far other people are able to differentiate the op-amps at G=10 as well. He also said that when the study is complete he will have more to say publicly.
In addition, PMA organized a test of non-inverting unity gain op-amp buffers in a forum thread here at DIYAudio. I was given credit for sorting them in order by ear, although not perfectly error free, and not without great difficulty. I also listened primarily to only one very brief cymbal hit in order to do the sorting, which is obviously not normal music listening.
While these things don't constitute well established scientific evidence, they are suggestive that strongly worded statements to the effect that op-amps are completely distortion free to human listeners might someday be found to be a bit overstated.
Last edited:
Yes, of course. It is hard to alleviate it in discrete designs; Arto Kolinummi wrote (and I published) a whole book about it.
Lucky for us, in integrated parts can be very greatly minimized by design.
Jan
I would be happy to take a look if you would like to provide a reference to a specific post.
Also, I know JC sometimes states reasons he doesn't like the sound of op-amps in his circuits, and although the reasons he gives may not be correct, it may still be true he hears some very small, yet real distortion that he doesn't like.
I would be happy to take a look if you would like to provide a reference to a specific post.
Really.
I don't care what he hears, his arguments are misleading and bordering on dishonest. He knows that. You know that.
Jan
Yet he has stated repeatedly that the differences disappear (for him) when listening is unsighted. What are we to think of the stories?
Yes I remember that article. If you read the sidebox, he is quite critical (justified so) of the imbalances in the input stage. That lead to some harsh Letters to the Editor between Walt and Bongiorno (if I remember the correct name).
Jan
Yes and that is to him proof that controlled listening 'doesn't work'. The man has no concept of perception.
Jan
Sad to note that, after Walt put in all that effort they stopped selling the kit. I guess when even Walt blows a set of output devices its not quite ready for home assembly!
Having correctly heard very small differences myself, and having also been fooled occasionally myself, I believe sighted can help in the most difficult cases for hearing real differences (even if not necessary for some people), and and of course sighted is absolutely necessary for fooling one's self. So, it can go either way. Sighted could be necessary for him, as he says, or he could imagine difference sometimes. Probably some mix of both, would be my guess.
why you need to allow yourself to count above 4 - but carefully
regarding why a designer should fight
Because that excludes using precision matched arrays like the MAT-02s and MAT-04s, which are functionally 2 and 4 transistors in one package, respectively - more than 4 legs. Or using precision-matched arrays with 6 or 8 devices in one package. More legs still!
And yes, each of those 2,4,6 or 8 transistors are actually comprised of a plethora of paralleled devices, all on the same die, which some would cry foul about. "It's an integrated circuit!" No, just a bunch of identical transistors. But that's the point - you parallel that many devices fabbed on the same piece of silicon, that close together, and the math says the noise drops significantly - and measurements prove it.
Not to mention the advantages of the close thermal coupling.
And the fact that, without using such arrays, you have to do what whoever actually ends up assembling a product based on this concept, such as the Nakamichi 410 Preamp, has to do - when doing this with discrete transistors:
Get a whole lot of transistors from the same manufacturing run and tediously match and sort them. And you still can't use (or use as effectively) the little circuit tricks that can achieve truly astounding levels of performance.
Without such parts, try making a DC-controlled all-analog volume control or constant-volume panning circuit for stereo, or multichannel use, with a 90 dB or more adjustment range, with all channels accurately tracking, while maintaining THD at -112 dB, and a noise floor of -120 to -135 dB.
I'll wait here...but I won't hold my breath.
Or my personal favorite: an input-to-gain node topology that can be rigorously shown mathematically to cancel all terms in the distortion equations, up to the limits of the devices following the mathematical equations describing the laws of physics by which they work.
(Not the limit of their beta matching - the distortion is relatively insensitive to as much as 5% mismatch, which is guaranteed in off-the-shelf parts. Hand-picking for better matching is possible to less than 0.2%, at which point people will insist you're just showing off, the distortion is so low - at or below the quantum limits of the noise in the devices).
And no change of performance within a temperature range of, say, -25 F to +100 F.
But a little aside, here. Maybe not so little, but bear with me. Like one of Aesop's fables, there's a moral to the story.
There was (and probably still is) one particular 'boutique' designer, who there really is no need to actually name, who went running around the Bay Area in the mid 1980's hawking 'his' design of a 'discrete' op-amp replacement, because op-amps are bad, right? He even built them onto small custom circuit boards from which little legs stuck out the bottom, laid out so you could plug them into the PCB pad patterns of standard 8-pin op-amps. You know, so you could replace the evil op amps in a mass-produced preamp or line stage with the almighty discrete replacement. He made quite a point of going to audio expos, tech seminars or being an "invited Guest Speaker" at local audio enthusiasts' club meetings, always with a bunch of these for sale at $25 each. But you know how audio amateurs are - generally frugal. I let him sales-pitch me after an applications seminar, and pulled out $100 cash on the spot and bought 4 of them. A couple of reputable acquaintances looked at me like like I was balmy - and when I asked who wanted to go for drinks this guy chimed in, but everybody else politely declined, looking at me like they'd just watched me join a cult. I just winked at them, and told Derek Bowers I'd see him later at a pub we both enjoyed.
So I bought this guy a couple of drinks in the hotel bar, while I asked him questions about using the modules - what voltage rails would they tolerate, what about the compensation caps and gain-setting resistors in the original circuit layout, etc. Then, when I asked about the multi-legged TO-99 metal can buried in amongst the discrete transistors and 1% metal film resistors, he flat-out stated it was a 'transistor array'. "Oh - good - which one?"
He gave me a 'you don't really expect me to tell you that' look, but re-qualified his earlier assertion. It was a 'custom' part. So I put the two statements together, and point-blank, but innocently, asked "So it's a custom, matched transistor array, right?" and he said "yes", explaining that's why he had to charge $25. I told him I understood, paid the tab and left. The story has more details, but they're unimportant. I was doing apps consulting for Philips, Analog Devices, Burr-Brown and Crystal Semi at the time, as well as designing at NexGen, and didn't want what the guys had seen affecting their perception of my sanity or judgement, so I told them later when I ran into them that I bought them because I felt sorry for the guy. Which was true. I still do. I would get that same look from friends when I gave money to panhandlers on the street. Why give money away for free? Didn't I know that lady standing there with those two kids in the February drizzle at night was just putting on an act? Maybe, but if so, she was really living the part convincingly. Not an easy gig, either. Either way, she needed the money more than me. Call me gullible, but I have a soft spot for street performers. Except mimes. Something about the makeup, or costumes, or the fact that it requires no vocal talent or ability with a musical instrument. So sue me.
I ran into some guys from my audio club a week or so later at a restaurant, and they all wanted to know "how did they sound?". Well...
...hmmm....
About as well as an LM308 with the input pair replaced by external discretes, just like in one of National Semiconductors' application circuits for the part, was known to sound. Why it helps the performance, and the sound, is explained in the application circuit for the mod, written by the staff of National Semiconductor.Not bad - better than a 50-cent 741...just not worth $25 bucks. Why did I buy them? To keep my friends from being swindled.
And this guy really did sell the performance. Was it worth $100? You might think not, but knowing for sure what this guy was really like probably was.
He invited me and my girlfriend out to dinner with him and his girlfriend a while later. He'd heard a DAC I'd designed, and wondered if I wanted to do some design work for him, on a consulting basis. I wouldn't be paid right away, of course, until enough units sold to cover his costs in building them, and he was doing all the sales and marketing, demoing it to prospective customers and that involved his time and travel expenses, etc, etc. And I really didn't NEED the money up front, right? I had a good day job and all... but he WOULD need to see the schematic, and parts list, so he could evaluate what his costs to manufacture would be. And would I mind doing a custom PCB layout, to fit in the chassis he had in mind? And put the company logo in the silkscreen layer?
He would, of course, include that in what he was going to pay me, but again...later.
You could consider the LM308, technically, an 'array' of 'transistors'. The design WAS a kind of a 'custom configuration' - two of the transistors outside the can package, not inside. That was the 'standard' configuration.
John, I'm sure, remembers this guy, who actually managed to con a major, well-known high-fidelity company into paying him to design their first power amplifier product. And the end results were predictably poor, some would say disastrous, and the stuff of legend. It also made it extremely hard on independent consulting designers in the Bay Area for quite a while, until the dust settled, and some had managed to prove themselves with a track record of successful designs. Some, I know, had to go into other lines of work during the period the consulting gigs weren't available. A few never designed a commercial product again. A couple of them moved out of state, somewhere cheaper to live.
Maybe that's why John doesn't trust multi-legged parts.
regarding why a designer should fight
Because that excludes using precision matched arrays like the MAT-02s and MAT-04s, which are functionally 2 and 4 transistors in one package, respectively - more than 4 legs. Or using precision-matched arrays with 6 or 8 devices in one package. More legs still!
And yes, each of those 2,4,6 or 8 transistors are actually comprised of a plethora of paralleled devices, all on the same die, which some would cry foul about. "It's an integrated circuit!" No, just a bunch of identical transistors. But that's the point - you parallel that many devices fabbed on the same piece of silicon, that close together, and the math says the noise drops significantly - and measurements prove it.
Not to mention the advantages of the close thermal coupling.
And the fact that, without using such arrays, you have to do what whoever actually ends up assembling a product based on this concept, such as the Nakamichi 410 Preamp, has to do - when doing this with discrete transistors:
Get a whole lot of transistors from the same manufacturing run and tediously match and sort them. And you still can't use (or use as effectively) the little circuit tricks that can achieve truly astounding levels of performance.
Without such parts, try making a DC-controlled all-analog volume control or constant-volume panning circuit for stereo, or multichannel use, with a 90 dB or more adjustment range, with all channels accurately tracking, while maintaining THD at -112 dB, and a noise floor of -120 to -135 dB.
I'll wait here...but I won't hold my breath.
Or my personal favorite: an input-to-gain node topology that can be rigorously shown mathematically to cancel all terms in the distortion equations, up to the limits of the devices following the mathematical equations describing the laws of physics by which they work.
(Not the limit of their beta matching - the distortion is relatively insensitive to as much as 5% mismatch, which is guaranteed in off-the-shelf parts. Hand-picking for better matching is possible to less than 0.2%, at which point people will insist you're just showing off, the distortion is so low - at or below the quantum limits of the noise in the devices).
And no change of performance within a temperature range of, say, -25 F to +100 F.
But a little aside, here. Maybe not so little, but bear with me. Like one of Aesop's fables, there's a moral to the story.
There was (and probably still is) one particular 'boutique' designer, who there really is no need to actually name, who went running around the Bay Area in the mid 1980's hawking 'his' design of a 'discrete' op-amp replacement, because op-amps are bad, right? He even built them onto small custom circuit boards from which little legs stuck out the bottom, laid out so you could plug them into the PCB pad patterns of standard 8-pin op-amps. You know, so you could replace the evil op amps in a mass-produced preamp or line stage with the almighty discrete replacement. He made quite a point of going to audio expos, tech seminars or being an "invited Guest Speaker" at local audio enthusiasts' club meetings, always with a bunch of these for sale at $25 each. But you know how audio amateurs are - generally frugal. I let him sales-pitch me after an applications seminar, and pulled out $100 cash on the spot and bought 4 of them. A couple of reputable acquaintances looked at me like like I was balmy - and when I asked who wanted to go for drinks this guy chimed in, but everybody else politely declined, looking at me like they'd just watched me join a cult. I just winked at them, and told Derek Bowers I'd see him later at a pub we both enjoyed.
So I bought this guy a couple of drinks in the hotel bar, while I asked him questions about using the modules - what voltage rails would they tolerate, what about the compensation caps and gain-setting resistors in the original circuit layout, etc. Then, when I asked about the multi-legged TO-99 metal can buried in amongst the discrete transistors and 1% metal film resistors, he flat-out stated it was a 'transistor array'. "Oh - good - which one?"
He gave me a 'you don't really expect me to tell you that' look, but re-qualified his earlier assertion. It was a 'custom' part. So I put the two statements together, and point-blank, but innocently, asked "So it's a custom, matched transistor array, right?" and he said "yes", explaining that's why he had to charge $25. I told him I understood, paid the tab and left. The story has more details, but they're unimportant. I was doing apps consulting for Philips, Analog Devices, Burr-Brown and Crystal Semi at the time, as well as designing at NexGen, and didn't want what the guys had seen affecting their perception of my sanity or judgement, so I told them later when I ran into them that I bought them because I felt sorry for the guy. Which was true. I still do. I would get that same look from friends when I gave money to panhandlers on the street. Why give money away for free? Didn't I know that lady standing there with those two kids in the February drizzle at night was just putting on an act? Maybe, but if so, she was really living the part convincingly. Not an easy gig, either. Either way, she needed the money more than me. Call me gullible, but I have a soft spot for street performers. Except mimes. Something about the makeup, or costumes, or the fact that it requires no vocal talent or ability with a musical instrument. So sue me.
I ran into some guys from my audio club a week or so later at a restaurant, and they all wanted to know "how did they sound?". Well...
...hmmm....
About as well as an LM308 with the input pair replaced by external discretes, just like in one of National Semiconductors' application circuits for the part, was known to sound. Why it helps the performance, and the sound, is explained in the application circuit for the mod, written by the staff of National Semiconductor.Not bad - better than a 50-cent 741...just not worth $25 bucks. Why did I buy them? To keep my friends from being swindled.
And this guy really did sell the performance. Was it worth $100? You might think not, but knowing for sure what this guy was really like probably was.
He invited me and my girlfriend out to dinner with him and his girlfriend a while later. He'd heard a DAC I'd designed, and wondered if I wanted to do some design work for him, on a consulting basis. I wouldn't be paid right away, of course, until enough units sold to cover his costs in building them, and he was doing all the sales and marketing, demoing it to prospective customers and that involved his time and travel expenses, etc, etc. And I really didn't NEED the money up front, right? I had a good day job and all... but he WOULD need to see the schematic, and parts list, so he could evaluate what his costs to manufacture would be. And would I mind doing a custom PCB layout, to fit in the chassis he had in mind? And put the company logo in the silkscreen layer?
He would, of course, include that in what he was going to pay me, but again...later.
You could consider the LM308, technically, an 'array' of 'transistors'. The design WAS a kind of a 'custom configuration' - two of the transistors outside the can package, not inside. That was the 'standard' configuration.
John, I'm sure, remembers this guy, who actually managed to con a major, well-known high-fidelity company into paying him to design their first power amplifier product. And the end results were predictably poor, some would say disastrous, and the stuff of legend. It also made it extremely hard on independent consulting designers in the Bay Area for quite a while, until the dust settled, and some had managed to prove themselves with a track record of successful designs. Some, I know, had to go into other lines of work during the period the consulting gigs weren't available. A few never designed a commercial product again. A couple of them moved out of state, somewhere cheaper to live.
Maybe that's why John doesn't trust multi-legged parts.
Last edited:
- Status
- Not open for further replies.