Hugh wrote:
I am really surprised you wrote that, unless you were simply trying to be agreeable as is your manner, since you are someone who is reliant on listening to drive your design decisions.
I'm not sure what you mean by "reflects only high end market realities".
Brian
Hugh, your response is the antithesis of my design values. When people start appraising only with spectrum analysers and network analysers and stop asking "What's it sound like?" then they are on a well travelled road to failure.
I am really surprised you wrote that, unless you were simply trying to be agreeable as is your manner, since you are someone who is reliant on listening to drive your design decisions.
I'm not sure what you mean by "reflects only high end market realities".
Brian
Nelson wrote:
😎
Nevertheless, from what I know of what Pass Labs has made public about its designs I somehow feel that you would consider it naive of me were I to assess your products only by their THD+N measurement as has been prescribed in this thread.
😎
Jcx, that was quite funny. 😀 I may have chosen to read a little more between the lines than you have. 
Brian,
I'm not that opaque; I'm merely being ingenuous - we Australians love to say one thing but mean quite the opposite. It is a benign, comedic form of sarcasm.
With respect, you underrate my design approach!!
I am simply saying that if distortion, distortion spectra and THD are all exemplary, then it should be possible to predict perfect audio reproduction. The fact this is apparently not always so would point to gaps in our knowledge, and the 'What's it sound like' mindset would tend to support this, despite specs being used almost universally to sell equipment, such as the Halcro, for example.
I apologise if this medium did not convey my wink.......
Cheers,
Hugh
Yes. 🙂 The way I would put it is that the correlation between sound quality and the THD measurements that are generally being used is poor. These particular measurements are not comprehensive. That is not to say that comprehensive measurement is not possible. Is that a fair summary?
It is my personal experience of correlating bench measurements with sound quality and design choices with sound quality that leads me to ask these unpopular questions.
It is commonly held that the science of measurement and it's correlation with sound quality represents the highest abstractive skillset in audio design.
I cannot disagree with this, but would add that an even higher abstractive skill is the multivalent correlation of topology, component choice, layout and subsequent measurement with sound quality. But this requires huge empirical experience, which not everyone aspires to as measurement is generally more appealing than wiring up prototypes.
The fact is that the customer makes a decision when he buys. A good proportion will choose entirely on their perceptions, which are influenced by a host of factors, but which do give lip service to a listening test, regardless of its subjectivity. To attack this position from a purist standpoint is probably legitimate, but it ignores the commercial reality. And this is where marketing and engineering diverge, as not everyone looks for, or indeed hears, the same things.
I don't know if this helps, but I would be willing to bet this is 1 magnificent sounding amplifier, even if, as a general rule, I shy away from EC regimes.
Cheers,
Hugh
I cannot disagree with this, but would add that an even higher abstractive skill is the multivalent correlation of topology, component choice, layout and subsequent measurement with sound quality. But this requires huge empirical experience, which not everyone aspires to as measurement is generally more appealing than wiring up prototypes.
The fact is that the customer makes a decision when he buys. A good proportion will choose entirely on their perceptions, which are influenced by a host of factors, but which do give lip service to a listening test, regardless of its subjectivity. To attack this position from a purist standpoint is probably legitimate, but it ignores the commercial reality. And this is where marketing and engineering diverge, as not everyone looks for, or indeed hears, the same things.
I don't know if this helps, but I would be willing to bet this is 1 magnificent sounding amplifier, even if, as a general rule, I shy away from EC regimes.
Cheers,
Hugh
Oh sure, Hugh. There is a gulf between marketing motives and devices and purist goals. I'm not considering the marketing angle, as you say, as it does not concern me.
Strictly speaking, "how it sounds" was not in Edmond's list of performance goals. Those guys did set some difficult, measurable goals and met them and that is an excellent achievement in itself.
If you knew what I know I'm not sure you would be quite so bold with your money. I haven't heard the Halcro. I have heard some of the best sounding products out there and I'm not sure if any of them approach Halcro's THD figures. From what I know about other designs and what I know about Halcro's, if Halcro is really the best sounding then it isn't entirely due to its THD figures. There is a lot more to it than that.
Brian
PS: why do you shy away from EC regimes?
Strictly speaking, "how it sounds" was not in Edmond's list of performance goals. Those guys did set some difficult, measurable goals and met them and that is an excellent achievement in itself.
If you knew what I know I'm not sure you would be quite so bold with your money. I haven't heard the Halcro. I have heard some of the best sounding products out there and I'm not sure if any of them approach Halcro's THD figures. From what I know about other designs and what I know about Halcro's, if Halcro is really the best sounding then it isn't entirely due to its THD figures. There is a lot more to it than that.
Brian
PS: why do you shy away from EC regimes?
You expose the soft underbelly......
I shy away from EC because I don't fully understand the finer points, it is best with mosfets according to many and thus requires highish bias, and won't build stuff unless I fully understand the subtleties and can justify the additional R&D which is very, very costly.
Furthermore, I've found Self's Type II EF with charge suckout to be incredibly good sounding, and until pushed by the commercial realities (this has not happened yet and in any case the market realities tend to dominate when you lay down R&D dollars) I will concentrate on improving the front end and VAS, where I believe most of the gains are made.
Recent work with my Soraya power amp has revealed that huge sonic gains can be made in the input and VAS stages, and that a sensible, well tweaked EF II output stage can be extremely transparent, with most of the Vbe distortion picked up well by the nfb loop.
Lastly, the layout for an effective EC is no simple matter, and I try to design in modules, since a good layout can take literally hundreds of hours.
Nevertheless, I think EC output stages will become de rigeur in years to come, particularly if the small signal devices can be integrated into a readily available, cheap chip.
Hugh
I shy away from EC because I don't fully understand the finer points, it is best with mosfets according to many and thus requires highish bias, and won't build stuff unless I fully understand the subtleties and can justify the additional R&D which is very, very costly.
Furthermore, I've found Self's Type II EF with charge suckout to be incredibly good sounding, and until pushed by the commercial realities (this has not happened yet and in any case the market realities tend to dominate when you lay down R&D dollars) I will concentrate on improving the front end and VAS, where I believe most of the gains are made.
Recent work with my Soraya power amp has revealed that huge sonic gains can be made in the input and VAS stages, and that a sensible, well tweaked EF II output stage can be extremely transparent, with most of the Vbe distortion picked up well by the nfb loop.
Lastly, the layout for an effective EC is no simple matter, and I try to design in modules, since a good layout can take literally hundreds of hours.
Nevertheless, I think EC output stages will become de rigeur in years to come, particularly if the small signal devices can be integrated into a readily available, cheap chip.
Hugh
Re: PGP
Thanks Hugh!
It seems that you belong to the happy few who really understand our circuit. I'm glad you did not made silly remarks like sooooo complicated. You hit the nail on the head by saying "I don't believe there's anything redundant here"
For example, we didn't waist silicon on cascodes where they are not really needed. Maybe only two trannies in the front-end are less essential: the emitter followers (Q13/14) in the NDFL stage. You might omit them at the expense of a slight increase of the distortion, but then you'll need two diodes (or a Vbe multiplier) for temperature compensation of the next two trannies. So no big deal. Besides, we like to be on the safe side. Redesigning and buying new PCB's is a thousand times more expensive than two small signal trannies. That would be a perfect example of false economy
Cheers, Edmond.
Thanks Hugh!
It seems that you belong to the happy few who really understand our circuit. I'm glad you did not made silly remarks like sooooo complicated. You hit the nail on the head by saying "I don't believe there's anything redundant here"
For example, we didn't waist silicon on cascodes where they are not really needed. Maybe only two trannies in the front-end are less essential: the emitter followers (Q13/14) in the NDFL stage. You might omit them at the expense of a slight increase of the distortion, but then you'll need two diodes (or a Vbe multiplier) for temperature compensation of the next two trannies. So no big deal. Besides, we like to be on the safe side. Redesigning and buying new PCB's is a thousand times more expensive than two small signal trannies. That would be a perfect example of false economy
Cheers, Edmond.
traderbam said:
Hi Brian,
That "parameter" was not in my list of performance goals. Rather obvious, don't you think so? One more thing I'll give away: we did everything so that our amp sounds as worst as possible, 13 on scale of Curl. 😀
Cheers, Edmond.
Edmond,
Ok, now I get it. Those who see no shortcomings become one of "the happy few who really understand our circuit".
You are definitely in the right industry.
🙂
Ok, now I get it. Those who see no shortcomings become one of "the happy few who really understand our circuit".
You are definitely in the right industry.
🙂
AKSA said:
Hi Hugh,
You make a lot of good points, and it is true that in many cases the relationship between a single-number THD figure and perceived sound quality is tenuous at best. However, when proper measures are taken to drive down all sources of nonlinearity, and THD and many other distortions get very small as a result, I believe that those linearizing techniques used to get there play a role in making a better-sounding amplifier.
Note that Halcro doesn't just spec THD. They are one of the few that spec extremely low numbers on the 19+20 kHz CCIF spectrum, which I believe to be very important. If you get the CCIF spectral components all below about -120 dB, there is very little wiggle room left for any other kinds of nonlinear distortions to creep in.
I think that the Halcro may be a good example of a case where doing all the right things to get nonlinearity way down results in a very good amplifier. It was extremely well reviewed. Although Halcro has chosen to market their amplifier based on specs, and we may quibble with that, it does apparently perform extremely well sonically.
Cheers,
Bob
Bob,
Let me reveal my ignorance further by saying I've never heard a Halcro, that's why I chose the brand.
I continue to display a shocking ignorance of what is out there in the market. I'm really only interested in the technical problem, and how to make better sound.
On that note, and in support of Edmond's POV (and probably Brian's as well, I see the thread is getting a little rough at the edges), I would say that I have never improved the linearity of a power amp without hearing an improvement in sound quality. Linearity is not the same as THD, of course, but I believe it's a better indicator, particularly in that vital first watt, than most other measures.
I cannot stress enough that the vital spatial information in most music is contained in the first watt. If you lose control of the output stage around the dead zone of crossover, the fine detail goes AWOL and the sound of the amp is much degraded. I believe this is where EC is so very useful; it all but scotches the crossover glitches.
I assess linearity with triangle waves and differential traces. Primitive, but seems to correlate well with perceived sound quality, which is as much psycho-acoustic as mathematical.
Cheers,
Hugh
Let me reveal my ignorance further by saying I've never heard a Halcro, that's why I chose the brand.
I continue to display a shocking ignorance of what is out there in the market. I'm really only interested in the technical problem, and how to make better sound.
On that note, and in support of Edmond's POV (and probably Brian's as well, I see the thread is getting a little rough at the edges), I would say that I have never improved the linearity of a power amp without hearing an improvement in sound quality. Linearity is not the same as THD, of course, but I believe it's a better indicator, particularly in that vital first watt, than most other measures.
I cannot stress enough that the vital spatial information in most music is contained in the first watt. If you lose control of the output stage around the dead zone of crossover, the fine detail goes AWOL and the sound of the amp is much degraded. I believe this is where EC is so very useful; it all but scotches the crossover glitches.
I assess linearity with triangle waves and differential traces. Primitive, but seems to correlate well with perceived sound quality, which is as much psycho-acoustic as mathematical.
Cheers,
Hugh
traderbam said:
Hi Brian,
Shortcomings? Well, actually our amp has one shortcoming: it has no vacuum tubes in the VAS. We don't like the warm sound as much as you do. 🙂
Cheers, Edmond.
AKSA said:
Hi Hugh,
I agree completely with what you have said, and, yes, my choice of the word linearity (rather than THD) was intentional. I also agree that what goes on in the crossover region, where control of the signal is handed off from one set of transistors to the other, is very important. I also tend to believe in a smoother handoff as opposed to an abrupt handoff, and one handled with fast devices that will not introduce any glitch artifacts in the process. That is part of the reason I lean toward MOSFETs with error correction. The handoff accomplished by bipolars is more abrupt and may be accompanied by charge-storage artifacts.
Cheers,
Bob
traderbam said:
Brian,
If there are any shortcomings in our design or implementation, we would love to hear about. Please understand that we are sincerely looking into whatever level of criticism regarding our amp. Also, we are here to explain every detail on what and how was done and eagerly digest any suggestion for improvements. In this respect, your input is as welcomed as any other suggestions/opinions/etc...
OTOH, we are expecting mostly analytic comments rather than sarcastic comments. While I personally acknowledge the value and importance of the "good sound" concept (and I'm struggling in mapping any of the related psychoacoustic concepts to electrical, measurable variables) I can't imagine an amp spec that looks like "It should sound good, according to XYZ taste and experience".
Thanks again for your input. BTW, we are not in the right industry, we are on the right DIY forum.
syn08 said:
It seems to me you have been diligently sweeping all corners of the art picking the best - sometimes overlooked - bits and pieces, which quite naturally leads to rather complex implementations.
Others may prefer an all encompassing conceptual approach in search of the elusive goal of a design breakthrough leading to simple, elegant solutions.
By nature, breakthroughs cannot be forecasted so it is well worth pursuing them at the risk of no results. Meticulous attention to details through the first path as you have shown, surely must pay off, and congratulations for your work.
Ultimately, objective performance parameters as long as well established cannot fail to correlate with listening, at least for people who can distinguish good sound because of fidelity from good sound because of nice adjustments for one's preferences.
Rodolfo
ingrast said:
Hi Rodolfo,
I fully agree with your statements, in particular the last one!
As for 'simple, elegant solutions' , did you had a look at TMC?
Cheers, Edmond.