I am struggling with the idea of simple THD factor addition from input to output. My NAD CD player is rated at 0.008% THD, my Marantz power amp at 0.05%. Assuming simple addition, this would amount to 0.058%.
But is that really so?
The baseline for THD measurement for the CD player is created by the CD and what's on it. Let's forget for the moment how much THD is recordedon CD. However, the baseline for the amp is the already distorted output signal of the CD player relative to the disc.Additionally distorting an already distorted signal has a higher baseline that the CD player faces. I suspect the overall THD, disc to amp output will be higher than the simple sum of the two.
If so, then the classic setup of a TT, preamp and power amp is in real neck high trouble. The stylus alone quite normally has a IM factor declared as "<1%". Than that signal has to pass through a pono RIAA eq/amp, then through the line amp in the preamp to get to the power amp. God alone, if even He, knows what the total distortion buildup may be at the terminal of the loudspeakers. Yet we have a slew of people who think all that sounds better than a CD.
And while I agree with John that the loudspeakers will exhibit the highest distortion factor in the entire system, this is not necessarily as bed as it sounds. Yesterda, I was hunting for some data and ran across a reVox ad, which also mentiones their speakers. The biggest model, a 3 way classi, 12", mid and treble domes, has its THD quoted as "<0.2%, 50-20,000 Hz". I presume, because it's not explicitly stated, this applies to an SPL of 96 dB/4 Ohms, simply because most other thigs are related to this value. The speakers are nominally 4 Ohms. That's not so bad now, is it?
But is that really so?
The baseline for THD measurement for the CD player is created by the CD and what's on it. Let's forget for the moment how much THD is recordedon CD. However, the baseline for the amp is the already distorted output signal of the CD player relative to the disc.Additionally distorting an already distorted signal has a higher baseline that the CD player faces. I suspect the overall THD, disc to amp output will be higher than the simple sum of the two.
If so, then the classic setup of a TT, preamp and power amp is in real neck high trouble. The stylus alone quite normally has a IM factor declared as "<1%". Than that signal has to pass through a pono RIAA eq/amp, then through the line amp in the preamp to get to the power amp. God alone, if even He, knows what the total distortion buildup may be at the terminal of the loudspeakers. Yet we have a slew of people who think all that sounds better than a CD.
And while I agree with John that the loudspeakers will exhibit the highest distortion factor in the entire system, this is not necessarily as bed as it sounds. Yesterda, I was hunting for some data and ran across a reVox ad, which also mentiones their speakers. The biggest model, a 3 way classi, 12", mid and treble domes, has its THD quoted as "<0.2%, 50-20,000 Hz". I presume, because it's not explicitly stated, this applies to an SPL of 96 dB/4 Ohms, simply because most other thigs are related to this value. The speakers are nominally 4 Ohms. That's not so bad now, is it?
Distortion is an important factor, yes. Now, the question is: "why two amplifiers, with HD < 0.01% can sound so different ?
Why some amplifiers just products a "wall of sound", while others will reveal hidden details, like attacks on guitar strings and drums that you can feel easily? That i call "micro dynamic".
Why, on doubled voices, some amp will give-you "one composite source", while you can separate the two on an other one ?
Why some amplifiers just products a "wall of sound", while others will reveal hidden details, like attacks on guitar strings and drums that you can feel easily? That i call "micro dynamic".
Why, on doubled voices, some amp will give-you "one composite source", while you can separate the two on an other one ?
Demian, I guess only measurements can tell you, since you don't know what the composition and polarity of the distortion products are from a bare THD number.
You could even imagine having two stages that each produce an equal, arbitrarily large distortion percentage, but if the polarities of the distortion products are reversed, the combined train will have no distortion at all.
You could even imagine having two stages that each produce an equal, arbitrarily large distortion percentage, but if the polarities of the distortion products are reversed, the combined train will have no distortion at all.
Last edited:
It seems (to me) that both CFAs and "simple" can produce those details ease of reproduction. And i tend to work on slew rates as a priority when i have reached a good enough distortion factor in a amp project.
When i talk about "simple", i refer to 3 stages and poles in the feedback loop, like in some Hirraga, Nelson Pass, or Lazy Cat (VSSA) designs.
Looking at some of the OS CFA attempts in this thread: http://www.diyaudio.com/forums/solid-state/248105-slewmaster-cfa-vs-vfa-rumble.html#post4289318 i can see added poles, with the use of Cascodes, Super pairs, etc.
Yes they decrease further the distortions, but don't they decrease slew rates, and this 'details' reproduction ?
When i talk about "simple", i refer to 3 stages and poles in the feedback loop, like in some Hirraga, Nelson Pass, or Lazy Cat (VSSA) designs.
Looking at some of the OS CFA attempts in this thread: http://www.diyaudio.com/forums/solid-state/248105-slewmaster-cfa-vs-vfa-rumble.html#post4289318 i can see added poles, with the use of Cascodes, Super pairs, etc.
Yes they decrease further the distortions, but don't they decrease slew rates, and this 'details' reproduction ?
Quite so, agreed.
The point being that until you do, you can only guess, it's not a simple excercise in addition.
People here are obviously struggling with how to deal with this business of reproducing the sense of a live experience - in some cases making it extremely complicated. Again, fortunately, none of this is necessary, and high SPLs also aren't needed - the simple answer is clean reproduction, which is all about reducing to inaudibility the right set of distortion artifacts. I have no idea what the numbers are that express that level of distortion in terms of THD or anything else - but I know when I have got to that level by the quality of the subjective experience. One thing is certain: amplifying stages having nominally relatively mediocre distortion spec's can do an excellent job of reaching, or getting very close to the required performance levels IF managed properly.
This doesn't mean it's easy!! It was a huge struggle in the years of my earlier efforts, trying to get a handle on what was required - these days, the equipment in raw form is often intrinsically better, and better knowledge and instinct for what has to be done has accrued.
One way of describing what needs to be achieved, is that when the volume control is adjusted up, that the subjective impression is not that the sound gets louder, but that it becomes more intense. This is the difficult part, getting a system to evolve from the former to the latter behaviour.
This doesn't mean it's easy!! It was a huge struggle in the years of my earlier efforts, trying to get a handle on what was required - these days, the equipment in raw form is often intrinsically better, and better knowledge and instinct for what has to be done has accrued.
One way of describing what needs to be achieved, is that when the volume control is adjusted up, that the subjective impression is not that the sound gets louder, but that it becomes more intense. This is the difficult part, getting a system to evolve from the former to the latter behaviour.
Not necessarily so, but it is possible. I am reminded of what Bob Stuart of Meridian once said, that each and every transistor in the signal path imprinits some of its sonic signature on the passing signal. Hence, the less, the better.
This is one of those generalities that, with a bit of thinking, can be shown to be wrong.
The output of a single-transistor amplifier shows a lot of the 'personality'*, if you will, of that transistor. Add a second transistor, and use some of its gain to improve the transparency of the combo, and the 'personality' of the transistors gets orders of magnitude lower.
*Things like distortion profile, bandwidth, non-linear capacitances, etc.
Jan
Esperado, Jan,
Please don't take it too literally. I think Bob was alluding to wildly complex designs which do not seem to sound better than other, simpler designs.
I am not a friend of the oversimplification brigade, just as I don't much care about wildly complex circuits which ultimately deliver about as much as a simpler design. I do not plan how mcu material will I use, instead I prefer to plan on what effects should it all produce And in the dilemma between complexity and effects, effects win hands down.
Please don't take it too literally. I think Bob was alluding to wildly complex designs which do not seem to sound better than other, simpler designs.
I am not a friend of the oversimplification brigade, just as I don't much care about wildly complex circuits which ultimately deliver about as much as a simpler design. I do not plan how mcu material will I use, instead I prefer to plan on what effects should it all produce And in the dilemma between complexity and effects, effects win hands down.
I'm heartened that Bob Stuart said that. It is what I believe as well. I've known Bob for a long time, since 1972 when we met at a Baxandall lecture in London. He was just starting then, and we actually had a demonstration of problems with output protection at a party of his, where he went in and cut it out during the party! Yes, there was a difference. Bob has done some really good work on distortion evaluation as well. Check him out in WW in 1973.
Last edited:
Well.... more than a 2 transistor circuit is going to be better..... but less than a hundred is also going to be better. There are optimally efficient circuit topology's that gets it all done. Isnt that what we are looking for? JC's circuits come to mind as do those mentioned by Esperado for good examples.
THx-RNMarsh
Last edited:
Yes, for example cascaded grounded-cathode stages, perhaps with unbypassed cathode resistors, with the inverting gain at the plates adjusted to cancel even-order at the second plate. It works, although nowhere as easy to get good numbers as using feedback---but for the feedback-averse, a way.
With third we are usually dealing with the approach to limiting at high levels, so comparable cascaded distortion reduction is not likely. Everything is limited in voltage (or perhaps current) swing sooner or later, and if it is symmetrical, that entails production of third.
Exactly! In some situations, adding a transistor may make it better, in other situation, it works backwards. So, general 'less is better' is not correct. Another agreement - boy I'm having a field day!
Jan
Occam's electric shaver
The general philosophical attitude about complexity is prevalent, but frequently arises from strawmen. One such: cascade more and more stages, then apply global feedback to reduce the gain, with the gain being determined by passive components that are more ideal in their adherence to terminal equations than comparable equations for active devices.
This might have been a strategy in the early days of solid-state. As good transistors got cheap it could be explored, and if single-number THD marketing was connected to buying customers, it might prevail for a while. Although defects beyond static THD could be present and even audible, like the notorious TIM under appropriate conditions in a poorly designed product, you were supposed to ignore this, or attribute the odd unpleasantness to something else*.
With increasing sophistication in circuit design, rather than just a simple cascade of stages, various approaches to enhancing linearity emerged. These might entail additional active devices, but acting in a supporting role as it were. It took a while before the design freedom available with complementary polarity parts was exploited. So we saw VFAs with NPN and PNP long-tailed pair input stages, cascoding, use of constant-current generators, complementary emitter followers at the output. Eventually circuit topologies devised for other fields were rediscovered and applied to audio, like the Boxall (Baxandall and Larson ten years later) and Aldridge (Hawksford cascode much later) compound-stage ones that recycle the control electrode current. There are a few left probably not yet in currency from oscilloscope vertical amplifiers. Also JFETs got better and could be deployed where it made sense.
The problem with all of this is: you have to really know what is going on. Conjecturing that more parts are prima facie adding more phase shift, for example, is at best a half-truth.
Brad
*I had someone tell me I had to get a much more powerful amplifier when I was complaining about fatigue listening to an early and mediocre CD player---advice that was complete rubbish.
The general philosophical attitude about complexity is prevalent, but frequently arises from strawmen. One such: cascade more and more stages, then apply global feedback to reduce the gain, with the gain being determined by passive components that are more ideal in their adherence to terminal equations than comparable equations for active devices.
This might have been a strategy in the early days of solid-state. As good transistors got cheap it could be explored, and if single-number THD marketing was connected to buying customers, it might prevail for a while. Although defects beyond static THD could be present and even audible, like the notorious TIM under appropriate conditions in a poorly designed product, you were supposed to ignore this, or attribute the odd unpleasantness to something else*.
With increasing sophistication in circuit design, rather than just a simple cascade of stages, various approaches to enhancing linearity emerged. These might entail additional active devices, but acting in a supporting role as it were. It took a while before the design freedom available with complementary polarity parts was exploited. So we saw VFAs with NPN and PNP long-tailed pair input stages, cascoding, use of constant-current generators, complementary emitter followers at the output. Eventually circuit topologies devised for other fields were rediscovered and applied to audio, like the Boxall (Baxandall and Larson ten years later) and Aldridge (Hawksford cascode much later) compound-stage ones that recycle the control electrode current. There are a few left probably not yet in currency from oscilloscope vertical amplifiers. Also JFETs got better and could be deployed where it made sense.
The problem with all of this is: you have to really know what is going on. Conjecturing that more parts are prima facie adding more phase shift, for example, is at best a half-truth.
Brad
*I had someone tell me I had to get a much more powerful amplifier when I was complaining about fatigue listening to an early and mediocre CD player---advice that was complete rubbish.
Brad, this is something that connects also to what I noticed at the last ETF in Berlin. There were three presentations about, roughly, improving tube amps, by Frank Blöhbaum, Menno Vanderveen and Guido Tent.
Frank came up with a 100W SE tube amp with 0.01% THD at full power; Menno introduced a transconductance stage combined with a common cathode tube stage and Guido produced what must be the perfect cathode follower.
What these tube designs had in common was that the large leaps in performance came from judicious use of solid state stuff and negative feedback. Once you start to think outside the box of 'this is how it should be done' the sky is quite literally the limit.
But yes, you must know what yo are doing.
Jan
Frank came up with a 100W SE tube amp with 0.01% THD at full power; Menno introduced a transconductance stage combined with a common cathode tube stage and Guido produced what must be the perfect cathode follower.
What these tube designs had in common was that the large leaps in performance came from judicious use of solid state stuff and negative feedback. Once you start to think outside the box of 'this is how it should be done' the sky is quite literally the limit.
But yes, you must know what yo are doing.
Jan
I must say, I'm surprised at your reaction. From the several models I am playing with posted here it was very easy to see that a nominal 100W/8 Ohms amp in my project has at lesat 26 transistors, of which at least 6 (3 pairs) are the output stage proper. That hardly qualifies as a minimalist project, although I do feel it uses the minimum required to meet its design objectives. My high expectations from it simply require anything from 26 to 32transistors per side.
And even if 3 pairs of MJL 3281/1302 would be enough to get there, I don't like pushing it, so I typically use 4 pairs. Since I purchased them directly from ON Semi/Motorola, I feel their prices are quite reasonable, even good for what they offer in real life, so I have no incentive to use them frugally, but quite the opposite, if it's needed, just use it.
Remember, I have always been a current freak, so I don't "save" on the output stage, the capacitors and the power transformers.
- Status
- Not open for further replies.