DrG said:
Maybe this can shed a little light on the murky subject:
http:www.gedlee.com/distortion_perception.htm
And a slight highjacking question for Joe: I've seen a few posts regarding valve buffering for the NIGC (or is that rightly called a GC?), but not much discussion. Can you point me to any reasonable implementations of same? And thanks for your very interesting contributions.
Sheldon
Oops! Ignore the question Joe. I just spotted your response to the very same in another thread.
This study makes some sense with respect to loudspeakers. With respect to amplifiers it's just leaves the question open. Notice the table that shows the THD of some of the WAV files. The lowest is .1%. The rest are humongous! Even a headphone amp based on the lowly LM386. Were are looking a distortion levels at the boombox level. At those levels, it's not much of a revalation that spectral content affect perception. This is clearly applicable to loadspeakers and how they are percieved, but I don't see how the conclusions can be transfered to amplifiers without more work. I wonder though at the validity of the conclusions unless the various WAV files are normalize to have equal THD figures. Otherwise you have an uncontrolled variable.
I would like to see this repeated using analogous WAV files only with much lower THD figures. What happens for instance when the test is repeated using files where the WAV file THDs are no grater than .01% and no greater than .001%? These are levels regularly attainable by SS amps. At those levels are the differences perceptible regardless of how the WAV file is constructed? Is there some threshold at which a transition from audible to inaudible occurs?
Another question that needs examination: Is there some form of audible distortion that is prefered to the absense of distortion? I stronly suspect there at least for some portion of the population.
I would like to see this repeated using analogous WAV files only with much lower THD figures. What happens for instance when the test is repeated using files where the WAV file THDs are no grater than .01% and no greater than .001%? These are levels regularly attainable by SS amps. At those levels are the differences perceptible regardless of how the WAV file is constructed? Is there some threshold at which a transition from audible to inaudible occurs?
Another question that needs examination: Is there some form of audible distortion that is prefered to the absense of distortion? I stronly suspect there at least for some portion of the population.
sam9 said:
well, some of us have been "known" to be able to hear the difference between non-inverting and inverting configurations (that's about <0.0001% thd), who's to say that some of us cannot hear gigaheart?
🙂
Empty claims are nothing but empty claims.
Sheldon - thanks for the link. I'm always on the lookout for "new" ways of characterising distortion in a perceptually relevant manner (almost too many big words there for a Sunday morning).
My broken record is still playing on this thread - it basically goes "gainclones have extremely low high order harmonics for a SS amp, better than many high end designs, and similar to a single ended tube amp". For example the 5th harmonic (somewhere between the lower order harmonics and higher order crossover distortion components) is always below 0.001% at any output level, even on a 4 ohm load, for the LM1875 and LM3886. This equates to a power level ten billion times lower than the fundamental. Here's a link to my article on the subject:
http://www.adx.co.nz/techinfo/audio/gainclone1.htm
For anybody who's confused, by "higher order harmonics" I don't mean the "higher order frequency responses" in the GedLee presentation (which model the actual distortion mechanisms using polynomials) - although these things are closely related. I'd like to do more work on this field myself, but I haven't got time to unwrap my AES journals from their plastic wrapping, let alone do a paper.
A good parallel is the refresh rate on CRT monitors. In theory 50Hz is good enough - but only for the central part of vision, provided the image isn't too bright. 75Hz is ok if you're up close with a reasonably bright image, but you still might see flicker if you look far enough off to the side, so crank it up to 85Hz. That might look rock steady, then you could see a brief flicker only on transients (such as when a window is opened over a dark background), so crank it up to 100Hz. But with eye tracking movement, artifacts might be noticeable in certain situations up to 1000Hz or even more. And so on. There's definitely a law of diminishing returns involved, and how far you go depends on the big picture - is it a screen on a server in a back room, or a CGI graphics workstation working on the latest blockbuster? Is the money better spent on improving colour accuracy or resolution?
My broken record is still playing on this thread - it basically goes "gainclones have extremely low high order harmonics for a SS amp, better than many high end designs, and similar to a single ended tube amp". For example the 5th harmonic (somewhere between the lower order harmonics and higher order crossover distortion components) is always below 0.001% at any output level, even on a 4 ohm load, for the LM1875 and LM3886. This equates to a power level ten billion times lower than the fundamental. Here's a link to my article on the subject:
http://www.adx.co.nz/techinfo/audio/gainclone1.htm
For anybody who's confused, by "higher order harmonics" I don't mean the "higher order frequency responses" in the GedLee presentation (which model the actual distortion mechanisms using polynomials) - although these things are closely related. I'd like to do more work on this field myself, but I haven't got time to unwrap my AES journals from their plastic wrapping, let alone do a paper.
I would like to see the same thing too. Many "engineering types" think that 0.1% of THD is inaudible, no matter what the cause. Audibility of 0.1% of crossover distortion on a pure tone is easy to demonstrate, I can hear it down to about 0.01% but as you get lower the situations where it is audible become more rare. There is actually no absolute limit because there is a lot of variability in the system, which means it becomes more of a statistical analysis.sam9 said:
A good parallel is the refresh rate on CRT monitors. In theory 50Hz is good enough - but only for the central part of vision, provided the image isn't too bright. 75Hz is ok if you're up close with a reasonably bright image, but you still might see flicker if you look far enough off to the side, so crank it up to 85Hz. That might look rock steady, then you could see a brief flicker only on transients (such as when a window is opened over a dark background), so crank it up to 100Hz. But with eye tracking movement, artifacts might be noticeable in certain situations up to 1000Hz or even more. And so on. There's definitely a law of diminishing returns involved, and how far you go depends on the big picture - is it a screen on a server in a back room, or a CGI graphics workstation working on the latest blockbuster? Is the money better spent on improving colour accuracy or resolution?
Of course its easy to spot a low refresh rate, but claiming to hear into the megahertz is like saying you can see infra-red light.
sam9 said:"I tell you this truthfully, it is possible to hear slew rate problems measured at hundreds of kilohertz."
How many hundreds of kilohertz? Ten, that would be a Megaherz? Can I hear Megahertz? Gigahertz?
Oh my!
I don't think too many claim that, but what you can hear is slew rate induced distortion at much, and I do mean much, higher or if you like, above the normal range of human hearing (whatever that is deemed to be).
What happens is that an external signal finds itself entering an amplifier circuit which has a rise time which greater than that of the circuit itself. Often it is also related to feedback since it is possible that the rise time exceeds the ability of the feedback to correct. Hence the open loop distortion of the circuit can be laid bare... and this is only one of several possible scenarios.
Basically if it distorts then it distorts, and not necessarily at the equavalent frequency the rise time represents. This is quite commonly known, and for this reason designers of SS power amps usually incorporate a low pass filter on the input. Indeed I do so with the VIGC before the 3875, and voila... many can hear the improvement.
Even low level slew can cause 'greyness' or tonal 'flatness' and even 'edginess.' The other thing that regularly suffers is a reduced soundstage. These are very audible. They are also, in more severe cases, the classic characteristics of bad SS sound. BTW, tubes are much less likely to suffer this malaise.
Joe R.
sorry to resurect old thread but i did some test for myself to check.
using creamware scope 32 bit float @ 98khz sampling I set up an oscillator that can do sin and square wave. i put my el cheapo 30mhz o'scope on the output of the amplifier to make sure that the signal was still there when i couldnt hear it anymore. i was using event 2020 bas studio monitors. these monitors are noit great but for this test they seem to have worked.
the speakers response comes in to this as well but the max of these speakers does not roll off as fast as my ears.
(i know idealy i should have used a mic with response up to the very high end AFTER the speaker)
I turned the frequency as high as i could hear it while the oscilator was on sine wave.
(for me about 17.5khz) i notice the roll off is pretty fast on my ears as well, ince i get to 18 khz the sound level is almost gone
then i switched to square wave on the oscilator. the scope showd a squareish wave (rounded with a bit of overshoot for sure)
the difference is very audiable. supposedly you can not hear all those extra harmonics. but you can. i also made sure there were no sub harmonics that could be affecting my perception, there were none at amp level, possibly in speakers.
The way the human ear works is not like an amplifier or speaker, there are tiny hairs that move, as well as the eardrum. i can imagine with such a setup, a more sudden rise time affecting how you hear things that can not be tested with a sine wave.
creamware scopeDP is a synth/mixer/effects design program that uses creamware DSP cards. they use analog devices "SHARC" chips . my setup has 19 of them. happens you can make test instruments with it as well
using creamware scope 32 bit float @ 98khz sampling I set up an oscillator that can do sin and square wave. i put my el cheapo 30mhz o'scope on the output of the amplifier to make sure that the signal was still there when i couldnt hear it anymore. i was using event 2020 bas studio monitors. these monitors are noit great but for this test they seem to have worked.
the speakers response comes in to this as well but the max of these speakers does not roll off as fast as my ears.
(i know idealy i should have used a mic with response up to the very high end AFTER the speaker)
I turned the frequency as high as i could hear it while the oscilator was on sine wave.
(for me about 17.5khz) i notice the roll off is pretty fast on my ears as well, ince i get to 18 khz the sound level is almost gone
then i switched to square wave on the oscilator. the scope showd a squareish wave (rounded with a bit of overshoot for sure)
the difference is very audiable. supposedly you can not hear all those extra harmonics. but you can. i also made sure there were no sub harmonics that could be affecting my perception, there were none at amp level, possibly in speakers.
The way the human ear works is not like an amplifier or speaker, there are tiny hairs that move, as well as the eardrum. i can imagine with such a setup, a more sudden rise time affecting how you hear things that can not be tested with a sine wave.
creamware scopeDP is a synth/mixer/effects design program that uses creamware DSP cards. they use analog devices "SHARC" chips . my setup has 19 of them. happens you can make test instruments with it as well
According to latest and greatest research, the ear has active amplifier systems. This means that things get selectively amplified and filtered by your ears, nonwithstanding additional selection by the brain. The system appears to be asymetric ("single ended" 😉 ) and most sensitive at the crossover point.
This could explain some amplifier characteristics - some amps' shortcomings may matter more than others, because their deficiencies get selectively amplified by the ear-brain system.
MBK
This could explain some amplifier characteristics - some amps' shortcomings may matter more than others, because their deficiencies get selectively amplified by the ear-brain system.
MBK
such as distortion on the supposedly inaudiable range?
or special attention to the "human speech" range which we definately hear better.
or special attention to the "human speech" range which we definately hear better.
sorry to ressurect it again but now i realise that the supposed difference between my maximum hearing frequency at sine and square may have had something to do with distortion in the tweeeter.. who knows what it might be doing at its max frequency at a high level and with a square wave...duh. the diifference may have just been from the speakers.
so it is pretty much invalid test.
so it is pretty much invalid test.
It should remain valid so long as there is no subharmonic distortion or major nonlinearities affecting the level of the fundamental coming out of the tweeter (which is possible when things start to flap around). A small amount of harmonic distortion won't hurt.neutron7 said:
The main thing I was thinking about was level matching. Did you check that the level of the fundamental remained the same? (the p-p voltage should be the same on both waveforms, or ideally you'd measure it with a mic and spectrum analyser)
I've done a similar test with headphones, but it was impractical to match levels so I gave up. I've got a prototype ribbon tweeter laying around, I might give this a go some day.
DrG said:
I agree with very much of what you say.
Most chip amp builders will end up discussing some little remote component (what brand of resistor/capacitors are to prefer?)
that has no effect whatsoever on place where amplification takes place.
And this is inside chip - and can not be modified, however creative you are.
What is very good with chip amps, is they are nice for beginners.
A suitable level to start learning what diy audio is like.
We should be grateful for this interest in chips.
They can for many be the frontport into where your real challenges starts
and where you will need a lot of creativity.
🙂
Doesn't match up to my tube amps
Well here is my experience. I was curious. I have been building amps for like 20 years plus. I build one of Peter's kits, using the premium kit parts and recommended components.
Initially it sounded pretty ordinary, but after running in for a couple of weeks it started to sound pretty good. It was better than a standard-type mosfet amp.
After now running for 4 weeks or more (mostly into a dummy load for running in purposes) it sounds good. Probably as good as some Meridian 605s I used to have (which Martin Colloms used to rate highly)
BUT just nowhere near my SE 211 amps. Still chalk and cheese there. No contest. Nada.
It probably is one of the best solid state amps I have tried, if not the best. BUT still a long way from all the tube amps I have tried against, in at least 3 different systems. Tube amps that destroy my gainclone are:
-- my SE parallel 211 monsters tweaked to the max
-- push/pull 70W chinese amp
-- SE EL34 (quite close, but the tube still has it)
-- SE EL84 (as long as you only use low volume as the tube amp only makes 5W)
I've heard Peter say that he's tried his gainclone against some good tube amps and his friends thought the g/c sounded better. Well either (a) his gainclone sounds a hell of a lot better than mine, or (b) his tube amps don't sound as good.
Because when I tried this, the difference was simply laughable. Compared to the good tube amp, the g/c sounded 2-dimensional, lifeless, no sparkle to the music, no slam in the bass, no air, etc. All the things that make me use tubes, despite the fact that they are a complete pain in the *** with all their hassle.
I really wanted them to be good as I am such a fan of low cost goodness (my two top tips in this regard are : iPod : Anti-cables)
But I will keep the g/c as a spare for when the tube amps decide to break down.
--bob
Well here is my experience. I was curious. I have been building amps for like 20 years plus. I build one of Peter's kits, using the premium kit parts and recommended components.
Initially it sounded pretty ordinary, but after running in for a couple of weeks it started to sound pretty good. It was better than a standard-type mosfet amp.
After now running for 4 weeks or more (mostly into a dummy load for running in purposes) it sounds good. Probably as good as some Meridian 605s I used to have (which Martin Colloms used to rate highly)
BUT just nowhere near my SE 211 amps. Still chalk and cheese there. No contest. Nada.
It probably is one of the best solid state amps I have tried, if not the best. BUT still a long way from all the tube amps I have tried against, in at least 3 different systems. Tube amps that destroy my gainclone are:
-- my SE parallel 211 monsters tweaked to the max
-- push/pull 70W chinese amp
-- SE EL34 (quite close, but the tube still has it)
-- SE EL84 (as long as you only use low volume as the tube amp only makes 5W)
I've heard Peter say that he's tried his gainclone against some good tube amps and his friends thought the g/c sounded better. Well either (a) his gainclone sounds a hell of a lot better than mine, or (b) his tube amps don't sound as good.
Because when I tried this, the difference was simply laughable. Compared to the good tube amp, the g/c sounded 2-dimensional, lifeless, no sparkle to the music, no slam in the bass, no air, etc. All the things that make me use tubes, despite the fact that they are a complete pain in the *** with all their hassle.
I really wanted them to be good as I am such a fan of low cost goodness (my two top tips in this regard are : iPod : Anti-cables)
But I will keep the g/c as a spare for when the tube amps decide to break down.
--bob
Re: Doesn't match up to my tube amps
... quote ...
Better than standard-type mosfet.
But ...
BUT just nowhere near my SE 211 amps.
Still chalk and cheese there.
Because when I tried this, the difference was simply laughable.
... end quote ...
========================================
beau2317
Hehe, nice to have a spare chip amplifier at home,
in case your better amp has a nervous breakdown.
--------------
And exactly what I can imagine:
A few gainclones perform at a good and high quality level.
Even can be a bit better than some mosfet amps.
But,
they are somewhat limited in performance.
That little extra something,
you can not get out of a chip amp.
.... unless you open up the LM3875 chip and change a few things
beau2317 said:
... quote ...
Better than standard-type mosfet.
But ...
BUT just nowhere near my SE 211 amps.
Still chalk and cheese there.
Because when I tried this, the difference was simply laughable.
... end quote ...
========================================
beau2317
Hehe, nice to have a spare chip amplifier at home,
in case your better amp has a nervous breakdown.
--------------
And exactly what I can imagine:
A few gainclones perform at a good and high quality level.
Even can be a bit better than some mosfet amps.
But,
they are somewhat limited in performance.
That little extra something,
you can not get out of a chip amp.
.... unless you open up the LM3875 chip and change a few things
Re: Doesn't match up to my tube amps
Maybe combining g/c and tubes... ?
Never heard PD's gainclone, cannot make any comment there. But have I heard a gainclone sound better than SE tube? Sure have. But over-generalising is a sin of it own? Some SETs have limitations not acceptable to everyone.
Joe R.
Sydney - the home of Rod Elliott 😉
beau2317 said:
Maybe combining g/c and tubes... ?
Never heard PD's gainclone, cannot make any comment there. But have I heard a gainclone sound better than SE tube? Sure have. But over-generalising is a sin of it own? Some SETs have limitations not acceptable to everyone.
Joe R.
Sydney - the home of Rod Elliott 😉
I've never much liked discrete power amps made with mosfets, but then again, I've also never heard a modern valve SE design. GCs are unbeatable with the right system if built with care, but on some systems don't work at all.
Don't know where that gets us, but I thought I'd add it to the mix...😉
Don't know where that gets us, but I thought I'd add it to the mix...😉
Joe Rasmussen said:
I keep telling myself I must be, otherwise why would I be here?
a/ feeling existential angst...
lineup said:But,
they are somewhat limited in performance.
That little extra something,
you can not get out of a chip amp.
.... unless you open up the LM3875 chip and change a few things
I don't know if you know what you are talking about, but it certainly doesn't look like you do.
It takes a lot of experience building chipamps be able to say something like that.
Doesn't look like you have it.
No, you don't need to open the chip.
For instance, like Linn does with their Chakra amps: (TDA7293) chip working up to a certain current output, and then high current transistors kick in.
If it wasn't worth it, then why go to all this trouble?
Btw those amps soud good.
With great care and experience you can make a very good sounding chipamp, even with a single LM3886 per channel.
Everything counts: PCB layout, PSU, grounding arrangement, etc.
I just wanted to add that if you can't stand the smell of pee, maybe you should get out of the kiddies pool, we are all haveing fun here.
- Status
- Not open for further replies.