I think this observation is borne of years and years of careful listening. A depressed midrange might also be some sort of phase cancellation phenomenon such as an incorrect polarity on a midrange driver in a three way speaker system.
However, I do feel that any lack of H2 will affect the midrange aguably more than the bass and certainly less than the top end; describing it as a suckout seems extreme, but philosophically it makes sense.
The full complementary input stage is electrically elegant and pleasing to the eye; but since both half cycles are theoretically dealt with in similar fashion it does mean that non-linearities are more likely to be symmetrical, which in turn leads to odd order harmonic generation. The aesthetic aspect of electronic design are very powerful; a kind of private appreciation of the designer, who derives pleasure from a neat and cultured look in addition to a worthy specification! The connection here defies intuitive notions, particularly to those of us who have spent long hours gazing at oscilloscope waveforms.
If we accept that harmonic distortion is inevitable, it makes sense to foster more even order distortions than odd, and the best way I know is to process the signal with as many single ended stages as possible, right up to the push pull output stage.
With a blameless output stage and single ended input and voltage stages, it should be possible to make an amp with as near as dammit even order distortions, or a least even order distortions which predominate over odd. Certainly my designs have been described as Single Ended Triodes with Grunt, so the idea might be worth considering.
Cheers,
Hugh
The "suckout" phenomenon is real. I like this term for it. A long time ago I built such a symmetrical circuit and is sounded very odd: extremely "clean", sort of "dark" and the midrange was so polite as to be mediocre.
It is indeed to do with spectral distortion.
The thing is that Krell and others use such a circuit and get away with it. Naim and many others use a single LTP and get away with it. It is not an inherent problem with the choice of architecture, but it is an implementation problem.
On balance I would recommend using the single LTP configuration because it avoids some difficulties which are harder to solve than for the single LTP distortion and I'm not convinced it buys much if you balance your input transistors carefully.
Crd
Guys,
I don't think you can say that a SE design has more odd order components. For sure, a push-pull will cancel to a large extend the even order harmonics, but it will generally still have the remaining odd order harmonics lower then the run-of-the mill SE.
Not to want to start a flame war, but if PM says that he hears a midrange suck-out on a system that has objectively less harmonics, maybe he should say that the other systems have a mid-range bulge! I mean, if you listen for years to flawed systems, the moment you listen to a better system, you'll tend to say it's bad....
As to the need to match the two parts of a full compl design, the object would be to have an amp as linear as possible before the application of fb. That would mean, Hugh, that ideally then output transistors of a power output stage should be matched, yes. That will both decrease the even AND odd order harmonics, making it easier for the fb to get it really low!
Just my 2 euroct worth of personal opinion.
Jan Didden
jam said:
Guys,
I don't think you can say that a SE design has more odd order components. For sure, a push-pull will cancel to a large extend the even order harmonics, but it will generally still have the remaining odd order harmonics lower then the run-of-the mill SE.
Not to want to start a flame war, but if PM says that he hears a midrange suck-out on a system that has objectively less harmonics, maybe he should say that the other systems have a mid-range bulge! I mean, if you listen for years to flawed systems, the moment you listen to a better system, you'll tend to say it's bad....
As to the need to match the two parts of a full compl design, the object would be to have an amp as linear as possible before the application of fb. That would mean, Hugh, that ideally then output transistors of a power output stage should be matched, yes. That will both decrease the even AND odd order harmonics, making it easier for the fb to get it really low!
Just my 2 euroct worth of personal opinion.
Jan Didden
Jan,
You have a good point about the midrange of a single-ended design being slightly fat, but somehow I find the single ended approach more musically correct. This is not to say that comp. designs are bad as there are many good comp. designs out there as Bam has pointed out.
I don't believe that heroic mataching of devices reduce even and odd products equally, Besides the fundemental make up of P and N channel devices is quite different.
Which brings us to the arguement should the output stage be made up of the same type of device. Like the Aleph?
Regards,
Jam
P.S. Euroct don't count here, we need good old American cents...
..
You have a good point about the midrange of a single-ended design being slightly fat, but somehow I find the single ended approach more musically correct. This is not to say that comp. designs are bad as there are many good comp. designs out there as Bam has pointed out.
I don't believe that heroic mataching of devices reduce even and odd products equally, Besides the fundemental make up of P and N channel devices is quite different.
Which brings us to the arguement should the output stage be made up of the same type of device. Like the Aleph?
Regards,
Jam
P.S. Euroct don't count here, we need good old American cents...
..Has anyone ever looked at the actual distorition products to back up any of the claims being made? I mean, the input DA stages are generally designed to be quite linear and I would think most any distortion would be very low. I can understand the theory that a full comp. symm. design, in which there are mismatches, can create some distortions while cancelling others, but I would think these distortion products ought to still be very, very low.
As was alluded to, it appears that there are good and bad examples of most any topology. It does seem that these differences are due to either the preference of the human hear to hear moderate amounts of second order distortion, or that things such as matching and componets are at least as important as the topology.
As was alluded to, it appears that there are good and bad examples of most any topology. It does seem that these differences are due to either the preference of the human hear to hear moderate amounts of second order distortion, or that things such as matching and componets are at least as important as the topology.
jam said:Jan,
You have a good point about the midrange of a single-ended design being slightly fat, but somehow I find the single ended approach more musically correct. This is not to say that comp. designs are bad as there are many good comp. designs out there as Bam has pointed out.
I don't believe that heroic mataching of devices reduce even and odd products equally, Besides the fundemental make up of P and N channel devices is quite different.
Which brings us to the arguement should the output stage be made up of the same type of device. Like the Aleph?
Regards,
Jam
P.S. Euroct don't count here, we need good old American cents...
Wait till next year when my 2 eurocents are worth a dollar or two, hah!
On a serious note: matching will ensure that the pos and neg parts of the signal are more equal in shape, that in itself will reduce even order harmonics. You are right that it doesnot help the odd order distortion.
But I do like very much the concept of just one type of device in the output stage. Reminds me of the circlotron I built with 807 tubes, way, way back.
Note to JeffR: DA stages are INHERENTLY non-linear. Only judicious amounts of emitter degeneration or very high gain (which translates to high fb factor) to keep the differential input voltage low will let them work reasonably linear in practise.
Jan Didden
Yes
And yes.
Two of the reasons valve/tube amps have such an appealing sound quality in many areas, IMHO, is that valves are all n-channel and the electrical characteristics depend largely on mechanical dimensions, which are quite large and easier to control than the parameters in silicon transistors, like doping. This is why betas can vary hugely among the same transistor type. Electricity is asymmetrical, there are no "positrons" - there is no good reason why an amp circuit should be symmetrical to the eye.
janneman said:
Jan, and bam, thanks for sharing.
Of course the DA stage would have lots of emitter degeneration - sorry I didn't make that clear. Linearity is even better with a cascode DA.
I still don't see how matching of the top and bottom stages is all that important. Maybe in Class A operation I can see it, but in B operation, looking at the output stage at higher power levels, only one device or the other will be conducting, so there will be only distortion products from one or the other. True, there will be distortion during the transisition from B to A, but that will be true for any such complementary output stages regardless of the input VA stages.
What you guys are saying does make sense - though the elegance of the fully complementary symmetric amplifier is very appealing, as Hugh indicated. I don't mean to be boneheaded, but it is tough to give up on this topology! Plus this is a very interesting and educational thread.
Jeff,
If I am not mistaken (I could be wrong) the distortion components are different between both output stages at zero-cross which would be more apparent in a class B stage, and feedback does not get rid of all of it.
The cirlotron might not be a bad idea after all, BAT use this topology in their power amps.
Regards,
Jam
P.S. We probably need to move to a different thread?
If I am not mistaken (I could be wrong) the distortion components are different between both output stages at zero-cross which would be more apparent in a class B stage, and feedback does not get rid of all of it.
The cirlotron might not be a bad idea after all, BAT use this topology in their power amps.
Regards,
Jam
P.S. We probably need to move to a different thread?
Just back on stream after loading up XP.
Hmmm...... We'll see how good this OS really is in a day or two...
Can't see anything I'd disagree with here, and certainly agree that some solid measurements of distortion would clear things up nicely, but sadly I can't offer them.
I will throw in a tempter here. How about Single Ended Push Pull?
This is an interesting configuration where one rail holds the active output device, and the other a current source (or sink, as you wish). Both devices are controlled artfully so that neither turns off at any time. Thus one device effectively controls load voltage, while the other controls standing current and pulls the current of one half cycle under control of the other device.
Self has something on this in his design. see here: http://www.dself.demon.co.uk/cfp.htm
This is effectively SEPP, and it offers potential of H2 only distortion, with superior efficiency to Class A, a little lower than AB.
Cheers,
Hugh
Hmmm...... We'll see how good this OS really is in a day or two...
Can't see anything I'd disagree with here, and certainly agree that some solid measurements of distortion would clear things up nicely, but sadly I can't offer them.
I will throw in a tempter here. How about Single Ended Push Pull?
This is an interesting configuration where one rail holds the active output device, and the other a current source (or sink, as you wish). Both devices are controlled artfully so that neither turns off at any time. Thus one device effectively controls load voltage, while the other controls standing current and pulls the current of one half cycle under control of the other device.
Self has something on this in his design. see here: http://www.dself.demon.co.uk/cfp.htm
This is effectively SEPP, and it offers potential of H2 only distortion, with superior efficiency to Class A, a little lower than AB.
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.