Generalized attributes of amplifier technologies

I have been contemplating going to the "dark side"  (Purify based amp) on my main stereo.  Do I still need the masking distortion of my MOSFET amp after resolving the payer and DAC issues?  Still needed due to less than optimal source quality? (It does measure .003% @ 1W, 1K or better as that is about my test residual) This led me to thinking about the various reasons tubes, MOSFETS, BJT, or Class D sound different. 

These are very generalized attributes as I have experienced them. I have owned and built tube, bjt, and MOSFET amplifiers. Owned class D ( cheap for the garage)   

For any technology, the execution matters probably more than the technology. Personal preference matters. This is NOT a better or worse, just thoughts on differences. I am sure there are additional attributes.  It would be nice if we could quantize them with how our brain interprets the sounds so we could rely (design to) more on objective measures than subjective. I am assuming competent design and components.  

Tube amps have a rolled off top end due to transformer output. Fewer ultrasonic artifacts. We sense ultrasonic and they can cause tweeter ringing and IM issues. 

Tubes may have more even order HD than odd. It may be higher providing some masking. Our brain favors even low order. 

Tubes have higher but steady noise floor. Again, masking a lot of ills. 

Tubes tend to "extended decay" as reviewers say, ringing is the reality. Euphonic masking?

Tubes, for the same measures steady state power, tend to sound more dynamic. Sufficient power supply or less compression?

Tubes tend to have much lower DF so deep bass control is loser. 

MOSFETS have less compression with amplitude than BJT.  More linear transconductance.

MOSFETS typically had higher bandwidth than BJT and way higher than tubes. Possible differences with IM. Or TIM as the fad in the 80's. No longer a big difference with BJT.

Any technology, Differences in IM with gain.  IM being non harmonic our brain does not like it. 

Any technology, insufficient dynamic power supply current causes compression and higher distortions.

Solid state, and maybe tubes too, the more parallel outputs, the lower the current in each and the more linear. ie, less compression. 

Dynamic power supply current. Here is where budget amps fall flat.  Cap arrays about 4X calculated with very low ESR make a big "compression" difference.  Dips in the rails can have nastier effects to the IPS and VAS than the outputs. 

Class D used to have  serious load invariance issues. Combined with switching power supplies way too small.  Not sure that is true anymore. 

Class D used to have a pumping noise floor.  Now so low, maybe irrelevant

Class D tends to drive low Z loads easier. BJT and MOSFET stability get difficult under 4 or 5 Ohms.

Tubes miss-match of the transformer output  and load raises both linear and harmonic distortion. Never seen a 2 or 3 Ohm tap.  Tough on widely varying load impedances. 

Differences in clipping behavior. Inherent but implementation can mitigate somewhat. 

Architecture differences, local vs global feedback ratios, Miller vs dominant pole compensation, differential vs se IPS and VAS, and on and on. 


FWIW, I have also considered going back to a tube on my desk just for fun.  I do love my giant 2W Schiit amp though.  It's huge improvement over the Creek is what got me thinking I can use a cleaner amp in the main stereo. 




Do I still need the masking distortion of my MOSFET amp after resolving the payer and DAC issues? ... It does measure .003% @ 1W, 1K or better as that is about my test residual ...

It’s very unlikely that you can hear distortion at that level.

... I am assuming competent design and components ... Tube amps have a rolled off top end due to transformer output.

Some do. Many don’t. I don’t think your generalizations are going to lead you anywere.

Well, many SS amps have in excess of 50 Khz BW.  Most transformers do not. They are by definition an inductive LP filter.  What is up there is only harmonic distortion. This is not a bad thing, only a difference.  I am not suggesting roll-off in the classic audio band as even the 6W Chinese wonder chassis I built a Red Light Special on  held up pretty well into the mid teens. I do see McIntosh has continued to make transformer improvements extending their BW.  Kind of their hall-mark! The current ones seem to sound less "classic tube" to me than the earlier ones though I know amps like the old CJ or Dyna better. Could be transformers, could be larger grid isolation resistors. I am sure the higher quality passive components help a bunch. 

MOSFET amps sound different from BJT. They all can measure what single tone lab testing says is inaudible and I agree that with a 1K tone, I hear no difference and if I did, I would blame it on the speaker that is probably in the .3 or .4% range there.  Unfortunately, that is only part of the story. If it was, My Fosi V3 would be the killer amp.  Fine in my garage but I put it on my main system for a lark and with music it fell flat on it's face. Terrible distortion at what had to be less than 10W. This also  brings up why on earth we can hear tiny differences with music in electronics when the speakers are 100 times worse?  What is our brain doing!? Toole does not yet and if He doesn't, for sure I have no clue. But we can. 

Single tone measurements are valid, but limited. MOSFET amps, or at least the majority that are based on the original Erno B. Hitachi design ( mine is highly modified but similar character) all have a false smoothness to them while beating the dynamics of a BJT .  General rule may not hold to the Atoll which in a store sounded more BJT than I expected and the Hegel sounded more FET-ish. 

I have three DACs right now. All with classical measurements as one might see on ASR  should be well, make that way,  below audibility. They should sound identical. They do not.  I have some suspicions on the implementation and on how they handle certain input signals that explain some of it. Reducing through JRiver -3 dB did make them sound much closer, close enough I question if I hear differences between two of them, so digital cliping from filter overshoot may well be a valid issue. I note RME and Chord both deal or at least talk about that. Pre-ringing and the amount of post ringing from the various filter algorithms may be related. I have an idea that may be a lot of why some prefer R2R NOS DACS . 

I recently saw a suggestion that some class D and A/B handle broad band noise differently. This may be only in the dirt cheap range of class D and fully resolved as is load invariance in the current higher quality units. Hypex, Purify, maybe ICE. Not picked up in two tone IM measures. 

The point is, from these generalizations and more I am sure I am missing, can we extract some parameters that may eventually lead us to a multiple attribute plot with subjective preferences we could graph devices on using objective measures?   If you are a detail freak, pick from this quadrant, if buttery midrange, pick from that quadrant . More weight and so on.  We need to quantify what causes these attributes to measure them. We could then design to them. 

Generalizations are a place to start. I'm retired and not designing another amp.  My goal is to get some thought that may lead someone else further down the road so maybe I can buy the fruits. What can bring the classic objective measure camp and the pure subjective camp to common ground?  Both right, both wrong. 

So do I buy the Vidar, Buckyeye or Denafrips amp?  Keep searching for a Stasis 7?Am I thinking of it with only nostalgia? Aragon? Older Classe?  Early Cary? Or are my old 60's and 70's music remastered on early CD so crappy I should stick with my Mosfet? If I get it wrong, the level of domestic distortion will skyrocket. 


It is less meaningful to talk about advances in class D design nowadays without mentioning GaNfet technology. Relative to Mosfet, aside from faster switching, higher efficiency, and being able to operate in higher voltage/temp environment, the GaNfet results in more accurate PWM waveform (see below) in general and hence cleaner sound (less THD/improved SNR, etc.) The bybrid configuration that combines tube and SS is worth mentioning too.

Not meaningless, as it follows under innovation. Improvement in application we will have to see, but yea, an omission that we should watch.   Unfortunately, still quite rare and expensive only in boutique products as of yet.    Hmmm, seems Peachtree has a product.  I had ignored them as until now, I hated every one I heard for classic "Digitates".  I need to track one down.  Progress I hope!

Looked up the Peachtree GaN400.  Specs wise, mid-range so that makes it into probably "need to listen" for the other half of the sound.  I see they still make the old 500 that I did not like.  

VTV is an interesting product but as I only have USB, not viable.  I would have a lot of questions about their D2D front end. It is probably the way of the future.  The only DAC being the output stage. Less can be more.  Falls under innovation for sure. 

Curious side note.  In every class D amp I have seen internals of, the output networks are ferrite core and electrolytic caps , yet we go to great length and expense to remove them from our speakers.  Something to investigate. 

@tvrgeek Tube amplifiers do not have to be rolled off. A lot depends on the design of the output transformer. Harmon Kardon made the Citation series amplifiers back in the 1960s. Their Citation 5 had 100KHz bandwidth. Seems to me the Citation 2 was not far behind. I am running a 5 Watt tube amp in my bedroom which has full power bandwidth to 100KHz. Bandwidth is not why tube amps seem to have less energy in the highs than a lot of solid state.

Distortion is the culprit. It may not be much, but the ear uses higher ordered harmonics to sense sound pressure- that brightness is caused by higher ordered harmonics; the ear interprets them as brightness (and also harshness).

In case its not clear at this point, the distortion signature of any amplifier is its 'sonic signature'. The class of operation doesn't have an enormous effect on that- the topology and execution of the circuit does.


If you read my post, I said exactly that. A general statement and some did better. Most did not. I have a hard time believing 100K though. 

Also agree, the distortion profile is driven by the topology. Going to large grid resisters "hardens" the sound of a tube. David Manley demonstrated this. The myth of even or odd harmonics is purely a topology factor even in a tube amp. SE vs P-P if I remember.  I can miss-balance the LTP of a SS amp and get high even harmonics. I sold my tube books so I can't give you a direct reference. Cordell and Self did talk about even vs odd in their SS books. 

And absolutely! Execution is often more important than the technology.  My generic statements refer to well executed. You can screw up the best parts and technology. 

For innovation, I was just looking at the Sajab amplifier.  It takes in the USB data stream,  processes in DSP for level eq.,  generating the PWM it feeds to the output class D stage. So no conventional DAC in the front end. Only the integrator on the output.   This makes sense to me. How it sounds I haven't a clue and have not seen traditional objective specs. 

I have a hard time believing 100K though. 

@tvrgeek our amps have all had 100KHz bandwidth but they are OTLs. The little 5 Watt amp I mentioned uses output transformers made by Edcor. The spec the transformer to 50KHz. But I measured the amp and it did 100KHz with ease.

Google returns this on 'Harmon Kardon Citation 2 specs":

  • Manufactured between 1959-1962.
  • Power Output: 60 watts/channel.
  • Frequency Response @ 60W: 18Hz to 40,000 Hz +0, /-1.0 dB.
  • Frequency Response @ 20W: 18Hz to 60,000 Hz +0, /-1.0 dB.
  • Frequency Response @ 1W: 2Hz to 80,000 Hz+0, /-1.0 dB.
  • Frequency Response (Power Amplifier): 10Hz to 80,000Hz +/-1.0 Db.

I measured the output transformers of the Citation 5 myself, on which I based my comments. Tube amps that have that sort of bandwidth are not that weird. Its even more common for them to to to 50KHz. I hope you're not suggesting that a 50KHz bandwidth results in the amp sounding rolled off.

The myth of even or odd harmonics is purely a topology factor even in a tube amp. SE vs P-P if I remember.

A PP amp, if fully balanced, will express a cubic non-linearity; even orders are cancelled from input to output. The amplitude of the higher ordered harmonics will fall off more rapidly than they do in a single-ended circuit, where a quadratic non-linearity is expressed (resulting in a large 2nd order harmonic, usually with succeeding harmonics falling off on an exponential curve).


Maybe why the old Cary OTLs sounded so nice.

Yea, thanks for the reminder on topology. I forgot which was which. 

You've been down a few more paths than I have.  I still haven't warmed up to Class D, but have otherwise been through countless tube and ss amps in class A and AB topologies.  Separated the treasure from the trash in both cases.  Powered "difficult" Magnepan and Thiel speakers with tube amps with delightful results.  Well, the ss amps still won with the CS3.5, but it's a push with the Maggies.  OTOH, I'm torn between the SE45 amp I built (~1.7 wpc @ clipping) and my vintage Craftsmen and Fairchild amps into Altec 604-8G.  From a technical perspective, all are flawed in some way yet absolutely delightful to listen to.  There are many tube OPT that have extended bandwidth if that becomes a pursuit for you.  Google John Atwood's transformer test report from Dec 1990 for a partial list of vintage OPTs and their measured attributes.