The Sound of the Machine The Hidden Harmonics behind THD

[WD5JKO]

  I ran across a link written by Lynn Olsen that describes many facets of vacuum tube audio amplifiers with design input, and details of why one amp may sound different than another when both may have similar frequency response and similar THD figures.

   Many of us hams might chime in with the "audio-fools" comment, but after careful read Lynn seems to make some pretty good points with data to back them up. One example he gives is what happens to a standard triode amplifier when we remove the cathode bypass capacitor.

  He also makes the case where a solid state Crown DC-300 will look almost perfect on the scope, but will be almost un-listenable with music as the source when the amp is driving quality speakers. Those same speakers and music will sound much better with an old Dynakit ST-70 stereo tube amplifier. The article makes a good case for why this is. This is not a case of SS versus tubes, but more a case against high gain and huge amounts of global negative feedback.

This is a long read, but I think a worthwhile read.

Regards,
Jim
WD5JKO
http://www.nutshellhifi.com/library/FindingCG.html

[KD6VXI]

  I ran across a link written by Lynn Olsen that describes many facets of vacuum tube audio amplifiers with design input, and details of why one amp may sound different than another when both may have similar frequency response and similar THD figures.

   Many of us hams might chime in with the "audio-fools" comment, but after careful read Lynn seems to make some pretty good points with data to back them up. One example he gives is what happens to a standard triode amplifier when we remove the cathode bypass capacitor.

  He also makes the case where a solid state Crown DC-300 will look almost perfect on the scope, but will be almost un-listenable with music as the source when the amp is driving quality speakers. Those same speakers and music will sound much better with an old Dynakit ST-70 stereo tube amplifier. The article makes a good case for why this is. This is not a case of SS versus tubes, but more a case against high gain and huge amounts of global negative feedback.

This is a long read, but I think a worthwhile read.

Regards,
Jim
WD5JKO
http://www.nutshellhifi.com/library/FindingCG.html

I must admit, I didn't read the article, but your summary sounds a lot like the same things Bob Carver said back in the 80s.

Of course, he did a double blind comparison of the audio phools with 5 or 6 amplifiers.  After modifying the neg feedback and output Z, the audio phools couldn't figure out which was the t00bed amplifiers and which where the SS.

Most thought all where t00bed, when in fact, only one was.

He was poo pooed as an audio phool, as well, but who knows.

Shane

[K1ZJH]

I'd think a scope offers very limited dynamic range to analyze a signal. You need to use a log scale, not linear, when looking for harmonic distortion, etc.

Also,  I'd imagine the reflected load that the amplifier presents to the speakers must also play a very big part--it provides the dynamic braking for the speaker back EMF to work against. Wouldn't removing the cathode bypass on a triode greatly affect that one parameter? Without getting into a lot of others?

Pete

[W2XR]

I would have thought that you  needed an audio spectrum analyzer, so as to see the even and odd harmonic content of the audio output signal, particularly the higher order harmonics.

I agree with Lynn Olson's assessment. I recall reading this information in a now-defunct publication called "Sound Practices"; it was a great audio magazine that catered to the DIY/homebrew vacuum tube audio enthusiast, of which I am one.

It is the higher-order odd harmonics that sound particularly dissonant to human hearing, such as the 3rd, and to an even greater extent the 9th, the 11th harmonic, etc. An 11th harmonic of the fundamental, even when it is only .01% of the fundamental, can be very irritating to listen to and the human ear is very sensitive to this kind of distortion.  And consider too, the intermodulation products created by these higher order harmonics and you can imagine what this does for the sonic purity of an amplifier.

Unfortunately, transistors tend to generate much more of the higher order odd harmonics than vacuum tubes, and this is most probably the factor responsible for the so-called "transistor sound". Vacuum tubes, on the other hand,  when operating linearly, generate mostly even-order harmonics, and these harmonics are much more consonant to human hearing,  embellishing the sound with a "warmth" that many listeners find appealing. The even harmonics can be suppressed to a large degree with push-pull operation.

Interestingly too, excessive negative feedback, normally used to correct for the non-linearities inherent in any amplifier, actually increases the magnitude of the higher-order odd harmonics, making such circuits even more painful to listen to. Feedback really only reduces the magnitude of the lower order odd and even harmonics. Some of the better designed all-triode audio amplifiers utilize zero to only 8 dB of global negative feedback to improve the linearity of the design. Highly regarded audio amplifiers such as the older McIntosh and Dynaco tube units employed as much as 20 to 23 dB of voltage feedback.

Just my 2 cents worth. Not trying to start any kind of a raging debate on this topic, which has been going on for the last 45 years.

73,

Bruce

[W3RSW]

Oh here we go.
I agree; a lot's been hashed over.

Regarding 'hi end' speakers sounding awful on certain amps. ...-anyone remember Damping factor? DF as regards transistor vs. tube amps?  The acoustic loaded speaker debates vs. the older bass reflex and tuned port designs.
 
My dad and I (guess who did all the wood screwing) even built a labarynth, two column woofer. This was waaaay before Bose put little plastic replicas in their radios.  It was a plywood box about 7 ft. long by 2 ft. wide by 8" deep box with two intertwined sound columns placed upright in a closet behind the living room with the port into the room.

-I used to love to read of "The Audio Amateur," particularly when they had numerous construction articles.  I began to loose interest when they split into "TAA" and "Speaker Builder."  TAA used to have numerous articles about converting Dynaco SP series preamps, amps, etc.

One of the best amps I ever built was Reg Williamson's twin 40 amplifier.  Parts and boards came from Old Colony Sound Lab.  Very clean audio to once younger ears. This amp had a novel power supply using a capactance value amplifier to make a smaller electrolytic cap act as a much larger one.

Audiophools used to be respected; they built their own stuff, some so different even we could hear the difference. 

[WD5JKO]

Shane said:

"I must admit, I didn't read the article, but your summary sounds a lot like the same things Bob Carver said back in the 80s."

Reply by Jim, WD5JKO:

   Please link to the Bob Carver stuff if it is online. I once worked in a R&D lab and we obtained a Carver amp, circa 1980. The specs were unbelievable (too good), yet the thing was small (abt a 8" cube) and was very light. The marketing blurb also touted it as having a "digital computer" to protect the amp from a mis matched load.

  The marketing dept. trumped the engineering dept. What they did was take a 250 watts / channel amp, and remove the heatsinks, and remove the power transformer. To keep the amp from over heating they added the "digital computer" which was a thermal Klixon switch. On the bench we noticed that this thing was as safe as a 5 tube AC/DC table radio, and capable of 10 watts / channel (sine wave). Any more power than that and it would thermal off and on like a Christmas tree light.

   After this experience, anything said by Bob Carver was not listened to be me.

  The article I linked to seems to try to describe much more than tubes versus transistors. A special emphasis is made towards gain stage linearity with moderate gain, instead of high gain controlled with feedback amounts approaching 40 db. Somewhere from the mid fifties to the mid 80's we relied on way too much feedback to overcome gain stage non-linearities.

73,
Jim
WD5JKO

[KD6VXI]

Shane said:

"I must admit, I didn't read the article, but your summary sounds a lot like the same things Bob Carver said back in the 80s."

Reply by Jim, WD5JKO:

   Please link to the Bob Carver stuff if it is online. I once worked in a R&D lab and we obtained a Carver amp, circa 1980. The specs were unbelievable (too good), yet the thing was small (abt a 8" cube) and was very light. The marketing blurb also touted it as having a "digital computer" to protect the amp from a mis matched load.

  The marketing dept. trumped the engineering dept. What they did was take a 250 watts / channel amp, and remove the heatsinks, and remove the power transformer. To keep the amp from over heating they added the "digital computer" which was a thermal Klixon switch. On the bench we noticed that this thing was as safe as a 5 tube AC/DC table radio, and capable of 10 watts / channel (sine wave). Any more power than that and it would thermal off and on like a Christmas tree light.

   After this experience, anything said by Bob Carver was not listened to be me.

  The article I linked to seems to try to describe much more than tubes versus transistors. A special emphasis is made towards gain stage linearity with moderate gain, instead of high gain controlled with feedback amounts approaching 40 db. Somewhere from the mid fifties to the mid 80's we relied on way too much feedback to overcome gain stage non-linearities.

73,
Jim
WD5JKO

I don't have any of it, hence the reason I didn't link to it in the beginning.

This was text referenced from the brain, circa mid 80s

HOWEVER, if my memory serves me well enough (below 40 yrs old, but lotsa parties in the teens -n- twenties), it was either a Sharper Image catalog, or one of the Popular series of magazines.  My stepfather had an ongoing subscription to them, the Heathkit catalogs and a bunch of robotic stuff.

The radio I picked up from Gramps, may he be resting in peace every time I stick my hands in this stuff

Shane

[WD5JKO]

Shane,

    Yes the 1980's are a bit of a blur to me as well!  There is some stuff about Bob Carver on Wikapedia. I am quoting one part of his amplifier here:

"A power transformer connected to an amplifier, said transformer being adapted to operate at maximum power output at a relatively high frequency, in the order of 20 KHz. There is a control circuit which produces control pulses at a frequency of 20 KHz. The voltage level at the output of the transformer is compared to the amplitude of the audio signal to be amplified to produce a control signal related to the difference between the two. This control signal acts through a modulator to pass portions of each control pulse, with the duration of the pulse portions that are passed being generally proportional to the magnitude of the control signal. These pulse portions in turn open and close a switch connected to the primary winding of the transformer to control the duration of current pulses which flow through the transformer at a 20 KHz frequency. When the power requirements of the amplifier are either higher or lower, the duration of the current pulses to the transformer are made longer or shorter, respectively, to match the power requirements of the amplifier."

  What this was in reality was a low voltage 60 hz power transformer to power the amp for 5-10 watts output. Beyond that there was some form of switching network to bypass the transformer to support the load and while doing so to lose the isolation from the power line. This was revolutionary and I wonder if the UL rules as written then would have caught this issue and deemed it a problem.

   The company I was working for was really into switching regulator design at that time. It was about 1980 when IR came out with the first of their power FET's, the IRF-100. These power FET's, and some others from Siliconix started a revolution. Today we have class D and E AM transmitters from their descendants.

The Carver approach although clever sure seemed to be a method prone to trouble and controversy. We looked elsewhere for innovative designs. The Carver amp ended up on the shelf and I'm sure eventually discarded.

Jim

[Tom WA3KLR]

I used to get Speaker Builder and The Audio Critic.

Audio Critic is now an on-line 'zine:

http://www.theaudiocritic.com/cwo/Web_Zine/

According to Peter Aczel the editor, the two critical parts today are the recording acquisition and the speakers.  The amps today should all basically be similar and transparent.

Take the time to read though his stuff, including back issues.

[WD5JKO]

Tom said:
"According to Peter Aczel the editor, the two critical parts today are the recording acquisition and the speakers.  The amps today should all basically be similar and transparent."

Reply by Jim:

  I suppose if the amps are all similar in design, and vintage, and without tone controls, or equalizer, then that might be true. Otherwise changing from one amp to another can make a huge difference in the sound. Again this is not tube versus solid state. Here is one example where an amp with fabulous specs (THD and freq response) sounds awful with complex waveforms (music):
http://www.stereophile.com/solidpoweramps/1292crown/index1.html

For most of my lifetime engineers have concentrated on frequency response and THD. Those variables are only part of the picture while Intermodulation distortion (IM), Transient Intermodulation distortion (TIM), and Power supply rejection ratio (PSRR) often take a back seat to THD. Addressing these issues works against profit margins, and the bean counters come into the situation too and fire the engineers so they can start production.

The article I linked in my first post needs to be read in full. I had to read it 3 times!

73,
Jim
WD5JKO

[WBear2GCR]

The Carver designs worked fine. The basic idea was to float a small (10 watt or so) amp on top of a rather crude amp that used voltage switching rails. The feedback around the little amp was supposed to clean up the nasty rail hops... it works in as much as you can make a very lightweight amp that puts out a ton of apparent power, but really doesn't sound all that good. Lot's of PAs used various Carver incarnations. The original "cube" did work, but had some problems, I don't recall exactly what - but the basic idea continued into other models.

As far as the harmonics of an amp, anyone who doubts all this should search online for Dr. Earl Geddes, and the "Geddes Metric". He is a highly qualified Phd, who is the antithesis of an "audiophool". Read what he has done.

My opinion of Peter Aczel is that he is the equivalent of "Consumer Reports"...
While he is correct that the ultimate determining factors (other than one's ears) is the original recording and the speakers, he is dead wrong that all amps are essentially transparent and the same. See what Earl Geddes says, if you doubt.

There is a neat site somewhere online where a fellow did some nice tests comparing a NAD to a SE Triode amp. Interesting bit that one. The SE Triode actually had LESS distortion than the lower spec'd NAD as the power level went up. Things are not all that they appear to be on the surface.

I suspect that Lynn Olsen and most others are using Audio Precision gear and/or high performance PC soundcard based FFT analysis, not a simple scope. Some folks may be using HP or similar spectrum analyzers also...

I don't know of any respectable high-end amp mfr today that is NOT concerned with TIM, THD, IM and PSRR, and more... fwiw.

If you want to know what people are thinking about in terms of solid state amp design, I strongly urge you to follow almost any thread in http://diyaudio.com in the Solid State section to see just how much minutia is being considered!! (Lynn is a frequent contributor in the Tube section there, btw.)

                     _-_-bear

[Tom WA3KLR]

Jim,

We've had threads going on receiver rf dynamic range measurements and found that the best today is really no better than the best receivers of 50 years ago.

I guess what you are saying is that today's amplifiers are not any better than 50 years ago also.  Is this right?

[Tom WA3KLR]

Bear,

What's wrong with Consumer Reports type of reporting?  What's wrong with identifying a $100 tuner that is as good as a $1000 tuner, a $200 amplifier that is as good as a $3000 amplifier?

[Tom WA3KLR]

Can this be separated out - does a tube amp sound better than a solid-state amp really, or is it just that one grew up listening to the tube amps?

[Tom WA3KLR]

Jim,

When an amplifier output stage has no feedback there will be great interaction between the output stage and the speakers.  That means that a given speaker cabinet/system will have a different frequency response and sound with a different feedback-less amplifier. I don't consider this a good situation at all.

[Tom WA3KLR]

The speaker's distortion is so much greater than the amplfier's.  When you believe you hear IMD products (which is most probably the difference in the "good" and "bad" sounds) this is most likely from the speakers.

Bear, you agree that the critical points in the audio system are the recording session and the speakers.  Even with the most mediocre listening system you can  point out a poor recording session.  We can't do anything about the recording.  Speakers is what the listener need so concentrate on.  This is what Peter is saying.  Yet people are obsessed with the amps!  Go figure.

P.S.  I can't hear the 18th harmonic of 1 kHz, and neither can anyone else.  But all of us can hear the 2nd and 3rd harmonics of 1 kHz very well.

[W3SLK]

Tom said:

Can this be separated out - does a tube amp sound better than a solid-state amp really, or is it just that one grew up listening to the tube amps?

Now there's an interesting concept!

[Steve - WB3HUZ]

If the snake oil salemen selling the so called esoteric $1000 amps are exposed, they won't make money. That's what is wrong with it.

Until solid system engineering principles are applied to a sound system, all this talk about amps is missing the forest for the trees.

Bear,

What's wrong with Consumer Reports type of reporting?  What's wrong with identifying a $100 tuner that is as good as a $1000 tuner, a $200 amplifier that is as good as a $3000 amplifier?

[W3RSW]

Considering growing up with tube amps..

One area where tube amps were considered superior is in the overdriven state. ...(which a lot of youngsters did ... I know I did.) The harmonics are usually more musical rather than what you get from solid state due to tube operating characteristics.

Agree with Steve that 'solid engineering principles' need to be applied, further there needs to be convergence between perceived sound testing and engineered sound testing. 

All systems should be double blind tested to a set of standards.  The standards can be 'gold' or 'silver,' just as long as they're consistant.  After all my 'gold' may be your 'lead.'  Double blind speaker tests have to be in the same acoustic environment, right down to spacing from walls, room dimensions and form factor (deviation from rhombic, approaching spherical, etc.) , acoustic absorbant material placing, faux furniture, listener postion.. all 3 dimensions for both source and listener, etc.  This is sometimes hard to do given that some testers place one set of speakers outside of another, etc.  Perhaps a revolving lazy susan of speaker sets behind an acoustically damped wall might be in order. 

[Opcom]

I listen to what sounds best to me. Usually it is a tube amp. The speakers also have alot to do with the way an amp sounds. I've heard it both ways, tube amp sounds better on these speakers, ss sounds better on those speakers. But that is subjective anyway. I don't buy my hi-fi gear in the stereo store but if I did it would no be by specs but by listening.

When I owned a repair shop in the 80's, a guy brought in a Carver cube. Upon examining it, the circuit was too unusual to fix easily without a schematic and Carver would not sell me one. I told the customer it was special circuitry and suggested he take it to the authorized repair center for a prompt repair. He moaned about what they'd charged him last time, but what could I do?

[W1EUJ]

Sorry folks, listening to rock music on this equipment is gay.

[WD5JKO]

Tom,

   I agree with you to a point. The audio source and speakers are important. The best amp and speakers cannot overcome a poor recording. That said, an amp can certainly "colorize" the sound in many ways. One example I already linked to was the Crown reference amplifier that the reviewer gave a thumbs down to. The amp was tops driving woffers to high SPL, but poor with complex music where it suffered from IM distortion, and TI distortion problems. That amp like many is not transparent. With IM we create distortion products right where our ears can hear them. Here is an example:


From: http://www.pmillett.com/Books/bbc_feedback.pdf

"The two terms, known as the sum and difference components,
cause the unpleasant sound from an overloaded amplifier. The process
by which they are introduced is known as intermodulation, the two input
signals acting as the carrier and modulating signal, to produce sum and
difference terms which are equivalent to upper and lower sidebands res-
pectively in amplitude modulation.
As an example, suppose two input signals are applied to an amplifier,
the frequencies being 256 cfs and 320 c/s. These frequencies are those of
middle C and E and blend harmoniously to give a chord known as a major
third. Due to non-linearity in the amplifier, the output contains components
with frequencies of 64 cfs (the difference term) and 576 c/s (the addition
term]. The difference frequency is that of C two octaves below the input
signal and harmonises with 256 c/s and 320 cfs, but the addition frequency
is that of D one octave above the input signal and produces a discord with
both 256 cfs and 320 cfs."


I have another example that is more relevant to us hams. I once built a push pull plate modulator with a pair of 8417's. I had a good 550v plate supply and a regulated 300v screen supply. I set the bias for 50ma total, and proceeded to modulate a 4D32 with it. On the scope I could modulate the rig very well, and the waveforms looked quite good with no evidence of clipping. On the air the reports were terrible saying my voice was muffled. The reports got much better when I increased the idle current to a little more than 100ma total. My conclusion was that the IM distortion of the high gain Beam Power tubes and pentodes (EL34, 6550, 8417) is very high unless you run these tubes very hard at or near the plate disippation limit.

  Negative feedback reduces THD (2nd 3rd, 5th) more than the higher harmonics, and does little to reduce IM. For low IM we need to go back to making the amp as linear as possible prior to adding the negative feedback. When we do this we don't need 20 - 40 db of NFB. Maybe we can get away with 6 - 10 db NFB instead. With proper design we can still have an acceptable source versus load impedance, and damping factor.

Another factor with audio amplifiers is that when they clip, they momentarily go open loop, and with a high degree of negative feedback, that gain reduction goes down depending on the extent of clipping. The human ear is very sensitive to this, and since a solid state amp clips more abruptly, it is much more important to keep a SS amp away from the clip point.

The speaker impedance is all over the map too. Between the cross over networks and speaker and box resonance points, the amplifier has a formidable job to do. The amp may see anything from a few ohms to 80 ohms over the audio range and some reactance too. Many amp designers ignore this and instead use a 4 or 8 ohm resistive load while tweaking away for max power at min THD. This is why amps are not transparent, and are every bit as important as the audio source and the speakers chosen.

I respectfully disagree with your point that amps don't matter.

Regards,
Jim 
WD5JKO

[Steve - WB3HUZ]

The two terms, known as the sum and difference components,
cause the unpleasant sound from an overloaded amplifier.

Now there's a statement of the obvious. Overload any amp and it will distort. This issue is not how much IM is created when overloaded but how much is created when not. More BS from the snake oil salesmen.

[WBear2GCR]

Steve,

What in the world are you talkin' about??

You need to read up some on the more recent thinking, engineering and testing on these things. Find and read what Dr. Earl Geddes says.

The way an amp clips or even a modulator is indeed important, but that's not the main discussion, since the main discussion revolves around unclipped amplifiers.

However, if you consider that many CDs are recorded with peaks that are 20dB up from the average level, and then you go to play back said CD on your home system at say 90dB SPL at the listening position how much headroom do you need?

Assume to simplify the matter that you manage to get 90dB SPL at 1 watt because you listen very close to ur speakers (yeah, right    ) then we need to reproduce unclipped 20dB up from there. That means we need 100 watts. But if it is a real world speaker and it takes say 4 watts to get to that level, then you need 400 watts, and if it is a typical high-end speaker with only mid 80s reference sensitivity, or you like to play at a slightly higher average level, like 93dB SPL average, you then need to double all those powers - you gotz a 200 watt amp in ur system? You gotz a 1000watt amp? Whatcha got?

I'll bet dollars to donuts that most amps in most systems are clipped far more often than people would like to believe - like all the time.

What's the spectra of distortion of ur amp, no matter what the power rating, as the power gets within 6dB of clipping, how about 3dB of clipping? Betcha that you start to see much larger differences between amps when you look at them this way, and you can hear the same differences just as well...

If one doesn't care, doesn't listen that critically, doesn't notice any differences, that's fine, but there are measureably, objectively, scientifically, and audibly present differences.

                           _-_-bear

                 

[WBear2GCR]

Bear,

What's wrong with Consumer Reports type of reporting?  What's wrong with identifying a $100 tuner that is as good as a $1000 tuner, a $200 amplifier that is as good as a $3000 amplifier?

Nothing at all, IF you could do such a thing.

One can't.

If you simplify, or reduce the criteria sufficiently then it is all equal.
The analogy I've used over and over is that of a full bore Formula Car vs. your "daily driver". If you never run the Formula Car over 55mph, I suppose that they're pretty equivalent. Imho, that is what Aczel and Consumer Reports both do, in their own way.

It is very rare that a cheap and inexpensive bit of audio gear offers equal or better performance than more expensive gear. There are some exceptions but very few.

A good example might be Behringer - good stuff, very cheap, right?
But not as good as most better grade or top-of-the-line gear that does the same job.

Those who are more discerning (and those who pretend to be as well) want the best performance available, not the best bang for the buck. (although it is good to know about both)

                   _-_-bear