Tailoring distortion spectrum

It would be interesting to hear other people´s experience on tailoring of the distortion spectrum of an amplifier.

I pay more attention to the distortion spectrum rather than absolute levels. 0,005% pure 2nd harmonic sounds much better than 0,005% 3rd and 5th. Over the years I learned a lot about how to tailor the distortion spectrum, and today I am able to design solid state amps that generates pure 2nd harmonics. Sonics are close to tube gear - but more transparent and with superior bass.

In general, odd numbered harmonics will often dominate the spectrum if the classic differential topology is used. These harmonics should be masked by generating even harmonics of higher amplitude (2nd stronger than 3rd, 4th stronger than 5th etc.). The best is of cource not to generate any odd harmonics, but especially a class AB power amp is a pain getting "clean".

The methods I use to generate even harmonics are basically the following:
- jfet´s and mosfet´s.
- resistive loading of the voltage amplifying stage. A somewhat crude method that easily can give audible distortion if not used with care.
- asymmetric current loading in the signal path. Think of how tubes work, with the positive half period having different gain than the negative half period. This generates even harmonics in a nice and controlled manner. This is the mechanism I try to copy in solid state. For example, increasing the quiescent current in a bipolar push-pull line stage output will increase the asymmetric loading of the preceding stages. This is easily verifed with PSpice. As the current gets too high the 3rd harmonic will begin to increase. Listening tests correlate very well with the PSpice simulations.
- Increase the open loop even harmonics with the methods above and then increase the negative feedback. The odd harmonics are pushed below the noise floor and you are left with even harmonics only - voilá. But I don´t use this very much, I fancy modest feedback i.e. open loop linearity.

Of course the circuit topology matters a lot, but once a topology is selected the distortion spectrum can be tailored for best performance.

Any similar experiences ?
 
Hi Syl,

Very nice post. I try to do this too.

It is, in my view, THE way to improve hifi.

A few points:

1. I agree with resistively loading the voltage amplifier, but use a boostrap to keep collector current variation within +/-10%. This has the advantage of loading up the collector at very high frequencies where the boostap cap has rising impedance, and this in turn makes it easier to stabilize the amp.

2. Don't particularly like the jfet and MOSFET as the former has quite wide tolerances, even within the same batch, and the latter has strong tendency to self-oscillation.

3.
 
To continue; hit the wrong button:

3. I think feedback networks require care. The ubiquitous AC feedback voltage divider is flawed, I suspect because of the traditional electrolytic used as a DC blocker. While servo systems have other problems and JFETs are flawed, neither require a blocking cap and thus intermodulation is not so much of a problem.

4. Something has to be done to ameliorate switching spikes at the crossover disjunction.

Thus I agree with your points very strongly.

Cheers,

Hugh R. Dean
www.printedelectronics.com
 
Hello AKSA and Syl, interesting postings.
3. I think feedback networks require care. The ubiquitous AC feedback voltage divider is flawed, I suspect because of the traditional electrolytic used as DC blocker.
In my experience choosing low esr smps electrolytic cap and fitting multiple parallel smaller caps in reducing value, down to s/mount and tubular construction gives very beneficial cleaning of sound - a single electrolytic is not good enough - no measurements, sorry.
and thus intermodulation is not so much of a problem
How come intermods and not harmonics ? - Technical question.
4. Something has to be done to ameliorate switching spikes at the crossover disjunction.
Doug Self in Electronics World magazine has done a pretty good job in his investigations - well worth a read. He uses the concept of a "blameless amplifier" - low measured distortions, but unfortunately he does not give any aurally correlating subjective evaluations .........Anybody?.............
0,005% pure 2nd harmonic sounds much better than 0,005% 3rd and 5th.
SYL, I'm interested to know how audible and if not loudly audible, what nuance or change of flavour do you find with each of these distortion levels and harmonics to be ?.......and at what sort of levels .......and what sort of music ?
Over the years I learned a lot about how to tailor the distortion spectrum, and today I am able to design solid state amps that generates pure 2nd harmonics. Sonics are close to tube gear - but more transparent and with superior bass.
Well done! - I agree very strongly with you also - if you cannot entirely eliminate distortion, then it is much better to steer the amp distortion products to those distortions that cause the least aural offense, or indeed as a subtle tone control to make it sound 'nice', or involving or seductive even......etc..
Do you have circuits and corresponding evaluations that you are willing to share ?

Regards, Eric.
 
Hi Eric,

Thank´s for your kind words. Here is a try to expand on your questions.

*************************
Eric question:
SYL, I'm interested to know how audible and if not loudly audible, what nuance or change of flavour do you find with each of these distortion levels and harmonics to be ?.......and at what sort of levels .......and what sort of music ?
*************************

0,005% is -86dB, well above the noise floor of the system. The percepted sum of distortions in a signal chain is additive, a bit like adding up noise. Even if a power amp delivers at about 0,05% (which is OK for a power amp) the much lower distortion of a line stage is still detectable in the music. If the music is played so that the noise floor is heard in quiet passages this level of distortion is audible with any kind of music, but in particular with acoustic instruments. Intertransient silence and imaging has a lot to do with what happens below -70dB.

The sonic perception of the different harmonics is described in litterature, but unfortunately I don´t have a link for you but Google.com should help you find something.

This is my subjective perception of what happens:
- The 2nd gives warmth in much the same way as tubes.
- The 3rd at the same level increases perceived loudness but the attack of the music is adversaly affected and things sound dissonant (like there as tones in the violin that should not be there).
- The 4th and 5th gives "bite" to the music, about the same type of difference as between a trumpet played quietly or loud. A quietly played trumpet still sounding kind of loud is a sure sign of higher order distortion.

Depending on preference it may not be optimal for some listeners if the distortion is 2nd only. If a certain loudspeaker is developed together with electronics with "bite", that sonic signature has been tailored into the loadspeakers sonic qualities. Connect a clean amp to the same speaker and the owner might say - there is no life, no bite. Sure he is right, but the amp should not be blamed. I know guys that prefer gear with this "bite" to bring life to their dead speakers. Do I need to say that they dismiss tube gear ?


*************************
Eric question:
Do you have circuits and corresponding evaluations that you are willing to share ?
*************************

As a DIY addict since long ago I have the bad habit of not documenting very well (schematics sketches on paper etc.). The perfect engineering is reserved for my work in the automotive safety electronics business. That is the virtue of DIY - just make it work and be happy with that. But the PCBs are carefully designed and well documented.

I have thought about starting my own site and present my designs together with the philosophy behind, but I never seem to find the time. Besides, there are tons of sites out there, who needs another one ?

The topology I am playing with right now is a fully complementary push-pull amp with jfet input, bipolar output and current feedback. Start with the J. Hiraga 20W Class A, use complementary input jfets instead and play around with the output stage topology. This topology suits both line level circuitry and power amps.

I hope you found something useful.

Regards

Syl
 
Hello SYL thanks for your reply.
My theory and experience agrees with all the above that you say.
I get to repair and hear a lot of high power stage pa amps - most of them are pretty damm clean, but they do exhibit the harmonic differences you describe.
I am more into high power stuff, so the Hiraga design is not quite so usefull to me, though I should go and study it.
Summers here are too hot to live with bigish class A, and my engineering morals turn me away because of the inefficiency.
Also I have lived with amps that bump up the monthly power bill markedly!

Along the lines of tube amps seeming to go louder than the output power rating suggests, do you get a similar effect with class A - I have had many listenings with Aleph 0 and it seemed to go quite loud.

Regards Eric.
 
You know Eric, what is interesting with the Hiraga topology is that with different kinds of output stages it can be used for either Class B, AB or A. The topology of the input stage and the VAS could be the same for all three. An obstacle I see for pro power is the start/stop of the amp - there is no single point to pull in order to start up or bring down the amp. It is a typical DIY design.

Nice to hear that you confirm my thoughts on distortion.

Regards

Syl
 
Hi Eric,

Pro power = power amplifiers requiring extensive protection circuitry and fail-safe mechanisms.

What I meant was the possibility for protection circuitry to shut down the amp in a simple way. A long-tailed pair can be shut down by controlling the current source. The fully complementary topology used here does not allow for such simple solutions.

Of course it is possible to design a shutdown path, it just takes more components to do it. The design will be more vunerable to faults in the protection circuitry, either bringing down the amp when it shouldn´t or not bringing it down when it should (which could be fatal for the speaker). Simplicity is a virtue when it comes to protection circuitry.

Regards

Syl
 
Hello SYL,
You have make good points about methods of protection.
Yes pro audio needs both amplifier and load SOAR protection.
I fix a lot of audio gear and pretty much only the high power stages ever fail, and virtually never in line level gear.
Main culprits are dry joints and dry electros, so the reliability of the protection stages is not really and issue I feel, although controlled startup and shutdown are.
I remember reading info years ago about different harmonics caused by different nonlinearities - eg assymetrical half cycle gains or something causing 2nd harmonic or something.
Do you have a link to a relevant page ?

Regards, Eric.
 
ideas for line level signals?

Great topic. I'm wondering if you guys apply similar techniques to solid state line level circuits such as phono pres, line amps, or active crossovers. I know that the idea for hi-fiers is to keep all distortion as low as possible at these stages, but as was already stated, all the distortion in the path does add up and in the case of line level, is greatly amplified later. I'm personally also intersted in circuits that introduce tailored distortion for ...er.. musical purposes.

Of course you would have more flexibility with discrete designs, but perhaps there are some things one could do to tailor the distortion spectrum of op amp circuits? I guess it would depend largely on the specific device. Are there specific op amps or op amp topologies that stand out as producing distortion tailored toward the 2nd order harmonics?
 
Hi GrantB,

I certainly apply these ideas to the complete signal path, not just power amps. I use simulations to explore different topologies and how they behave. The simulations, where even harmonics are compared with the odd, correspond very well with listening tests, so I nowadays trust the computer to give me the basic topology. Final tuning and transistor selection is made by ear when the circuit is built. Class A line level amps can be designed to generate 3rd harmonic below -120 dB. Such a circuit should have the 2nd harmonic tailored around -100dB. As the 3rd creeps closer to the 2nd the sound shifts toward the typical solid state sound we all know of. It is interesting to note that many topologies does not allow for this kind of optimization - the 3rd is dominant whatever you do.

I have a unity gain buffer / active crossover in pipeline right now, and it is optimized like this. Takes special transistors to do it though (need 2SK246-GR and 2SJ103-GR), and I have not found a source yet. Anyone who knows where I can find them ? Y or BL is no good, I need the GR.

On OP-amps I can´t give you any real guidance (never used them for audio). I think there is a general problem getting access to the mechanisms that can tailor the distortion spectrum. The only I can think of is to bias the OP-amp in Class A with a CCS at the output and play with the bias current.

Regards

Syl