Opamp vs. Buffer vs. descreet Class A

[abbey road d enfer]

Don't worry. That's part of the learning experience, you ask to us old farts and first we sneer, then we explain (sometimes). Hold on! It's gonna be worse (we're getting older and older everyday).

 

[ricardo]

I checked Pass's view on NFB. He concludes that he prefers a low THD stage with no NFB than a high THD stage with a lot of NFB. Doh!

I prefer circuits with inherently low THD, around which I apply some NFB.  Feedback is useful when you need defined gain and response.

However, I've never jibed at circuits without NFB if they will do what I want.  eg my lowest noise effort at 0.28nV/rtHz and VERY low distortion has no NFB.

It's not NFB that's evil but that you have to use resistors to apply it.  Resistors ... well they have been known to resist the flow of electrons.  Transistors on the other hand, transist ...  8)

 

[abbey road d enfer]

I prefer circuits with inherently low THD, around which I apply some NFB.

Well that is just common sense (when one has developed his "common sense" with years of formal AND practical training). Now Mr. Pass putting this on a website doesn't make him a genius - I rather think it makes him a guru.

It's not NFB that's evil but that you have to use resistors to apply it.  Resistors ... well they have been known to resist the flow of electrons.  Transistors on the other hand, transist ...  8)

It makes sense on low-Z, low-noise, low level circuits. Now on a power amp, this is debatable.

 

[audiomixer]

One of the things I wanted to know in the beginning was whether there is some opinion on the question are there alternatives to the opamp route, and not so much if NFB is good or not, and if it would make sense to go into the wide area of discreet transistors to seek for an alternative. the NFB isn't so much of concern to me, at least not at an religious level. I do positively know you can make excellent sounding gear with opamps. I wave heard Pass designs in the past and it was a memorable thing. to call him a audioguru I find quite inappropriate.

now I seem to have started exactly what I had not intended, a heated debate about something we all can't prove and only have some opinion on.

dirkwright - I will check the possibility of testing your simple buffer and maybe some other available out there, but not in the main line of my mixer project. no other came up with a useful tip what I could / should test so far, a tad sad....

cheers,

- Michael

 

[ricardo]

One of the things I wanted to know in the beginning was whether there is some opinion on the question are there alternatives to the opamp route.

Michael, all successful designs start with a clear definition of what is the aim of the exercise.  Arno and others asked some pertinent questions which you haven't quite answered.

Properly applied negative feedback isn't just a way 'to get rid of harmonic distortion'.  It's main purpose is to make the signal chain more accurate.  This is what I strive for and I'll use whatever method is needed.  The reason I don't use NFB on my ultra low noise amp is cos it will reduce accuracy.  But this is a special case.  Usually, properly applied NFB improves accuracy.

If accuracy is important to you, it's hard to do better than very simple circuits with NE5532/4.  Even the new uber OPAs show little practical difference except in very specialised applications.  NE5532/4 is an example of a circuit with inherently low THD around which you can apply a little or a lot of NFB and generally get excellent results.

If accuracy is low on your priorities, the ground rules are different.  You are really asking what type of milk shake is best.  Making up PCBs for a circuit you haven't auditioned yourself is like laying up a year's supply of milk shakes (when you've never tasted a milk shake) based on some chance recommendation by some stranger, guru or not.

You are better off devising some way to conduct Double Blind Listening Tests (which will cost A LOT of money to do properly) and be prepared to try loadsa stuff to find one you really like.

I can tell you from experience that when you do the above, you often find accurate may be the most musical after all.  There's a lot of garbage spouted about the sound of this or that by people who are deaf.  I know cos I've tested a lot of these self declared Golden Pinnae over more than a decade.

An essential tool in this investigation is to have a good properly working accurate comparison.  Why don't you make one of these first, then compare this with more musical (?? ) versions?

All the above doesn't mean I don't think some small (??) departures from accuracy may sound better.  eg I've had some success making an accurate digital channel sound like good vinyl playback.

My field is speakers & microphones and I'll admit that some old valve classics like AKG C12/24 do sound nice.  But IM very biased opinion, a Calrec Soundfield Mk4 and TetraMic sound even better.
But to hopefully help your Quixotic quest, I'll recommend

• Avoid the DOAs with LTP i/ps.  They are only a poor attempt at NE5534.
• This leaves stuff like the old Neve & Calrec DOAs with single BJT i/p.  But even then, their designers would say the same thing as above
• One DOA type, I've had some experience of which DOES have small distortion AND sound goods & different is JLH's Class A power amp topology.  Avoid any improvements except by the man himself

 

[audiomixer]

thanks ricardo

your precise characterization of my quest (seems i wear a strange hat here) and your statement about what to avoid and were to look for is exactly what I was hoping for.
I can and will weigh in comments and tips by the rest of the forums threads, be assured. I know that there are many people on forums with a 'made' opinion on things - but I can really differentiate between the interesting post and the less interesting ones... and specially this forum seems full of rather nice people very dedicated to sound equipment, freaks and nerds that can read a schematic, build a circuit etc. so I feel comfortable to listen to your advice.

your suggestion about what to avoid and where to look for clues is helpful, probably meets my own opinions pretty much. listening test would obviously be 'the way to go' but besides costs and feasibility who does them anyway (real double blind tests...)? we all think we have golden ears and can hear all sort of voodoo stuff, no? so I can't rely on that 'objective' criteria either, but I'll ask some of my friends to give me a unbiased opinion when I have at least two different variants.

Arno's questions cannot be answered, as he actually puts it himself - try to define good sound, I can't. I can just pinpoint my wow moments when listening to real gear and specifically asked for your (subjective) opinion and experiences on the two questions I asked,  not about perfect sound.  Now I have heard a strong voicing against low NFB.... and regarding signal quality (measured in THD THD-N, IM etc.)  I wanted to know if someone advocates for the less perfect. but that seems not to be the case either. I will definitively try other topologies - good point, but lots of work in front of me.... I am thinking of two other variations, one Neve like with BJT and transformer and one with simple buffer where possible and maybe a discreet gain stage....


cheers,

- Michael

 

[dirkwright]

But to hopefully help your Quixotic quest, I'll recommend

• Avoid the DOAs with LTP i/ps.  They are only a poor attempt at NE5534.
• This leaves stuff like the old Neve & Calrec DOAs with single BJT i/p.  But even then, their designers would say the same thing as above
• One DOA type, I've had some experience of which DOES have small distortion AND sound goods & different is JLH's Class A power amp topology.  Avoid any improvements except by the man himself

Interesting list, thanks. I should study the JLH design so I can better understand it.

What I have noticed in my sim's is that the single ended LTP usually creates a resonance at some really high frequency and feedback often makes it worse. It's a convenient way to apply NFB though.

Speaking of that, I'm having trouble applying NFB to single ended circuits w/o LTP's. I wind up usually having to use a huge DC blocking capacitor in the FB loop and really low resistor values in the loop as well, neither of which are desirable. Maybe the JLH circuit can give me some clues about that.

For example, if I apply shunt feedback to the base (or gate) of a common emitter (or source) transistor, noise is increased. In order to reduce the noise, I try to lower the resistor at the base but then the other feedback resistor is too small for the output stage to drive. Like the attached circuit. It has a gain of 60dB as shown. The first stage is an SRPP using the LSK170 jfet then directly coupled to an emitter follower. The SRPP is an inverting stage, so shunt feedback to the gate is required for a NFB loop. If I increase the 100ohm resistor, I get more noise, if I reduce the 1k resistor, the emitter follower gets loaded down too much. So, I'm kind of stuck on that.

 

[ricardo]

What I have noticed in my sim's is that the single ended LTP ...

Duu.uuh!  Was is das?

...  usually creates a resonance at some really high frequency and feedback often makes it worse. It's a convenient way to apply NFB though.

Got some examples?

For example, if I apply shunt feedback to the base (or gate) of a common emitter (or source) transistor, noise is increased.

This applies to all shunt feedback.  That's why for most applications, series feedback is quieter.

There is one exception; when the input resistor = 0R.  ie a virtual earth.  This is one of the classic cofigurations for ultimate low noise and should be better known.

In order to reduce the noise, I try to lower the resistor at the base but then the other feedback resistor is too small for the output stage to drive.

This applies to all feedback using evil resistors. 

 

[gentlevoice1]

Resistors ... well they have been known to resist the flow of electrons.  Transistors on the other hand, transist ...  8)

;D  Interesting thought, Ricardo 

 

[Matt Ward]

P.S. I see that you mention an A/D converter. If it is a switched capacitor converter Bruno Putzeys of DSD A/D converter fame (and Hypex switched amps) recommends a 4.5 nF capacitor to ground just prior to the A/D converter so that the capacitors in the A/D converter has something to charge from (as I understand it). I might be able to find a link if you are interested ...

I have found a link to a paper by Bruno Putzeys which I would hope you are interested in.  Unfortunately it's on the less liquid and less spacious negative feedback and the control theory you seem to think is so damaging for your audio.

http://www.linearaudio.net/images/onlinearticlesPDF/volume1bp.pdf

 

[glorymorris]

I'm too lazy to search out Pass' opinions on this, but what you get from low or no NF is the transfer function of the basic circuit topology without error correction. It is impossible to generalize "no feedback" as having a single sound characteristic. The results will be as variable as number of different topologies available. 

I've been less lazy than you ; I checked Pass's view on NFB. He concludes that he prefers a low THD stage with no NFB than a high THD stage with a lot of NFB. Doh!

Perhaps I should have said I wasn't interested enough..  8)

The anti-NFB crowd was one of the sundry alternate religious cults kicking around the fringes of audiophillia over the decades. There wasn't any there-there then, still none now.

If it wasn't for negative feedback (invented by a guy named Black early in the 20th century), they would have never made a workable telephone system and many other technologies we take for granted.

JR

I second this sentiment. Discussions about NFB "sounding bad" are unquantifiable and if you consider yourself an Engineer, I think you'd do well to stick with what's measurable. Terms like "more liquid" and "open" don't state anything objectively.

 

[audiomixer]

I've read your paper (well not yours...bruno's) and that might be a valid view, I don't now for sure. basically it says: do use more, much more feedback! so it's industry trend....

I  tested the suggested simple fet buffer, does work kind of, but I had  to swap some components, so thats difficult to know if it's near the simulated version, got to swap the input fets to something better and give it another go. nothing that I could really use right now. I also got some LME49610 audio buffers, they are as close to what I would like to use (besides being muchos to big @ 250mA!, but hey, bigger's better, no?). I didn't find the time to put them into a working circuit right now, but I will try soon.

the current feedback opamp does look like its for high frequency stuff... is there an advantage at audio frequencies?

next steps will be to build and compare two or three versions - of course with 20+ panel ABx tests  ;-)  or just my ears, hehe...

cheers,

Michael

 

[dirkwright]

I  tested the suggested simple fet buffer, does work kind of, but I had  to swap some components, so thats difficult to know if it's near the simulated version, got to swap the input fets to something better and give it another go. nothing that I could really use right now.
Michael

The corresponding transistors in that circuit have to be matched for best performance because the DC output offsets will not match otherwise. Those are not ordinary diodes either, but CRD's (CLD's). I built one it appears to function fine except that I forgot to match the transistors. Since that mistake lead to some DC current in the output transformer, the bass response was not good in one channel. The LSK170C has an Idss of some 10mA or so. The 2SC2705 is a "power" TO-92 NPN transistor with max 900mW dissipation. When there is no loop feedback to lower the output impedance, I have to really juice up the output stage to reduce distortion and provide good drive to a load.