Opamp vs. Buffer vs. descreet Class A

Hi,

when designing my mixer project I resorted to use conventional opamp circuits. First I had a high voltage class A opamp in mind (think SPL), then i found that this would only give me headroom above 24dBu - so with my AD/DA converter this is of no use... then I looked into lowering the THD and noise of my design to allow for the best performance. this is nicely achievable with todays modern opamps, their measured specs are really excellent. but hey, the neg feedback involved in these is really large. I looked into the voltage follower buffers as I need only little gain and could put all gain I use in one discrete stage at the input, but again, will it sound any better?
then there is this large group of DOA API (et. al) users out there. by todays measurement  standard a medium quality opamp, but it appears to sound nice for some of you....
then there is the quite large fraction of users that have buffers and discrete gain stages in the design. lower overall feedback appeals to me, I never known why, maybe because the feedback always is about canceling THD, but is it also about getting good sound?

so two questions:
any experiences with buffers, discrete class A, high voltage opamps? why (can) they sound better? did you A / B some designs?

does the measured specs really reflect the optimum regarding nice sounding gear? I know we sometimes like the added sound - so should we aim for more distortion then the 0.005% you can get with these modern monsters?


would be glad to hear from your experiences, but hey, no need to bash at each other, no flame war, ok...



- michael

PS: I think I will have to build my mixer with a 'less feedback' topology to test  A  vs. B - kind of have to know the sound that could have, huh!

"Sound better to you" is a personal subjective decision only you can answer.  Argue with yourself.

JR

what are you trying to achieve?

Signal "integrity", ie a purely functional output amplifier that drives long cables, low impedances and manages to handle high output levels with as little distortion as possible added

or distortion, which you subjectively find pleasing?

you have to define this for yourself, before you talk about "good sound". otherwise you'll be running in circles...
no one will be able to help you unless you clearly state what YOU define as "good sound"

Yea, the discrete opamps that people like so much is for the distortions....not something I would necessarily want on converters.. (using the whole idea of keeping convetors and "clean"  as possible and then using preamps, compressors for color) But then again everything is adding something. Everything has a sound....subjective of course. Some 990s might be nice as they are actually pretty clean.

There may be some merit, in addressing the transfer function around electrical saturation in the analog domain before conversion. A soft clipping that takes several dB to turn hard, could be useful for some scenarios in increase effective headroom or dynamic range. For clean linear operation don't overdrive the convertor, if you want to get a little nasty, hit it hard. 

This approach made more sense back before we had so many bits, but might be of some interest. (IIRC DBX may have done something like this in some products, probably to make noise floor look better).

JR

audiomixer said:

this is nicely achievable with todays modern opamps, their measured specs are really excellent. but hey, the neg feedback involved in these is really large.

Click to expand...

What's wrong with negative feedback?

-a

that's a good question I'd like to ask.....

negative feedback is an excellent method of getting rid of harmonic distortion, but is it 'musical' in terms of sound (again, not trying to define that in a definitive way).

I designed a Bass Amp (5W class A, no feedback) around the (very inefficient)  ZEN amp concept and it's sound is great (to my taste...) - Nelson Pass describes his view of feedback vs. distortion on his site - passlabs.com. Now I am still kind of ripped apart between the two extreme - just wanted to know if any has made the effort to A / B that kind of question. can we get better sounding result with simple circuits in a mixer application, is it worthwhile trying?

to JR:
It's not about my specific music taste, as this is really not a criteria that can be defined in a discussion - valid point.

- Michael

audiomixer said:

that's a good question I'd like to ask.....

negative feedback is an excellent method of getting rid of harmonic distortion, but is it 'musical' in terms of sound (again, not trying to define that in a definitive way).

Click to expand...

You need to define "musical" is a way that relates to signal transfer functions. to get an answer in that context.

NF allows designers to trade voltage gain for accuracy, i.e. output will be the same as the input at some lower voltage gain or whatever intended transfer function you define.

I designed a Bass Amp (5W class A, no feedback) around the (very inefficient)  ZEN amp concept and it's sound is great (to my taste...) - Nelson Pass describes his view of feedback vs. distortion on his site - passlabs.com. Now I am still kind of ripped apart between the two extreme - just wanted to know if any has made the effort to A / B that kind of question. can we get better sounding result with simple circuits in a mixer application, is it worthwhile trying?

Click to expand...

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. 

to JR:
It's not about my specific music taste, as this is really not a criteria that can be defined in a discussion - valid point.

- Michael

Click to expand...

JR

JohnRoberts said:

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.

Click to expand...

Excellent John, beautifully succinct.

For me, it's all about the "voicing" of the entire chain. I have IC op-amp's in there, both regular and CFB, a push-pull tube stage with zero NFB, discrete op-amps.. They all sound different. Nothing wrong with NFB, and sometimes nothing wrong with not having any feedback at all.

IC op-amps are a grat starting point. There's plenty, they sound different are easy enough to swap around. If you put a 2520 footprint in there, the choice gets even wider. You could also design just about any kind of a discrete line amp in the 2520 footprint.

I suggest experimentation with the actual circuit. Takes sime time, though. My chain is in a constant development.

Viitalahde said:

For me, it's all about the "voicing" of the entire chain.

Click to expand...

Of course, what else...

But let me post a smart-ass remark: whenever you have a "no nfb" single-ended gain-stage (f.e. common cathode, source or emitter), there is a chance that in reality, it isn't a "real" no-nfb gain-stage. Any cathode, source or emitter degenerative resistance is going to introduce some form of nfb, even if it is "bypassed" with the baddest low-esr cap in the world.

Heck, even the lead self-resistance introduces some nfb.

0.0000000009dB.

0.000000000009dB with SMD parts.

But that's the price of progress!

tv said:

whenever you have a "no nfb" single-ended gain-stage (f.e. common cathode, source or emitter), there is a chance that in reality, it isn't a "real" no-nfb gain-stage.

Click to expand...

Obviously, you're right. And frankly I don't even care as long as things sound good.

It's wide variety of active parts I have in my chain, and some of it doesn't measure too well but sounds just right. The extreme example would be the Neumann W495stb EQ, I think there are 4741 op-amps in there, which to my understanding is a quad 741 op-amp.

The NTP discrete op-amps, they sound wonderful. I have no idea how they measure up, but they aren't an 990 for sure. The Forssell 993's don't have the lowest distortion either.

Then there are modern IC's. I love the 49713 CFB op-amp. Sounds just right to me.

47913 - me too. If you have the metal-can, TO99, you can put on a small star-heatsink and load it with some nice fat resistor to smooth it out nicely.

741 metal-cans can be treated similarly if you're into vintage vibe.

audiomixer said:

Hi,

when designing my mixer project I resorted to use conventional opamp circuits. First I had a high voltage class A opamp in mind (think SPL), then i found that this would only give me headroom above 24dBu - so with my AD/DA converter this is of no use... then I looked into lowering the THD and noise of my design to allow for the best performance. this is nicely achievable with todays modern opamps, their measured specs are really excellent. but hey, the neg feedback involved in these is really large. I looked into the voltage follower buffers as I need only little gain and could put all gain I use in one discrete stage at the input, but again, will it sound any better?
then there is this large group of DOA API (et. al) users out there. by todays measurement  standard a medium quality opamp, but it appears to sound nice for some of you....
then there is the quite large fraction of users that have buffers and discrete gain stages in the design. lower overall feedback appeals to me, I never known why, maybe because the feedback always is about canceling THD, but is it also about getting good sound?

so two questions:
any experiences with buffers, discrete class A, high voltage opamps? why (can) they sound better? did you A / B some designs?

does the measured specs really reflect the optimum regarding nice sounding gear? I know we sometimes like the added sound - so should we aim for more distortion then the 0.005% you can get with these modern monsters?


would be glad to hear from your experiences, but hey, no need to bash at each other, no flame war, ok...



- michael

PS: I think I will have to build my mixer with a 'less feedback' topology to test  A  vs. B - kind of have to know the sound that could have, huh!

Click to expand...

What's the load impedance for your circuit? Is your source balanced or single ended? I may have something you can use if you want to.

Source is balanced, 15dBu. impedance in the signal path quite low, with 3k3 suming resistors. I would love to see your design and maybe test it.

- Michael

audiomixer said:

Source is balanced, 15dBu. impedance in the signal path quite low, with 3k3 suming resistors. I would love to see your design and maybe test it.

- Michael

Click to expand...

So that's 4.4Vrms input, with a load impedance of 3k3, and you want a 0 dB gain buffer? I don't think I have something that can handle that right now. Let me think about it some more. I'm just an amateur and I'm not completely familiar with pro audio.

OK, I'm surprised. The sim for this shows good performance with 4.4Vrms balanced input driving a 3k3 load. It's a very simple buffer. The diodes are CRD's (current regulating diodes). It's a bridge circuit with just a couple of cascaded followers on each side of the bridge. I just ordered some boards for this from Pad2Pad.

Since the OP was talking about a buffer to hypothetically drive an A/D, how do you model the non-linear loading of a converter input S/H switching? I remain apprehensive about SIM to evaluate high performance designs, unless you fully characterize the design constraints.

JR

JohnRoberts said:

Since the OP was talking about a buffer to hypothetically drive an A/D, how do you model the non-linear loading of a converter input S/H switching? I remain apprehensive about SIM to evaluate high performance designs, unless you fully characterize the design constraints.

JR

Click to expand...

They have spice models of A/D converters. I've never used them. I assume they behave in a similar way to real ones. I don't know the characteristics of the non-linearity of the converter input, so I don't have a clue how the circuit would behave. Since there's no loop feedback, I assume it would be stable and limited only by the amount of current it can dump into the load. This design is not suitable for higher power applications.

Obviously, this design is limited by transformer performance, since it's rare to find one with less than 0.001% distortion.

Hi,

Without going too much into your talk here I'd say that in my years of audio simplicity and an unusual attention to detail has (practically) always been my way to obtain superb results. In my experience anything & everything matters if the goal is to be really superb. I don't know if that is where you are heading but some of my experiences have been:

- no feedback (more liquid and spacious sound)
- no complementary circuits (one transistor clapping as I remember Nelson Pass once said). I guess this may have to do with transistors never being identical
- no parallelling of transistors - same consideration.
- best possible components (as decided by listening tests). Some components work "here" but might not work elsewhere. I have had very good results with Panasonic HFZ capacitors, RFZ resistors (I think they were called - used for hearing aids), Panasonic copper foil capacitors (low value), and some also mention the new Rubycon capacitors as something special. Plugs, connectors, wire insulations, PCBs, etc. have a HUGE impact in my experience. I personally hardwire my circuitry (no PCBs where possible) and use uninsulated copper.
- simplest possible circuitry with high current in the transistors. As a point in time I did some listening tests and found that most all small signal BJTs I listened to (Japanese types) sounded best (to my ears) when I divided their max current current with 13-14 times.
- I have never experienced ICs to be as open & transparent & fluid as discrete circuitry - my guess is that probably this is because most all of the include feedback, complementary circuits & low currents.

Well, a few comments on my choices in this. Good luck with your endeavors ...

Jesper

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 ...