[ssuh]

Hi everyone, I'm hoping one of the more "electrically advanced" members could explain in layman's terms why an op-amp based amp is "bad" and/or why an amp designed like the Gilmore Lite is preferred. I mean, don't Pimeta based designs use op-amps and aren't Pimeta's supposed to sound excellent?

I've read tons of amp threads and several technical papers that I dug up online, but I can't seem to get just a simple answer (at least one I understand) on the sonic pro's and con's of opamps, other than op-amp based amps are cheaper.

Anyway, thanks for any info and I hope I'm not the only scratching my head at opamps.

 

[kevin gilmore]

All monolithic opamps made by any manufacturer are subject to a few
design rules that must be adhered to. One is that you really can't
have the best pnp and npn transistors on the same substrate at the
same time. If you optimize for the best npn possible which is usually
the case (vertical npn's) then the pnp's end up being lateral. As the
pnp is already slower to begin with due to minority carrier travel time
the chances of perfectly matched pnp and npn transistors is very
difficult to obtain. Usually a pnp is tied to a npn transistor making
a composite pnp which still is not as good as the corresponging npn.

In addition the significant amount of open loop gain may be good for
some applications, but is 40 to 50 db more than necessary for a good
headphone amplifier whose gain is usually less than 10db. The additional
gain increases TIM and other slew rate induced distortions.

For audio use a well designed discrete amplifier will beat any opamp
based unit.

My opinion of course.

 

[thatopampguy]

Hi,

I believe he is referring to "Process Corners" in developing transistors.

I think I can explain this in layman terms, given my recent experience as a frustrated grad student in electrical engineering (vlsi circuits no less).

When you design an opamp, they have to be "fab'ed" at a clean room like MOSIS in Marina Del Rey, CA. The CMOS (twin-tub, p, n) fabrication process involves the laying down of layers of metal or "substrate" to create a particular circuit design. Pits in the substrate are formed via etching via chemical washes. The end result is a tiny chip with all the proper transistors and switches. Care must be taken not to create parasitic current-drawing npn or pnp junctions in the substrate where they were not intended to be.

Now, there's another process called "doping". Doping in simple terms is where you infuse the substrate with a particular ion that is either an electron donor or electron acceptor. The doping process affects the performance in a transistor. There are two types of CMOS transistors, NMOS and PMOS. NMOS pass logic "0" well and PMOS pass logic "1" well. The switching speed is heavily dependant on the doping process.

Often times, you'll see what are called "Process Corners" marked with the letters SS, FF, FS, TT. "TT" means that both NMOS and PMOS are "typical" in their switching speed, i.e. compromises were made so that the switching speed of both transistor types were equal. FS means that the PMOS switches Fast but the NMOS is slow.

In general, if you optimize the process so that the NPN (nmos) transistors switch very quickly and are high performance, you will sacrifice the performance of the PNP transistors, and vice versa.

High gain negative feedback op amps often have harmonic distortion. The measure of this is the IIP2/3 TIM figures. I believe IIP3 is intermodulation for small signal AC and TIM is large signal . . . not sure.

Slew Rate is a figure that is calculated by finding the speed at which the voltage changes per second during a logic transition from 0 to 1 (10% to 90% of swing).

Anyway, hope that clears things up

Disclaimer: I've never built an amp for audio applications in my life. My experience is solely in the field of VLSI design, RF Circuits, and oddly enough chemistry. . .

Thanks,
thatopampguy.

Quote:

Originally Posted by kevin gilmore All monolithic opamps made by any manufacturer are subject to a few design rules that must be adhered to. One is that you really can't have the best pnp and npn transistors on the same substrate at the same time. If you optimize for the best npn possible which is usually the case (vertical npn's) then the pnp's end up being lateral. As the pnp is already slower to begin with due to minority carrier travel time the chances of perfectly matched pnp and npn transistors is very difficult to obtain. Usually a pnp is tied to a npn transistor making a composite pnp which still is not as good as the corresponging npn. In addition the significant amount of open loop gain may be good for some applications, but is 40 to 50 db more than necessary for a good headphone amplifier whose gain is usually less than 10db. The additional gain increases TIM and other slew rate induced distortions. For audio use a well designed discrete amplifier will beat any opamp based unit. My opinion of course.

 

[ssuh]

Thanks for everyone's input. I appreciate it and one day it may make sense. But I've got to say, most of what was posted (at least the abbreviations) went way over my head. It's kind of an ego blow, really, as I used to think I was technically savvy. I may have to consult "Electricity stuff for overachieving dummies."

In the meantime, I'll take everyone's word that discrete designs like the Gilmore Lite are the way to go for amps. But here's one question someone may be able to answer. Why are the Pimeta op-amp based designs held in such high regard if they're sonically "no good"?

-SSuh

 

[Sovkiller]

Quote:

Originally Posted by kevin gilmore For audio use a well designed discrete amplifier will beat any opamp based unit. My opinion of course.

In theory this is the general consensus among the audio gurus, but it is not what happen at all in real life, there are a lot of very good OPamp based amp designs, that evne will beat a lot of discrete amps out there (and sound better) and are more versatile, just look around for the Corda amps, for the Emmeline amps, for the PPA, Headroom amps, etc.....the Blockhead is OPamp based and it is considered one of the best SS amps out there, same as the Prehead....
Of course there are also very good discrete and tube amps, case of your designs Dr. Gilmore, Rudistor amps, Singlepower, etc...among the ones that crossed my mind right now, but IMO you could get a good design and a very good sound eitherway....

And I'm not talking of personal preferences, just objectively and unbiased talking, personally I have my own experience, that I preffer not to talk about it....

 

[Tyll Hertsens]

Quote:

Originally Posted by ssuh but I can't seem to get just a simple answer (at least one I understand) on the sonic pro's and con's of op-amps, other than op-amp based amps are cheaper.

Actually, op-amp designs are not necessarily cheaper. We are about to do some more op-amp listening tests and some of these parts are $15 ea. --- and that's a single op-amp package. In our module we use four op-amps per channel (input buffer, two in the cross-feed, summing amp) making our cost in op-amps alone about $120.

I think the big advantage of using op-amps in a headphone amp is consistency and reliability. It seems to me that, if you buy a very good part, you will get consistent performance from unit to unit in manufacture, and any drift that may occur over time is usually compensated for in the part; so they seem to work well over time.

Discrete designs, OTOH, have to possibility of being built with a minimalist approach, which can result in a very clean and pure audio presentation. But often parts must be selected and matched (which can be costly) and drift over time is not usually compensated for in the circuit design.

Both design approaches can result in exquisite or poor performance depending on the design and manufacture.

We almost have to use op-amps in our design as we include the cross-feed circuit that require all-pass, phase-delay, active filters that would be crazy complicated to do in a discrete design. I very much like some of the op-amps we've found out there; and I think it may be hard to beat them for performance over time with discrete designs. But I do agree with Kevin that, as a fundamental rule, a very good designer could build a better discrete amp than an off-the-shelf integrated circuit.

I think it also might be worth noting that I don't think that it is appropriate to drive headphones directly with most op-amps. They just don't seem to have the umph to do the job well. So some distinction should be made between headphone amps with op-amp outputs and headphone amps that may use op-amps but have real miniature power amps at the output. For a long time HeadRoom used an op-amp based design with a discrete power amp output stage. We've switched to a small power buffer at the output, and we think it is much better sounding than our old design. That's not to say that a better discrete output couldn't be built, but we think it's appropriate for us to make the trade-off for the consistency and reliability of an integrated circuit.

 

[EdipisReks]

Quote:

Originally Posted by TURBO The answer my friend is blowing in wind, the answer is blowing in the wind.

don't be cruel

. speaking of Idiot Wind, i used to quote it after every post by a particular person at another forum i frequent. the person never got it.

 

[WmAx]

Quote:

Originally Posted by Tyll Hertsens Both design approaches can result in exquisite or poor performance depending on the design and manufacture.

Op-amps are evil.

--- it's a standard audiophile rant.

I recommend people to actually review the result of a good circuit using opamps in a headphone amp(BlockHead), as measured by a 3rd party on an Audio Precision Analyzer:

http://www.stereophile.com/amplifica...14/index6.html

Am I supposed to believe(as is claimed by some) that the type of performance indicated here(one thousandeth of one percent total harmonic distortion + noise at 1kHz, 2 whole thousandeths of one percent total harmonic distortion + noise into 20kHz at over 6 volts, or about four thousandeths of one percent of total harmonic distortion plus noise into 1.2 volts as may be used into HD600 or AKG501 during typical listening) is supposed to be somehow inferior to what can be achieved with a discrete design? The residuals, in this measurement set for example, are of such small contribution, that they are immaterial for any practical circumstance. Several magnitudes below what would be audible levels.

People that have an issue with op-amps, I believe, are either just fixated on discrete as some idea of what should be ideal or just swallowing the standard audiophile B.S., hook, line and sinker. Either way, their is no valid/credible data to suggest that a properly designed op-amp circuit will be audibly inferior in any aspect as compared to a properly designed discrete circuit. The only data suggesting otherwise is that born from speculations and audiophile rants. If one considers that credible, God help them. But I , for one, require substantiation derived from a logical perspective, to demonstrate a high probability that a claim is valid, before I'll give it consideration.

-Chris

 

[Sugano-san]

Quote:

Originally Posted by Sovkiller [...] there are a lot of very good OPamp based amp designs, that evne will beat a lot of discrete amps out there[...]

Those discrete amps that are "beaten" by opamp based designs probably don't meet Kevin's definition of "well designed discrete amplifier".

 

[Tim D]

A lot of DIY arguments are made on 'paper', but not often matched up mano a mano to precision measurement gear. And I'm not talking a soundcard and RMAA.

The Benchmark DAC 1 was slammed for using 'cheesy' opamps and looking bad according to 'experts', but when you consider the sheer amount of gear out there that is unable to come even close to 16 bit, not to mention a true 24-bit noisefloor you have to question how expert these opinions are. And if you make the slightest mistakes in your DIY DAC even though it looks great on paper and uses good parts, I am pretty sure you won't come close to achieving the near 24 bit resolution either.

Not saying there aren't truly great bargains in DIY, but I feel it sometimes goes a bit overboard and qualified more by purely 'technical' discussions as opposed to technical measurements not to mention plain listening. It was obvious for example that Benchmark knew what they were doing, and the product clearly performs as a 'benchmark' via qualified 3rd party measurements to make some 'expert' opinions quite suspect. And I'd rather trust that benchmark, than the kid that is familiar with the cost of some DIY parts and has RMAA and at most a $200 sound card.

Whatever happened to the Headroom offer of letting some guys send in their gear to see how the stuff really measures up?

 

[jpr703]

Different designers come from different schools of thought, and they're entitled to their opinions. They have reasons for thinking these things and their opinions are what allow them to take a certain type of design and push it to new heights. Whether it's Kevin's fantastic work with discreete designs like the Dynahi or Ray's work with opamp designs like the well respected SR-71, there are good things that come out of the different design philosphies that these guys take the time to develop. I for one am glad that they have opinions and I hope that they continue to push the envelope in the areas that they choose to specialize in.

Thanks Kevin, Ray, Phil and all the others out there who work hard to make the music sweeter to everyone.

 

[Sovkiller]

Quote:

Originally Posted by Sugano-san Those discrete amps that are "beaten" by opamp based designs probably don't meet Kevin's definition of "well designed discrete amplifier".

In your shoes I won't be so sure of that...!!!

 

[kevin gilmore]

Performance measurements, especially those done with a sound card
these days are suspect at best and are just one of the many tests
that could be performed on an amplifier. Opamp or otherwise. I still
cannot get the same distortion numbers from a sound card (even very
expensive ones) and my st100a which is almost 20 years old. The
st100a consistently reads 10 to 20 times more distortion. And its not
the unit itself as i have tested it with pure reference oscillators that
don't even budge the meter.

Many power amps in the late 1970's bench tested very well yet sounded
like crap. Extremely harsh and nasty. Many of the best sounding amplifiers
of today actually have higher distortion figures than some of that old
stuff.

Captain Crossfeed speaks the truth with respect to opamps directly driving
headphones. Of the available buffer boosters most people like the 5002
part the best. Still it is limited to the amount of power it can deliver.
Same thing with a buf634. Wonderful for portables. Or stack a few
and make a very nice home amp.

I build discrete stuff. Uncle phil builds discrete stuff. Some use opamps
and discrete output stages. Some use tube front ends for solid state
buffer amps. Pick your poison.

 

[Canman]

It's all in the implementation of each respective design.

 

[WmAx]

Quote:

Originally Posted by kevin gilmore Many power amps in the late 1970's bench tested very well yet sounded like crap. Extremely harsh and nasty. Many of the best sounding amplifiers of today actually have higher distortion figures than some of that old stuff.

Bench tested well in the 70's? The same 70's, before anyone even bothered to plot THD vs. Frequency or other critical detailed analysis parameters as a standardI(usually just a 1kHz fundamental), and thus missing obvious slew induced distortion(TIM) by-products at higher frequencies, etc.? If the amplifier sounded 'wrong' under double blinded, level matched conditions, then it will measure 'wrong', at least that is the only probable conclusion that can be made according to established perceptual data; I.E.: no one has conclusively demonstrated a *mystery* parameter to exist to this date, of which, I am aware.

-Chris