Op-amps integrate at least 4 stages of gain into one package. When you try to put feedback around that much gain and phase shift, you have to haul down the "dominant pole" way down where the rest are out of sight. This usually means serious slew limitation, so early op-amps were notorious for slew induced distortion. Then there is the lack of good PNP transistors on an N-type silicon chip, and asymmetric outputs with ~class-B crossover distortion. Complicated circuits tend to have nasty surprise behavior like overdrive phase inversion and latch-up. Even op-amps commonly used in audio circuits have issues like supply sequencing and decoupling. A lot of gotchas like these scare away amateurs or leave them disenchanted. Common chips are limited to about 40 VDC and they are not fast enough to be part of feedback around power transistors. The parts (transistors) needed to adapt an op-amp into a power amp over ~20 Watts are as much or more than a discrete circuit. It takes about 4 transistors to replace the op-amp and the results are better, and the voltage, therefore power limitation goes away.
There are virtually no op-amps with 4 gain stages, a buffer does not have gain.
I think you missed about 4 decades of progress here. Jung's SID article was from, what, 1972?
One of the best audio opamps available has just one gain stage. 100V/uS is routinely available in a very small package for a buck or so.
You really need to read up.
Jan
Hi Steve,
Much of what you say was largely true of early op amps of the 1970's and before, especially in respect to the need to use lateral PNP transistors because verticals were not available at the time. However, beginning largely with the availability of the legendary 5534, most of these limitations were overcome. The 5534 was an especially remarkable accomplishment, in my opinion, because it achieved its performance without vertical PNPs.
I tend to agree with some of your arguments regarding the use of op amps in power amplifiers. It is usually not worth it, given the voltages usually present in a power amplifier when an ordinary op amp is used as the IPS/VAS portion of a power amp. There are a number of exceptions, of course.
I think the main emphasis in this thread is in regard to the use of op amps in line-level applications, however. There, modern op amps are hard to beat.
Cheers,
Bob
Bob, an under-appreciated feature of using an opamp chip inside a power amp is: you can actually buy super low noise, JFET input opamps. It's getting harder and harder to buy low noise discrete JFETs.
Once you've decided to build a two part amp with IC opamp first part, you can use a very very low impedance feedback divider in the discrete circuit power amp (for lowest noise). This low Z feedback network drives the IN- port of the discrete amp. For input offset matching you want the same low impedance on the IN+ port of the discrete amp ... and because the opamp has lots of output current, the opamp is easily able to drive the low impedance IN+ port. Eureka.
Partitioning the gain between two amplifiers also permits greater closed loop bandwidth of the overall assembly. Instead of asking the discrete amp to have a gain of 20X, and thus achieving a closed loop bandwidth of (funity / 20), you can arrange the opamp stage to have a gain of 4X and the discrete amp to have a gain of 5X. Now the closed loop bandwidth rises to (funity / 5). More HF feedback --> more HF distortion reduction --> more better.
You could fool around with inverting amplifier topologies too, if interested in exploring common mode distortion or the lack thereof.
Once you've decided to build a two part amp with IC opamp first part, you can use a very very low impedance feedback divider in the discrete circuit power amp (for lowest noise). This low Z feedback network drives the IN- port of the discrete amp. For input offset matching you want the same low impedance on the IN+ port of the discrete amp ... and because the opamp has lots of output current, the opamp is easily able to drive the low impedance IN+ port. Eureka.
Partitioning the gain between two amplifiers also permits greater closed loop bandwidth of the overall assembly. Instead of asking the discrete amp to have a gain of 20X, and thus achieving a closed loop bandwidth of (funity / 20), you can arrange the opamp stage to have a gain of 4X and the discrete amp to have a gain of 5X. Now the closed loop bandwidth rises to (funity / 5). More HF feedback --> more HF distortion reduction --> more better.
You could fool around with inverting amplifier topologies too, if interested in exploring common mode distortion or the lack thereof.
As far as line level preamps go I'll chime in. I have been using mostly tube preamps for years now with great results. Recently I wanted to experiment with a solid state design and built the Doug Self preamp using opamps. It uses many opamps in the signal chain but I think it sounds fantastic. I have had it in my system for going on a year now without going back to my tube front end. I'm not saying it is better sounding but it is just as enjoyable as the tube unit. I would not hesitate to build with the modern day opamps that are out there.
Hi Mark,
These are some good points. I am definitely a fan of JFET inputs, as I have outlined in my book. However, I generally use the Linear Systems LSK389 and LSK489 dual monolithic JFETs. They are readily available and have quite low noise. For amplifiers with higher rail voltages, they do need to be cascoded. For fully complementary input stages, they also have the LSJ689 P-channel dual monolithic JFET pair.
We should also not overlook the possibility of putting JFET source followers in front of a BJT IPS LTP. The source followers can be powered from only a +15V supply.
If one is using an op amp as part of the input portion of a power amp, there are, of course many ways to do it, including whether the op amp is inside or outside the global feedback loop. Using an op amp up front to contribute some of the closed loop gain is not necessarily a good thing, since in many amplifier topologies reducing the closed loop gain while maintaining the same ULGF will degrade the achievable slew rate.
Cheers,
Bob
I can believe that. The majority of tube preamps i've seen published here really suck.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.