Simplest discrete op-amps

[bcarso]

[quote author="JohnRoberts"][quote author="ruffrecords"]
This is completely wrong. Global feedback does not compromise bandwidth.

Ian[/quote]


No, but the internal compensation required for stability with negative feedback does indeed.

I'm no student of scope schematics so the following is mostly WAG. They probably use less overall NFB and apply it locally rather than globally to minimize this constraint. They may also favor phase and amplitude response preferentially over extreme linearity as we are not going to eyeball nonlinearities more than say 30-40 dB down. or not...YMMV

JR[/quote]

Thank you John.

Let me be a bit more precise. Scope amps are an example of wideband lowpass circuits which attempt to get acceptable transient response, usually aiming for Thomson filter shape, i.e., low overshoot and fast settling, with the maximum gain bandwidth from the active devices that are available at a given time.

Negative feedback in the sense of Black's invention is typically viewed as a method of reducing gain in exchange for reduction of distortion and improvement in gain stability. And in order to assure stablity against oscillation the open loop response in terms of gain and phase has to meet certain criteria, the details of which are well-represented in books and courses. Additional considerations must be met as well in terms of the feedback network itself, and the overall resultant topology is what is studied to determine the transient behavior.

If I am building a scope amp and want it to have significant gain and reasonable stability, I will more likely get there by cascading gain blocks with relatively simple gain paths, and not chaining a bunch of stuff together and applying negative feedback around the lot of it. And that's what I was alluding to when Lazy mentioned his scope differential amplifier as being something that might approximate an opamp. It is not likely to do so, and the accumulation of phase shifts if it follows typical scope amp practice will not lend it to use as an opamp per se---that is, it will not tolerate the application of global feedback of the typical sort.

 

[LazyOne]

your probably right.
the only thing i see to prevent it form oscillating is that at the beginning when te signal enters the amp there are two coils in series with the input before the first attenuator circuit (one for + and one for -) they are 1.8µH.
probably to stop HF entering the circuit.
then at the out there are som small caps at the cathode followers and it is DC coppled. all the stages are dc coupled (it has to be to use it as a dc amp when the switch is in dc)
altough if i probablt wanted to use it as an opamp it can start oscillating. maybe i'll try see what it gives
maybe a nice experiment. i can wire it and i can see the output right on the screen.
wire an extra pot for the feedback so that i can adjust it and see what it does.

there was an opamp with tubes that was sold in those days. it was the phillbrick kw2 or something.
maybe also a good start. altough the gain is less

 

[ruffrecords]

[quote author="JohnRoberts"]Sure, if the closed loop gain is high enough stability criteria may be met without additional internal roll-offs.

If we are going to wrestle over semantics indeed "internal compensation" is not a specific constraint for stability, just a common technique to roll-off HF gain before it becomes problematic when using lots of negative feedback. [/quote]
My apologies. I came into this thread rather late and did not look closely enough at preceding posts. It's just I have seen so many unqualified 'global feedback is bad' posts elsewhere that they just get up my nose. I had hoped this group was better versed in stability criteria and from your post is clearly is. Again, apologies for going off the deep end.

Cheers

Ian

 

[PRR]

> Define opamp.

Slew that. Why are we even considering Operational Amplifiers for audio?

The op-amp is a universal building block for synthesizing any mathematical function with lossy parts. It is for a computer. It can add values, integrate varying functions, approximate LOGs, and implement break-functions such as rectification.

In Audio, we need a reasonably stable gain, usually low. We need poles/zeros but they rarely have to extend to infinity. LOGs and breaks are generally Bad Things. And we need Low Noise. And we do NOT want DC response until all stray sources of DC are elimiated from all sources.

> diff amp--..resistive "tail"..voltage amplifier...output section.

Yeah. 741. Still a good amp (once you lose the lame power-PNP from 1969). The current source in the volt-amp is only to get a stupendous DC Gain number, and has little audio advantage. The current source in the long-tail is needed for low-gain followers, not for high-gain or inverters. And arguably the diff-input is a gilded-lily for practical audio use. If you can drive a load, you can NFB to the input emitter, lose a noise source, and cap-couple your DC troubles away.

If you must design ONE building-block to cover MANY jobs (chip-making demands such silly notions), then yes the op-amp is the block to build. That argument holds no water when I solder-up one or two boosters.

If you must design one module to cover any Audio problem... you know the answer already. 35dB mike preamp with beefy output. Langevin, RCA, etc.

> wise to be a little cautious of any ...design that drives one specification into the dirt as it may be at the expense of another.

Such as the absurd DC-Gain specs on early IC op-amps. Older bottle and module op-amps could barely turn a gain of 20,000, which makes gain error a real issue. Make transistors super-cheap, you can get DC Gain over a million, WOW! Or because the bits are so small, you can force power consumption way-way down, WOW! 709, 301, 741, 709: thin, gritty, hissy audio junk.

> What is a little hard to explain is why the integrated opamps don't seem to perform "sonically" as well as the discrete opamps

It is easy to explain. But too many people (even some very wise people) can't hear and won't listen. And when understanding dawned, the audio market wasn't large enough for more than token interest. The exceptional 5532 may have damped development of even better types. Oddly what has helped in recent years is the queer ADSL market: meeting ADSL specs seems to give nearly fault-free audio.

> a rather good GP discrete opamp ...by Deane Jensen

"Rather good" indeed. Unsurpassed for nearly 30 years. One of the current licencees has better documentation of the basics (not the hand-tweaking). There's a non-Jensen FET variant with large DC offset which is good too. NYD could do a lot worse than cobbling-up a bushel of NYD-990s.

If you must DIY non-heroic Audio opamps: Doug Self covers the waterfront. There's no basic difference between his canonical power amp and an opamp except scale. All the basic considerations (and common subtle mistakes) are in there. Just scale-back.

> gyrators in the Avedis EQ are 3-transistor based.

An awful lot of graphic EQs use one-transistor gyrators.

> these days the GP opamps cover so many bases so well

Quite true. I actually wouldn't sit and fret "are 3 transistors enuff? Is 7 too many?" I can drop a chip with fewer pins than 3 transistors, be done, and get better test numbers.

OTOH, I just spent a week soldering an Octal socket and accessories. There is a geeky satisfaction in putting my finger on a grid wire and feeling the actual signal in the middle of a 3-tube amplifier. It isn't the same as poking the COMP pin on a LM301. Be One with the Audio, grasshopper. Anyway, I can still see Octal lugs. My DIP days are behind me, and SOICs are just grains of dirt. (And you know the old 4-pin socket fans whined about Octal, then Miniature....)

> Cost/benefit is at the heart of all engineering.

Ah, but you can't engineer Art. We have some artists here, and some wannabes like me.

Cost/Benefit? My 3-tube amp makes 13 Watts, gulps electricity, will cost over $500 before decoration. On cost/benefit analysis, I should trot over to the pawn shop and get a slightly-used 15-Watt guitar amp for $50. If they're out, Banjo World has the chip-amp for $99, a 2-bottle gitar amp for $130, or a 15W bottle-amp for ~$300. I'm an economic fool. However, so far I feel broke-even just from the odd fun of sniffing rosin (and puzzling a novel variation on a theme which was done to death 50 years ago). If I turn it loose on a good guitarist and s/he finds some "soul" in it not found in Banjo World's slapped-together amps, it'll be an emotional profit.

> maybe it's not even necessary to stay at exactly 0 volts DC at all times.

Hmmmmm..... (he ponders while feeling the plate of a 12AX7).

> the mindset in consumer hifi where they tend to ignore the common knowledge of decades ago expecting to stumble upon some near magical breakthrough.

The audio community has forgotten far more than they know.

Loudspeaker "gurus" really peeve me off. There's always a new breakthrough which defies basic physics clearly outlined in 1925. Followers of Theile/Small manage to completely miss the point.

And advances in vacuum tubes (other than cost) after 1922 have been very few. Maybe just one: heater-cathode construction. (Yes, I do like the pentode as a power tube for guitar, but for "good audio" the pentode is a dubious thing.) The Bell Labs boys didn't leave much ground un-covered, and what crumbs they left were cleaned that decade. (Apropos of opamps: the 1922 paper shows the long tail pair and describes its DC advantage.)

> If you want to advance the art, stand on the shoulders of those who have gone before.

The way things are today: if more of us would just stand on the shoelaces of the pioneers, we'd have less BS and more good sound. I don't mean using type '27 vacuum tubes; chips have real practical merit. Learning how to hear, what we can hear, what pleases the ear, rather than focusing on mere parts.

 

[ruffrecords]

[quote author="PRR"]If you must DIY non-heroic Audio opamps: Doug Self covers the waterfront.[/quote]
Doug is definitely the master when it comes to finding out what is really going on.

There's no basic difference between his canonical power amp and an opamp except scale. All the basic considerations (and common subtle mistakes) are in there. Just scale-back..

And better still run the output stage class A.

Ian

 

[JohnRoberts]

[quote author="PRR"]
> Cost/benefit is at the heart of all engineering.

Ah, but you can't engineer Art. We have some artists here, and some wannabes like me. [/quote]

While perhaps not art per se, there is an elegance in accomplishing a circuit function well with a minimal number of parts. This is often consistent with economy and compactness, but modern manufacturing has tilted the cost calculus in favor of more expensive larger scale integration to reduce insertions and save real estate. IMO circuit design is a little like the film makers craft. It's machinations should be invisible to consumer while delivering sundry benefits.

[quote author="PRR"]
Loudspeaker "gurus" really peeve me off. There's always a new breakthrough which defies basic physics clearly outlined in 1925. Followers of Theile/Small manage to completely miss the point.

[/quote]

Loudspeaker technology seemed frozen for decades with only incremental improvements in materials (new cone materials, square VC wire, higher temperature adhesives in VC, etc.). There seem to be a few competing recent developments in horn design that look to me like they are nudging performance forward again. (Danley- Unity horn technology, and Gunness @ EAW with DSP correction of horn distortions).

These are both related to larger scale sound reinforcement rather than home hi-fi. Just like power amp design is a good model for opamp theory, live sound reinforcement may offer a similar parallel. As a friend of mine likes to say, "it's hard to ******** thousands of people at the same time."

JR

 

[Svart]

That has to be one of the best PRR replies yet. :green: I always look forward to his postings.. :thumb:

As far as ADSL opamps are concerned, speed and type of feedback are of the most importance. something that most tend to overlook or give no credit to when designing. 5v/us might be "good enough" for CD audio but there is a whole other world of issues that this could create.

I'm now working in the RF world and have already found incredible problems with slow opamps. Recently I was working on a peak hold circuit that clearly was not catching certain peaks (vertical sync vs. horizontal sync) in a video signal. the output was bouncing a couple of db and throwing the calibration off. the parts were opa340, 6v/us which technically was fast enough for the signal, refresh at 60hz and vsync tips at 1khz. To make a long story short, the hsync tips were only 1us wide and the opamp couldn't slew fast enough to recreate clean edges and simply rounded them off. The result was an average that bounced because the syncs were mismatched even though the input amplitudes were the same. Current drive from the opamp was also an issue but less so than the sheer speed.

The fix? different opamps with 4 times the bandwidth and 3 times the slew. And this was to get an opamp to follow 1us pulses spaced at 1khz.

Imagine if your opamp needs to recreate a complex signal at 20khz. Lets say that you have a few very tiny transients on the 20khz carrier. If we follow my findings that would mean that you would want to slew at least 60v/us just to get in the ballpark of recreating a signal at 20khz with some of it's transients and upper harmonics.

 

[Martin B. Kantola]

[quote author="Svart"]That has to be one of the best PRR replies yet. [/quote]


It's a Piece Of Art. Still digesting, totally unable to respond to the actual issue. But I will. Soon. Have to stop giggling first.

Martin

 

[JohnRoberts]

[quote author="Svart"]That has to be one of the best PRR replies yet. :green: I always look forward to his postings.. :thumb:

As far as ADSL opamps are concerned, speed and type of feedback are of the most importance. something that most tend to overlook or give no credit to when designing. 5v/us might be "good enough" for CD audio but there is a whole other world of issues that this could create.

I'm now working in the RF world and have already found incredible problems with slow opamps. Recently I was working on a peak hold circuit that clearly was not catching certain peaks (vertical sync vs. horizontal sync) in a video signal. the output was bouncing a couple of db and throwing the calibration off. the parts were opa340, 6v/us which technically was fast enough for the signal, refresh at 60hz and vsync tips at 1khz. To make a long story short, the hsync tips were only 1us wide and the opamp couldn't slew fast enough to recreate clean edges and simply rounded them off. The result was an average that bounced because the syncs were mismatched even though the input amplitudes were the same. Current drive from the opamp was also an issue but less so than the sheer speed.

The fix? different opamps with 4 times the bandwidth and 3 times the slew. And this was to get an opamp to follow 1us pulses spaced at 1khz.

Imagine if your opamp needs to recreate a complex signal at 20khz. Lets say that you have a few very tiny transients on the 20khz carrier. If we follow my findings that would mean that you would want to slew at least 60v/us just to get in the ballpark of recreating a signal at 20khz with some of it's transients and upper harmonics.[/quote]

The smoothing or rolling off of out of band edge rates is a good response for audio processing. For audio you only need to pass full audio power bandwidth while harmlessly LPF the "10-4 good buddy" out of band signal. Of course there may be some debate over what the appropriate audio bandwidth is. For nonlinear crunching different strokes.

That sounds like an argument for perhaps using an application specific opamp. While focus on slew rate over rise time is almost code for questionable audio performance (IMO), for nonlinear applications like S & H you are less concerned in how it gets from here to there and more on getting there fast with good settling characteristics after you arrive.

An opamp with non-degenerated bipolar input devices while less than optimal for audio may be better for your app. To wit a bifet opamp with a couple volts of Vth won't deliver full slew rate for only 1v p-p video (just guessing about the 1v I haven't messed with too much video). Also in addition to raw speed look for anti-sat clamps on inter stage circuitry that will allow max slew rates with quick recovery to prevent excessive overshoot (look at large square wave response on data sheet).

There may be a way to search for data sheets that cite S/H in their target app. As I recall there were several parts back in '70s- '80s that looked like they were designed for working with early A/D conversion which often involved rolling your own S/H front end to give SARs time to finish their conversion.

JR

PS: you may learn more than you want ot know about capacitors too (like dielectric absorption).

 

[Svart]

Very true John, I was meerly asserting that to create an identical signal, sometimes we need a device that performs well over the accepted criteria. With rise time, I also expect that the swing up/down will also capture the small details as well.

Smoothing of the signal as well as distortion, pleasing or unpleasing is something I wasn't factoring into the equation just yet due to the grossly wide array of opinions on these matters. Sometimes the sound you capture is accurate but it just doesn't sound good. I find that changing an accurate but bad sounding audio signal is a job for effects.

As for the dialetric absorbtion, we moved over to a polystyrene cap for the hold. works well now.

 

[JohnRoberts]

[quote author="Svart"]Very true John, I was meerly asserting that to create an identical signal, sometimes we need a device that performs well over the accepted criteria. With rise time, I also expect that the swing up/down will also capture the small details as well.

Smoothing of the signal as well as distortion, pleasing or unpleasing is something I wasn't factoring into the equation just yet due to the grossly wide array of opinions on these matters. Sometimes the sound you capture is accurate but it just doesn't sound good. I find that changing an accurate but bad sounding audio signal is a job for effects.

As for the dialetric absorbtion, we moved over to a polystyrene cap for the hold. works well now.[/quote]

Sorry I was not very clear as I was applying linear audio considerations to a non linear HF application. By "good" smoothing, I simply meant LPF of out of band signal. This is not distortion in the analog sense of added harmonic components, but perhaps an unacceptable distortion to the wave shape for nonlinear crunching (or more precisely LPF too low for the app).

Yes, polystyrene is a sweet dielectric, my only criticism is it's poor robustness in manufacturing. I've had problems with heat tolerance (same with some other plastic film types) and with polystyrene even had some corrupted by wash water when circuit boards were dried off with too much air pressure (had an old business partner who thought there was no such thing as too much air pressure :roll: ).

I am trying to convey that there are some opamps designed to be operated transiently at full slew with minimal artifacts as compared to general purpose audio opamps where if you encounter slew rate limiting it's the wrong part for the gig.

I've even seen some power amps (opamps on steroids) that disable their input low pass filter just so they can measure and publish a slew rate specification (and keep consumers happy). Perhaps this is just my personal problem but for audio I see slew rate limiting as clipping in the "rate of change" domain. Since you're describing a non-audio application this criteria doesn’t apply.

I hate to lean on cliché analogies but perhaps this is a little like the difference between a dragster with drag parachute for brakes, compared to an offy race car with more nuanced rate of change. A dragster that can stop well is like a high slew rate opamp with little overshoot. While an offy car will beat any passenger sedan at the strip, a dragster is the right tool for that job.

JR

PS: To beat this anaolgy to death,,, audio is not a drag race...

 

[Guest]

[quote author="PRR"]> Define opamp.

Slew that. Why are we even considering Operational Amplifiers for audio?

[/quote]

Hmmm... Excellent devices if to use them to get rid of DC offsets and drifts (servo amps/integrators), to rectify/amplify control/indicating signals, etc...
And it is reliable, despite of internal complexity: a SINGLE crystal with 8 only legs!

 

[bcarso]

As discussed in about sufficient detail above, the requirements on amps for data acquisition are clearly rather different than for audio reproduction. In the former we have the ability to not look at a signal until we know it has settled to some desired accuracy, usually, whereas in audio in principle our perceptual apparatus is working all the time.

I agree with JohnRoberts about the judicious filtering of fast transients to avoid taxing the audio signal chain with extraneous material. The judgment as to where to draw that line varies.

A somewhat-related digression:

Years ago Keith Johnson allowed as how he felt one of the worst defects of a lot of audio equipment was its asymmetrical treatment of positive- and negative-going transients. He belived that the ear etc. was extraordinarily sensitive to this, and strove to make his equipment symmetrical. One approach was to keep signals balanced throughout the chain.

When, many years ago I was introduced by a mutual friend and allowed to hear his home system of that time, I noticed that he had built his preamp into the base of the rickety old AR turntable ("where it belongs!" he said, when I asked about it). And the cartridge signals were amplified differentially from the outset, that treatment persisting all the way to his custom loudspeakers, which had his power amps built in. It was a fine-sounding system I will admit, particularly when we listened to master tapes he'd recorded on the tape machine he'd recorded them with, the semi-custom unit that was to be used for the stuff he did for Reference Recordings. The tape heads were of his design. Keith tends not to do things by half-measures :shock:

 

[CJ]

I bet he had a good stash of direct to disk vinyl.

 

[Svart]

I agree with JohnRoberts about the judicious filtering of fast transients to avoid taxing the audio signal chain with extraneous material.

Oh I'm not disagreeing one bit, I just wonder if keeping the signal as unmolested as possible through the EQ and gain until your output where you can stick it through a nice transformer to smooth it out could help preserve a great degree of the detail and nuance instead of sticking it through tons of opamps that just mush the signal over and over.

 

[bcarso]

I agree---once it is in a system you know, there shouldn't be any need to further "safety-filter".

 

[JohnRoberts]

[quote author="Svart"]

I agree with JohnRoberts about the judicious filtering of fast transients to avoid taxing the audio signal chain with extraneous material.

Oh I'm not disagreeing one bit, I just wonder if keeping the signal as unmolested as possible through the EQ and gain until your output where you can stick it through a nice transformer to smooth it out could help preserve a great degree of the detail and nuance instead of sticking it through tons of opamps that just mush the signal over and over.[/quote]

Using an opamp to LP filter your audio path could certainly be the wrong tool for the job. If the opamp isn't fast enough it could still screw up. This can also vary somewhat depending upon the filter topology used as some are friendlier than others wrt very high frequency inputs but I digress.

If you're going to use a transformer as your filter it's best used in front of the audio chain not after. If the circuit runs out of gas and rectifies some out of band garbage, the noise generated will very likely fall inside your passband and happily pass through your transformer along with the valid audio signal.

This pretty much needs to be dealt with in the earliest stage and if a passive filter is inappropriate, device selection will make a difference. FET inputs with their high Vth offer decent slew rate and intrinsic rise time limiting at higher closed loop gains. Sorry for the too short drive-by explanation it's obviously a little more complicated than that but I'm dealing in the broad strokes.

By this I do not suggest that every circuit block needs to be strictly band passed 20-20k.That isn't acceptable because real filters are not brick wall so cascaded stages will accumulate and interfere with the desired passband. IMO the real criteria is that everything a signal block passes, audible or above is done linearly. Some people discount THD above 20 kHz based on higher harmonics being above human audibility, but that same distortion mechanism means complex HF signals (like a cymbal crash) could generate audible IMD down in the passband. If you roll it off, before it ever gets nonlinear, however high that needs to be, the output will always be clean.

While I am philosophically a straight wire with gain guy, I am not beyond a little prophylactic bandpassing so other audio gear I may share an audio path with might benefit from my cleanly taking out the out of band trash so they don't have to deal with it.


JR

 

[Svart]

It's funny that you mention Jfet opamps, I used to rave about them and their abilities to reject DC on their inputs and the fact that most of them are much "faster" than others as well as being more "clean" but I am finding more and more that I prefer the sound of the BJT opamp. Most of the Jfet opamps tend to have a tinny and thin upper mid range to my ears. It gives the illusion of having greater top end but it's actually the scooped mids that give this illusion. As I stated before, pile a bunch of these on the audio path and you get sonic mush.

It's actually sort of aggrivating, having to choose one over the other. Honestly, opamps that are musical are becoming less available. We really need someone like THAT to step up and bring us the next best IC opamp.

 

[JohnRoberts]

[quote author="Svart"]It's funny that you mention Jfet opamps, I used to rave about them and their abilities to reject DC on their inputs and the fact that most of them are much "faster" than others as well as being more "clean" but I am finding more and more that I prefer the sound of the BJT opamp. Most of the Jfet opamps tend to have a tinny and thin upper mid range to my ears. It gives the illusion of having greater top end but it's actually the scooped mids that give this illusion. As I stated before, pile a bunch of these on the audio path and you get sonic mush.

It's actually sort of aggrivating, having to choose one over the other. Honestly, opamps that are musical are becoming less available. We really need someone like THAT to step up and bring us the next best IC opamp.[/quote]

Are you talking about some objective measurable deviation? I haven't found any demonstrable difference between jfet or bjt linearity when properly applied. Of course they all have their design quirks. So I wouldn't speculate on what you're hearing.

BJT opamps (unless degenerated like LM308) get in trouble with much smaller step impulses than FET inputs so that is clearly not the mechanism you perceive.

JR

 

[Svart]

Yes it is totally objective, I don't mean to imply that I have tested this professionally. They just sound strange in the same circuits.