AFAIK it's still available, CH prefers the AD844 for access to the gain node. These are both text book CFA's the simplified schematics are about it, no tricks.
My comment was in reference to things like nested feedback (not used in any amp in our catalog) and FB around low gain stages causing added harmonics (old news). Choose a path quote numbers or cobble and listen. JC loved a tube PA at last(?) years RMAF that had considerable distortion including high orders that did not gradually roll off, others love Cary and its double digit distortions.
Charles SY said it, the TI simplified schematic outlines the xDSL style (I think they borrowed it from that group) output stage. You can get that stage to do -80dB at 400mA output current with <1mA bias @1MHz, at audio frequencies who knows. It uses feedback so have at it, I don't get the "sound" of feedback arguments. I don't hear it, that's all plain and simple.
Scott,
As I mentioned this is a tutorial! There was lots of hissy stuff going and I though actual measurements would shed light on some basics that seem to be confused with religious beliefs.
The very simple triode circuit I used showed harmonic distortion that is similar to that produced by stringed instruments. (Yes stringed instruments harmonics or overtones if you prefer, vary greatly with bowing technique, strings, etc.) This leads to the problem of distinguishing say a violin from a viola. Although this triode distortion is not punch your eardrum unpleasant.
Ganging stages did not create additional harmonics and actually brought up the issue of cancellation. Adding local loop feedback reduced the distortion as expected by most. It did not create additional harmonics which was a surprise to some. Adding global feedback to three stages was a bust as expected due to a number of issues. Pointing out that what is inside the feedback loop cannot be ignored as just a black box.
The issue of "Are stages of local feedback better than overall global feedback" has presented the interesting issue of cancellation.
John, Nelson and others use parallel constructions to reduce distortion, a now well known technique. There is also the method of serial cancellation also well known as "pre-distortion." That actually may be an interesting technique to use, although I do not recall seeing it used as series feedback loops deliberately.
Then there is the issue of deliberate coloration, which has not been touched on yet.
I think the issue of noise has shown it's head, but has not been discussed.
Now the statement that it is easier to design good sounding equipment with tubes rather than discrete semiconductors, I think has been supported.
The issue of "Do op amps magically destroy sound quality" I think can be considered a religious issue. If you experience a difference you are welcome to believe. If you do not you are welcome to scoff (Politely, a word many will have to wiki!). Of course if new measurements show a difference that changes the issue to an engineering one.
I promise Jan an article on how feedback behaves as this seems to have controversy. Measurements are a handy method to show what is happening and reduce the heat. Math is based on theory and while it is handy to use to design and predict there can be a difference depending on how good your model is and if there is an issue you forgot to include!
As I mentioned this is a tutorial! There was lots of hissy stuff going and I though actual measurements would shed light on some basics that seem to be confused with religious beliefs.
The very simple triode circuit I used showed harmonic distortion that is similar to that produced by stringed instruments. (Yes stringed instruments harmonics or overtones if you prefer, vary greatly with bowing technique, strings, etc.) This leads to the problem of distinguishing say a violin from a viola. Although this triode distortion is not punch your eardrum unpleasant.
Ganging stages did not create additional harmonics and actually brought up the issue of cancellation. Adding local loop feedback reduced the distortion as expected by most. It did not create additional harmonics which was a surprise to some. Adding global feedback to three stages was a bust as expected due to a number of issues. Pointing out that what is inside the feedback loop cannot be ignored as just a black box.
The issue of "Are stages of local feedback better than overall global feedback" has presented the interesting issue of cancellation.
John, Nelson and others use parallel constructions to reduce distortion, a now well known technique. There is also the method of serial cancellation also well known as "pre-distortion." That actually may be an interesting technique to use, although I do not recall seeing it used as series feedback loops deliberately.
Then there is the issue of deliberate coloration, which has not been touched on yet.
I think the issue of noise has shown it's head, but has not been discussed.
Now the statement that it is easier to design good sounding equipment with tubes rather than discrete semiconductors, I think has been supported.
The issue of "Do op amps magically destroy sound quality" I think can be considered a religious issue. If you experience a difference you are welcome to believe. If you do not you are welcome to scoff (Politely, a word many will have to wiki!). Of course if new measurements show a difference that changes the issue to an engineering one.
I promise Jan an article on how feedback behaves as this seems to have controversy. Measurements are a handy method to show what is happening and reduce the heat. Math is based on theory and while it is handy to use to design and predict there can be a difference depending on how good your model is and if there is an issue you forgot to include!
Last edited:
How so? I don't get that from your data, and my experience has sadly comprised a vast number of wretched-sounding tube amps (excessive distortion, blocking, bandwidth limiting, bass compression...). I built a few of those before I learned how to do things better.
Did you try designing discrete semiconductor amplifiers with the same level of naivety?
Even the basic grid leak bias triode sounds not horrible. Try that with a transistor and two resistors.
Of course if you get the impression hybrids may take advantage of both that is a different issue.
Also I have yet to go into, biased tube amplifiers, basic bipolar junction transistors, better basic bipolars, current mode bipolars, basic FET, and very basic op amps. There is also an illustration or two about funny stuff inside a feedback loop.
After all I promised Jan a short article and I always lie.
Jan,
The overlay in the small deatil is horrible, I use this, but only on the screen where the selected curve is highlighted, making for good comparisons. But printed on paper or as JPG? Brrrhhhhh...
Ciao T
The overlay in the small deatil is horrible, I use this, but only on the screen where the selected curve is highlighted, making for good comparisons. But printed on paper or as JPG? Brrrhhhhh...
Ciao T
The level of misinformation that you're presenting is in the stratosphere. Nobody got lynched. Relax. This forum welcomes anyone who can entertain a civil debate.
The pre-output driver is not that interesting here. What is is the biasing on the output devices, of extra interest is that the output stage can provide additional voltage gain. You will notice the commutation of bias currents top to bottom. This can be arranged so that neither side shuts off (geometric mean biasing). The details are in a patent assigned to National, sorry don't remember the name (see was that Carson, Bill Carson... no it was a C name though I think).
Attachments
Here's another free one. This is the AD model 50 brick discrete op-amp, one of the best ever work horse amplifiers in the olden days. This is a complete schematic with all component types and values. No guarantee that SK170's will work at the input but BF862's might and would give you a quieter amplifier (though second stage effects might step in).
Attachments
Now that is an interesting circuit. I assume the current sources are more in the line of mirrors to keep the output swing or is there another technique you can use?
As you are aware gain in the output stage has been around a long time. I seem to think at least '60's maybe earlier, but that was of course discretes.
Thanks Scott. Yes the output stage has gain, as i said above, this gain depends on the load the output sees. The biasing is interesting, less for line level discrete circuits, as they can easily be biased into class a, but rather for power amps.
The current source values are chosen as are the areas of the transistors to set the output bias current (by forcing the output device Vbe). This would be difficult with discretes. This all acts also as common mode stabalization since a fully differential amp would have no mechanism to set the "shoot through" current at the output. The shoot through is hard to control under ALL dynamic conditions, I assure you failure is dramatic (exploding packages).
It can be done well enough. I used the same technique to get Ed's multi-curve graphs in his resistor distortion article come out legible. Even in gray scales. Not JPG, of course; that's strictly for entertainment.
jan didden
Scott,
We do live in different worlds. Dramatic... There was the electrician wiring the AC service to a TV transmitter. The input voltage to the transmitter power supply was 4160 3 phase at 100 amps or so. The way high current wires are connected is with "Bugs" or screw on clamping devices. The power supply required the AC leads to be a precise distance apart to meet and clamp properly. The electrician took out his Stanley tape measure to adjust the AC leads to match. Well the wires were hot, the Stanley metal tape turned into plasma and the resulting explosion and fire destroyed the TV station.
Then there was the cautious electrician who made sure the power was disconnected from the 4160 supply to the transformer he was servicing. (It is good practice to clean out dust and dirt to avoid fires). Turns out the guy who had wired the system put the switch in the neutral leg, after all that would break the circuit... So as those in the U.S. know the neutral is also tied to earth. The result was a closed casket service.
Then there was the Tornado that picked up a crane...
Or the crane that had the sling slip on the load of lumber...
So exploding IC's may wake you up, even hurt a finger or two, but safety is always an issue.
I forget who mentioned losing a finger on a table saw as stupid... Use one often enough and you will understand why mine is the fully guarded European safety standard model or maybe you have never burned yourself with a soldering iron?
I wear eye protection when clipping wires and soldering, one hand in my pocket when the voltage goes above 70, and earmuffs are placed by every loud power tool.
We do live in different worlds. Dramatic... There was the electrician wiring the AC service to a TV transmitter. The input voltage to the transmitter power supply was 4160 3 phase at 100 amps or so. The way high current wires are connected is with "Bugs" or screw on clamping devices. The power supply required the AC leads to be a precise distance apart to meet and clamp properly. The electrician took out his Stanley tape measure to adjust the AC leads to match. Well the wires were hot, the Stanley metal tape turned into plasma and the resulting explosion and fire destroyed the TV station.
Then there was the cautious electrician who made sure the power was disconnected from the 4160 supply to the transformer he was servicing. (It is good practice to clean out dust and dirt to avoid fires). Turns out the guy who had wired the system put the switch in the neutral leg, after all that would break the circuit... So as those in the U.S. know the neutral is also tied to earth. The result was a closed casket service.
Then there was the Tornado that picked up a crane...
Or the crane that had the sling slip on the load of lumber...
So exploding IC's may wake you up, even hurt a finger or two, but safety is always an issue.
I forget who mentioned losing a finger on a table saw as stupid... Use one often enough and you will understand why mine is the fully guarded European safety standard model or maybe you have never burned yourself with a soldering iron?
I wear eye protection when clipping wires and soldering, one hand in my pocket when the voltage goes above 70, and earmuffs are placed by every loud power tool.
Hi,
Actually, if you use a lot of averaging (say 256) you can lower the noisefloor and observe more of what goes on. Also, the way you daisy chained stages still made one stage dominant in it's contributions to final outcome, moreso given the middle stage (the one that is out of phase with either first or last) has the lowest level and HD.
Well, I recall seeing it in both german and american tube amplifiers of the 1930's. The legendary (perhaps justly so) WE 91 Amplifier with the 300B Output tube in single ended mode used both series cancellation (quite agressive as well) and looped feedback, while the German studio amplifiers used to monitor the first stereophonic magnetic tape recordings used series cancellation of harmonics and an open loop, "current source" output stage.
Incidentally, that german Amplifier was connected to a 12" Coaxial speaker, time aligned, 2nd order series crossover and using mechanical means to damp the fundamental resonances of both drivers, probably what is in concept still the most advanced speaker developed since
Maybe we really want to "don't go there girlfriend"...
I was gonna comment, but I found it hard to work with the polts in two scales and really see what is going on. Also, please average out more noise, it takes longer to run the measurement, but you see clearer.
I guess so. And I agree, it is easy to make good sounding tube gear DESPITE the relatively high measured THD and so on, while I find making decent sounding gear with solid state, even with Op-Amp's if I must rather harder work. And as it is said, "work is the enemy of the drinking classes"...
Actually, Samuel Groner has done some really interesting work on the subject of Op-Amp distortion. The intimidating size and detail of the datasets per Op-Amp can make this tome (now some 436 pages long and a 35MB download) very hard going. Plus Samuel very carefully avoids making comments of any subjective nature and comments minimally on the objective performance, so you really DO HAVE TO analyse these thousands of graphs...
Still I find it quite interesting, especially seeing that measurements of his discrete Op-Amp's are also included (as well as those of many other discrete Op-Amp's).
Ciao T
Actually, if you use a lot of averaging (say 256) you can lower the noisefloor and observe more of what goes on. Also, the way you daisy chained stages still made one stage dominant in it's contributions to final outcome, moreso given the middle stage (the one that is out of phase with either first or last) has the lowest level and HD.
Well, I recall seeing it in both german and american tube amplifiers of the 1930's. The legendary (perhaps justly so) WE 91 Amplifier with the 300B Output tube in single ended mode used both series cancellation (quite agressive as well) and looped feedback, while the German studio amplifiers used to monitor the first stereophonic magnetic tape recordings used series cancellation of harmonics and an open loop, "current source" output stage.
Incidentally, that german Amplifier was connected to a 12" Coaxial speaker, time aligned, 2nd order series crossover and using mechanical means to damp the fundamental resonances of both drivers, probably what is in concept still the most advanced speaker developed since
Maybe we really want to "don't go there girlfriend"...
I was gonna comment, but I found it hard to work with the polts in two scales and really see what is going on. Also, please average out more noise, it takes longer to run the measurement, but you see clearer.
I guess so. And I agree, it is easy to make good sounding tube gear DESPITE the relatively high measured THD and so on, while I find making decent sounding gear with solid state, even with Op-Amp's if I must rather harder work. And as it is said, "work is the enemy of the drinking classes"...
Actually, Samuel Groner has done some really interesting work on the subject of Op-Amp distortion. The intimidating size and detail of the datasets per Op-Amp can make this tome (now some 436 pages long and a 35MB download) very hard going. Plus Samuel very carefully avoids making comments of any subjective nature and comments minimally on the objective performance, so you really DO HAVE TO analyse these thousands of graphs...
Still I find it quite interesting, especially seeing that measurements of his discrete Op-Amp's are also included (as well as those of many other discrete Op-Amp's).
Ciao T
Yeah, I'm really looking forward to spending another 200 hrs or so taking graphs apart and then reassembling them again. Not!
jan
- Status
- Not open for further replies.