I am not talking about ME, Bob, I am talking about output stages in general, open loop, looking at the distortion generated by any open loop approach. I know that both you and Halcro took Hawksford's local feedback circuit and put it to good use. It is a good approach, but Charles and I don't use it. It confuses the issue BECAUSE it adds an additional feedback loop to the output stage, and therefore is not a fair comparison between class A and class B output stages. However, I would bet you dollars to peanuts, that YOUR circuit would measure somewhat better in full class A mode than Class AB-2 mode, or even Class AB-1 mode, like I use, IF the measuring level is high enough to turn off one of the devices while testing.
What is wrong with you, Bob?
What is wrong with you, Bob?
PMA said:
Hello PMA
Sounds like you know of an amp which is unknown to Edmond and others on this thread that uses an alternative method which is a non error corrected class B output stage and achieves 1ppm distortion , maybe you could post a THD graph ( at 100 watts into 4R ) of this unknown amp showing its performance . I would be personally very interested in this .
Regards
Arthur
john curl said:
Nothing. I just like pulling your chain when you say something stupid about other designers.
Where recently here have I said anything about error correction? You are the one clouding the issue.
BTW, I agree that my circuit would likely measure better in Class A. Class A is a good thing, it is just expensive.
BTW, Class AB2 is not a very good term for describing over-biased class AB. It harks back to the tube days when class AB2 was when you drove the cr@p out of the grids into conduction and sacrificed quality for power. I prefer to call it class AAB.
Cheers,
Bob
Hi Joshua_G,
LOL!
You really don't add much here, except for comic relief. You are acting like you've always wanted to moderate a forum, but you haven't a clue when to moderate and when to
be quiet.
I understand you want to be John's little cheering section, but he really doesn't need one. He can stand on his own two feet, all by his little self.
Hi John,
BTW, I have seen too much feedback used and an unintentional oscillator sprang to life.
I also thing Bob Cordell made a valid point in his last post. Please, pay attention to how you address others. It would make you nicer to talk to.
-Chris
LOL!
You really don't add much here, except for comic relief. You are acting like you've always wanted to moderate a forum, but you haven't a clue when to moderate and when to
be quiet.I understand you want to be John's little cheering section, but he really doesn't need one. He can stand on his own two feet, all by his little self.
Hi John,
Another post where you belittle others. Mind you, not too bad considering your standards.
BTW, I have seen too much feedback used and an unintentional oscillator sprang to life.
I also thing Bob Cordell made a valid point in his last post. Please, pay attention to how you address others. It would make you nicer to talk to.
-Chris
anatech said:LOL!
Hi Chris and others,
(begin political announcement)
On the subject of annoying troll posters, I've proposed something which might help curb the problem - which is as much one of people responding to the poster as it is the poster themselves. That proposal is here. I added it to the Feature Request List also.
(end political announcement)
Sorry for the OT.
RE: A and AB distortion..
What follows may seem elementary, and I apologie in advance if it is to any individual reader, but we DO have people reading this thread with less experience and theory for whom the following may be of some benefit, and it is in the spirit of education that I offer these comments, NOT one of condescension.
I don't know if this will help or confuse, but there is an important concept of complimentary curve distortion cancellation ...also applicable to SRPP topology [ which is sort of similar to PP with opposite channel devices except made with both same channel devices] , and differential topologies.
When your in class A with complimentary devices, you have the entire voltage sweep of both channel devices with hopefully near-complimentary transfer curves available [ depends on accuracy of the compliment character of the parts] to sum together and to a large degree cancel the even order distortions....in AB, once you leave the A range, you are in ONE channel device only delivering current, and get the full non-linearity of that device with no complimentary opposite cancellation from the other opposite device.
This was an important concept when it was hammered into my skull about 40 years ago, and, in MY view, it still is... It is part of the classic trade off between lower open loop distortion with much higher energy waste, and increased efficiency, higher voltage swing, at higher distortion.
Lets drop power amp thinking for a moment and think of op-amps for a minute....In today's market, there is a huge need for low power consumption due to battery operation in hand held gadgets, and hence a lot of new op-amps are being developed with ultra low idle power consumption. These usually employ class B second stage and output stage to keep the idle currents really low.The published measurements at large signal swing with feedback all look nice, but at low signal levels the switching noises-crossover notch distortions are horrid, and not well addressed by global feedback.
A great deal of professional audio equipment had for DECADES employed all class A discrete op-amps that are free of the crossover notch and class B problem. Of course, it is not unusual to find idle currents in the range of 30 to 50 mA, and some even higher, and also running at up to plus and minus 24 volt rails! Not a good application for AAA or tiny flat battery power, but great for good sound with big power supplies. All the classic API, NEVE, Quad-8, SSL and other major brands of consoles, EQs, compressors and such uses large scale discrete op-amps because the SOUND BETTER and measure better too than high voltage variants of the vast majority of monolithic op-amps. The list of op-amp module builders is not too long, but meaningful in that there are plenty to choose from with different types of active devices, different open loop gains, different open loop bandwidth, etc, resulting in different sonic signatures and differing appropriateness for different applications. AS an example, for a summing junction amplifier you need a lot of open loop gain, and for simple small amount of gain-make-up you want lower open loop gain with great open loop bandwidth so less NFB is needed to arrive at the moderate amount of gain you need.
A few makers of discrete op-amp builders....lot a complete list.....
Forssell Technologies: several varieties, different open loop gains, all J-Fet
Millennia Media: simple topology, J-FET, first gen. by Fred Forssell when he worked at MM
API: bipolar
NEVE: bipolar
Great River: bipolar, modern design
Quad 8: likely all bipolar vinatge design
Jensen 990: bipolar vinatge design
and there are more....
If for cost and efficiency, you HAVE to choose a mass produced monolithic to use in some electronic widget, there are few good sounding/measuring monolithic op-amp choices, the National LM-4562-in the metal case, and Scott Wurcer's Analog Devices AD-797, are usually pretty good..Both of these I personally rank as among the best sounding monolithic in most applications, but both are also smoked to dustsonically and in really low signal level distortion measurements by much more expensive and less efficient discrete op-amps, such as the Forssell 990-3, an all J-Fet class-A design with moderately higher open loop gain and bandwidth.
Anyway, the methods of implementing and rationalization for complementary distortion cancellation encountered in designing op-amps are largely the same as encountered at higher voltage and currents in power amps. It is the same, just scaled different/bigger.
CONCEPT:
Two nearly mirror image opposite curves sum together to make only a VERY SLIGHT remaining curve, a sum that is much straighter than the individual part's curves. You can't get any cancellation with only one channel of device in the P-P stack delivering current. This is EASY to see on an spectrum analyzer as you increase input signal from small to larger, going from A region to AB region. The spectrum and amplitude of distortion products totally changes for the worse!!! ...and most people can hear it too! [ making room here for those with poor auditory discrimination who seem to pervade the audio biz, and make a lot of noise about everything sounding so similar- Alas another topic altogether...]
PLEASE, think about it....I think this is why John Curl prefers to choose class A as much as possible...It is at least why I like to choose class A as much as possible.
I have to get back to work now....
What follows may seem elementary, and I apologie in advance if it is to any individual reader, but we DO have people reading this thread with less experience and theory for whom the following may be of some benefit, and it is in the spirit of education that I offer these comments, NOT one of condescension.
I don't know if this will help or confuse, but there is an important concept of complimentary curve distortion cancellation ...also applicable to SRPP topology [ which is sort of similar to PP with opposite channel devices except made with both same channel devices] , and differential topologies.
When your in class A with complimentary devices, you have the entire voltage sweep of both channel devices with hopefully near-complimentary transfer curves available [ depends on accuracy of the compliment character of the parts] to sum together and to a large degree cancel the even order distortions....in AB, once you leave the A range, you are in ONE channel device only delivering current, and get the full non-linearity of that device with no complimentary opposite cancellation from the other opposite device.
This was an important concept when it was hammered into my skull about 40 years ago, and, in MY view, it still is... It is part of the classic trade off between lower open loop distortion with much higher energy waste, and increased efficiency, higher voltage swing, at higher distortion.
Lets drop power amp thinking for a moment and think of op-amps for a minute....In today's market, there is a huge need for low power consumption due to battery operation in hand held gadgets, and hence a lot of new op-amps are being developed with ultra low idle power consumption. These usually employ class B second stage and output stage to keep the idle currents really low.The published measurements at large signal swing with feedback all look nice, but at low signal levels the switching noises-crossover notch distortions are horrid, and not well addressed by global feedback.
A great deal of professional audio equipment had for DECADES employed all class A discrete op-amps that are free of the crossover notch and class B problem. Of course, it is not unusual to find idle currents in the range of 30 to 50 mA, and some even higher, and also running at up to plus and minus 24 volt rails! Not a good application for AAA or tiny flat battery power, but great for good sound with big power supplies. All the classic API, NEVE, Quad-8, SSL and other major brands of consoles, EQs, compressors and such uses large scale discrete op-amps because the SOUND BETTER and measure better too than high voltage variants of the vast majority of monolithic op-amps. The list of op-amp module builders is not too long, but meaningful in that there are plenty to choose from with different types of active devices, different open loop gains, different open loop bandwidth, etc, resulting in different sonic signatures and differing appropriateness for different applications. AS an example, for a summing junction amplifier you need a lot of open loop gain, and for simple small amount of gain-make-up you want lower open loop gain with great open loop bandwidth so less NFB is needed to arrive at the moderate amount of gain you need.
A few makers of discrete op-amp builders....lot a complete list.....
Forssell Technologies: several varieties, different open loop gains, all J-Fet
Millennia Media: simple topology, J-FET, first gen. by Fred Forssell when he worked at MM
API: bipolar
NEVE: bipolar
Great River: bipolar, modern design
Quad 8: likely all bipolar vinatge design
Jensen 990: bipolar vinatge design
and there are more....
If for cost and efficiency, you HAVE to choose a mass produced monolithic to use in some electronic widget, there are few good sounding/measuring monolithic op-amp choices, the National LM-4562-in the metal case, and Scott Wurcer's Analog Devices AD-797, are usually pretty good..Both of these I personally rank as among the best sounding monolithic in most applications, but both are also smoked to dustsonically and in really low signal level distortion measurements by much more expensive and less efficient discrete op-amps, such as the Forssell 990-3, an all J-Fet class-A design with moderately higher open loop gain and bandwidth.
Anyway, the methods of implementing and rationalization for complementary distortion cancellation encountered in designing op-amps are largely the same as encountered at higher voltage and currents in power amps. It is the same, just scaled different/bigger.
CONCEPT:
Two nearly mirror image opposite curves sum together to make only a VERY SLIGHT remaining curve, a sum that is much straighter than the individual part's curves. You can't get any cancellation with only one channel of device in the P-P stack delivering current. This is EASY to see on an spectrum analyzer as you increase input signal from small to larger, going from A region to AB region. The spectrum and amplitude of distortion products totally changes for the worse!!! ...and most people can hear it too! [ making room here for those with poor auditory discrimination who seem to pervade the audio biz, and make a lot of noise about everything sounding so similar- Alas another topic altogether...]
PLEASE, think about it....I think this is why John Curl prefers to choose class A as much as possible...It is at least why I like to choose class A as much as possible.
I have to get back to work now....
andy_c said:
Me too, but while we're on the subject...
I'd like to see a place attached to their profile where I could make a quick note about the member, to remind myself of past head buts. Many here hold a grudge forever and I find I'm at a loss for the source of animosity when I've completely forgotten the individual and incident.
Hi audiowolf,
You also have to keep in mind the temperature rise. Imagine a large SSL console full. Now try to keep it, it's power supply and digital controller, cool. Then there is the control room ... and they haven't showered.
But seriously folks, often times we can not afford the heat burden with larger class A power amps. Even class A preamplifiers can throw off a ton of heat. Then the consumer "gets it" twice. Once through increased utility bills and twice through decreased reliability.
I think we are forced as a society to try and get the sound right while resorting to trickery, rather than brute force via class A. We simply can not afford the energy costs from taking the easy way out. Not unless you can consider classifying other schemes that do switch (more nicely perhaps) but retain current flow in both polarities of output and driver devices.
-Chris
...hard to record.
You also have to keep in mind the temperature rise. Imagine a large SSL console full. Now try to keep it, it's power supply and digital controller, cool. Then there is the control room ... and they haven't showered.
But seriously folks, often times we can not afford the heat burden with larger class A power amps. Even class A preamplifiers can throw off a ton of heat. Then the consumer "gets it" twice. Once through increased utility bills and twice through decreased reliability.
I think we are forced as a society to try and get the sound right while resorting to trickery, rather than brute force via class A. We simply can not afford the energy costs from taking the easy way out. Not unless you can consider classifying other schemes that do switch (more nicely perhaps) but retain current flow in both polarities of output and driver devices.
-Chris
audiowolf said:
Ok, got it.
Now, do yourself a favour and read+compare the measurements of both monolithic and your beloved and much praised discrete opamps (including Forsell, Sound Skulptor, etc...):
http://www.sg-acoustics.ch/analogue_audio/ic_opamps/pdf/opamp_distortion.pdf
Then repost and tell us if you changed a iota your opinion.
syn08 said:
OK, lets measure candidates at open loop at 20 uV output and 2 mV output, actual amplifier behavior not masked by feedback and large signal swing...
Most of what we listen to is -50, -60 and -70 dB or more, from max signal..measure here, not at peak outputs, everything important gets masked at high swings, and measures reasonably good, even IF it performs very poorly at the lower signal levels....
audiowolf said:
Huh? I don't follow...
Not to say that discrete opamps are necessary bad. But as long as I can get the same (or better) performance (again, read the results) at 5-10% the cost I am having a hard time justifying that solution.
Of course, if I would be in the audio business I would not even blink before getting the discrete opamps. Everybody knows ICs are made out of dirty sand and therefore they can't sound anything but dirty.
I have done a quick, and I do mean quick, scan of the op amp distortion paper, in no way a complete read, but just a fly-through-it scan.....already I find that on both the Forssell 993, and the TI OPA2604, the devices are measured at far lower voltage rails than the manufacturers specified environment. In both cases, a note is given that the test is made at low rails, but the measured results suffer, so they are totally irrelevant.
Its like putting a 215/60 tire on an 11 inch wide rim.....You can, just barely, get the beads to seal and hold air, but the real performance holding up a car will be terrible...
Why bother to even test these in "out of intended" voltage environment....
I also noted that Fred Forssell himself was one of the contributors to the paper.
I still contend that small signal errors are a dramatic portion of what we can actually hear as opposed to what we can easily measure. Some people do disagree, and that is fine too....
The crux of the issue is that for decades on decades, there have been easily heard differences in audio devices that measure similarly, or that have contrary measurements and sound. NO matter, we do need benchmarks that can be measured and repeated as some aspect of evaluation, but no-one has come up with a measurement that totally and totally predicts the sonic nature of a stage, device, or apparatus. What we are left with is a search for measurements that have at least in part, some correlation to the quality in hearing. We are as a whole farther along than what existed 40 or so years ago when I was just a beginner, but the same arguments were being made by some that if it measured better [THD+N] it HAD to sound better.....It was a simple task to demo a DB Systems pre-amp at 0.0003 [ Gobs of feedback] distortion VS a no FB tube unit at 1%....the local meter readers always picked the tube unit in blind testing of A-B VS X. ... A device, and B device, compared to no device, then got mad as H when the units under test were revealed. The problem was not that the measurements were bad, it was that we were not measuring what the ear-brain perceives as important to comprehension. Some designers have balanced listening and measuring, to try and find solutions that both sound good and measure at least reasonably good. Some folks have determined that very low signal, open loop distortions, are a general predictor of how something might actually sound with music in real world conditions. If all we wanted were clean steady state tests, the game would have been concluded long ago......Some of you will totally disagree, fine..some people love NASCAR, some love F1, some like hot curries from India, some like bland rice....We have many evaluation tools and techniques at our disposal now, and a little bit better chance of obtaining some correlation between some tests and actual sonics, compared to years ago, but we STILL do not have "THE TEST", so we have to use what we have, and work on evolutionary methods to move ever forward. Try to keep an open mind about learning what some of the folks on the bleeding edge are doing....if they will share at all....For some, the goal is better music, not SIMPLY better existing test measurements.
A few weeks ago the thread was taking about resistors and the different sound that they create...Some said it was not important, some said differences are clearly audible. Many of the physical characteristics that create the different sonic signatures are still very hard to measure, if at all, and yet some people can reproducibly detect the exchange of one type for another. Clearly there are problems with insufficient measurements, and with some people lacking the neuro/auditory skills, or the psychological predisposition to trust their actual perceptions if they do hear differences. These problems can lead to conflict against those with differing perceptual capabilities and value systems, and yet the TRUTH for both sides of the camp is indeed the TRUTH, with no room for appreciation that the other side has any validity at all. Most engineers have little interest in, or training in philosophy and ethics, and hence will fall back on the simple facts as they currently see them, closed minded old way that all is known and everything else is unimportant and illegitimate, or the opposite that all knowledge is subject to evolution, of not revolution as some seek the crossing of currently understood boundaries into exploratory realms.
In my view, some measurements are good, in fact critical, as a PORTION of an audio evaluation. I DO recognize that the perfect audio test does not yet exist, but that is important to measure what we can, and then ALSO evaluate for what the darn thing actually sounds like too....Is this not why audio designing is still a technical art, and not simply a technical exercise?
Almost every good audio design of the last 70 years combines competent engineering, AND and individual approach that is a reflection of the individual designer, their biases, preferences, and state of technical practice. I don't see this changing much, though some folks would argue that large signal steady state spectrum and fast rise time just about tell it all.....So be it, if you do....Just be mindful that some designers think that there is more going on in the undercurrents that IS relevant, and that indeed, for both camps, the perceived truth is both different and yet wholly valid....so .............fight about it, or learn about it....simple choice to make. Fighting can be FUN for some, can't it.... Just look at all the ad-hominum attacks in this thread in the past 3 weeks..really productive stuff, yup, really productive stuff that was....
Its like putting a 215/60 tire on an 11 inch wide rim.....You can, just barely, get the beads to seal and hold air, but the real performance holding up a car will be terrible...
Why bother to even test these in "out of intended" voltage environment....
I also noted that Fred Forssell himself was one of the contributors to the paper.
I still contend that small signal errors are a dramatic portion of what we can actually hear as opposed to what we can easily measure. Some people do disagree, and that is fine too....
The crux of the issue is that for decades on decades, there have been easily heard differences in audio devices that measure similarly, or that have contrary measurements and sound. NO matter, we do need benchmarks that can be measured and repeated as some aspect of evaluation, but no-one has come up with a measurement that totally and totally predicts the sonic nature of a stage, device, or apparatus. What we are left with is a search for measurements that have at least in part, some correlation to the quality in hearing. We are as a whole farther along than what existed 40 or so years ago when I was just a beginner, but the same arguments were being made by some that if it measured better [THD+N] it HAD to sound better.....It was a simple task to demo a DB Systems pre-amp at 0.0003 [ Gobs of feedback] distortion VS a no FB tube unit at 1%....the local meter readers always picked the tube unit in blind testing of A-B VS X. ... A device, and B device, compared to no device, then got mad as H when the units under test were revealed. The problem was not that the measurements were bad, it was that we were not measuring what the ear-brain perceives as important to comprehension. Some designers have balanced listening and measuring, to try and find solutions that both sound good and measure at least reasonably good. Some folks have determined that very low signal, open loop distortions, are a general predictor of how something might actually sound with music in real world conditions. If all we wanted were clean steady state tests, the game would have been concluded long ago......Some of you will totally disagree, fine..some people love NASCAR, some love F1, some like hot curries from India, some like bland rice....We have many evaluation tools and techniques at our disposal now, and a little bit better chance of obtaining some correlation between some tests and actual sonics, compared to years ago, but we STILL do not have "THE TEST", so we have to use what we have, and work on evolutionary methods to move ever forward. Try to keep an open mind about learning what some of the folks on the bleeding edge are doing....if they will share at all....For some, the goal is better music, not SIMPLY better existing test measurements.
A few weeks ago the thread was taking about resistors and the different sound that they create...Some said it was not important, some said differences are clearly audible. Many of the physical characteristics that create the different sonic signatures are still very hard to measure, if at all, and yet some people can reproducibly detect the exchange of one type for another. Clearly there are problems with insufficient measurements, and with some people lacking the neuro/auditory skills, or the psychological predisposition to trust their actual perceptions if they do hear differences. These problems can lead to conflict against those with differing perceptual capabilities and value systems, and yet the TRUTH for both sides of the camp is indeed the TRUTH, with no room for appreciation that the other side has any validity at all. Most engineers have little interest in, or training in philosophy and ethics, and hence will fall back on the simple facts as they currently see them, closed minded old way that all is known and everything else is unimportant and illegitimate, or the opposite that all knowledge is subject to evolution, of not revolution as some seek the crossing of currently understood boundaries into exploratory realms.
In my view, some measurements are good, in fact critical, as a PORTION of an audio evaluation. I DO recognize that the perfect audio test does not yet exist, but that is important to measure what we can, and then ALSO evaluate for what the darn thing actually sounds like too....Is this not why audio designing is still a technical art, and not simply a technical exercise?
Almost every good audio design of the last 70 years combines competent engineering, AND and individual approach that is a reflection of the individual designer, their biases, preferences, and state of technical practice. I don't see this changing much, though some folks would argue that large signal steady state spectrum and fast rise time just about tell it all.....So be it, if you do....Just be mindful that some designers think that there is more going on in the undercurrents that IS relevant, and that indeed, for both camps, the perceived truth is both different and yet wholly valid....so .............fight about it, or learn about it....simple choice to make. Fighting can be FUN for some, can't it.... Just look at all the ad-hominum attacks in this thread in the past 3 weeks..really productive stuff, yup, really productive stuff that was....
- Status
- Not open for further replies.