Now replying in a funnier way to the first post of this thread:
- The efficiency of class D is something easily measurable (even comparing temperatures with a finger
one heatsink burns, the other doesn't).
- The purity of class A is a purely subjective quality. Most exotic designs tend to be strong distortion generators actually.
- Class D roughness is again a subjective quality, I was stuck on that way of thinking until I tried it a few years ago. There are many bad designs out there, too. It's easy to think that way when you don't understand how it works.
Probably half of the people that I could indirectly blame for indirectly teaching me to think in this and many other ways, that later proved to be wrong, write regularly on this forum. Fortunately some are evolving towards a more open mind too.
- The efficiency of class D is something easily measurable (even comparing temperatures with a finger
one heatsink burns, the other doesn't).- The purity of class A is a purely subjective quality. Most exotic designs tend to be strong distortion generators actually.
- Class D roughness is again a subjective quality, I was stuck on that way of thinking until I tried it a few years ago. There are many bad designs out there, too. It's easy to think that way when you don't understand how it works.
Probably half of the people that I could indirectly blame for indirectly teaching me to think in this and many other ways, that later proved to be wrong, write regularly on this forum. Fortunately some are evolving towards a more open mind too.
Evan, Eva,
Thanks for your insights, which, as I understand, come from experience.
As I mentioned before, I just started to read about Class D and possible combinations and of course have a lot to learn. Having heard only class A, AB amplifiers, I might be biased in taking a plunch into fully class D designs. I have played with the idea of trying a couple of Hypex modules, but never actually did.
As technology progresses very fast, the Audio Physic AD approach may be outdated. As I understand, it was a way of combining the best of both worlds into a single design, which was quite ok and maybe succesful for a short while.
As you mention Eva, there are a lot of crappy class D designs available and for me it makes it even more difficult to distinguish good or best from mediocre. In that regard some advice as what to read or look for is appreciated. I have already done some reading in the class D thread, but may not have hit truly valuable information yet.
Thanks so far.
Thanks for your insights, which, as I understand, come from experience.
As I mentioned before, I just started to read about Class D and possible combinations and of course have a lot to learn. Having heard only class A, AB amplifiers, I might be biased in taking a plunch into fully class D designs. I have played with the idea of trying a couple of Hypex modules, but never actually did.
As technology progresses very fast, the Audio Physic AD approach may be outdated. As I understand, it was a way of combining the best of both worlds into a single design, which was quite ok and maybe succesful for a short while.
As you mention Eva, there are a lot of crappy class D designs available and for me it makes it even more difficult to distinguish good or best from mediocre. In that regard some advice as what to read or look for is appreciated. I have already done some reading in the class D thread, but may not have hit truly valuable information yet.
Thanks so far.
courage said:
I don´t believe that technically achievements and improvements in sound quality happen in an strict "evolutionary" way. Marketing, fashion and production cost all have an important in the process of manufacturers decisions and consumer´s behavior. With time we may have a general improvement for those who cares as well
as a negative scenario for the wast majority....
BTW. Evolution takes time and contrary to many peoples belief Charles Darwin didn´t get the idea on evolution while visiting the Galápagos Islands but much later after returning to England and under the influence of currents of the minds of others e.g. Thomas Malthus and Adam Smith.
An while the evolution in nature goes on and is by no means finished so does the evolution in audio... But my remark as a believer in evolution is that evolution dosen´t necessarily means that something better manifests itself....better or worse; these concepts belong only to human evaluation and are hard to decide upon.
An while the evolution in nature goes on and is by no means finished so does the evolution in audio...
But my remark as a believer in evolution is that evolution dosen´t necessarily means that something better manifests itself....better or worse; these concepts belong only to human evaluation and are hard to decide upon.Evan Shultz said:
IMHO it was never (more so in the past) a particularly good idea as a class D + class A concoction will still lag behind the overall technical performance and efficiency that can be achieved with well established linear class B design techniques.
There are ways of significantly reducing the dissipation in class A designs, for the goal in itself of chasing ever better measured performance specs at ever increasing power levels, but they do not include class D.
Class B and AB with their glitches and abrupt discontinuities in open loop gain near zero current crossing (and not zero voltage) are the worst of all Try to match and flatten low-level open-loop-gain vs. current and voltage plots for class B and AB and you will get plenty of headaches, while class D inherently results in a straight line. To make it worse, the shape of these discontinuities and glitches is signal frequency dependent.
In class D, amplification is done on the "time domain" rather than on "voltage/current domains", and it's time linearity what matters, which is usually excellent. In the end, MOSFET switching times are a linear function of drain current, particularly when body diodes are not allowed to store substantial charge during on-time.
Time discontinuities may arise due to dead time (or body diode conduction) but they don't happen at or near zero crossing, and zero (or almost zero) dead time operation is possible (some of us are doing it that way).
In my biggest amplifier (170Vpk, 70Apk), discontinuities (actually narrow glitches on a straight line) due to slight dead time don't happen until +8.5A and -8.5A, and open loop linearity in this range is already excellent. Distortion products due to these discontinuities lay typically outside the audio band. Then add some dozen dB of negative feedback too. Then add output filter "glitch" attenuation above 20Khz, which you don't have on class B or AB.
Controlling switching timing rather than linear voltage or current has unique advantages.
It seems that you have a long way to go until you can really understand these facts.
BTW: Class B and AB efficiency with typical full-range music signals at low to moderate output levels is not much better than class A efficiency. It only improves substantially (up to 65%) when output level approaches clipping.
Try to match and flatten low-level open-loop-gain vs. current and voltage plots for class B and AB and you will get plenty of headaches, while class D inherently results in a straight line. To make it worse, the shape of these discontinuities and glitches is signal frequency dependent.In class D, amplification is done on the "time domain" rather than on "voltage/current domains", and it's time linearity what matters, which is usually excellent. In the end, MOSFET switching times are a linear function of drain current, particularly when body diodes are not allowed to store substantial charge during on-time.
Time discontinuities may arise due to dead time (or body diode conduction) but they don't happen at or near zero crossing, and zero (or almost zero) dead time operation is possible (some of us are doing it that way).
In my biggest amplifier (170Vpk, 70Apk), discontinuities (actually narrow glitches on a straight line) due to slight dead time don't happen until +8.5A and -8.5A, and open loop linearity in this range is already excellent. Distortion products due to these discontinuities lay typically outside the audio band. Then add some dozen dB of negative feedback too. Then add output filter "glitch" attenuation above 20Khz, which you don't have on class B or AB.
Controlling switching timing rather than linear voltage or current has unique advantages.
It seems that you have a long way to go until you can really understand these facts.
BTW: Class B and AB efficiency with typical full-range music signals at low to moderate output levels is not much better than class A efficiency. It only improves substantially (up to 65%) when output level approaches clipping.
Quote;"In my biggest amplifier, discontinuities (actually glitches) due to slight dead time don't happen until +8.5A and -8.5A, and open loop linearity in this range is already excellent. Distortion products due to these discontinuities are typically outside the audio band. Then add some dozen dB of negative feedback too. Then add output filter "glitch" attenuation above 20Khz, which you don't have on class B or AB."
Outside the audio band, but maybe it can influence performance inside the audio band? Just a hypothesis. Negative feedback is a servo system that is intended to correct what has already happened in an unwanted way so this is partly a time domain problem. A more or less sharp filter at 20 Hz may be the course
of less than perfect treble reproduction, again I say maybe.
Outside the audio band, but maybe it can influence performance inside the audio band? Just a hypothesis. Negative feedback is a servo system that is intended to correct what has already happened in an unwanted way so this is partly a time domain problem. A more or less sharp filter at 20 Hz may be the course
of less than perfect treble reproduction, again I say maybe.
ClassAD, IR make one and said they like the sound result
http://www.google.com/patents?id=tUubAAAAEBAJ&dq=us+2007/0063764
http://www.google.com/patents?id=tUubAAAAEBAJ&dq=us+2007/0063764
Energy "wasted" is a bad thing. I am positive to the ongoing improvements in class D design and as I triamp my speakers the heat generated by the linear amps is nothing I´m proud of nor is it comfortable on a hot day. I like the treble reproduction of my big and heavy (about 100 lbs) class A amp, luckily it´s possible to use it in Class AB mode. Otherwise it idles at 550 W and worst case is when it doesn´t get a signal at all. Then (as you know) all the energy is wasted. Weird!
Big heatsinks is hopefully soon a very uncommon sight.
Big heatsinks is hopefully soon a very uncommon sight.
Eva said:Class B and AB with their glitches and abrupt discontinuities in open loop gain near zero current crossing (and not zero voltage) are the worst of all
In class D, amplification is done on the "time domain" rather than on "voltage/current domains", and it's time linearity what matters, which is usually excellent. In the end, MOSFET switching times are a linear function of drain current, particularly when body diodes are not allowed to store substantial charge during on-time.
Time discontinuities may arise due to dead time (or body diode conduction) but they don't happen at or near zero crossing, and zero (or almost zero) dead time operation is possible (some of us are doing it that way).
In my biggest amplifier (170Vpk, 70Apk), discontinuities (actually narrow glitches on a straight line) due to slight dead time don't happen until +8.5A and -8.5A, and open loop linearity in this range is already excellent. Distortion products due to these discontinuities lay typically outside the audio band. Then add some dozen dB of negative feedback too. Then add output filter "glitch" attenuation above 20Khz, which you don't have on class B or AB.
Controlling switching timing rather than linear voltage or current has unique advantages.
It seems that you have a long way to go until you can really understand these facts.
BTW: Class B and AB efficiency with typical full-range music signals at low to moderate output levels is not much better than class A efficiency. It only improves substantially (up to 65%) when output level approaches clipping.
Oh, very good
I said nothing to dispute these "facts" and you do not seem to have the slightest interest in what I actually said.
As for class AB / B linearity, amplifiers with less than 0.001% THD at 20kHz and any power level (right down through the class A-B cross over region, which has you in such a bother for some reason) have been aound since the 80's and designs with over 10 times less distortion (<0.0001% THD-20kHz) have been built and demonstrated on the SS forum.
If you think that class D is currently at this level of linearity then you are living on another planet.
courage said:
Combining class D with class A is a waste of time IMHO because ultra-low distortion can be done better and easier with just plain class B.
My comments in my previous post were specifically in relation to the Class D + class A amplifier with tracking rails, as described in this threads opening post.
Once the class A linear output stage is operating with sufficient (tracking) rail voltages (at least +/-5V, as discussed already) the idle dissipation will be at or above that of a class B amplifier of the same power rating.
So what benefit is there in terms of efficency? - none!
The best example I know of, of the rail tracking concept is Technic’s A+ topology. There is a patent and an AES preprint but I don’t have the numbers or publication dates on the top of my head ATM. Google.
I tried this with a 250KHZ class D amp and a very Good mosfet AB amp
The Class D input comes off the predriver for the AB amp. As the linear amp current demand goes up , the class D drives harder.
Distortion stays very low (.0005 ) until clipping where the linear and Class D mildly fight. Sonically it is OK but it will get Hot.
I think it is the Pumping aspect of the Class D as Both are run off the full Rail voltage.
My latest design gets around this problem by cranking the class D frequency to 1.35 Mhz. where the performance is equivalent.
My only problem is over current protection. The damn thing dies instantly if the output exceeds 15 Amps.
btw - class D exhibits no crossover distortion
The Class D input comes off the predriver for the AB amp. As the linear amp current demand goes up , the class D drives harder.
Distortion stays very low (.0005 ) until clipping where the linear and Class D mildly fight. Sonically it is OK but it will get Hot.
I think it is the Pumping aspect of the Class D as Both are run off the full Rail voltage.
My latest design gets around this problem by cranking the class D frequency to 1.35 Mhz. where the performance is equivalent.
My only problem is over current protection. The damn thing dies instantly if the output exceeds 15 Amps.
btw - class D exhibits no crossover distortion
G.Kleinschmidt said:
Glen,
Thanks for your comments and suggestions. It certainly gives me something to digest on. Maybe I'm looking for the "holy grail" in the wrong place, but it is always good to enter a sideway and see what's going on there and why different choices are made.
Will Google a bit more. Have done some exploring allready. If, by chance, you bump into someting that may be of interest for me, please let me know through the forum or PM. Thanks in advance.
Have a nice day
I've been looking for my copy of the AES "A+" rail tracking amplifier paper but I can't find it right now.
I remember getting it from Andy C. If you send him a PM I'm sure he'd be happy to send you a copy.
It shows how to use a class B amp to drive the ground of a floating low voltage power supply for the power output stage of a class A amplifier (as well as the relevant points to make the concept work properly).
You could do this with a class D amplifier if you really wanted to though.
I remember getting it from Andy C. If you send him a PM I'm sure he'd be happy to send you a copy.
It shows how to use a class B amp to drive the ground of a floating low voltage power supply for the power output stage of a class A amplifier (as well as the relevant points to make the concept work properly).
You could do this with a class D amplifier if you really wanted to though.
G.Kleinschmidt said:
To achieve this in class B or AB high power amplifiers (over +/-45V rails), 60dB or more of negative feedback are being applied to the output stage taking advantage of fast modern bipolar transistors that were not available 20+ years ago, at least not to the general public. You couldn't do that with early 1980s bipolar transistors (or MOSFET).
In class D this is not yet possible with good efficiency due to limitations in MOSFET technologies currently available to the general public (it's only doable in low or medium power amplifiers, limited to +/-45V rails). However, there are new MOSFET families soon to be released with lower gate charge, lower capacitances and lower Qrr body diodes for given Vds and Rds-on requirements. These will allow to use higher switching frequencies and more negative feedback (like 500Khz-1Mhz with rails well over +/-45V) without efficiency loss, and do the same: "Blameless" class D
The point that I was trying to make is that in class D you have a much more linear open-loop characteristic to start with than in class B or AB because the parameter that drives the output is time rather than voltage or current (which are subject to all kind of transistor non-linearities).
Remember that I'm currently dealing with 180V full bridges providing over 60A. At these voltage and current levels, things have changed a lot since IRFB4227 and IRFB4127 were released by IR (2-3 years ago), and will further improve once better transistors are available. These power levels are quite far from practical with class B or AB (or any exotic mixture).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.