Every one knows "global feedback" is not a good thing in any amp, it’s a
"sound sterilizer" and should only be used very sparingly if possible,
just to clean things up a little, and preferably use only "local
Which won’t help your plight as you’ll need a stack of
"global "feedback" that has to incorporate even the speaker output
filter into it, to try to remedy this kind of phase shift problem down
to 1khz 40 degrees out of phase, and 70 degrees!! at 10khz (in red). https://ibb.co/vvwzGV5
Actually 'everyone' does not know this about feedback and I think this is where your misunderstanding originates.
For many years, feedback did do exactly as you said. This isn't to say that 'every one knows "global feedback" is not a good thing in any amp' but it is to say that there was a known problem but no good solutions. This had been true for about 60 or 70 years, so you shouldn't (and probably won't) feel too bad about being ignorant of advances in the technology.
Now I've mentioned this before and I'll wager it went over your head. But nevertheless: It has to do with something called Gain Bandwidth Product. The math gets tricky, but in a nutshell you can sort of equate GBP to feedback as gasoline is to a car. When you run out of gas, the car stops, when you reach the limit of the GBP, you run out of feedback. This means that where ever that limit is, your feedback goes to zero.
You need GBP well into the MHz region to support feedback going up to 20KHz if you're going to use a lot of feedback. The more feedback you use, the more GBP you need to have. Well guess what? Nearly all solid state and tube amplifiers made up until about 15 years ago or so lacked the GBP to allow for a lot of feedback at higher frequencies. Put another way, this causes the distortion to rise as frequency rises.
To get around this problem most amps are measured at 100Hz; this sweeps the dirt under the rug. You don't see how the distortion is in fact higher at more critical frequencies like 5KHz. So at 100 Hz the feedback is adequate in almost any solid state amp, which is why they have a reputation for good bass, and this creates the idea that they must be OK elsewhere too. But we've been hearing the brightness and harshness for decades- in fact that is why tubes are still around.
Now how it works otherwise with feedback is if you add a little, it seems to help, you add more and the benefit diminishes, as higher ordered harmonics are added by the process of the feedback itself- and this is the cause of harshness and brightness in most solid state amps (and tube amps too- this is why feedback has a poor reputation). But this does not happen to infinity. It turns out that if you can get enough gain, if you run about 35dB of feedback this will allow for the circuit to compensate for the effects of the feedback itself. IOW, you'll have low distortion at higher frequencies.
Normally getting this sort of gain is hard, because your open loop gain is killed off by the feedback you have- and any power amp will need about 25 dB of voltage gain to be practical, and higher powered amps (which are likely to be used on low efficiency loudspeakers) might need 30 or 35dB. Well if you have 35dB to start with (and 40dB is better) that means you need at a minimum 60 dB of gain in the circuit open loop. That's a lot- and you run into something called 'Phase Margin' that limits how much feedback you can use before the amp goes into oscillation.
The semiconductors needed to make this sort of gain possible in a traditional design have not existed until fairly recently. But you have to know what you're doing even if you have the devices in your hand. One amplifier of 'traditional' design that actually does have the GBP and feedback to pull this off is the Benchmark. It is a very rare exception in this regard; obviously John Siau knows his stuff.
Class D amps have an advantage here though. If you give it so much feedback that it oscillates, you can use that oscillation as the switching frequency. All you have to be careful about is that it always find the same 'solution', the same switching frequency. But now at audio frequencies you have consistent feedback even at 20KHz; and this easily corrects for phase shift caused by the output filter (which is included in the feedback loop so as to allow the feedback to do its job).
This is all standard control theory. I don't expect you to know it because its clear you don't have an engineering degree. But you don't (and I'm sure you won't) have to trust me on this- its easy enough to find on the web if you cause your hand to move and search for it.
As said in my first post, they don’t give it as they either turn off
going into protection mode, or they are current starved into 2ohms, or
the worst oscillate.
Your speakers present a very hard load especially the combined impedance together with - phase angle (EPDR)
And for @justmetoo here’s a link to "try" and understand how speakers torture amps, especially EPDR. (page2)
This is fake news.
If you have a class D amp, and a speaker that dips to 3 ohms or even 2 ohms, don't worry- just hook it up. EPDR won't be an issue- the output impedance of any modern class D amp is only a few milliohms; what's happening at 2 or 3 ohms is thousands of times higher impedance and the amp won't care.
It may not double power into 2 ohms though, but that's not important except at clipping. What's important with most speakers is the amp behave as a voltage source, which it will do just fine.
With regards to that bit about 'oscillate': most modern class D amps already oscillate as part of their modus operandi. You can't somehow make them oscillate *more*; the idea is rubbish.
It may not double power into 2 ohms though, but that’s not important That more fake than anything!
Class-D’s, independent tests, can’t better the 4ohm wattage with the
2ohm load, if they don’t turn off first, let alone double it. That’s how
"strangled" they are for current.
With regards to
that bit about ’oscillate’: most modern class D amps already oscillate
as part of their modus operandi. You can’t somehow make them oscillate
*more*; the idea is rubbish. No you are. You know exactly
what was meant by that, don't play the arrogant card, just another *****
side step to protect an upcoming Class-D from you
In case you didn't mean that as a joke (because its pretty funny):
Think about it- if an amp has an output impedance (Purifi for example) that is lower than a foot of 14gauge wire, how is it really going to be affected by a 2 ohm load? The answer is it isn't. Now the power supply might not support the current the amp would demand at full power, but all that means is you would clip the amp at some point. When that happens, simply turn down the volume. Why is this so hard to understand??
The idea that you can make a self-oscillating class D amp oscillate when in fact they already oscillate for a living is rubbish. Maybe you can explain that??
When an amplifier is oscillating, another way to put that is its Phase Margin has been exceeded. This simply means that there is so much phase shift at some high frequency that the negative feedback applied is now positive feedback- and presto! The amp oscillates. If you don't believe me on this, read the works of Norman Crowhurst. You can find his books on Pete Millet's excellent resource site:http://www.tubebooks.org/technical_books_online.htm
Now I doubt you'll be doing you're homework, but I provided the link so you or others can, so you or others can follow along. So try this on:
In a self oscillating class D amp, the phase margin is exceeded ***on purpose***. The oscillation is used as the switching frequency. Its got a certain elegance, but now that you brought it up, one of the advantages is that its quite literally impossible to get it to oscillate any more than it already is. So its impervious to load conditions in that regard.
because of my obsession with GaN with it’s ability to have 1.5mhz switching speed the Achilles heel of Class-D
This really isn't correct. You can get vanishingly low distortion with less than half that speed; alternatively you can go a 1.5MHz using MOSFETs as they are plenty fast enough.
GaNFETs have other advantages completely unrelated to their speed, although you can run them with slightly less deadtime. The thing is, if you run at 500KHz, the deadtime (being a constant) has 1/3rd the significance that it has at 1.5MHz.
Then the "switching noise output filter" can also be set higher and so eliminates most of the switching frequency noise left at the speaker terminals. Without introducing phase shifts back down into the audio band, like what happens with lower 600-800khz switching frequencies/filters.
This is fake news. Let’s, once again, unpack and debunk this rubbish to set the record straight.
The ’switching noise’ of any class D amplifier is a sine wave at the switching frequency. If the filter is properly designed, the sine wave is quite small and low power- it has to be so because the amplifier can’t interfere with other services, like AM broadcast (1.5MHz is at the top of the AM band). But an amp that is switching at 600KHz will have the same amount of noise and 600KHz is in the AM broadcast band too- the same rules apply.
Phase shift is indeed a consequence of using a filter. But it is incorrect to assume that because a filter is used, even one of a low frequency like 80KHz, that there will be phase shift in all examples of class D amps that have a filter set this low. This is because some class D amps are of the self-oscillating variety, and self oscillating class D amps run a prodigious amount of feedback. They do this as part of their switching scheme but that is not why so much feedback is used.
The thing about feedback is that its application causes distortion while suppressing the innate distortion of the circuit in which its used. The distortion generated by feedback tends to be higher ordered harmonics, which is audible as harshness and brightness, due the ear being keenly sensitive to these harmonics. It is that sensitive because it uses them to sense sound pressure.
But if you use enough feedback (north of 35dB) you can not only suppress the distortion caused by the application of feedback, but you can also correct phase shift. This is why self oscillating class D amps with a switching frequency of 500KHz and a filter set at 70KHz can have less than 1 degree of phase shift at 20KHz- quite unlike the amp depicted at the link which George put up in his last post- and I might add, knew that this was the case.
Amps that have less then 35dB (which includes about 99% of all solid state amps ever made) will have some brightness and harshness due to that lack of sufficient feedback. This is why class D.
@mijostyn We’ve been working on a class D project of our own design for about 4 years. In November of 2019 we were awarded a patent in the field. Our prototypes make 100 watts into 8 ohms with 200 watts into 4 ohms and 400 watts into 2 ohms (sorry George) no worries, if given enough current in the power supply (the GaNFET output devices are rated 35 amps). When testing traditional solid state amplifiers the FTC requires that the amp be preconditioned at 1/3rd power for an hour and then the amp can be tested for as short a time as possible to get the full power specs. This is done because traditional solid state amps will overheat if you run them at full power for a long period of time and usually that’s only a few minutes. Our amp (and I assume for most class D amps this is true) can be operated at full power all day long and you can hold it in your hand- they get warm but not hot. We don’t know where this will lead, but we have a goal and now we’re seeing if it can be achieved.
Once again, straight into product protection mode
This statement is false. I've simply been engaged in correction of misinformation. That is why I used the phrase fake news. But I could use others- ignorance comes to mind. Misleading. Deceptive. And so on. But since I don’t know/understand why you insist on posting incorrect information, I don’t use the word ’deceptive’ although due to the tenor of your posts I’m often tempted.
Again for the umpteeth time: READ UP ON THIS STUFF. Parading the same rubbish over and over isn’t winning you any accolades and isn’t impressing anyone that you know what you’re talking about.
Its easily shown that its not product protection, for the simple fact that whether my company produces an amp or not, what I have stated remains factual. Its not like our amp somehow makes it true.
It may be that you won’t get this simple fact as this has been gone over and over again and you seem to hang on to the same mythology in the face of facts (just like that same person you alluded to in your post above). Please consider just reading and educating yourself. You might find it worthwhile. If you need links to articles I can provide them. OTOH if you can find anything that debunks my comments, please produce them; I’d be interested to see. But don’t send that tired link about an amp that is obviously decades behind us.
The 1200AS 2 is still the current Class-D module from ICEPower, what have you been sniffing.
Nothing but air. However if you are suggesting that this amplifier is representative of all class D amps you would be committing an excellent example of the logical fallacy known as Guilt By Association, and it works like this: 'Because this particular amplifier has significant phase shift at 20KHz and is class D, therefore all class D amps have significant phase shift at 20KHz.' Because its a logical fallacy, this idea is false.
Obviously ICEPower has chosen a different set of parameters for what they feel is important, which with a short look at their website, appears to be public address. At any rate there are class D amplifiers that have less than 1 degree of phase shift at 20KHz.
OK George- that one is debunked. Please proceed to the next.
Cyrill Hammer is a good designer but entirely mistaken in his comment you quoted.
When class D amps (Purify) are making distortion that is 120dB down, it is serious cutting edge; lower distortion than the Soulution amps.
Again, you need extraordinary proof to support your extraordinary claims.
Yes like the ones and their lackies pushing yet to be released
commercial products, with fanciful claims that will only be proved right
or wrong "if" given access to by independent measurement testers.
To be clear here, the 'fanciful claims' were made by Bruno Putzeys, a good 15 years ago with the introduction of his UcD module. And again with his Hypex and Purifi modules. These devices have been measured many times. Any of the claims I've made are a subset of those by Bruno.
So really the extraordinary claims are at this point coming from George. And to trot out yet another worn trope, he needs extraordinary proof to support his claims (beyond just wilful ignorance). So far, not seen it.
I don't see anything in Cyrill Hammer's bio that would suggest the
extensive knowledge in signal processing or advanced control theory or
even ultra high speed switching for analog reconstruction that would
make his opinion any more valid than many many others, and perhaps less.
That GHz comment really makes me suspect his depth of knowledge on the
topic is really quite weak.
It makes more sense that his comments are that of marketing rather than an engineering background. Obviously George missed my debunking of his use of Cyril's comments earlier on this page. The idea that you need GHz switching speed in a class D amp is rubbish. I'm sure the guy that designed the Soulution amp is cringing at that bit of marketing hype!
I'm going to deal with the Technics issue that George keeps bringing up; not to take them down (I've heard their amp and its pretty good) but to debunk some of the things George says about it (that no other class D amp has); to wit: that because of its higher switching frequency, it has no phase shift because its output filter is set higher.
This is true, but there is an important distinction here! The Technics website claims that the Technics amp is **zero feedback**. When any amp has no feedback (we've been making zero feedback amps for decades), wide bandwidth is required in order to minimize phase shift. That is why Technics opted for a higher switching speed. Its also quite likely that they were wary of Bruno Putzeys' patents in the field of self-oscillating class D amplifiers.
So here we see that there are in fact two methods of preventing phase shift at audio frequencies. The first is the old school accepted method of wide bandwidth (this is the means we have used for 45 years in our OTLs). The other means is to apply so much feedback that even though you don't have bandwidth much past the audio passband, the phase shift is nevertheless controlled to the same degree.
No because it's only to get the low order output filtering on the
speaker output up high also, so then you can get rid of "all switching
noise" without effecting the phase integrity down into the audio band.
This has been explained ad nauseum; this statement is rubbish.
All I can say Ralph after all your put many downs on anything mentioned
threatening the release your upcoming Class-D. And all the technical
praise you give it.
1: Better be a real good one
2: Designed by you.
3: Have no after market "tweaked" BS modules like, ricevs, pascal, Rowland, red dragon etc etc etc, it may exposed for doing it.
4: Be reasonably priced
OK- we got 2 and 3. Working out the math for the oscillation criteria was a real female dog.... How 'good' it is- we'll leave that to others. We are planning about $5,000 for a pair of 100 watt monoblocks. That's beta production; the actual power might be increased by the time we see production.
The amount of power the filter has to contend with in Class-D amps
compared to a dac, now your showing how ridiculous your statements are.
This statement is wildly incorrect!
Although digital and class D amps both employ switching, it is important to understand that one is digital and the other analog. However the primary difference between the two is the the order of ones and zeros has significance with digital as does the word length. With class D it does not (there is no word length for starters....) and it is always switching states at its switching frequency. In a class D amp its the width of the on and off states that has significance, as these widths describe the audio signal.
As a result the filters are vastly different. The filter in a digital system is usually operating at a *much* lower frequency than a class D amp. So this isn't just about power!!
It’s more about power, because you can’t see the forest for the trees,
with what Mark Levinson tried to do by series up "low order" output
filters (so they don’t burn out) and trying to get a steeper roll off as
not to introduce phase shift into the audio band in the upper
mid/highs, which they achieved, but it had then other problems
This comment really says you don't understand what's going on here. ML didn't do what you're describing. Its obvious from the internal photos that the circuit employs what is known as an 'H' bridge output (which tends to be common in higher powered class D amps). Its also common to see two chokes as a result- one for each half of the bridge circuit. Without knowing further details of the circuit its unwise to speculate further, but the fact of the amp being an H bridge is easy to see if you know what to look for.
The waveform at the link is a 10KHz squarewave. There is a small amount of overshoot, and what looks like a bit of an oscillation on top of the waveform as it is consistent with each iteration so does not appear to be a measurement error. As square waves go, that's not a bad looking one- most power amps will round the leading edge (if open loop/zero feedback you need 100KHz bandwidth to make that leading edge look right).
So I don't see the 'other problems' to which you referred by including that link. I might add though that more modern class D amps do make a better showing of this sort of thing. When you attack a technology based on older examples, it really doesn't help your argument!
Please don't be stupid. It is stupid to respond to George.....he is a
rock....he will never move from his position.....and he has to have the
last word.......so it will always go on and on and on. @ricevs
I get what you are saying, and regret the diversion that this has often caused. Generally speaking, if there is going to be misinformation on any given thread about class D on this website, the chances are very high that it will be coming from George.
Now I could sit back and watch him do that; I have a theory that a rising tide raises all boats. So knowing the truth of the matter should I sit back and do nothing? We've all seen in the past when someone passes misinformation as truth that it is possible that everyone will suffer. OTOH if I do respond, it looks like I have a personal beef with George which I don't - I've never met him. So this has been an on-going dilemma for which I've never felt I've had a good solution.
@ricevs what happens to the residual when you do this?
That's why it's seems only fixable by moving the switching frequency
and filters way up higher so there's no effect of phase integrity down
into the audio band, and the only way to do this is move the switching
frequency up much higher to 1.5mhz and the filter as Technics did with
the SE-R1 and hopefully with the SU-R1000 also.
And you can't fix
it but throwing "more global feedback" at it, as that is a sound
destroyer in itself. The only way is the > 1.5mhz switching
frequency, so the filter can do it's job without introducing phase
problems down into the audio band.
Rubbish. Bruno Putzey's UcD module, which is quite old at this point (and predates NCore and Hypex), has less than 1 degree of phase shift at 20KHz. It switches at 400KHz and is self-oscillating, with about 40dB of feedback.
I did not measure the amount of residual after removing the secondary
coil and caps. Since the caps to ground were just 1000 pf....I imagine
that the residual looked much the same.
Controlling the residual is pretty important. Its not the sort of thing I'd leave to imagination.
it's Ralph that got hostile calling rubbish on me first
This is not out of hostility. Just calling out obvious misinformation. I see though that you did not refute my comment about phase shift. You might not like the UcD module, but it does not have phase shift. You'll find that the Purify and Hypex don't either, and for the same reason. I've heard the UcD sound quite good; clearly its a matter of execution.
Regarding feedback, you either don't run any or you run really a lot. Its sort of like a bell curve with 0dB on one end and 35dB at the other. The worst place to be is 15-20dB as that's in the middle of the curve. Of course I'm oversimplifying here but to really dive into it is very hard to do on this site since it does not support a lot of mathematical symbols.
The other problem is current delivery into very low impedance speakers, like the classic linear high end amps can, that could take some more time, but they can drive the speakers that aren’t such a savage/nasty load like Wilson Alexia etc, so it’s not an overhaul big problem.
George, you might want to know something about testing in the US.
Back in the 1970s many companies were making some pretty outrageous claims of amplifier power. So the Federal Trade Commission (FTC) stepped in and made some rules. The first was that the output power of any amp sold in the US had to conform to something called ’RMS Power’ which has no meaning outside of the FTC rules, but is essentially the output power using RMS volts to calculate the power.
The second was temperature pre-conditioning of the amplifier prior to testing it. This was done because in a traditional solid state amplifier if you run it at full power for a few minutes you’ll overheat it. So the pre-conditioning rule is where the amp is run at 1/3rd full power for one hour after which it is then tested at full power, which might only take a few seconds. This was simply to avoid damaging the amplifier during its testing.
It might interest you to know that our prototypes, which are making 200 watts into 4 ohms, can run at full power on an indefinite basis! This is only because our circuit uses GaNFETs and a minimal heatsink design. You can hold the amp module in your hand while this is done and it does not get hot. I have to assume that is true of any module using GaNFETs although I’ve not checked that out.
But this is **quite different** from traditional solid state amps which will blow up if you treat them like that! Additionally its worthy of note that distortion does not change as the module heats up, since bias of the output section is not affecting the distortion.
Since our output devices have a rating of 35Amps, we can easily double power (400 watts) into 2 ohms no worries. We do reach a limit of doubling power into 1 ohm- the safety margin of 50% is exceeded into that impedance at full power (meaning that the module has a chance of surviving that but generally with any solid state design its good practice to not exceed 50% margins of maximum specs). Of course if the amplifier is not being pushed to full power then doubling its output will go on even if the load is less than 0.125 ohms! If we wanted to make more power into such low impedances that could be done with more output devices. Generally when such high powers are needed, the output circuit employs what is known as an ’H’ bridge. This allows the output devices to make more power without their voltage margins being exceeded.
This however is true of any class D amp; the limits **are the same as seen in any traditional design**:
1) the current capacity of the output section
2) the heatsink capacity and
3) the current available in the power supply.
IOW the statement I quoted from you above is outright false. This is very easy to determine. Again you don’t have to take my word for it; Google is your friend.
Willful ignorance is stupidity; misinformation for its own sake is a heinous act.
So you calling Rubbish! instead of calmly trying to explain yourself, is not hostile.
Its simply factual; I have in fact often calmly explained myself. I've often received complements for my patience in dealing with your insults and mis-information, but since I've never met you, its hard to take any of the slings and arrows personally so I've no reason to be angry. I do feel a need to correct misinformation where it occurs, as a rising tide raises all boats.
Regarding feedback. George, do you realize that all the solid state amps of traditional design you've pointed to employ feedback? How do you explain that they are alright and a class D isn't? In common vernacular this is known as 'talking out of both sides of your mouth'. In legal terms its known as 'dirty hands'. 'Dirty Hands' is the legal idea that one cannot reserve for themselves a right that they do not also confer to others.
At this point all the objections you've raised have been debunked, including the one about me being in 'product protection mode'. I think that you do not understand my motivations: right after high school, the very next day, I was employed by Allied Radio Shack as a service technician in their 5-state service facility. I enjoyed working there for years and used that income to put myself thru engineering school at the University of Minnesota MIT. In the late 1970s I started Atma-Sphere. As you know that involved tubes for a very long time, but I missed the troubleshooting I performed in the service industry. So I'm active on this and other sites, often answering questions that have nothing to do with my business. IOW I do it because I enjoy it.
Now I can see that this is hard for some people to understand. But that's how it is. I do this because I enjoy it. Because of my experience as both technician and engineer (which can sometimes be at odds, technicians often find out about engineering errors before the designer does), I do feel a certain compulsion to correct misinformation. For reasons unknown to me (as I've no idea of your motivations) this means that you and I cross swords quite often. I see that as unfortunate. But from an ethical point of view I'm a bit between a rock and hard place when it comes to mis-information; if I sit back and say nothing, people will often flush perfectly good money down the loo. Its a bit like what happens when tyrannical forces come into power:
First they came for the socialists, and I did not speak out—
Because I was not a socialist.
Then they came for the trade unionists, and I did not speak out—
Because I was not a trade unionist.
Then they came for the Jews, and I did not speak out—
Because I was not a Jew.
Then they came for me—and there was no one left to speak for me.
Of course this is about education rather than power. FWIW though, willfully ignorant is almost indistinguishable from just plain stupid.
And Ralph search is your friend, you might want to be reminded, that you
were anti GaN fets way back at first, they are still the same ones
You are quite correct. When you first commented that GaNFETs didn't need a heatsink in an evaluation circuit, I didn't believe you. But I followed the link you provided and promptly went down the rabbit hole learning more about it. So you can rightfully take credit for that.
What is puzzling here is that even though you knew that bit about heatsinks, most of the rest of it is a complete unknown to you. For example, if I were to ask you, 'what is the **main** advantage of GaNFets?', what would be your answer?
If you answered 'speed', you would be incorrect. GaNFets have lost their speed advantage over MOSFETs, which have continued to improve since GaNFETs were introduced.
Stop trying to divert that you were very anti GaN, and now pro just because it will be "the format" and suit your sales/profit modeling.
Actually you’re hard pressed to find me being ’anti GaN’; that statement is false. What is true is that I did not believe that they could work without heatsinks and it was your heckling that caused me to take a look.
Its true that for evaluation you don’t need heatsinks and most of our GaNFET prototypes didn’t use them. But in practice, ***they do in fact actually for real do need heatsinks*** which. again, was my original belief. Otherwise, you can expect them to blow right off the circuit board when subjected to low impedance loads at full power! Ask me how I know this :)
The primary advantages of GaNFETs (compared to mosfets) are low on resistance and low input capacitance.
The low input capacitance is a major deal! It makes the driver situation easier to manage. The lack of a body diode is the other issue. Between these two, that’s a good portion of why they are faster. These things interact with the choke in the output filter; GaNFETs won’t switch right if the filter isn’t there to give them a bit of an inductive kick to initiate the switch. Strange but true. We’re talking about something called ’inverse conduction’.
Allowing for less deadtime is of course important. But since both GaNFETs and MOSFETs are operated far below their maximum switching capacities I don’t see this as the big advantage; you can run MOSFETs at the same speeds we’re seeing now with very similar deadtimes. After embarking on this project at this point I think Technics is missing a bet by switching so fast. Deadtime is a constant and does not change with the switching frequency; so the faster you switch, the greater the percentage of the supposed ON time is taken up with deadtime, which increases distortion. Put another way, if Technics simply reduced their switching speed they could get lower distortion.
I don't see the potential dead times being similar as Qrr is a
significant contributor of the dead time requirement outside of basic
MOSFET switching performance.
That's the body diode thing I mentioned.