Hi AW
1..25 Ampere peak, so you should disable the protection when playing on Scintillas. Other than that, it will work fine together!
2..I have looked though my mailbox, and found no mails from you.
Are you sending to: lc@lcaudio.dk?
Best regards
Lars Clausen
1..25 Ampere peak, so you should disable the protection when playing on Scintillas. Other than that, it will work fine together!
2..I have looked though my mailbox, and found no mails from you.
Are you sending to: lc@lcaudio.dk?
Best regards
Lars Clausen
Three times the bandwitdth of the UcD should allow much steeper transient output for your amp than the UcD could ever achieve (when fed with a rectangular signal).
Looking at the scope output this is definitely not the case. This should give you some food for thought, Lars !
Regards
Charles
Jan-Peter said:
Ah, finally, steep edges,
Interesting to note is that both on the 4 ohm and 16 ohm examples, the upper graph takes nearly 20us to settle, but for different reasons. At 4 ohm because the edge is slow, at 16 ohms because it overshoots and needs to ring out. The lower plot shows about 10us settling time invariant of load.
The impedance of most dynamic loudspeakers rises well above 16 ohms at higher frequencies.
Hearing the effect of "overshoot and ringing" on actual music is interesting. For example, on a drum hit, an "overshooting" amp/speaker combination will greatly exaggerate and thicken the initial transient, completely masking the natural ring-out of the drum that follows. You hear a loud, sharp "bang" but information about the size, stereo depth and even tuning of the drum is largely lost in the time your ear spends recovering. To the casual listener, the exaggerated transient sounds "more dynamic" until it becomes clear that when the overshoot (and corresponding HF boost) is corrected, you can play louder without feeling as if your eardrums are bursting, get actually more realistic dynamics (well at least if you know what a drum kit sounds like in real life) and have a much more life-like, "3D" presentation of the drum kit.
It is quite surprising this should have such an effect on band-limited material like CD recordings (and using band-limited sensors like human ears), but the effect can't be missed. This shows you're much better off having a somewhat narrower but fully controlled bandwidth than to reproduce higher frequencies with an unknown frequency (and hence time!) response.
(Suggested test disc for percussion: "Touch" by Maarten van der Valk, published by Siltech)
Bruno Putzeys said:
Ah, finally, steep edges,
Interesting to note is that both on the 4 ohm and 16 ohm examples, the upper graph takes nearly 20us to settle, but for different reasons. At 4 ohm because the edge is slow, at 16 ohms because it overshoots and needs to ring out. The lower plot shows about 10us settling time invariant of load.
The impedance of most dynamic loudspeakers rises well above 16 ohms at higher frequencies.
Hearing the effect of "overshoot and ringing" on actual music is interesting. For example, on a drum hit, an "overshooting" amp/speaker combination will greatly exaggerate and thicken the initial transient, completely masking the natural ring-out of the drum that follows. You hear a loud, sharp "bang" but information about the size, stereo depth and even tuning of the drum is largely lost in the time your ear spends recovering. To the casual listener, the exaggerated transient sounds "more dynamic" until it becomes clear that when the overshoot (and corresponding HF boost) is corrected, you can play louder without feeling as if your eardrums are bursting, get actually more realistic dynamics (well at least if you know what a drum kit sounds like in real life) and have a much more life-like, "3D" presentation of the drum kit.
It is quite surprising this should have such an effect on band-limited material like CD recordings (and using band-limited sensors like human ears), but the effect can't be missed. This shows you're much better off having a somewhat narrower but fully controlled bandwidth than to reproduce higher frequencies with an unknown frequency (and hence time!) response.
(Suggested test disc for percussion: "Touch" by Maarten van der Valk, published by Siltech)
Not too long ago I read an article that explained a lot. A test was done on a group of listeners with the goal of measuring their sonic time arrival discrimination. The researchers thought this was a vital clue to determining where sound came from therefore a survival trait.
They did routine testing and classified the listeners in 2 general categories. Average people and professional people that dealt with sound including so called golden ear types. Their measurements of the ability of average people was in the 10-15 microsecond range and pro types routinely detected 5 microseconds. Just how this could be is amazing with our limited bandwidth. I don’t remember if this was just leading edges or both rise and fall time that was measured. The ear/brain mechanism is a remarkable instrument.
This means that our equipment must maintain timing relationships to 10 times better or 500 ns. I think this goes a long way to explaining how conventional measurements don’t tell much about the sound staging ability of audio equipment. It also explains things like why we can hear a 1% cap imbalance between channels in a RIAA circuit.
Point I am making is that actual rise time ability may not be that important but it must be uniform and the same between channels. Of course, ringing totally screws up our perception of both edges.
Roger
Bruno: No no you are missing the point again! My amplifier sounds more open and dynamic, simply because of the low feedback delay, and the lack of load interference in the feedback loop.
It's that simple.
When you compare the signal in the feedback loop with a signal that is 3-5 uS old from the output, you lose precision and definition in the top. (It is obvious!) When the EMF from the woofer intereferes with the signal in the feedback loop, then you lose bass control.
Works likewise for conventional and Class D amplifiers.
Much of the commercial audio industry never paid attention to these simple factors, much like they never paid attention to the noise level of clock generators in their CD players. (Because engineers couldn't measure the difference in THD with a low noise or high noise clock, and so - you would most probably agree - the factor didn't exist). Well it does!
Jocko: I know bandwidth limiting does away with a lot of stability issues, but then i think it's worth while doing an effort to make the best amplifier, and get the best sound. Even if it means the amplifier isn't fool proof under all circumstances. (Like when a phono plug is only halfway inserted in the socket, so only the hot wire is connected). In that case a bandwidth limited amplifier will not take notice, but a wideband amplifier may go into unintended oscillation.
It's that simple.
When you compare the signal in the feedback loop with a signal that is 3-5 uS old from the output, you lose precision and definition in the top. (It is obvious!) When the EMF from the woofer intereferes with the signal in the feedback loop, then you lose bass control.
Works likewise for conventional and Class D amplifiers.
Much of the commercial audio industry never paid attention to these simple factors, much like they never paid attention to the noise level of clock generators in their CD players. (Because engineers couldn't measure the difference in THD with a low noise or high noise clock, and so - you would most probably agree - the factor didn't exist). Well it does!
Jocko: I know bandwidth limiting does away with a lot of stability issues, but then i think it's worth while doing an effort to make the best amplifier, and get the best sound. Even if it means the amplifier isn't fool proof under all circumstances. (Like when a phono plug is only halfway inserted in the socket, so only the hot wire is connected). In that case a bandwidth limited amplifier will not take notice, but a wideband amplifier may go into unintended oscillation.
Hi Lars,
Your posts are getting more amusing by the day...
And so obviously wrong. It has been explained at length before, by myself and by charles, that the signal from the output filter is not delayed, and that a simple phase lead network is sufficient to get over it.
http://www.diyaudio.com/forums/showthread.php?postid=467294#post467294
http://www.diyaudio.com/forums/showthread.php?postid=656464#post656464
Most interestingly, both were in reply to repeated statements from your side mentioning delay... You still haven't learned, have you?
If the output filter produced a straight delay, how would I be getting 400kHz switching frequency (and I can get much higher if for I wanted), with all feedback post-filter? If there even were a delay of only 3us, switching frequency couldn't possibly be higher than 1/(2*3us)=166kHz. Worse still, how on earth am I getting anything like a flat frequency response if the delay story were even remotely correct? In other words, UcD wouldn't even work if your view of things were correct. If you really think of the phase shift of a 2nd order lowpass filter as a straight delay, it's high time to go back to school. You might finally learn that there is a crucial difference between "group delay" and "delay":
Group delay is defined as phase shift divided by frequency (phase shift in radians, freq in rad/s), and is only another way of looking at phase.
Phase shift of an output filter above the cut-off frequency converges to a constant 180 degrees. Therefore, group delay converges to ZERO. There is NO delay in an output filter. When you apply a step at the input of a 2nd order lowpass filter, the response appears immediately, if small initially.
Delay is used only to denote systems that have a fully constant group delay, also called "linear phase" (phase shift directly proportional to frequency).
The difference between delay and group delay is not subtle, my dear, and not understanding it can be quite an embarrassment.
Or maybe you do know better, and you're only repeating the same old flawed argument hoping it will stick with some people.
I forgot which propaganda chief it was that said "Lie, lie, until the lie becomes truth."
Indeed, since confusing "delay" with "group delay" is so easy to do, and many people can't be bothered with lear understanding how wrong the delay story is,
the idea that an output filter produces "delay" is easy to sell, most people hate having to look beyond gross oversimplifications, so the one who repeats his point is most likely to be believed, not necessarily the one is right. Usually the truth is just a bit harder to understand.
In propaganda, you are at a clear advantage, because all you have to do is make a short, simple but false statement, whereas I have to take time to point out all the errors to show how false it is. This will wear me out quicker than it will do you. This, for instance, is why creationists get such following. Serious scientists simply give up because they can't explain the complex truth as quickly as the creationists can spew easily digestible nonsense.
What this means is that you might well not be trying to make me "see" your point, but to convince your possible clientele, through constant repetition, that your view is the correct one.
Sorry Lars, appreciation of sound may be open to opinion, the operation of electronic circuits certainly is not. Electrons don't respond to opinions, only to the laws of physics. It's not because you think, or can convince people that a lowpass filter has a constant group delay beyond its bandwidth that it suddenly be so.
All while super-feedback amplifiers like the old Krells are considered to be the ones that have the best bass control
Sorry Lars, you've still got quite some work to do, both on the theoretical front and on learning how to listen to audio.
Your posts are getting more amusing by the day...
Lars Clausen said:
And so obviously wrong. It has been explained at length before, by myself and by charles, that the signal from the output filter is not delayed, and that a simple phase lead network is sufficient to get over it.
http://www.diyaudio.com/forums/showthread.php?postid=467294#post467294
http://www.diyaudio.com/forums/showthread.php?postid=656464#post656464
Most interestingly, both were in reply to repeated statements from your side mentioning delay... You still haven't learned, have you?
If the output filter produced a straight delay, how would I be getting 400kHz switching frequency (and I can get much higher if for I wanted), with all feedback post-filter? If there even were a delay of only 3us, switching frequency couldn't possibly be higher than 1/(2*3us)=166kHz. Worse still, how on earth am I getting anything like a flat frequency response if the delay story were even remotely correct? In other words, UcD wouldn't even work if your view of things were correct. If you really think of the phase shift of a 2nd order lowpass filter as a straight delay, it's high time to go back to school. You might finally learn that there is a crucial difference between "group delay" and "delay":
Group delay is defined as phase shift divided by frequency (phase shift in radians, freq in rad/s), and is only another way of looking at phase.
Phase shift of an output filter above the cut-off frequency converges to a constant 180 degrees. Therefore, group delay converges to ZERO. There is NO delay in an output filter. When you apply a step at the input of a 2nd order lowpass filter, the response appears immediately, if small initially.
Delay is used only to denote systems that have a fully constant group delay, also called "linear phase" (phase shift directly proportional to frequency).
The difference between delay and group delay is not subtle, my dear, and not understanding it can be quite an embarrassment.
Or maybe you do know better, and you're only repeating the same old flawed argument hoping it will stick with some people.
I forgot which propaganda chief it was that said "Lie, lie, until the lie becomes truth."
Indeed, since confusing "delay" with "group delay" is so easy to do, and many people can't be bothered with lear understanding how wrong the delay story is,
the idea that an output filter produces "delay" is easy to sell, most people hate having to look beyond gross oversimplifications, so the one who repeats his point is most likely to be believed, not necessarily the one is right. Usually the truth is just a bit harder to understand.
In propaganda, you are at a clear advantage, because all you have to do is make a short, simple but false statement, whereas I have to take time to point out all the errors to show how false it is. This will wear me out quicker than it will do you. This, for instance, is why creationists get such following. Serious scientists simply give up because they can't explain the complex truth as quickly as the creationists can spew easily digestible nonsense.
What this means is that you might well not be trying to make me "see" your point, but to convince your possible clientele, through constant repetition, that your view is the correct one.
Sorry Lars, appreciation of sound may be open to opinion, the operation of electronic circuits certainly is not. Electrons don't respond to opinions, only to the laws of physics. It's not because you think, or can convince people that a lowpass filter has a constant group delay beyond its bandwidth that it suddenly be so.
InterestingLars Clausen said:
All while super-feedback amplifiers like the old Krells are considered to be the ones that have the best bass control Ehhhh I don't think Jocko said that. Band-limiting the input to a marginally stable amplifier does not make it stabler. Another popular myth dashed.Lars Clausen said:
Sorry Lars, you've still got quite some work to do, both on the theoretical front and on learning how to listen to audio.
I would really like to see that phase shift of your output filter on a plot.
30uH and 680 nF, that's an Fc of around 35000 Hz. So you have a phase shift of 0 - 180 degrees from 0 to 35000 Hz. (Or am i also complete wrong about the math here?)
I don't think the old Krells (Like the KSA-100) had more or less feedback than most other amplifiers. What it did have though was a huge power bank, and loads of output transistors. Maybe that had some effect??
I think we both are. And we can keep on as long as you like. For example, your curves, are supposed to show your 'clientele' that you guys have the best product. Well yes as long as you can make people believe that what's on paper is the whole truth.
But why don't Jan-Peter go to the local flea market, and buy a Technics amplifier from 1982. It should cost maybe 20 Euro. I am sure if he made all the same curves on that one, they will all look a lot nicer than both of our amplifiers.
But does it sound better?
Below the phase plot of a second order lowpass filter, such as found in both UcD and ZAPpulse.
The yellow curve is the filter response of 30uH/680n.
The white curve is the phase response of that same filter.
The green curve is the phase response of the filter in ZAPpulse.
Attachments
Lars Clausen said:
According to my simple calculations, a phase shift of 180 degrees at 400kHz (at oscillation frequency) corresponds only with a delay of 1.25us. I don't see where you get the 3-5us from????? I think there is no significant difference between ZAP and UcD on this feedback time stuff as both use the output signal of the amp, one before the filter, one after the filter, but they both oscillate at about the same frequency. In the ZAP case, an integrator is used (1st order roll off) and the UcD uses a 2nd order low pass (the output filter) and then a phase lead in the feedback which makes the combination of the filter and phase lead network a kind of 1st order roll off like an integrator. As a result, both seem to oscillate at about the same frequency (due to similar delay or more correct phase shift), so even if we could speak about something like delay, it would be the same for both ZAP and UcD. Or am I missing something here.
Also don't get your remark on back EMF of woofers. An ideal amp is a voltage source (at least considered by most), especially for bass amps, and such an amp has to absorb the back EMF. To be able to do that, the feedback loop has to see it. Am I missing something here?
I think you started posting the first pictures for UcD and ZAP comparison, now the UcD team strikes back and I think your defense is pretty weak. Although as Bruno said, people who do not have any electronics knowledge may fall for your marketing like stories. I would say that it would probably be pretty easy for the UcD team to use the same trick by showing the measurements data. Anybody without any electronics knowledge could immediately see that UcD is "better" by just looking at the square waves as it would be easy to convince people that a better looking square wave must also sound better. Also distortion that is flat over the frequency range looks very nice in the eyes of amateurs if you use it as a selling argument.
Note that I'm not saying that your amps sound better or worse than UcD, I have never done a real good 1 to 1 comparison, just don't see how your arguments make sense.
In fact, one thing I could counter against pre-filter feedback (if I may use your own logic) is that since the output filter gives delay, the feedback loop in ZAP is actually correcting the signal before it even has arrived at the output, how can you correct something that does not yet exist? I know, you will probably easily counter this, as you basically say that is the feature of the ZAP and not a bug since you say that pre-filter feedback is better since it does not have the filter delay, and then we are back at the beginning of this mail.
So maybe I'm too stupid to understand your theory (could be) and I may see the light sometime but that will not be anytime soon I'm afraid.
Best regards
Gertjan
Ideally you can consider the amplifier to be a perfect voltage source, it's the so called 'black box' theory. That theory was popular back in the start of the 1980's but nowadays many high-end manufacturers consider the amplifier a part of the chain, where it must cooperate as well as possible with the loudspeakers, and source, to make good performance as a system.
The black box theory does not work in real life. But some amplifiers have low THD in a resistive load, and higher THD when a speaker is used as the load. Others have high THD with a resistor load, and significantly lower THD with a real speaker as load. Lab or listening room ......
Ghemik: you know better than that ...
I will repeat my self for the 50th time: i do not use distortion as a selling argument. But i would say Jan-Peter does, as did most of the japanese audio industry in the 1980's Anyway i will say that Bruno and Jan-Peter's amplifier sounds a lot better than most of the stuff you could buy back then Exactly my point!
And if you were to take the feedback directly from the output, what good does it do to correct something wrong, that has already been sent to the loudspeaker??? (To stay in your terminology). Youhave to correct the flaws of the signal before they reach the output. As does many other engineers in the industry.
I will again repeat that i am not saying either principle is better than the other. It would be like arguing that Jazz is better than Rock. Yes it is for some people, and for others it's the other way around. They are different.
I'm also trying to explain my way of doing the Class D compared to Bruno's. And of course i find my way is better, in light of which principles i believe have effect in true high-end audio design. I would not make something that i believed was second best, of course
Who would?And Bruno is of course doing the same, it's nothing personal, i am sure if Bruno, Jan-Peter and myself were to meet in real life, we could have a couple of Heineken together, and a good friendly chat. So i think even if this thread at some points seem to be a little edgy, maybe, it still suits a good purpose of bringing out the differences of design philosophy on each side. After all it is a discussion board, and so it's kind of the purpose of the site to make discussions. And i think this one is mostly with a good technical content, and interesting points of view. That's why i'm participating.
And of course i will also like to thank Bruno for participating in the conversation in here, as i have done already to Jan-Peter by e-mail.
Lars Clausen said:
Ideally you can consider the amplifier to be a perfect voltage source, it's the so called 'black box' theory. That theory was popular back in the start of the 1980's but nowadays many high-end manufacturers consider the amplifier a part of the chain, where it must work as well as possible with the loudspeakers, and source, as a system.
The black box theory does not work in real life.
Ghemik: you know better than that ...
Exactly my point!
As does many other engineers in the industry.
I will again repeat that i am not saying either principle is better than the other. It would be like arguing that Jazz is better than Rock. Yes it is for some people, and for others it's the other way around. They are different.
I'm also trying to explain my way of doing the Class D compared to Bruno's. And of course i think my way is better, in light of which principles i believe have effect in true high-end audio design. I would not make something that i believed was second best, of course
And Bruno is of course doing the same, it's nothing personal, i am sure if Bruno, Jan-Peter and myself were to meet in real life, we could have a couple of Heineken together, and a good friendly chat. So i think even if this thread at some points seem to be a little edgy, maybe, it still suits a good purpose of bringing out the differences of design philosophy on each side. After all it is a discussion board, and so it's kind of the purpose of the site to make discussions. And i think this one is mostly with a good technical content, and interesting points of view. That's why i'm participating.
And of course i will also like to thank Bruno for participating in the conversation in here, as i have done already to Jan-Peter by e-mail.
It seems our views (technically, not talking about sound quality)are far apart, even with respect to beer we seem to disagree, if it has to be a Dutch beer then I prefer Grolsch. But to be honest, I prefer the beer from Belgium, I'm afraid Bruno would as well.
Cheers
Gertjan
Will my amplifier performance increase if I drink belgium beer vs dutch beer? Do I have to drink it while listening or is it OK to drink it before? What about the choice of glass - will that affect the tonal balance...? (Should I use a premier choice of glass or will a standard glass be ok?)
Thanks for any insights
Ed W
Thanks for any insights
Ed W
Lars Clausen said:
Exactly my point!
Lars,
I have to take exception to this point. All feed back is after the fact, the important thing is how long after. Nano seconds are not going to mater. This is fast enough to be considered as instantaneous, at least as far as our ears are concerned. So that doesn’t work as a point for me. What also doesn’t work is the fact that all inductors and capacitors are far from perfect. Having them outside the loop means this non ideal behavior is presented to the speaker as distortion. Working on crossovers taught me how bad inductors can be.
Roger
Uhh yes, your maths are quite far off, as exemplified in your own plot. The phase shift hits 90º at fc, converging towards 0º at low frequencies and towards 180º at high frequencies.Lars Clausen said:I would really like to see that phase shift of your output filter on a plot.
(Or am i also complete wrong about the math here?)
That's a common misconception. Once the output impedance is significantly below the speaker impedance, it makes no difference if it's 50milliohms or 50 micro-ohms, because you get the voice coil and crossover and cable resistance in series, which is in the several ohms region. Therefore a large bank of power transistors will not make any difference. Power supply ripple translates into small amounts of modulation with the audio, especially at high frequencies, muddying the stereo image. Because of this, having low ripple on the power supply (and hence a large storage bank) helps stabilise the stereo image on many amplifiers, but it does not provide better bass control.Lars Clausen said:I don't think the old Krells (Like the KSA-100) had more or less feedback than most other amplifiers. What it did have though was a huge power bank, and loads of output transistors. Maybe that had some effect??
What happens with most amplifiers (as exemplified by Krell) is that distortion at low frequencies is low but it rises at high frequencies. Because the high frequencies are distorted relative to the low frequencies, the low frequencies will be perceived as "cleaner", even though it's rubbish at high frequencies that's actually causing the effect. Solving the distortion problem at high frequencies will initially give the impression bass is "weak" until the listener discovers s/he can discern much better detail and stereo imaging in the bottom end.
Far from it.Lars Clausen said:
1) Distortion figures were often lied about in those days, because good distortion measuring equipment was not generally available. Claiming 0.003% was easy, if most labs could hardly measure down to 0.03%.
2) The distortion spectrum was particularly rich in high harmonics.
3) THD at the top of the audio range was particularly abysmal.
Lars Clausen said:
Ehm, what are you trying to demonstrate there, apart from the fact that your conception of the phase response of a 2nd order LPF was wrong? I know what the phase response of the UcD filter looks like, because I closed a loop around it
All it does is show that indeed the group delay of the filter at 400kHz is not quite 3us.
Also, the phase plot of the zap post filter looks rather suspicious, because it levels out at 90º. This suggests that you would be using a 1st order output filter. This is obviously not the case, so something funny must have happened in the plot expression
Neither do we. We use distortion plots to show we know what we're doing. Then we send our [OEM] customers a box containing the amplifier for them to try out. Talking about sound is like reading musical notes out loud.Lars Clausen said:Ghemik: you know better than that ...
That's very cool of you. In return I'll say your modules sound better than the measurements alone would suggest. You have done a good job selecting the passives.Lars Clausen said:
ROFLOL, you're not suggesting we should post measurements of UcD vs Zap when loaded with a real speaker, are you? I could do that if you wish...Lars Clausen said:
Lars Clausen said:
Same here. I also think it's time to call it a day. It was fun, in a way.
I'm surprised to see that lars clausen is not understood here on the delay.
Ok, he made a mistake with the phase plot, of course the 30μH/680nF lowpass network has 90 degrees phase shift at 35000 Hz and only approaches 180 degrees at the highest frequencies, like Bruno said.
BUT, when you have a load varying between 16 and 4 ohms on this network, group delay is between 1.9 and 7.5 μsec across the entire audio band.
I suspect that is exactly the delay that lars clausen mentions. NOT the delay at much higher frequencies where the group delay is much lower.
So please note that with a realworld loudspeaker load which varies with frequency, the open loop delay in the audioband is also varying with frequency and roughly between 2 and 10 μsec depending on the load on the lowpass network.
Ok, he made a mistake with the phase plot, of course the 30μH/680nF lowpass network has 90 degrees phase shift at 35000 Hz and only approaches 180 degrees at the highest frequencies, like Bruno said.
BUT, when you have a load varying between 16 and 4 ohms on this network, group delay is between 1.9 and 7.5 μsec across the entire audio band.
I suspect that is exactly the delay that lars clausen mentions. NOT the delay at much higher frequencies where the group delay is much lower.
So please note that with a realworld loudspeaker load which varies with frequency, the open loop delay in the audioband is also varying with frequency and roughly between 2 and 10 μsec depending on the load on the lowpass network.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.