Why should I start with a 1 kHz square wave?? Is that what you listen to?
If a 20 kHz sine has a slew rate of 0.15V/us/V, a 1 kHz sine wave has 1/20 of that, or a puny 0.0075 V/us/V. You amp that has 50V/us for 50V output has 1V/us/V, which is a safety factor for the 1 kHz of 133 times.
If you think this is wrong, please do me the courtesy to explain me where my reasoning is wrong. 'Because I say so' is is no explanation.
Jan
Last edited:
Jcx, you did not read my input properly. (bandwidth limited) means that an extra rolloff HAS to be put in between 20KHz and 100KHz, depending on how serious you are about high speed measurements. Of course, without this extra rolloff, the slew rate tends toward infinity, although it will be limited by the test signal.
Now, I can use (if I wanted) a 1KHz, 5V square wave with a 1ns rise-time. This is an intrinsic slew-rate of 5V/1ns or 5000V/us. Kind of high isn't it. Yet put a 30KHz, 6dB/oct filter and you get something much more reasonable. Same source, just externally bandwidth limited.
Now, I can use (if I wanted) a 1KHz, 5V square wave with a 1ns rise-time. This is an intrinsic slew-rate of 5V/1ns or 5000V/us. Kind of high isn't it. Yet put a 30KHz, 6dB/oct filter and you get something much more reasonable. Same source, just externally bandwidth limited.
square-waves in normal pop music
I took a picture of a square-wave type shape in normal music. Uploaded as an attachment.
Note that the square-wave here is a rhythmic beat, however as soon as she starts singing with the beat later it looks a lot less like a square-wave shape, even if it still is.
The song example here is Tata Young - El Niño
I usually see unnatural shapes in synthetic music like trance, dark electro, chiptune, etc.
I took a picture of a square-wave type shape in normal music. Uploaded as an attachment.
Note that the square-wave here is a rhythmic beat, however as soon as she starts singing with the beat later it looks a lot less like a square-wave shape, even if it still is.
The song example here is Tata Young - El Niño
I usually see unnatural shapes in synthetic music like trance, dark electro, chiptune, etc.
Attachments
Last edited:
I agree with you on most of this..... I need to clarify something.... I was not talking about CFB vs VFB amp as to which is better. Yes, both sides have learned they can do anything the other can do. the Point was the SR development there.
I even think JC's SR number of 100 might be low today. Certainly not less. Not for SOTA. From one of the Old Guys. The main reason I am here is to do a brain dump of experience and help point a direction. Perhaps the other Geezers are also. Well maybe not Demian.... he is just starting to get old enough.... not retired yet .... to be classified as a true Geezer. Maybe we can find the other Geezer designers who did the designs for products we all know and love(?). Get them to do a brain dump, also. I mean give a whole lot of details about what they learned about. If they are still making money on their designs then you will not get even Geezer status guys to tell much. Too bad it has to be that way. Most of what is known now wont get into print and will be lost art and knowledge. After all, no body will again devote a life to studying all these things when the world has changed and gone in a new direction.
THx-RNMarsh
Last edited:
This is very heavily clipped music. That's why it looks like a square wave, it runs into the max level limit, probably in the ADC. This must sound horrible!
See the attached similar example.
Jan
Attachments
And that's exactly the point, Richard.
This is a DIY forum and not everyone can be expected to have full electrical engineering background, so it has to be spiced with some explaining, for example, why you think this or that is important, what you think is the lowest acceptable level and what you would recommend up and above it, and pointers how to get there.
I do NOT expect you or anyone else to turn over your completed projects, the fruit of your labor, as is. Heck, I don't WANT you to do that, I want to do it myself if I can, but some pointers would be good. If I'm off course, tell me where I'm going wrong and give a general suggestion how to remedy the problem. Such as, for example: don't use a single transistor for the VAS, take time and work to make it a cascode stage. That sort of thing.
Because as you say and I absolutely agree with, you are the first and probably last generation who did the grunt basic research and development work. I'm sure that is still going on today, but I'm equally sure it's nowhere near to the volume of what it was in the 70ies.
And if the principle is sound (pun intended), it will stand firm today just as it did then.
Last edited:
Well guys I read all of that and remain totally unconvinced or more so. Sure I understand what you say. It's your science I don't care for. To say Vinyl records cause problems, I am afraid to say that doesn't wash. I know a bit about that. The 75 uS should do a lot to remove the problem especially if passive. The Slew rates called desirable are 50 times above expectations or 100 times when early estimates ( Peter Baxendale using Vinyl 0.55 V/uS if 100 watts ). I have no objection to people saying I do it because I can and I am proud of my work. I suspect they accurately heard something and we need the proper answer. If real science we would not be allowed to say what we do. Yes guys Hi Fi isn't real science.
I suppose John and others built high slew rate amps and like magic they sounded much nicer? They then had to simply say why that should be. The PR guys says, look chaps this will be way over their heads. Better to say a recent discovery says high slew rates sound better. Customer then using "a little knowledge is a dangerous thing" routine fixates on this ( or did ).
What I am trying to do is find the real answer. People have become victims of their own hype I feel.
I have drawn a very crude example of current limiting, I care not one fig if a real example. The limiter is a switch. Ideally if should be a FET with it's gradual turn on curve. Hand selecting devices might make it possible.
The next conjecture is the nasty Vinyl is nothing of the sort. It is the Atomic bomb whip crack from the protection circuit, in addition the output devices in AB loosing the loop. A fast/quick recovery circuit will make a very good attempt to restore the balance. As the spectrum will be VHF it is going to be fun. Being that the amplifier is more or less symmetrical in input No1 and No 2 where a problem comes from is unimportant. Remember also the output is mostly the input plus much more. Left to it's own choice it would sit the output at one rail if no feedback.
The Emitter resistors will help a bit and might suggest being generous ( 0R47). That will slightly increase distortion.
I am assuming Matti Otala to be right when saying currents approaching 30 amps might be realistic for a 50 watt amplifier. As we pursue very low output impedance that might be closer to the truth than we think. I have never asked myself if virtual low impedance is the same as low resistance. I guess it must be? Naim's 0R22 in place of the inductor looks sensible.
The question you have to ask yourself is on a scale of 1 to 10 how challenging is the input No 1 when real music ? 2 would be my answer. If my conjecture is right a score of 50 from the other darker side . That is identical input No 2, the feedback side. If it is 50 we need 25 times that the performance music demands. That might be a walk in the park to prove. Part of the conjecture is that current limited clipping is common and not at absolute clipping levels we think of.
Dejan the amplifier you follow should work if this is right. I saw an amp use fuses as emitter resistors. If DC protection and fuse blowing detector is fitted it might be an idea?
I show the IGBT output stage as it might be an economical answer. In general it retains the CE performance in everything it does, almost a clone. I doubt it needs drastic current limiting. Bias point is just as predicable as any familiar circuit. Doubtless stability more tricky. It is mildly better into 4 ohms as a bonus when looking at distortion changes. It might be possible to use cheap FET's. If so go for high Ron. N FET's at 500V have that and most P types. They appreciate similar to bipolar biasing so should be possible.
If my further conjecture is right no current limiting required if enough MOS FET's used. Then the Atomic Bomb never goes off.
If you build a valve amp it falls over in the audio band if seeing a square-wave above 2 kHz. In reality nothing bad happens , quite the reverse. Well there can't be two versions of the truth can there ? Now we can't invent rules to suit the picture we see. Trust me, valves excel on what looks to fail. The lack of feedback must be why. BTW. Nothing is more awful than a valve amp with high feedback. Not impossible, not worth trying if looking for instant results.
When testing the Hypex UCD 180 I found some universal truths emerged. All the things said to be bad have to be employed. The amp needs to roll off at 47 kHz. It doesn't sound dull. It has 20 + amps of current output. Guess what? It sounds a lot better than it should. I would suggest what Hypex had to cure we need to cure. The amps are not a lot different . Both Class AB and D have devices that switch. They don't need any extra complication.
I rest my case. If anyone doesn't like how it is drawn show us the correct way.
Last edited:
Futterman, Harman Kardon, and McIntosh (just to name three off the top of my head) might disagree.
Slew rate
I tried to put some numbers on slew rate distortion, not sure though I'm doing it the right way.
I took an AC current source, 20kHz, 1mA peak, into a capacitor, to simulate the output current of an LTP into a compensation cap.
See graphs.
Top to bottom: 100pF, 1nF, 10nF.
At first sight it appears that the larger cap gives larger distortion. But if you look at the scales, there's almost no change!
Am I missing something?
Jan
I tried to put some numbers on slew rate distortion, not sure though I'm doing it the right way.
I took an AC current source, 20kHz, 1mA peak, into a capacitor, to simulate the output current of an LTP into a compensation cap.
See graphs.
Top to bottom: 100pF, 1nF, 10nF.
At first sight it appears that the larger cap gives larger distortion. But if you look at the scales, there's almost no change!
Am I missing something?
Jan
I know that exactly as I am working on similar to drive headphones ( OTL valves). It is not what you would choose to do. Harvey Rosenberg said it all. Julius died the moment they got it right said H R. The output transformer when fitted doesn't help.
You might like this. Garrard 301.
http://howtospendit.ft.com/audiovisual/58943-turntables
You might like this. Garrard 301.
http://howtospendit.ft.com/audiovisual/58943-turntables
Last edited:
Nige, if you think about it, the amp I am thinking about is simply a repackaged version of the Otala/Lohstroh amp, same thing done a bit differently in terms of topology, the greatest difference being that it uses a triple stage output. And of course, it uses more modern semiconductors, which is boon unto itself. From the outside, I have the added benefit of higher voltages and lower distortion figures from that
I have tried to incorporate what we have learnt since, as for example shown on Samuel Groner's site, and added a bit o' spice by exploiting the fact that the buffer stage has a voltage drop across it of just 3.2V, meaning that I can easily use FETs rather than BJTs.
The rest both is and is not mine. Remember that the article mentions that for simplicity's sake it doesn't show the protection circuitry, while I do, and it makes mine seem to be much more complicated, when in fact it's not all that more complicated.
Mayn here, including you, will argue that electronic protection is a pain. I agree in term of added work required, but I will not give it up at any cost. No fuse ever made can hope to match transistors for reaction speed.
I have tried to incorporate what we have learnt since, as for example shown on Samuel Groner's site, and added a bit o' spice by exploiting the fact that the buffer stage has a voltage drop across it of just 3.2V, meaning that I can easily use FETs rather than BJTs.
The rest both is and is not mine. Remember that the article mentions that for simplicity's sake it doesn't show the protection circuitry, while I do, and it makes mine seem to be much more complicated, when in fact it's not all that more complicated.
Mayn here, including you, will argue that electronic protection is a pain. I agree in term of added work required, but I will not give it up at any cost. No fuse ever made can hope to match transistors for reaction speed.
even if you had magic slewrate amp, IMHO still does not exist such speaker to faithfully reproduce square signal (inertia,mass,...)
or been invented some audiophile magic grease for speaker cones, which improves this ?
- Status
- Not open for further replies.