Thanks for quantifying the feedback numbers, wave's tap dancing was impressive to say the least ....
Bam ...
Give the man a drink, we are back on topic ..............DVV , interesting comments , where are you seeing the biggest improvements with the caps, PSU...?
Err, ahhhh, Wave's a strange cat ...
He's From Siberia , but is afraid of the cold, part Ruskian and don't drink vodka, into tubes but only has SS ..
Ohh, he's a great tap dancer ...............
...... 
Copper , No ...
I have to ask , do you work alone ? is there anyone else you trust when evaluating, if so how did you come to the conclusion you had golden ears before this revelation ...
By popular request, I have placed this article online here:
Linear Audio | Online Resources
(2nd article down).
Opinionated perhaps, but Bruno is one who shores up his opinions with facts and figures. If someone would not agree on something with Bruno, he would have to show the flaw in his reasoning or his facts and figures. So far, no callers
jan didden
Jan,
Thanks for the link , interesting comments ......
"From the viewpoint of distortion, there is no difference
whatsoever between one macho feedback loop and local feedback with a simpler global loop around it. Designers who propose to use “mostly local feedback and only a little global feedback” are labouring under an illusion. It makes no difference. Whether you choose to use a nested loop or global feedback depends on other practicalities but has no bearing at all on actual audio performance."
Comments anyone ....?
Well, the 2N2222A'a by ST I got from Mouser measured hFE as about 190 where the On-Semi were 18 at 1mA. Very close to the original Motorola in the amp. Minimum spec is 50. I can use these!
"Regarding Golden Ears, my girlfriend has the best ears I know when it comes to low level sounds, very high frequencies (>16 kHz) or following a conversation across the room filled with 100+ people. It's almost scary what she can hear."
Not to worry. No relevance to measuring distortion. That's a girl thing. You will come to understand after you have been married for 35 years or so. Fair warning!
"Regarding Golden Ears, my girlfriend has the best ears I know when it comes to low level sounds, very high frequencies (>16 kHz) or following a conversation across the room filled with 100+ people. It's almost scary what she can hear."
Not to worry. No relevance to measuring distortion. That's a girl thing. You will come to understand after you have been married for 35 years or so. Fair warning!
By popular request, I have placed this article online here:
Linear Audio | Online Resources
(2nd article down).
Opinionated perhaps, but Bruno is one who shores up his opinions with facts and figures. If someone would not agree on something with Bruno, he would have to show the flaw in his reasoning or his facts and figures. So far, no callers
jan didden
Well done with this text! It's been YEARS since I saw a text so clearly written, and oh so true!
Well , this should get their juices flowing ....
"There are only advantages and no disadvantages to applying stratospheric amounts of negative feedback in an amplifier. The only hard part is figuring out how to do it. The more feedback, the better it sounds provided that it’s never less than 30dB at any audio frequency."
"There are only advantages and no disadvantages to applying stratospheric amounts of negative feedback in an amplifier. The only hard part is figuring out how to do it. The more feedback, the better it sounds provided that it’s never less than 30dB at any audio frequency."
@A.Wayne
I am hard pressed for an answer. I am tempted to say "everything matters", because it really is so.
But, if I had to choose, thw three most siignificant factors, in order of relevance, would be:
1. The power supply, its construction and quality,
2. The output stage, and
3. The driver stage.
Ad 1. Power is to amps like blood is to us. You can't ask a heart patient to run a marathon. If you want good sound, you have to make sure the amp is being fed with clean power and sufficient energy, while keeping a good ebergy reserve. We need those Joules badly.
Ad 2. I am NOT advocating rows upon rows of output devices, but I am saying that getting 100W/8 Ohms will take more than one pair of trannies (as im Onkyo 282 power amp). Some simple maths will determine how many, use the derating curve from the data sheet and compare that with your design goals.
I am heavily in the True Voltage Source trip, so I don't cut corners, since my 100W/8 Ohms also means 200/400W into 4/2 Ohms STEADY SТАТЕ (although the heat sinks will give way soon enough), at which time I will need to have those trannies working at no more than about 50% of their actual, real world capability if I don't want excessive distortion, so at 956 Watts of EFFECTIVE power at 65 degrees centigrade, 3 pairs of Motorola/ON Semi TO-3 MJ 21195/21196 will do the job just right.
Ad 3. I use a triple stage, meaning predriver -> driver -> output stage. Predriver is not critical if it is used properly, but the drivers stage I find is all too often being "saved" on.
While a single 50W trannie will do the job, TWO 50W trannies witl do an even better job, the lower the impedance of the speaker, the better.
Anyway, many years ago, I discovered that two parallel 50W trannies will do the same job better than a single 100W trannie.
I am hard pressed for an answer. I am tempted to say "everything matters", because it really is so.
But, if I had to choose, thw three most siignificant factors, in order of relevance, would be:
1. The power supply, its construction and quality,
2. The output stage, and
3. The driver stage.
Ad 1. Power is to amps like blood is to us. You can't ask a heart patient to run a marathon. If you want good sound, you have to make sure the amp is being fed with clean power and sufficient energy, while keeping a good ebergy reserve. We need those Joules badly.
Ad 2. I am NOT advocating rows upon rows of output devices, but I am saying that getting 100W/8 Ohms will take more than one pair of trannies (as im Onkyo 282 power amp). Some simple maths will determine how many, use the derating curve from the data sheet and compare that with your design goals.
I am heavily in the True Voltage Source trip, so I don't cut corners, since my 100W/8 Ohms also means 200/400W into 4/2 Ohms STEADY SТАТЕ (although the heat sinks will give way soon enough), at which time I will need to have those trannies working at no more than about 50% of their actual, real world capability if I don't want excessive distortion, so at 956 Watts of EFFECTIVE power at 65 degrees centigrade, 3 pairs of Motorola/ON Semi TO-3 MJ 21195/21196 will do the job just right.
Ad 3. I use a triple stage, meaning predriver -> driver -> output stage. Predriver is not critical if it is used properly, but the drivers stage I find is all too often being "saved" on.
While a single 50W trannie will do the job, TWO 50W trannies witl do an even better job, the lower the impedance of the speaker, the better.
Anyway, many years ago, I discovered that two parallel 50W trannies will do the same job better than a single 100W trannie.
Well , this should get their juices flowing ....
"There are only advantages and no disadvantages to applying stratospheric amounts of negative feedback in an amplifier. The only hard part is figuring out how to do it. The more feedback, the better it sounds provided that it’s never less than 30dB at any audio frequency."
My "magic number" is 26 dB; I find this to be well neigh perfect.
Opinionated perhaps, but Bruno is one who shores up his opinions with facts and figures. If someone would not agree on something with Bruno, he would have to show the flaw in his reasoning or his facts and figures. So far, no callers
I'm not surprised.
Easy to talk the talk, harder to walk the walk.
From the context of the article, the 30dB should be read as 'minimum for reasonable performance', and it is negotiable. So your 26dB would fill that requirement reasonably.
But the crux of the statement is of course that he says that stratospheric amounts of feedback can only make it better, if only you knew how to do it.
jan didden
@a.wayne
No, 3 pairs is more than enough, however, please bear in mind I was VERY specific about WHICH three pairs.
Work it out yourself. Their derating number is 1.43W/degree centigrade above 25. Let's assume you adjusted your overheat procetion to come on at 65 deg, on the heat sink. Assume you have an L-bar holding the trannies so that you can avoid lengthy wiring. Since the L-bar will have loss of its own, let's further assume a 10 degree loss on the L-bar (actually less if you, like I, make it of pure copper, but most won't, so 10 degrees is a fair assumption), and finally, les add another 10 degrees for whatever reason, just to be on tha safe side. That's 85 degrees all told, or 65 degrees over the ambient 25.
65 x 1.43 = 92.5 Wazzs below rated maximu, which is 250 Watts. In this case, the actual, real world available power is (250 - 92.5) 157.5 Watts per trannie, or 945 Watts for the whole stage.
That is enough for a 2 Ohm load with a -65 degreed phase shift, which would require "only" 800 Watts. What more do you want?
And that's steady state.You don't listen to pure sine waves at full power, do you Wayne? You do, I sincerely hope, take care to accomodate those transients, which means that you will never ever drive the amp harder than -6 dB of its nominal power, and even that is shaving below the skin? If not, you must have a back yard full of very dead speakers ...
Your REAL problem in such cases are definitely not the transistors, it's the whole system of power supplies. To deliver 400 Watts into 2 Ohms with a phase shift of -65 degrees is really like running the amp at 800 Watts into 2 Ohms. That is approximately 80 Joules of energy per channel, and all that juice has to come from somewhere. That means hefty toroidal power transformers separate for each channel, that means two hefty full wave bridge rectifiers per channel, and about 33.000 uF per supply line, or 66.000 uF per channel, or 132.000 uF for a stereo amp.
But I do go further and put in a fully reagulated power supply; actually, two regulated power supplies per channel, since I run my voltage amplifiers from higher regulated loads, specifically +/- 55V for the voltage section, and +/- 48 V for the surrent gain section. This includes about +/-4 V volts reserve, in case the power line drops, I don't want the amp to even know about it.
This lets me use lower voltages for the current section, which in turn lets me use those trannies more to the left of their SOAR curves.
And since this is an ego trip (i.e. not commercial), I don't stick to the old system of having one power trannie voltage regulate for two audio power trannies, rather I use exactly the same number of the same power trannies in the regulator. In effect, this is like having two power amps made for each channel, and the cost begins to be felt. But if you want great sound, you don't save and you throw the standard vulgarized KISS principle out to the doggs, where it belongs.
The kind of project you do just once in your life. With plenty of options, so everybody can find their own version.
No, 3 pairs is more than enough, however, please bear in mind I was VERY specific about WHICH three pairs.
Work it out yourself. Their derating number is 1.43W/degree centigrade above 25. Let's assume you adjusted your overheat procetion to come on at 65 deg, on the heat sink. Assume you have an L-bar holding the trannies so that you can avoid lengthy wiring. Since the L-bar will have loss of its own, let's further assume a 10 degree loss on the L-bar (actually less if you, like I, make it of pure copper, but most won't, so 10 degrees is a fair assumption), and finally, les add another 10 degrees for whatever reason, just to be on tha safe side. That's 85 degrees all told, or 65 degrees over the ambient 25.
65 x 1.43 = 92.5 Wazzs below rated maximu, which is 250 Watts. In this case, the actual, real world available power is (250 - 92.5) 157.5 Watts per trannie, or 945 Watts for the whole stage.
That is enough for a 2 Ohm load with a -65 degreed phase shift, which would require "only" 800 Watts. What more do you want?
And that's steady state.You don't listen to pure sine waves at full power, do you Wayne? You do, I sincerely hope, take care to accomodate those transients, which means that you will never ever drive the amp harder than -6 dB of its nominal power, and even that is shaving below the skin? If not, you must have a back yard full of very dead speakers ...
Your REAL problem in such cases are definitely not the transistors, it's the whole system of power supplies. To deliver 400 Watts into 2 Ohms with a phase shift of -65 degrees is really like running the amp at 800 Watts into 2 Ohms. That is approximately 80 Joules of energy per channel, and all that juice has to come from somewhere. That means hefty toroidal power transformers separate for each channel, that means two hefty full wave bridge rectifiers per channel, and about 33.000 uF per supply line, or 66.000 uF per channel, or 132.000 uF for a stereo amp.
But I do go further and put in a fully reagulated power supply; actually, two regulated power supplies per channel, since I run my voltage amplifiers from higher regulated loads, specifically +/- 55V for the voltage section, and +/- 48 V for the surrent gain section. This includes about +/-4 V volts reserve, in case the power line drops, I don't want the amp to even know about it.
This lets me use lower voltages for the current section, which in turn lets me use those trannies more to the left of their SOAR curves.
And since this is an ego trip (i.e. not commercial), I don't stick to the old system of having one power trannie voltage regulate for two audio power trannies, rather I use exactly the same number of the same power trannies in the regulator. In effect, this is like having two power amps made for each channel, and the cost begins to be felt. But if you want great sound, you don't save and you throw the standard vulgarized KISS principle out to the doggs, where it belongs.
The kind of project you do just once in your life. With plenty of options, so everybody can find their own version.
THE John Curl? Of Parasound fame?
I was listening to your 100W or so model, now say 10 or 12 years old, the other day. I used a word then, one I have used less than 10 times in my life, and that is "impressive", the way it tackled my old but refurbished AR94 speakers (the good original series, later models were junk). Just kept serenely filling them up like it was no big deal, and those speakers drained many an amp to an inch before tears.
Unfortunately, but understandably, the owner refuses to part with it ...
Perhaps so, but the question is why would you?
If you disconnect the NFB loop, feed your amp 1/20 th of its nominal signal at the input, and it remains stable up to say 80 kHz with no obvious anomalies, why would you want to apply more than 26 dB of NFB?
To be fair, my wife's Harman/Kardon HK 680 integrated amp, delivering 85/130 W into 8/4 Ohms, and using just 12 dB of overall NFB, plays some good sounds in her JBL Ti 600 floorstanders. And, being curious as I am, and testing it on the bench, I discovered it will pump out as much as 490 Watts into 1 Ohm (t=40 mS) and still remain completely stable, with some modest overshoot.
This just reiterates my initial question - why would you NEED more global NFB?
I think NFB is very alike to bias current. If you have an amp designed for class AB operation, you can usually raise its bias level. I have tried with several integrated amps and found that not a single one of them will sound better once you get to about 130 mA per output device. I think NFB is like that, there is a point after which you can pile it on but with no audible result to speak of.
After all, you do design for it, or you don't. If your gain levels are set to a certain point, adding more that its share will simply make the amp require more and more input signal for nominal power.
How do you know if you have reached a good individual stage gain point? Theoretically, we could now launch a big discussion, but empirically, it's easy to identify - once your amp is completely stable with no global NFB even way out of the audible range, like say at 50 kHz and above, you must be doing something right.
Last edited:
He forgot to mention which exact "performance" he measures: "has no bearing at all on actual audio performance".
Remember my example?
Actually, it was the same idea as Peter Walker patented in his famous Current Dumping patent. I did not know about him and his patent, so instead of working around slow opamp and power transistors implementing "The Bridge" like he did, I went differently, but the main idea is still the same: approximation of overall transfer function by summing 2 functions, adjusting summing result finally by additional feedback.
It would be interesting to see how The Author solved this problem without multiple feedback loops.
- Status
- Not open for further replies.