My experience shows you can get away with base stoppers on EF2, but, on EF3 though I have had protos break into parasitics without the stoppers. I do think layout has a big impact - keep base and collector trace inductances as low as possible. I always fit 4.7 ohm 2512 SMD 1W stoppers in all my designs - with time and money, they could be engineered out, but they provide solid benefits IMV.
Ferrites also work well - I use them in my headphone amp.
For Re, I've settled on 0.33 on all my designs. I tend to run the OPS bias on the high side (1.5 to 1.8x over the ideal) because it sounds a bit more euphonic to my ears. Of course distortion is a bit higher but I am not one for chasing 0 ppm. Just my design philosophy.
Ferrites also work well - I use them in my headphone amp.
For Re, I've settled on 0.33 on all my designs. I tend to run the OPS bias on the high side (1.5 to 1.8x over the ideal) because it sounds a bit more euphonic to my ears. Of course distortion is a bit higher but I am not one for chasing 0 ppm. Just my design philosophy.
Last edited:
Many audiophile designers don't like having ferrites or ferrite beads in their circuits because they presumably degrade sound quality. I avoid them just to be on the safe side. Maybe a 10-ohm 1/2 watt resistor with a goodly number of turns of #30 on it would be a decent choice.
Cheers,
Bob
But this is for base stop!!! For each individual transistor. Current can never be over 100mA of base current......I hope not.
Very good advice on inductance. I insist on using ground and power plane and have 0.1uF ceramic and 10uF in parallel right at the collect pin to ground at each EF transistor, no exception. I still use 2.2ohm base stop resistor.
That's the thing, here, we are talking about low distortion. BUT is distortion the only thing that matters? I don't have experience as I am a newbie in audiophile design. The real question is how much the 0ppm translates to superb sound. The very amp, YBA that got me into designing the "perfect" amp for myself is very primitive. People laugh at the schematic. But when I listened in the demo studio, it was jaw dropping compare to my Acurus which, by itself, is no dog. The store used that as demo amp for their high end speakers including the JMLab Utopia.
The circuit of the YBA is everything that is supposed to be "what not to do". It has complementary IPS but no cascode, 2EF output stage only.
So the better question is what sounds good. I can tell you I up the bias current of my Acurus and it sure improve the sound. I definitely have gm doubling as the Re is 0.5ohm and I ran almost 100mA per stage of the 4 stages parallel output. This is gm doubling big time.
Last edited:
From my understanding, ferrite bead over the transistor lead is just a simple single turn inductor. If all else's equal, you'll have less saturation problem as you only have one turn.
Of cause it depends on the material used. But they work at much higher frequency than audio. I think you can just calculate the inductance using one turn. My question is whether it can work down to 30MHz that effectively for the slow power transistor. I cannot even imagine the bead can have effect at audio frequency.
Besides, it's all depend on the material, for the same material, there should be no more distortion than a normal inductor that people use in analog filters.
Last edited:
better sound anyone?
There's a couple of points (eg current drive of O/P stage) that I'd like to cross swords with Messrs Cordell & Self but I'm resisting cos ..
... the last 10 pages or so of this thread .. Class XD, Oliver point, zillion A quiscent bias bla bla ... address THD (for any half competent design today) below 0.01%. There is NO reliable evidence that changes in THD at this level lead to better sound or even a difference.
I note Guru Cordell has studiously ignored my comments about amp 'features' that have actually been shown to result in better sound.
Alan is even prepared to accept a 6dB drop in output capability to allow his zillion A Class A current. This takes his amp into a region where I expect overload behaviour to be audible.
He might argue that he never listens at levels which require more than his castrated power rails .. and if he listens only to compressed modern music (??!) with 1dB dynamic range this might be true.
But if you make your own recordings and record certain types of traditional music, eg small, really good unaccompanied groups of singers or Beethoven piano, you may find, like me, that you clip 2x500W @ 8R at very reasonable spls even with 90dB/1m/2.83V speakers.
I'm happy with 2x200W @ 8R .. not cos I won't clip ... but cos the clipping is usually inaudible.
Of course everyone is different so to find out what power YOU need with YOUR speakers .. goto
test-how-much-voltage-power-do-your-speakers-need.html
_______________________
Bob, are you going to talk about Class G in your 2nd edition ... and other methods to get more power with the same thermal/PSU/cost?
There's a couple of points (eg current drive of O/P stage) that I'd like to cross swords with Messrs Cordell & Self but I'm resisting cos ..
... the last 10 pages or so of this thread .. Class XD, Oliver point, zillion A quiscent bias bla bla ... address THD (for any half competent design today) below 0.01%. There is NO reliable evidence that changes in THD at this level lead to better sound or even a difference.
I note Guru Cordell has studiously ignored my comments about amp 'features' that have actually been shown to result in better sound.
Alan is even prepared to accept a 6dB drop in output capability to allow his zillion A Class A current. This takes his amp into a region where I expect overload behaviour to be audible.
He might argue that he never listens at levels which require more than his castrated power rails .. and if he listens only to compressed modern music (??!) with 1dB dynamic range this might be true.
But if you make your own recordings and record certain types of traditional music, eg small, really good unaccompanied groups of singers or Beethoven piano, you may find, like me, that you clip 2x500W @ 8R at very reasonable spls even with 90dB/1m/2.83V speakers.
I'm happy with 2x200W @ 8R .. not cos I won't clip ... but cos the clipping is usually inaudible.
Of course everyone is different so to find out what power YOU need with YOUR speakers .. goto
test-how-much-voltage-power-do-your-speakers-need.html
_______________________
Bob, are you going to talk about Class G in your 2nd edition ... and other methods to get more power with the same thermal/PSU/cost?
Problem is there doesn't seems to be an agreement with all the experts here.
As I said, progress is painfully slow here. Just received the 0.22ohm today, but this debate make me put on hold again.......Not that I cannot use the time to build the preamp. I practically have nothing other than two pairs of speakers!!! Everything has to be build.
Ha ha, I just received the remote control motor drive volume pot and have been testing it. Next is to figure how to mount the IR detector on the front panel so it still looks good.....believe me, it's harder to make the thing sound good.
As I said, progress is painfully slow here. Just received the 0.22ohm today, but this debate make me put on hold again.......Not that I cannot use the time to build the preamp. I practically have nothing other than two pairs of speakers!!! Everything has to be build.
Ha ha, I just received the remote control motor drive volume pot and have been testing it. Next is to figure how to mount the IR detector on the front panel so it still looks good.....believe me, it's harder to make the thing sound good.
Actually I did not believe THD is the only thing before. But seems like most of the people here include Mr. Cordell and Self put a lot of emphasis on lowering the THD. Who am I going to argue against it? I am an EE, but from designing guitar amps, I truly believe listening is everything, but I was told here time after time that this is not the case here in audiophile. So here I am, confused.
On a typical speaker , you have an quite a complex electrical equivalent (below).
Not just a big power resistor or simple R/C load usually tested with.
An amp with low THD into a resistor is fine .... but would this
be degraded by all those L/C/R's ?
Not to mention damping factor (and speaker wire).
I'm sure you lose a whole lot of them PPM out at the other end of that
speaker wire , and most likely within the X-over.
Hey , my amps have a separate NFB wire - but would I want a 3'rd wire right
at the woofers (+) terminal ? PS - could at least eliminate the speaker
wire (with NFB).
OS
Not just a big power resistor or simple R/C load usually tested with.
An amp with low THD into a resistor is fine .... but would this
be degraded by all those L/C/R's ?
Not to mention damping factor (and speaker wire).
I'm sure you lose a whole lot of them PPM out at the other end of that
speaker wire , and most likely within the X-over.
Hey , my amps have a separate NFB wire - but would I want a 3'rd wire right
at the woofers (+) terminal ?
PS - could at least eliminate the speakerwire (with NFB).
OS
Attachments
Hi MiiB,
Yes, it is elegant. I don't care for the JC front end, but the rest is up my alley. Start with J-FETs, cascoded of course. I would do the servos differently, but they aren't required for most front ends. Unbalancing the front end to look after DC offset isn't the way to do this at all.
Hi David,
Did John do an output stage with no emitter resistors? I am surprised by that. I also missed those posts, but that's okay. It was surprising that that thread was still plodding along!
-Chris
Yes, it is elegant. I don't care for the JC front end, but the rest is up my alley. Start with J-FETs, cascoded of course. I would do the servos differently, but they aren't required for most front ends. Unbalancing the front end to look after DC offset isn't the way to do this at all.
Hi David,
Did John do an output stage with no emitter resistors? I am surprised by that. I also missed those posts, but that's okay. It was surprising that that thread was still plodding along!
-Chris
Anatech, why is unbalancing not the way to deal with DC.?? as I see it, altering the currents shifts the device to a different working condition, where you have an exact gain matching in the two symetric halves. I normally make my houskeeping by feeding the circuit through current mirrors, then I can servo shift the currents by injecting different currents up/down. (Normally the driving currents are very close of each other (uV) as the DC gain is high)
Last edited:
I have a question about the Re discussion:
I've noticed that in some older (power-) amplifiers for Re values of .5 Ohms to 1 Ohm are applied. Is there a special business-case for making such choises?
Simulating such an amp with Re = .22 or .10 seemed to work too, with even lower distortion. From what has been discussed here already, I would not consider using the .10 Ohms in real life, since the output-stage has no protection against exceeding SOA.
I suspect some design issue that's solved by using higher values for Re. Can somebody tell me if that assumption is correct and if not, why?
Thanks in advance.
Edwin
I've noticed that in some older (power-) amplifiers for Re values of .5 Ohms to 1 Ohm are applied. Is there a special business-case for making such choises?
Simulating such an amp with Re = .22 or .10 seemed to work too, with even lower distortion. From what has been discussed here already, I would not consider using the .10 Ohms in real life, since the output-stage has no protection against exceeding SOA.
I suspect some design issue that's solved by using higher values for Re. Can somebody tell me if that assumption is correct and if not, why?
Thanks in advance.
Edwin
I learned a lot about the Re. At this point, I think I am more comfortable to go below 0.22. I think if you match the Vbe and beta closely between the parallel transistors, you can definitely go lower than 0.22.
The equation 15.2 in p300 of Mr.Cordell's book is for calculation of a single transistor, which, assuming that transistor is hogging all the current. If you spread it out between parallel transistor, it's a lot better.
Key thing is the rail voltage, there is no other way around it unless you do something very special to lower the thermal resistance. So matching transistors is the alternative way.
At this point, I am tending to go with 0.15 or 0.18ohm resistor instead of 0.12 originally designed. I have 5 pairs, I matched the transistor to 1% easily and I use 40V rail, it's a lot more forgiving. I am going a lot more conservative than Self.
So as of now, it's 0.15ohm, at 175mA, dissipating 70W per channel at idle. Or,
0.18ohm, 150mA dissipating 58W per channel.
Both satisfy Oliver's condition.
Thanks for your reply, Alan.
Idling current and rail voltage are less than what you describe, in the range of 80mA and 30 Volts. Single (darlington) transistor topology...
So I guess changing the Re's from 0.5 Ohms to .18 or .22 won't do any harm... giving less THD...
Still got to run simulations at 10Khz and 20Khz...
If you use 80mA, you need to use 0.33ohm to get Oliver's condition. Idle power dissipation is 60V X 0.08=4.8W
If you use 0.22ohm, you adjust to 118mA to get Oliver's condition.Idle power dissipation of single stage is 60V X .118=7.08W
If you use 0.18ohm, you adjust to 144mA to get Oliver's condition. Idle power dissipation of single stage is 60V X .144=8.64W
Don't just change the resistor without adjusting the current.
A ferrite bead can be thought of as a very lossy inductor, but the loss is not series resistance, it is loss introduced by the ferrite. The impedance starts out very low, nearly zero, at dc, then at high frequencies it increases, somewhat like that of an inductor, then it levels off at very high frequencies, where it looks resistive. It can be roughly approximated by a small inductor in parallel with a resistor.
Cheers,
Bob
There's a couple of points (eg current drive of O/P stage) that I'd like to cross swords with Messrs Cordell & Self but I'm resisting cos ..
... the last 10 pages or so of this thread .. Class XD, Oliver point, zillion A quiscent bias bla bla ... address THD (for any half competent design today) below 0.01%. There is NO reliable evidence that changes in THD at this level lead to better sound or even a difference.
I note Guru Cordell has studiously ignored my comments about amp 'features' that have actually been shown to result in better sound.
Alan is even prepared to accept a 6dB drop in output capability to allow his zillion A Class A current. This takes his amp into a region where I expect overload behaviour to be audible.
He might argue that he never listens at levels which require more than his castrated power rails .. and if he listens only to compressed modern music (??!) with 1dB dynamic range this might be true.
But if you make your own recordings and record certain types of traditional music, eg small, really good unaccompanied groups of singers or Beethoven piano, you may find, like me, that you clip 2x500W @ 8R at very reasonable spls even with 90dB/1m/2.83V speakers.
I'm happy with 2x200W @ 8R .. not cos I won't clip ... but cos the clipping is usually inaudible.
Of course everyone is different so to find out what power YOU need with YOUR speakers .. goto
test-how-much-voltage-power-do-your-speakers-need.html
_______________________
Bob, are you going to talk about Class G in your 2nd edition ... and other methods to get more power with the same thermal/PSU/cost?
Hi kgrlee,
I don't think we disagree very much. I regard those features that you talked about as very important, and most of them are covered in my book. I am a strong advocate of soft clipping circuits like the Klever Clipper which I developed many years ago. See my chapter on civilized amplifier behavior.
Where we seem to disagree is that I choose not to get into the DBLT debate in my book.
I do agree wholeheartedly, and have demonstrated it at meetings, that even powerful amplifiers clip on well-recorded music when played at realistic, yet listenable, levels in ordinary-sized rooms. Of course, this depends on the efficiency of the loudspeakers being used. The loudspeakers I used in many of these demonstrations were my M3s, built with Morel drivers and having efficiency on the order of 85dB. I even designed and built my peak-average meter to demonstrate that. Most of this is discussed on my web site.
These sorts of things were demonstrated at workshops at RMAF and HE2007. I have cited the well-recorded Ricky Lee Jones "Getto of my Mind" on her Flying Cowboys album numerous times as an example of a recording with very high dynamic range and crest factors.
Low distortion is a good thing, but not the only thing. Amplifiers must behave well in all respects under many different playing conditions to sound good overall. If they clip, they should do it cleanly. They must never burst, even briefly, into parasitic oscillation. The list goes on. It takes a lot of attention to detail to achieve very low distortion, and this can lead to a very good amplifier in other respects as well - maybe respects that are not as well-measured.
I did cover class G in my book, but I did not devote a whole chapter to it.
Cheers,
Bob