I present this circuit for general discussion, it is based on the ETI project 470 from November 1979. My scaling to 20 watts (into 8 ohms) works and has worked flawlessly for several years.
Fire away!
Hi,
Just being soldering some cables 🙂
Nice old fashion amp.
One thing Im sure about: why not to connect base of K9 between R10 and R9 (and adjust both values to suit). That's old proven way to avoid faulty pot (if wiper opens) to burn amp down immediately, and has no negative sides
Second thing I'm not sure (no parts designation) , why K3 and K4 in parallel, isn't one enough?
Third also no sure; VAS has no Muller cap... not needed?
Cheers,
Drazen
Just being soldering some cables 🙂
Nice old fashion amp.
One thing Im sure about: why not to connect base of K9 between R10 and R9 (and adjust both values to suit). That's old proven way to avoid faulty pot (if wiper opens) to burn amp down immediately, and has no negative sides
Second thing I'm not sure (no parts designation) , why K3 and K4 in parallel, isn't one enough?
Third also no sure; VAS has no Muller cap... not needed?
Cheers,
Drazen
K12 & K14
TIP 142
K13 & K16
TIP 147
R13,15,14,16
1Ohm mox, 2W || 1Ohm mox, 2W
Set bias per BJT = 30mAdc, initial !
K9 and K3,4 is BD 139-16
K8
BC-550B
K10
BD 140-16
K11 and K7,6,5 is BC-560B
K1 & K2
BC-550C
PowerSupply
Rails are +/- 25Vdc
The optimum bias voltage for Class B has resulted from the manual determination of the input characteristics of the NPN and PNP Darlington at 3.1Vdc. Under no circumstances should it fall below 2.6Vdc, I read as a handwritten note in my own notes on this project. All my measurements confirm the original characteristics (see the corresponding ETI article) of this little "blameless power amp".
The sound is unspectacular, absolutely not demanding, even low-impedance speakers pose no challenge for this classic - it is without fault.
Greetings,
HBt.
TIP 142
K13 & K16
TIP 147
R13,15,14,16
1Ohm mox, 2W || 1Ohm mox, 2W
Set bias per BJT = 30mAdc, initial !
K9 and K3,4 is BD 139-16
K8
BC-550B
K10
BD 140-16
K11 and K7,6,5 is BC-560B
K1 & K2
BC-550C
PowerSupply
Rails are +/- 25Vdc
The optimum bias voltage for Class B has resulted from the manual determination of the input characteristics of the NPN and PNP Darlington at 3.1Vdc. Under no circumstances should it fall below 2.6Vdc, I read as a handwritten note in my own notes on this project. All my measurements confirm the original characteristics (see the corresponding ETI article) of this little "blameless power amp".
The sound is unspectacular, absolutely not demanding, even low-impedance speakers pose no challenge for this classic - it is without fault.
Greetings,
HBt.
That was my mistake with the PCB design that was lost in thought.
Of course you are absolutely right here, in a new version of the layout this fatal source of error should be corrected.
One is enough,
but I really wanted an absolutely guaranteed ability to handle all eventualities of various loads without strain - keyword 4-quadrant operation, or simply put, a low static output resistance ... and so on ..!
not needed!
regards,
HBt.
Cob of two paralleled BD139’s is probably enough to stabilize the amplifier. It has a miller cap there whether you want one or not. Usually one goes out of their way to minimize it, then add a linear cap in parallel.
There’s this poorly designed “PA amplifier” design floating around here that comes up in internet searches all the time. People periodically ask if it’s any good (and the answer is always the same) . It uses a TIP41 there (and a TIP42 above it). No miller cap needed there, it’s got plenty inside. But cr@p for gain. At least a BD139 would actually make it sound good - but wouldn’t work at all on the usual +/-50V.
There’s this poorly designed “PA amplifier” design floating around here that comes up in internet searches all the time. People periodically ask if it’s any good (and the answer is always the same) . It uses a TIP41 there (and a TIP42 above it). No miller cap needed there, it’s got plenty inside. But cr@p for gain. At least a BD139 would actually make it sound good - but wouldn’t work at all on the usual +/-50V.
I must write a little more
about this small hi-fi audio amplifier:
The abbreviation 20WPA stands for 20 Watt Power Amplifier, and not for a stage amplifier, which is a bit confusing and originally comes from the German magazine ELRAD. I came across this design proposal myself via the publisher's archive.
The original Canadian article describes the project as a “simple 60W low distortion amplifier”.
At the time (late 1970s), the catchphrase “TIM, TID and so on” was in vogue and on everyone's lips. This amplifier was intended to live up to this doctrine.
Take a very close look at the components around Q1 and Q3||Q4, i.e. the values 100Ohm, 13.7kOhm and the two 15Ohm resistors.
This is not a bad design, but a standard template. I use ON-Semi TIP 142 & TIP 147.
As I said before, it runs and runs, day in and day out, at my home.
#
OL
G >41dB
fh >302kHz
Any other voices?
about this small hi-fi audio amplifier:
The abbreviation 20WPA stands for 20 Watt Power Amplifier, and not for a stage amplifier, which is a bit confusing and originally comes from the German magazine ELRAD. I came across this design proposal myself via the publisher's archive.
The original Canadian article describes the project as a “simple 60W low distortion amplifier”.
At the time (late 1970s), the catchphrase “TIM, TID and so on” was in vogue and on everyone's lips. This amplifier was intended to live up to this doctrine.
Take a very close look at the components around Q1 and Q3||Q4, i.e. the values 100Ohm, 13.7kOhm and the two 15Ohm resistors.
This is not a bad design, but a standard template. I use ON-Semi TIP 142 & TIP 147.
As I said before, it runs and runs, day in and day out, at my home.
#
OL
G >41dB
fh >302kHz
Any other voices?
Last edited:
But this stage amplifier you are probably referring to here has nothing to do with “Trevor Marshall and Phil Wait”, who were responsible for the original “simple 60W low distortion amplifier”, or 60WPA for short.
?
Let's discuss the 20WPA at hand, openly and relentlessly.
HBt.
Relentlessly?
Relying on (voltage dependent) Cob intrinsically for overall loop stabilization is a recipe for distortion which rises with frequency, across the entire audio spectrum. Tends to produce lots of higher order products (distortion of distortion) as well - as the internal feedback capacitance is “non-linearizing” it more as frequency rises.
It might not be too objectionable with a high fT device like BD139, but stick in a TIP41 and it will drive you from the f*ing room. It was cited as an example of taking this too far.
Relying on (voltage dependent) Cob intrinsically for overall loop stabilization is a recipe for distortion which rises with frequency, across the entire audio spectrum. Tends to produce lots of higher order products (distortion of distortion) as well - as the internal feedback capacitance is “non-linearizing” it more as frequency rises.
It might not be too objectionable with a high fT device like BD139, but stick in a TIP41 and it will drive you from the f*ing room. It was cited as an example of taking this too far.
definitely and absolutely correct 👍
Now I understand your point.
Please let us relentlessly analyze the entire topology /design.
greetings,
HBt.
Douglas Self's "Blameless" has refined such amps. Important differences are:
1. LTP degeneration
2. Current mirror
3. Darlington VAS
4. Driver cross-coupling (not possible with Darlingtons)
Possible issues are:
1. K11 needs a capacitor like C1 or a base resistor to guarantee stability of the CCS.
2. R7 is probably not a good idea.
3. K5 and K6 provide a total of about 14.5mA to the LTP, probably 10x too much. And then K5 is not needed. R4 then needs to be changed but a current mirror is a better idea.
4. If you add gain with a current mirror and Darlington VAS then you will need a compensation capacitor where R7 is, and you might want to consider two pole compensation.
5. C12 and C14 are a bad idea, and the resistor networks set the gain at 9.3x0.9~= 8.3x, ~18dB total where 30dB gain would be more useful.
6. R8+C2 may be related to the other compensation choices, but I think it's established that they are a poor choice.
7. C6+C7 seem to be a Zobel network that avoids a separate ground wire to the supply. I would go with the wire vs an extra capacitor, or just one side, if an EF output needs a Zobel. In any case, you do need a series inductor+resistor to the load to avoid shorting the feedback.
8. As per above, Darlingtons are not a good choice for audio because they are slow to turn off and prone to shoot through current, because you have no access to the final bases for cross coupling.
9. Some may prefer to live dangerously without any output protection, but I think a current limit at minimum. VI could work if it muted the audio instead of shutting down the output. Typical VI limiting does have some serious stability problems.
10. A Baker clamp may be useful to avoid rail sticking.
11. C9 needs a voltage clamp to avoid latchup.
Nuff 4 now
1. LTP degeneration
2. Current mirror
3. Darlington VAS
4. Driver cross-coupling (not possible with Darlingtons)
Possible issues are:
1. K11 needs a capacitor like C1 or a base resistor to guarantee stability of the CCS.
2. R7 is probably not a good idea.
3. K5 and K6 provide a total of about 14.5mA to the LTP, probably 10x too much. And then K5 is not needed. R4 then needs to be changed but a current mirror is a better idea.
4. If you add gain with a current mirror and Darlington VAS then you will need a compensation capacitor where R7 is, and you might want to consider two pole compensation.
5. C12 and C14 are a bad idea, and the resistor networks set the gain at 9.3x0.9~= 8.3x, ~18dB total where 30dB gain would be more useful.
6. R8+C2 may be related to the other compensation choices, but I think it's established that they are a poor choice.
7. C6+C7 seem to be a Zobel network that avoids a separate ground wire to the supply. I would go with the wire vs an extra capacitor, or just one side, if an EF output needs a Zobel. In any case, you do need a series inductor+resistor to the load to avoid shorting the feedback.
8. As per above, Darlingtons are not a good choice for audio because they are slow to turn off and prone to shoot through current, because you have no access to the final bases for cross coupling.
9. Some may prefer to live dangerously without any output protection, but I think a current limit at minimum. VI could work if it muted the audio instead of shutting down the output. Typical VI limiting does have some serious stability problems.
10. A Baker clamp may be useful to avoid rail sticking.
11. C9 needs a voltage clamp to avoid latchup.
Nuff 4 now
Real "hot" (high Ic) VAS directly driving BJT (darlington) outputs. No NFB or miller ?
Circuit seems to be designed around specific devices , Why ?
My cheap "sonance 260" has far less semi's but does 20ppm 100W , again - why ?
"Relying on (voltage dependent) Cob intrinsically for overall loop stabilization is a recipe for distortion"
Perhaps some like it's "sound" ?
Ummm , one word - junk ? like some of the Alibaba kits ...
Wow , the Japanese were light years ahead around that time (Sansui , pioneer). Stable beautiful sounding amps ! Last 40 years between re-capping.
OS
The problem with this “junk” is the following:
when I came across this article in 2018, I was amazed or perhaps a little puzzled - after reading the article five or six times, I thought to myself “fire up the mill-drill-plotter and just build the circuit”.
No sooner said than done - since the end of 2020, two monos have been working perfectly -> THD with a flat curve of around 0.003%.
HBt.
PS
Every time I fire up these two boxes, I get big eyes - so what makes this ancient circuit proposal special, can it do magic?
From a purely intellectual point of view, I would give preference to an original Douglas Self “Blameless-Amplifier-Design".
But my 20WPA also seems to be a blameless thing, seems to be, build from the gut.
Why?
when I came across this article in 2018, I was amazed or perhaps a little puzzled - after reading the article five or six times, I thought to myself “fire up the mill-drill-plotter and just build the circuit”.
No sooner said than done - since the end of 2020, two monos have been working perfectly -> THD with a flat curve of around 0.003%.
absolutely!
HBt.
PS
Every time I fire up these two boxes, I get big eyes - so what makes this ancient circuit proposal special, can it do magic?
From a purely intellectual point of view, I would give preference to an original Douglas Self “Blameless-Amplifier-Design".
But my 20WPA also seems to be a blameless thing, seems to be, build from the gut.
Why?
Last edited:
30 ppm , hard to believe. 1K ? surely not 20k , prove it.
Similar simple 80's OEM amps typically average .05%- 20k .005 - 1k ,
with much better Sanken EF2's ,lead/ lag global NFB plus OP inductors .... Zobel.
No NFB ,No way , prove it....
OS
That's exactly the point, I can hardly believe it myself.
But as soon as I find time for this digression, I will carry out measurements again and document them here. At the moment, the only problem is the time window and other adversities of life.
I will deliver, don't panic.
I thought of this little distraction in connection with Nick S.'s ABBA thing. Something for our edification.
regards,
HBt.
Addendum
After all, somewhere in the background Trevor Marshall is standing up for this topology with his name - and Phil W.
HBt.
But that's not a qualified analysis of the present circuit, is it?
After all, somewhere in the background Trevor Marshall is standing up for this topology with his name - and Phil W.
HBt.
It's possible , it does have a beta enhanced VAS.
But , darlingtons tend to suck vs. a real modern EF3/2.
I believe in magic , like a audio modulated oscillator (hypex class D).
Any new thing you just need to "wrap your head around it"...
I wish I could model it , but I dont have a darlington model.
OS
But , darlingtons tend to suck vs. a real modern EF3/2.
I believe in magic , like a audio modulated oscillator (hypex class D).
Any new thing you just need to "wrap your head around it"...
I wish I could model it , but I dont have a darlington model.
OS
The models for the two ON-Semi TIP142 /147 Darlington can be found in the iNet.
I used two models back then, but I have to look for them at home first - then I can post them here.