mod to feedback?
Looking at Jarek's circuit in post 302 with the unusual feedback point from the Hawksford stage. Would it not be a better choice to connect R11 and R34 to the bases of Q10 and Q11? This seems to be a more neutral point as regards the internal correction signal of the Hawksford stage. The present setup seems to cause the input stage to take the Hawksford error feedback back out again. I haven't used the Hawksford circuit before, so don't take my word for it without trying the Sim on it.
I must say however that my interest is growing in the Hawksford corr. idea, seems straightforward and interesting having gain in the feedback path. I wonder how well it would work in a class AB push-pull tube circuit with a little adaptation for phase polarity (LTP or diff. amp stage for Q10 and Q11). Sure would be a new tube circuit.
Don
Looking at Jarek's circuit in post 302 with the unusual feedback point from the Hawksford stage. Would it not be a better choice to connect R11 and R34 to the bases of Q10 and Q11? This seems to be a more neutral point as regards the internal correction signal of the Hawksford stage. The present setup seems to cause the input stage to take the Hawksford error feedback back out again. I haven't used the Hawksford circuit before, so don't take my word for it without trying the Sim on it.
I must say however that my interest is growing in the Hawksford corr. idea, seems straightforward and interesting having gain in the feedback path. I wonder how well it would work in a class AB push-pull tube circuit with a little adaptation for phase polarity (LTP or diff. amp stage for Q10 and Q11). Sure would be a new tube circuit.
Don
I have been browsing quickly through this very interestion thread. I don't think, however, that anyone has mentioned Lars Clausens "The End". This kit has been very popular among diy'ers in scandinavia for many years. I have one of these myself and has always thought it looked like a very clever design.
Its latest incarnation looks like this: This one has emitter follower outputs.
http://www.lcaudio.dk/com/milldia.pdf
An older version is implemented with CFP's:
http://www.lcaudio.dk/temk2.jpg
Kind regards
Joachim
Its latest incarnation looks like this: This one has emitter follower outputs.
http://www.lcaudio.dk/com/milldia.pdf
An older version is implemented with CFP's:
http://www.lcaudio.dk/temk2.jpg
Kind regards
Joachim
Hi Jarek:
Maybe another more convenient feedback pickoff point for R19 would be to use the other end of R20? No need for extra resistors then. Again, back to the Sim.
I think I am going to build a P-P tube design now using the Hawksford corr. stage. Never seen anyone do that before. Could even be applied to Single Ended designs too, I think.
Hi Joachim:
Are any distortion specs published for the Lars Clausen designs? I would expect that the 4 sequential emitter followers in the output stage, without an error correction scheme like the Hawksford one, would add up to some odd order distortion. On the other hand, the EF distortion could be of the opposite sense to the input diff. pair odd order distortion, I think. Depends on the combined current gain curves of the EFs and operating point. If operating in the falling EF current gain region, then they would add distortions instead of cancellling. Perhaps at some output load impedance, and correctly chosen transistors, the two could be made to cancel, giving amazing results. But would need a well behaved speaker impedance characteristic to take advantage of this.
Don
Maybe another more convenient feedback pickoff point for R19 would be to use the other end of R20? No need for extra resistors then. Again, back to the Sim.
I think I am going to build a P-P tube design now using the Hawksford corr. stage. Never seen anyone do that before. Could even be applied to Single Ended designs too, I think.
Hi Joachim:
Are any distortion specs published for the Lars Clausen designs? I would expect that the 4 sequential emitter followers in the output stage, without an error correction scheme like the Hawksford one, would add up to some odd order distortion. On the other hand, the EF distortion could be of the opposite sense to the input diff. pair odd order distortion, I think. Depends on the combined current gain curves of the EFs and operating point. If operating in the falling EF current gain region, then they would add distortions instead of cancellling. Perhaps at some output load impedance, and correctly chosen transistors, the two could be made to cancel, giving amazing results. But would need a well behaved speaker impedance characteristic to take advantage of this.
Don
Looking at site below you can find some SPICE simulations of The End Millenium amp which comes from LC Audio.
http://geocities.com/researchtriangle/node/2356/
click Power amplifiers and then Amps without negative feedback (Another piece of power amps schematics)
http://geocities.com/researchtriangle/node/2356/
click Power amplifiers and then Amps without negative feedback (Another piece of power amps schematics)
Hi mandat;
Thanks for the link. Very interesting. I would be very surprised too if the advertised .0017% distortion figure could be met. However simulations are somewhat poor at predicting distortion accurately. Particularly in the present case, where transistor characteristics are important for any distortion cancellation without using feedback. The use of the non-linear shunt diode in the output is puzzling also, seems like one would typically like the opposite effect there.
The second circuit below it at that link has a similar shunt idea. But it doesn't seem to help much in the simulation. I do like the linear mirror-with-gain stage in the 2nd amplifier though. But I think this idea will not extend itself easily to work for the output stage of a non-feedback design well, since it will still suffer from current gain droop at high output levels or low speaker impedance. Perhaps one could substitute three terminal unity gain buffers (a la John Addis page 119 in "Analog Circuit Design" EDN/B-H 1991/4) for the output transistors in the mirrors to correct for this.
Don
Thanks for the link. Very interesting. I would be very surprised too if the advertised .0017% distortion figure could be met. However simulations are somewhat poor at predicting distortion accurately. Particularly in the present case, where transistor characteristics are important for any distortion cancellation without using feedback. The use of the non-linear shunt diode in the output is puzzling also, seems like one would typically like the opposite effect there.
The second circuit below it at that link has a similar shunt idea. But it doesn't seem to help much in the simulation. I do like the linear mirror-with-gain stage in the 2nd amplifier though. But I think this idea will not extend itself easily to work for the output stage of a non-feedback design well, since it will still suffer from current gain droop at high output levels or low speaker impedance. Perhaps one could substitute three terminal unity gain buffers (a la John Addis page 119 in "Analog Circuit Design" EDN/B-H 1991/4) for the output transistors in the mirrors to correct for this.
Don
Hawksford Tube design
1st a note on my last post:
My idea of using a 3 terminal unity gain buffer in the output of a current mirror is probably not working, the buffer is too linear to act as a mirror transistor. Better to just use the 3 terminal unity gain buffer idea for just replacing transistors in a normal (well... low feedback) design where high linearity transistors are required.
Here is my 1st shot at a Hawksford style error correcting Push-Pull tube output stage: (its fitted to a partial cathode feedback circlotron output stage, or "elliptron stage")
Arrgh! attachment didn't fit, trying again.
1st a note on my last post:
My idea of using a 3 terminal unity gain buffer in the output of a current mirror is probably not working, the buffer is too linear to act as a mirror transistor. Better to just use the 3 terminal unity gain buffer idea for just replacing transistors in a normal (well... low feedback) design where high linearity transistors are required.
Here is my 1st shot at a Hawksford style error correcting Push-Pull tube output stage: (its fitted to a partial cathode feedback circlotron output stage, or "elliptron stage")
Arrgh! attachment didn't fit, trying again.
>smoking-amp
In the mentioned simulation Russian DIYer has used BF471 and BF472 transistors instead of those marked with S1D and S2D in original The End Millenium amp. He has got rather great THD instead of claimed 0.0017 % in technical data.
I have found at Polish DIY forum information that
S2D is SMD transistor - SMBTA92 PNP 300V, 0.5A, 0.36W, case SOT23
and
S1D is SMD transistor - SMBTA42 NPN 300V, 0.5A, 0.36W, case SOT23
In the mentioned simulation Russian DIYer has used BF471 and BF472 transistors instead of those marked with S1D and S2D in original The End Millenium amp. He has got rather great THD instead of claimed 0.0017 % in technical data.
I have found at Polish DIY forum information that
S2D is SMD transistor - SMBTA92 PNP 300V, 0.5A, 0.36W, case SOT23
and
S1D is SMD transistor - SMBTA42 NPN 300V, 0.5A, 0.36W, case SOT23
mandat said:
The End Millenium will never measure 0.0017% THD past
anything but very low OP levels. The OP stage would have to
be biased at quite a few amps to get figures like that at
elevated power levels. It's plain physics. However, I'm not
sayng it's not a great amp, just the specs are questionable.
Terry
Hi Steven, Jarek,
It was an interesting thread for me in general and useful in particular, because I have NFB Tandberg 4036 amplifier. It has very similar output stage, more close to Steven’s schematic from post 33. DC offset servo implemented with termo-feedback loop.
One channel was exploded some time ago. I repaired it using different transistors for replacement because original parts are already not available. It is ok now, but I am not sure the error correction stage is in optimal state and would like to tune it up with the pot.
I did not catch the idea of you setup for adjusting the error correction stage with the pot R20 (on Jarek schematic). Could you please give more details or some kind of schematic of this setup. Is it true that square generator must be connected to the output of the amplifier?
It will be fine if you give also more comments on your oscilloscope scans.
Thanks Sergey
It was an interesting thread for me in general and useful in particular, because I have NFB Tandberg 4036 amplifier. It has very similar output stage, more close to Steven’s schematic from post 33. DC offset servo implemented with termo-feedback loop.
One channel was exploded some time ago. I repaired it using different transistors for replacement because original parts are already not available. It is ok now, but I am not sure the error correction stage is in optimal state and would like to tune it up with the pot.
Steven said:
I did not catch the idea of you setup for adjusting the error correction stage with the pot R20 (on Jarek schematic). Could you please give more details or some kind of schematic of this setup. Is it true that square generator must be connected to the output of the amplifier?
It will be fine if you give also more comments on your oscilloscope scans.
Thanks Sergey
ASA,
My remark to Jarek about testing the error correction with one end of the load resistror connected to the amplifier under test and the other end of the load resistor to a square wave oscillator (instead of to ground) was intended for the simulation. Then you have square wave oscillators that are 'infinitely' powerful. In this way you try to push current into the amplifier or pull it out of the amplifier. If the output impedance is zero (perfect error cancelling), you should not see voltage jumps on the output of the amplifier. The amplifier itself need not to be driven at the input.
In practice it is easier to just touch the output of the amplifier with the load resistor repeatedly. The other end of the load resistor is grounded as normal. The amplifier can be driven with a sine or square wave, also as normal. Big amplitude is not required, a couple of volts will do. The amplitude should be stable. With a scope on the output of the amp you will see that the output voltage drops when the output is loaded if the error correction is too less, it will rise when the error correction is too much and it will remain the same if the error correction is OK. It is good to do this for different frequencies of the input sine, to see if the adjustment of the error correction is OK for all frequencies in the audio band. This is for sines, using a square wave automatically a lot of discrete frequencies are tested (the harmonics).
Steven
My remark to Jarek about testing the error correction with one end of the load resistror connected to the amplifier under test and the other end of the load resistor to a square wave oscillator (instead of to ground) was intended for the simulation. Then you have square wave oscillators that are 'infinitely' powerful. In this way you try to push current into the amplifier or pull it out of the amplifier. If the output impedance is zero (perfect error cancelling), you should not see voltage jumps on the output of the amplifier. The amplifier itself need not to be driven at the input.
In practice it is easier to just touch the output of the amplifier with the load resistor repeatedly. The other end of the load resistor is grounded as normal. The amplifier can be driven with a sine or square wave, also as normal. Big amplitude is not required, a couple of volts will do. The amplitude should be stable. With a scope on the output of the amp you will see that the output voltage drops when the output is loaded if the error correction is too less, it will rise when the error correction is too much and it will remain the same if the error correction is OK. It is good to do this for different frequencies of the input sine, to see if the adjustment of the error correction is OK for all frequencies in the audio band. This is for sines, using a square wave automatically a lot of discrete frequencies are tested (the harmonics).
Steven
ASA, I think the amplifier should be able to deal with connecting and disconnecting a load when there is a few volt output signal, otherwise it would not be a very stable amplifier.
Jarek, I do not understand what you mean with a hundred Ohm resistor at the output. You mean as a load or in series with the load? The load resistor should be low value, like 4..8 Ohm, because then it is more clear on the scope whether the error correction stage is properly adjusted. If the output impedance of the amp is e.g. 0.1 Ohm, then loading with 100 Ohm will not make much difference at the output. You could even try a 1 Ohm load, if you keep the ouput voltage small, like 1..2V. But be careful, it is easy to heat up the amp too much if measurements take a long time, because these are steady signals and no music. Especially with a 1 Ohm load you can get already big output currents for low output voltages and then the dissipation in the output transistors is at maximum (still high Vce while having big Ic).
BTW, ASA, do you have a schematic diagram of the Tandberg amplifier? I would be interested.
Jarek, I do not understand what you mean with a hundred Ohm resistor at the output. You mean as a load or in series with the load? The load resistor should be low value, like 4..8 Ohm, because then it is more clear on the scope whether the error correction stage is properly adjusted. If the output impedance of the amp is e.g. 0.1 Ohm, then loading with 100 Ohm will not make much difference at the output. You could even try a 1 Ohm load, if you keep the ouput voltage small, like 1..2V. But be careful, it is easy to heat up the amp too much if measurements take a long time, because these are steady signals and no music. Especially with a 1 Ohm load you can get already big output currents for low output voltages and then the dissipation in the output transistors is at maximum (still high Vce while having big Ic).
BTW, ASA, do you have a schematic diagram of the Tandberg amplifier? I would be interested.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.