Would it be possible to modify the bias servo circuit,—instead of nudging the front-end—to adjust the actual current supplied by the positive and negative current source circuits?
I built up a nice tube phono preamp from a circuit on this forum by Sy Yaniger. The Equal Opportunity preamp. Cam out nice. But jeez lot os elements to that. Two signal boards, and two separate power supply boards. Two chassis..
On the 565 that is EXACTLY what I did. By adjusting R144 and R145 (in the current source), you can tweak the balance between the positive and negative Diff pairs on the front end, and adjust the overall offset very accurately
Yes, but what if the servo was disconnected from the front end, and was used to tweak the bias currents automatically? You could still put in your pots to tweak it to the center, and the servo would keep the amp at 0V.
It would be a lot of re-design though.
It would be a lot of re-design though.
Think about the Beta Matcher. it is designed to pull the same current through both transistors. You then measure the Vc voltage offset, and that voltage depends on the beta of the transistors. If the Vc values match, then that means the betas also match. The same idea applies to the PNP/NPN pairs. If you have the same tail currents, but the betas are different, then the Vc values will be different, and now you have offset to correct.
If you instead, adjust the tail currents to get the same Vc values (only opposite polarity), then you have essentially compensated for the effect of beta differences on the bias. In the grand scheme of things, the overall beta differences will affect bias much more than they will affect small signal gain.
The only better way to deal with this would be to painstakingly match both diff pairs, and PNP/NPN pairs.. And that would mean matching the dog NPNs with the champion PNPs...Not sure if that's a great idea.
Ahh, I see... a dynamically controlled differential current source..
Hmmm. I'll think about that. That would certainly be nicer than introducing voltage offset at the input.
Might be cool to do some sort of differential current source where the "virtual ground" was set by the servo output, so instead of offsetting the input signal, you woudl adjust the level of the virtual ground between the positive and negative current sources, and thereby balance the bias in the diff pairs.
Much more elegant!
@ Phloodpants:
Sir, am I not happy that you are up and about and here! I had thought you shut shop as a result of the health issues. Thank God you are fine and happy... yeah, we need you around here!
(BTW, I had some RSI issues about ten years back and they said surgery was the only option. But then a fortunate turn of events saw me taking some Homoeopathic medicines and I was fine in about three weeks--no recurrence too yet! My surgeon, my old college mate, was flabbergasted!)
@ Cogeniac:
Sure, will try the R620 pot idea with the worse amplifier and see how it behaves
Warm regards to all.
Sir, am I not happy that you are up and about and here! I had thought you shut shop as a result of the health issues. Thank God you are fine and happy... yeah, we need you around here!
(BTW, I had some RSI issues about ten years back and they said surgery was the only option. But then a fortunate turn of events saw me taking some Homoeopathic medicines and I was fine in about three weeks--no recurrence too yet! My surgeon, my old college mate, was flabbergasted!)
@ Cogeniac:
Sure, will try the R620 pot idea with the worse amplifier and see how it behaves
Warm regards to all.
BTW spent a lovely weekend at Spring Green once with my wife, before we were married (She lived Chicago then). Played golf, saw a cool play at the American Players Theatre, and toured Taliesin. Lovely area around Madison.
OK, so I did a little research. Looks like we can make a voltage controllable current mirror and control that with the servo output. Obviously this will need to be more thoroughly analyzed to get the component values, but it seems like it should work.
Conceptually, this is a pair of current mirrors, one for the negative side diff pair, and the other for the positive side diff pair. This part is not substantively different from the original circuit, except the current mirror "should" be a little moire stable over temperature than the LED stabilized circuit in the original.
The negative pair current mirror is formed by Q1 and Q3. Q1 fixes the base current in Q3 based on the resistances R1 and R3. These will drift together thereby making the current in Q3 stable over temperature. (this is a typical current mirror circuit).
The servo op amp drives a small voltage across the base of Q5, and this causes some current to flow through Q5. This added current causes a drop across R5 that changes the base current of Q3 slightly. As the voltage goes up, the current in Q5 goes down.
The positive pair side works the same, only with the opposite polarity.
Using this approach, the currents in the two diff pairs will change in the same way but in opposite directions as the servo voltage rises and falls.
I would probably use high quality trimmer pots so R1 and R2 and put a jumper between the output of the Op Amp and the RR11/R12 resistors. The way the offset inherent in the circuit because of the mismatch between the NPN darlington betas and the PNP darlington betas can be adjusts out without tearing up the circuit, and then the servo is there to make up for minor temp and aging changes over the life of the amp.
I thin I may STILL also use the servo output to shut down the amplifier power if the servo output goes too high in one direction or the other.
Conceptually, this is a pair of current mirrors, one for the negative side diff pair, and the other for the positive side diff pair. This part is not substantively different from the original circuit, except the current mirror "should" be a little moire stable over temperature than the LED stabilized circuit in the original.
The negative pair current mirror is formed by Q1 and Q3. Q1 fixes the base current in Q3 based on the resistances R1 and R3. These will drift together thereby making the current in Q3 stable over temperature. (this is a typical current mirror circuit).
The servo op amp drives a small voltage across the base of Q5, and this causes some current to flow through Q5. This added current causes a drop across R5 that changes the base current of Q3 slightly. As the voltage goes up, the current in Q5 goes down.
The positive pair side works the same, only with the opposite polarity.
Using this approach, the currents in the two diff pairs will change in the same way but in opposite directions as the servo voltage rises and falls.
I would probably use high quality trimmer pots so R1 and R2 and put a jumper between the output of the Op Amp and the RR11/R12 resistors. The way the offset inherent in the circuit because of the mismatch between the NPN darlington betas and the PNP darlington betas can be adjusts out without tearing up the circuit, and then the servo is there to make up for minor temp and aging changes over the life of the amp.
I thin I may STILL also use the servo output to shut down the amplifier power if the servo output goes too high in one direction or the other.
Been trying to understand the soft start circuit so I can determine how best to shut down the amplifier if the output offset rises too high.
I see that one side of the AC input is directly connected to the 0V transformer windings.
The other AC wire follows a very circuitous path: into the board at J505, out of the board at J506, through the power switch, and back to the board at J507.
It then goes to the relay, and also to a 4.7 ohm resistor (R506). I assume that's the big ceramic resistor on the board. The other side of the relay goes to J508. In the schematic, J508 is not connected to anything, but it turns out it is a pair of tab connectors on the relay. The odd looking "component" that is connected to point 6 in the schematic is actually a jumper that goes from J508 to point 6 on the board, and also to the transformer 120 V primary.
I see that the 100 volt winding of the lower primary feeds into the two diodes D501 and 502 forming a half wave rectifier. That then gets filtered by C501 to presumably from a positive DC voltage at the relay. Which presumably turns on power to J508 (wherever that goes).
I guess the idea is that the power comes up in the transformer through the big R506 resistor and the J506/point 6 jumper, and that slows down the inrush current to the big filter caps. The rectifier filter formed by R502 and C101 presumably charges somewhat slowly, so there is some delay until the relay closes. Once the relay closes, R506 is shorted by the relay, and the transformer gets full AC power. via J508.
The other rectifier appears to run a delay circuit that ultimately triggers and onto coupler that causes the current source for the input stage to power up. So this circuit mostly just delays the power on for the input stage, which doesn't really help much for shutting the amp down if a problem arises.
Seems like a best arrangement would be to re-make the soft start board and use a separate relay that just kills AC power entirely based on the signal from the servo.
I see that one side of the AC input is directly connected to the 0V transformer windings.
The other AC wire follows a very circuitous path: into the board at J505, out of the board at J506, through the power switch, and back to the board at J507.
It then goes to the relay, and also to a 4.7 ohm resistor (R506). I assume that's the big ceramic resistor on the board. The other side of the relay goes to J508. In the schematic, J508 is not connected to anything, but it turns out it is a pair of tab connectors on the relay. The odd looking "component" that is connected to point 6 in the schematic is actually a jumper that goes from J508 to point 6 on the board, and also to the transformer 120 V primary.
I see that the 100 volt winding of the lower primary feeds into the two diodes D501 and 502 forming a half wave rectifier. That then gets filtered by C501 to presumably from a positive DC voltage at the relay. Which presumably turns on power to J508 (wherever that goes).
I guess the idea is that the power comes up in the transformer through the big R506 resistor and the J506/point 6 jumper, and that slows down the inrush current to the big filter caps. The rectifier filter formed by R502 and C101 presumably charges somewhat slowly, so there is some delay until the relay closes. Once the relay closes, R506 is shorted by the relay, and the transformer gets full AC power. via J508.
The other rectifier appears to run a delay circuit that ultimately triggers and onto coupler that causes the current source for the input stage to power up. So this circuit mostly just delays the power on for the input stage, which doesn't really help much for shutting the amp down if a problem arises.
Seems like a best arrangement would be to re-make the soft start board and use a separate relay that just kills AC power entirely based on the signal from the servo.
I don't think you want to do much with a servo, and that is why I think the original design is so bad.
The LPT should be balanced without external help. If you end up with less than 50 mV DC offset, all is good. Most with matched transistors will sit at less than 10 mV. I honestly do not see the need for a DC servo when offsets are that low.
-Chris
The LPT should be balanced without external help. If you end up with less than 50 mV DC offset, all is good. Most with matched transistors will sit at less than 10 mV. I honestly do not see the need for a DC servo when offsets are that low.
-Chris
I agree, the idea is to get the circuit to balance without the servo, and then use the servo to keep it balanced as things age.
BTW, I am not suggesting anything for the 545 (other than using a pot for R620 to adjust the offset. I agree that that circuit is inherently offset, and it is hugely dependent on the bias points throughout the circuit. It looks almost like a classic push-pull output stage that is driven by a single ended class-A amplifier. In the olden days we would AC couple most of the class A gain stages, split that signal into two paths, invert one and then AC couple them to the Class B output stages. So no inherent offset because the output stages were basically biased at zero quiescent current. In the 545 it seem like someone was trying to avoid any AC coupling, and so they worked out all the bias points to make that work. The problem is that ANY error anywhere in the circuit can then upset the whole thing, and errors accumulate as you go from the input to the output stage.
The discussion above, about the servo and the current sources is for the 565, which, being double ended seems to have a much more stable bias setup, unless something causes a DC offset up at the front end. Seems to me the biggest thing that causes that (other than something failing, like a leaking cap) is that the PNP and NPN Darlingtons have different betas, so even if the pairs are matched, the collector bias points will be different.
I have had good results adjusting this offset by tweaking R144 and R145, which adjusts the bias current in the two Darlington pairs.
Chris Hoppe (Phloodpants) suggested that instead of using the servo to add DC feedback at the input of the amplifier, we could use it to adjust the relative bias levels on the positive and negative diff pairs, via the current sources.
BTW, I am not suggesting anything for the 545 (other than using a pot for R620 to adjust the offset. I agree that that circuit is inherently offset, and it is hugely dependent on the bias points throughout the circuit. It looks almost like a classic push-pull output stage that is driven by a single ended class-A amplifier. In the olden days we would AC couple most of the class A gain stages, split that signal into two paths, invert one and then AC couple them to the Class B output stages. So no inherent offset because the output stages were basically biased at zero quiescent current. In the 545 it seem like someone was trying to avoid any AC coupling, and so they worked out all the bias points to make that work. The problem is that ANY error anywhere in the circuit can then upset the whole thing, and errors accumulate as you go from the input to the output stage.
The discussion above, about the servo and the current sources is for the 565, which, being double ended seems to have a much more stable bias setup, unless something causes a DC offset up at the front end. Seems to me the biggest thing that causes that (other than something failing, like a leaking cap) is that the PNP and NPN Darlingtons have different betas, so even if the pairs are matched, the collector bias points will be different.
I have had good results adjusting this offset by tweaking R144 and R145, which adjusts the bias current in the two Darlington pairs.
Chris Hoppe (Phloodpants) suggested that instead of using the servo to add DC feedback at the input of the amplifier, we could use it to adjust the relative bias levels on the positive and negative diff pairs, via the current sources.
Last edited:
Chris' idea would reduce the audible impact of the op amp DC servo.
Personally, I strongly believe there is no use for a DC servo if the LPT is matched to begin with. The servo only existed to allow random transistors to be used in manufacture while eliminating an adjustment point. That's all. It was a 100% cost control step.
I watched designs change while doing these products under warranty. There was a strong movement to eliminate the input from technicians to help control warranty program costs. We would list correcting bias and DC offsets in any amplifier we looked at simply because it was the responsible thing to do. Bean counters interpreted that as an added cost when it really wasn't. The reason for the service encounter was sometimes misadjustment, but most often an external issue or baseless complaint. We just avoided future trouble by making certain equipment was properly adjusted. They did the same thing for CD players. Although a fine adjustment often improved a CD player's performance considerably. Often bias currents were way out causing distortion or excessive heat in amplifiers. Offsets could cause a "pop" in preamps and amplifiers.
-Chris
Personally, I strongly believe there is no use for a DC servo if the LPT is matched to begin with. The servo only existed to allow random transistors to be used in manufacture while eliminating an adjustment point. That's all. It was a 100% cost control step.
I watched designs change while doing these products under warranty. There was a strong movement to eliminate the input from technicians to help control warranty program costs. We would list correcting bias and DC offsets in any amplifier we looked at simply because it was the responsible thing to do. Bean counters interpreted that as an added cost when it really wasn't. The reason for the service encounter was sometimes misadjustment, but most often an external issue or baseless complaint. We just avoided future trouble by making certain equipment was properly adjusted. They did the same thing for CD players. Although a fine adjustment often improved a CD player's performance considerably. Often bias currents were way out causing distortion or excessive heat in amplifiers. Offsets could cause a "pop" in preamps and amplifiers.
-Chris
Yeah, I think Anatech is right here. If you null out the offset by adjusting the current sources, the output of the amp will still have some offset, and that will drift around a little, but if it's less than 10 or 20mV, no worries at all. Ditch the servo entirely!
That's an interesting historical perspective....
I definitely agree that if the input sections are balanced properly, offset should not be an issue. There are two concerns, however.
1) Things age, and that will cause the circuit to become unbalanced over time. And the servo can help compensate for that.
2) If there is a failure, the servo can't help, but some sort of servo driven power supply shutdown can prevent applying 85 volts to your speaker terminals.
ANd, I think you are right, the current source based servo approach seems to be less in-your-ears approach. It basically gets to the real cause of the problem instead of applying a band-aid over it.
I definitely agree that if the input sections are balanced properly, offset should not be an issue. There are two concerns, however.
1) Things age, and that will cause the circuit to become unbalanced over time. And the servo can help compensate for that.
2) If there is a failure, the servo can't help, but some sort of servo driven power supply shutdown can prevent applying 85 volts to your speaker terminals.
ANd, I think you are right, the current source based servo approach seems to be less in-your-ears approach. It basically gets to the real cause of the problem instead of applying a band-aid over it.