jkeny said:
Steve Dunlap said:
Steve, does the DC Servo, with LDRs 36/37, do as good a job as manually trimming R32/33? ...
what effect does it have re. matching of the output BJTs? ...
what is the maximum DC offset that can result using the DC Servo? ...
what is its effect re. THD? ...
thanks
🙂
KLe said:
If you use a good multi turn (25 turn) you can get results that are hard to beat. If the circuit drifts with temperature, your offset will drift also, but it should stay in an acceptable range. The servo will track the thermal changes and keep the offset low.
The servo has no effect on the matching of the output transistors. It does match the total current gain between the two halves of the output stage. This results in a more balanced load seen by the VGA.
Using the typical large signal gain of the TLO72 (200V/mV) the output from the op amp would drop below the max forward drop of the LED (2.5V) when the VGA offset reached 0.0125mV. With a worst case gain of 25V/mV that would increase to 0.1mV. Since the output stage servo operates with 0.6V, the offset will be about one fourth of these numbers.
I measure better THD when using either the servo or the trim pots verses not using them. The servo seems to have the edge, but at these levels I'm not sure it can be heard. Below 0.005% THD at 400W output into 8 ohms and 20KhZ is pretty good in my opinion. That goes down at lower powers or frequencies. I do not have the latest distortion meter and the lower limit I can test is 0.001% THD.
AndrewT said:
Sorry about that oversight. The pins are labeled on the schematic in my program, but I have to tell it to print them. I have attached the schematic with the inputs labeled on the op amps.
I see both opamps are operating as non-inverting Servos.
Would either or both perform better or inject less audio signal back into the low level stages if a single pole filter were added between the non-inverting input pin and the preceding resistor?
Would either or both perform better or inject less audio signal back into the low level stages if a single pole filter were added between the non-inverting input pin and the preceding resistor?
AndrewT said:
C31 and C32 do that.
No audio signal is injected back into the low or high level stages. That is why this method does not have any negative effect on the sound. With a conventional servo you do have the effect you describe. Since a conventional servo does feed back some frequencies, it has an effect on the low frequency bandwidth. An amp using a conventional servo will not pass a very low frequency square wave properly. My amp using my servo will.
I might be very wrong, but as I see it by modulating the VAS CCS current you are in effecting modulating the VAS and the output from it.
Similarly I think modulating the resistors between Q7 & 8 and Q10 & 11 will modulate the output signal.
The two caps give the integrating effect.
But you are feeding a full range audio signal into both opamps. If you pre-filter the signal to attenuate most of the audio content then the opamp has less signal to work with and should perform better. The pre-filter also reduces the level of audio signal coming out from each opamp.
Am I wrong? Please explain why.
Similarly I think modulating the resistors between Q7 & 8 and Q10 & 11 will modulate the output signal.
The two caps give the integrating effect.
But you are feeding a full range audio signal into both opamps. If you pre-filter the signal to attenuate most of the audio content then the opamp has less signal to work with and should perform better. The pre-filter also reduces the level of audio signal coming out from each opamp.
Am I wrong? Please explain why.
Some errors spotted....
Hi,
I've started the assembly of the 100W version and have spotted some mistakes / differences between schematics and actual board.
-main reservoir caps are numbered differently C13 and C22 are for the + rail, C14 and C23 are for the - rail
-the current souces of the output stage (D9-D10-Q12-R17 and D7-D8-Q9-R16 are connected to the regulated V+/- and not to the output stage supply
-R30 trough R34 have no provision on board (there is a direct connection instead)
-Q17 and q18 are labeled 19 and 20 on the board
More to come (hopefully not many ;-))
Ciao
Andrea
Hi,
I've started the assembly of the 100W version and have spotted some mistakes / differences between schematics and actual board.
-main reservoir caps are numbered differently C13 and C22 are for the + rail, C14 and C23 are for the - rail
-the current souces of the output stage (D9-D10-Q12-R17 and D7-D8-Q9-R16 are connected to the regulated V+/- and not to the output stage supply
-R30 trough R34 have no provision on board (there is a direct connection instead)
-Q17 and q18 are labeled 19 and 20 on the board
More to come (hopefully not many ;-))
Ciao
Andrea
The servo creates some NFB, reason for the better THD figures.
Steve I would like to know if there is any reason that you employ darlington type output, there are other topologies that are superior, well for me anyway.
I had a little laugh when destroyer x said this amplifier sounds good, he has in the past claimed that anything with a darlington or a current source ltp doesnt sound good.
Steve I would like to know if there is any reason that you employ darlington type output, there are other topologies that are superior, well for me anyway.
I had a little laugh when destroyer x said this amplifier sounds good, he has in the past claimed that anything with a darlington or a current source ltp doesnt sound good.
AndrewT said:
I will start by saying that I left off a resistor from each op amp input to ground. This makes a voltage divider that prevents the inputs from going over their limit.
If you don't mind slowing the response of the servo down to eliminate any possibility of modulating the resistors, then you can add the suggested filter. If you add a 1uF cap from each input to ground and change C31 and C32 to 1uF you have removed any AC portion of the signal above 1Hz. This is a 1 second wave form. The response time of the LDR is 5mS or about 200 Hz. The AC portion with the 0.1uF cap is 10Hz or 0.1 second. If you really want to push it, you can go to a 100uF or larger bipolar electrolytic. Since there will be almost DC on the cap, leakage should not be a problem. That will push your time constant down to 0.1Hz or 10 seconds.
The schematic with the 2 resistors added is attached. I also changed the caps to 1uF and added the suggested caps.
homemodder said:
The servo does not create feedback. It adjust the value of the resistor in the LDR which sets the DC offset to a minimum. This sets the amp up to operate in a more balanced operation current wise. That reduces the distortion. The same thing happens with the trim pots.
Darlingtons work. They are easy and they are stable. What topology do you prefer?
Hi Steve,
There was a discussion about servos creating some feedback, Ill look to see which one it is, anyway in the end it was proven to be the case, I was surprised too.
I havent built anything like this type of outputstage for large power outputstages. I have had many good results with cfp ef. Mine is a little more complex as I apply some techniques to it to reduce cross over distortion and high order harmonic cancellation, in reality it also doesnt allow switch off of the devices, there are so many ways to achieve this. What I like about this diamond buffer style is that it eliminates any loading of the vas which is very good indeed. it looks to me the darlington could easily be replaced with a cfp ef. Have you tried any other combinations and what results did you get ??
There was a discussion about servos creating some feedback, Ill look to see which one it is, anyway in the end it was proven to be the case, I was surprised too.
I havent built anything like this type of outputstage for large power outputstages. I have had many good results with cfp ef. Mine is a little more complex as I apply some techniques to it to reduce cross over distortion and high order harmonic cancellation, in reality it also doesnt allow switch off of the devices, there are so many ways to achieve this. What I like about this diamond buffer style is that it eliminates any loading of the vas which is very good indeed. it looks to me the darlington could easily be replaced with a cfp ef. Have you tried any other combinations and what results did you get ??
Re: Some errors spotted....
Starting with R30 through R34. This was implemented after these boards were made. This was discussed some time earlier in the thread. If you choose to use the resistor/trim pot adjustment you will need the cut the trace and insert them, probably on the bottom of the board. It would be rather crowded on the top side. If you can leave access to the bottom of the board it will make adjustment easier.
If you put the caps in the way they are labeled for polarity on the board they will be in correctly. I will change the numbering on the schematic.
I will also change the connection for the current sources on the schematic. With that extra current on the regulator transistors a small clip on or bolt on heat sink might be a good idea. These transistors will be dissipating about 0.75W each. They are rated at 2W and I haven't had failures not using heat sinks on them.
I will also change Q17 and Q18 to Q19 and Q20.
Please continue to post these mistakes here so everyone can benefit from your troubleshooting and my answers.
Andypairo said:
Starting with R30 through R34. This was implemented after these boards were made. This was discussed some time earlier in the thread. If you choose to use the resistor/trim pot adjustment you will need the cut the trace and insert them, probably on the bottom of the board. It would be rather crowded on the top side. If you can leave access to the bottom of the board it will make adjustment easier.
If you put the caps in the way they are labeled for polarity on the board they will be in correctly. I will change the numbering on the schematic.
I will also change the connection for the current sources on the schematic. With that extra current on the regulator transistors a small clip on or bolt on heat sink might be a good idea. These transistors will be dissipating about 0.75W each. They are rated at 2W and I haven't had failures not using heat sinks on them.
I will also change Q17 and Q18 to Q19 and Q20.
Please continue to post these mistakes here so everyone can benefit from your troubleshooting and my answers.
homemodder said:
Conventional servos do create feedback. That is why I do mine differently.
I haven't seriously considered other types of output stages. I get very low distortion and very wide bandwidth with what I use. And all without the output stage in the feedback loop. I think it sounds pretty good also.
Just musing to myself - would this kind of servo work with optocouplers instead of the LDR/LED arrangement, or do opto's have lousy linearity for this sort of thing?
Hi Steve,
thanks for taking the time to show this later version.
Which performs better? the earlier one or this later one?
Why did you change the time constants?
How well does the later version pass a square wave?
These DC servos create positive feedback, they do not help cancel distortion. That is why I believe that the pre-filter is necessary to attenuate as much of the audio signal as possible (simple single pole) before entering the opamps.
If we accept that modulating the LDRs in this way does modulate the output signal, then that surely is the same thing as feedback.
When you say the response time of the LDRs is ~ 5ms, is that the same thing as saying they are a low pass filter on the output of the opamp? Is it equivalent to a F-3dB=200Hz LP filter?
thanks for taking the time to show this later version.
Which performs better? the earlier one or this later one?
Why did you change the time constants?
How well does the later version pass a square wave?
These DC servos create positive feedback, they do not help cancel distortion. That is why I believe that the pre-filter is necessary to attenuate as much of the audio signal as possible (simple single pole) before entering the opamps.
If we accept that modulating the LDRs in this way does modulate the output signal, then that surely is the same thing as feedback.
When you say the response time of the LDRs is ~ 5ms, is that the same thing as saying they are a low pass filter on the output of the opamp? Is it equivalent to a F-3dB=200Hz LP filter?
jaycee said:
I tried optocouplers first and they do work. I abandoned them because I wanted as low a voltage drop as possible at this location. I can have as low as .15V drop with the LDR. I never got that low with the optocouplers. I never built a unit using the optocouplers, just a single channel prototype to measure voltages.
AndrewT said:
Hi AndrewT,
I did not claim that distortion was canceled, only reduced. I felt this was caused by the balance achieved just as with the trim pots.
Either version passes a square wave. I changed the time constant and added the input filter because of your comments.
I spent a fair amount of time thinking about what you said earlier, and you are correct on all points. The servo should be implemented as an inverting stage for best performance. The DC operating points were stabilized with the earlier version. The LDR may have been modulated at very low frequencies, but the resistor parallel with it insures that it will pass a square wave even if the LDR turns off.
The LDR does act like a LP filter and that explains why I didn't catch the modulation you pointed out in my measurements. The high frequency distortion measurements showed improvement. I never tested the low frequency distortion because it had always been very good. At high frequencies, the LDR was not responding and not modulating the signal.
No production units were ever built with the servo. I was still testing the original amp when my health went really fast. When I could no longer work, I withdrew into myself and gave up audio design - and pretty much everything else. It is nice to be back after 13 years. I have no doubt that AndrewT and others here will point out any other blunders I make.
I will rework the servo and post the changes.
- Status
- Not open for further replies.