Ylli,
I understand what you are saying. It's true - begfore the trace failure, I had replaced many of the transistors, and indeed, they test differently from the originals - particularly the first batch of KSA992, which I removed and replaced with the originals.
After the trace failure, which was on Q105, I had replaced Q105, Q107, Q155, Q157 with KSA1381, and Q108, and Q158 with KSC3503 on recommendation from a technician off-list. I repaired the trace with a short length of wire, and used conformal coating over it. Quadruple-checked for cont., shorts, etc.
Rail voltages are:
L+83.35
L-83.39
R+83.30
R-83.29
Thank you - I appreciate your input and comments. I'm learning from them all from each of you, which is more important to me than this one project!
I understand what you are saying. It's true - begfore the trace failure, I had replaced many of the transistors, and indeed, they test differently from the originals - particularly the first batch of KSA992, which I removed and replaced with the originals.
After the trace failure, which was on Q105, I had replaced Q105, Q107, Q155, Q157 with KSA1381, and Q108, and Q158 with KSC3503 on recommendation from a technician off-list. I repaired the trace with a short length of wire, and used conformal coating over it. Quadruple-checked for cont., shorts, etc.
Rail voltages are:
L+83.35
L-83.39
R+83.30
R-83.29
Thank you - I appreciate your input and comments. I'm learning from them all from each of you, which is more important to me than this one project!
Last edited:
Hi Mooly,
-Chris
Actually, it would be interesting to monitor the DC servo with matched and unmatched pairs. Personally, I think the design falls down due to the deliberate unmatching of the diff pair which is crucial for low distortion. In this design the servo is forced to function in one quadrant. It would have been better if the diff pair was balanced and the servo didn't have to do much in the way of correction. I wonder what they were thinking while designing this?
-Chris
Hard to say, perhaps a bit of subjectivism crept into the design and was found to sound better unbalanced. The greater the difference in Ic between the transistors in the pair and the more current the servo has to 'pull or push' as a correction into the base of the transistor.
A diff pair like that run with as much as 10:1 current inbalance (if that's what the designer wants) and the servo will easily pull the offset back to zero.
The servo IC could well be something like a TL071 which looks as though it would be perfect for the job... and in fact looking at the circuit shows it runs with a single rail supply. Could cost saving be one factor?
Design the amp deliberately with an offset that always falls in one direction and economise with a single rail servo. Now there's a thought.
A diff pair like that run with as much as 10:1 current inbalance (if that's what the designer wants) and the servo will easily pull the offset back to zero.
The servo IC could well be something like a TL071 which looks as though it would be perfect for the job... and in fact looking at the circuit shows it runs with a single rail supply. Could cost saving be one factor?
Design the amp deliberately with an offset that always falls in one direction and economise with a single rail servo. Now there's a thought.
Thank you for adding this discussion.
The installation of matched differential pairs was recommended to me by someone who routinely does this without it resulting in this problem, so I wonder if this is a confluence of too many component changes. I wonder how anyone could build a board with all new components without this being an issue?
Also, interesting comment about TL071; I had previously read comments suggesting AD820 was a close match. Nevertheless, I've reinstalled the Adcom parts (glad I soldered in sockets).
I'll be heating up the iron today or tomorrow to try to undo some of the changes and see if I can get back to correct operation.
Cheers
The installation of matched differential pairs was recommended to me by someone who routinely does this without it resulting in this problem, so I wonder if this is a confluence of too many component changes. I wonder how anyone could build a board with all new components without this being an issue?
Also, interesting comment about TL071; I had previously read comments suggesting AD820 was a close match. Nevertheless, I've reinstalled the Adcom parts (glad I soldered in sockets).
I'll be heating up the iron today or tomorrow to try to undo some of the changes and see if I can get back to correct operation.
Cheers
Last edited:
The diff pair shouldn't have any influence on your particular issue, matched or otherwise. I think your problem comes from having replaced the constant current sink Q103 and the LED driver Q104.
No faults, just slightly different parameters for a circuit that appears as though it could be sensitive to such things. A slightly lower Vbe in the current sink would increase the current in the diff pair slightly, and that would tend to turn on the LED driver. If that driver also has a lower Vbe then its already wanting to turn on more when there isn't a clipping situation.
That would be my take on it at the moment based on the history.
No faults, just slightly different parameters for a circuit that appears as though it could be sensitive to such things. A slightly lower Vbe in the current sink would increase the current in the diff pair slightly, and that would tend to turn on the LED driver. If that driver also has a lower Vbe then its already wanting to turn on more when there isn't a clipping situation.
That would be my take on it at the moment based on the history.
Changes in current sink Q108 can also cause a shift in the balance. To a lesser extent, changes in any of the active devices can shift the operating point.
Just going over your LTSpice tutorial. I'm learning a few new tricks. Great work!
Last night I had time to swap Q103/Q153 back to the original 2SC2240 parts. One of these has significantly lower hFE than the other (~264 v. 333), which is why I had subbed in KSC1845, which had nominal 332 hFE readings (not precise, as discussed previously).
The distortion indicators came on more slowly and glowed less brightly, but still did glow.
I didn't have time to measure voltage drop across R105, R106, and R114, but will today, as a comparison to the original recommended test.
The only remaining substituted active devices are KSC3503 for 2SC2912, and KSA1381 for 2SA1210. All passives test good and at correct values.
What is the role Q108 - is it current sink for Q307? Does it affect bias? Ylli mentioned Q108 could be a factor in the problem I'm having.
How does one update or replace bad components in these amps if the closest substitution part causes an imbalance like this? From all of the reading I've done here and elsewhere, the parts I used are common replacements.
Does it seem reasonable for me to work incrementally by first replacing Q108/Q158 with the original 2SC2912? Q104/154 are already back to original parts.
Thanks very much, gents!
The distortion indicators came on more slowly and glowed less brightly, but still did glow.
I didn't have time to measure voltage drop across R105, R106, and R114, but will today, as a comparison to the original recommended test.
The only remaining substituted active devices are KSC3503 for 2SC2912, and KSA1381 for 2SA1210. All passives test good and at correct values.
What is the role Q108 - is it current sink for Q307? Does it affect bias? Ylli mentioned Q108 could be a factor in the problem I'm having.
How does one update or replace bad components in these amps if the closest substitution part causes an imbalance like this? From all of the reading I've done here and elsewhere, the parts I used are common replacements.
Does it seem reasonable for me to work incrementally by first replacing Q108/Q158 with the original 2SC2912? Q104/154 are already back to original parts.
Thanks very much, gents!
Can you give me the component location codes for the ones connected to the trace please? You may have a burned out low value resistor that looks fine and you wouldn't expect. Often the only sign of damage is between the PCB and component, a small black circle on the body of the resistor. So just because a component looks fine, do not make that assumption. Measure and test everything that might have been damaged. Capacitors can also short and look fine too.
-Chris
Edit: Let's look at the resistors, diodes and capacitors right now. You might be able to leave the new parts installed.
-Chris
Edit: Let's look at the resistors, diodes and capacitors right now. You might be able to leave the new parts installed.
You have to look at and understand why the LED lights and then look at possible causes.
So... how does the LED light?
It does so by there having to be enough current flow in R106 to develop the required turn on voltage for Q104 (the LED driver). At a basic level we can say that Vbe will be around 600mv. That means a current of 0.6/365 has to flow. That's around 1.6 milliamps.
Armed with that information you could remove Q104 and measure the voltage across R106.
Why remove the transistor?
We remove it because the B-E junction will tend to clamp the voltage across the resistor to around 0.6v no matter what the current and we want to see what is there without anything else present that could skew the reading.
If the voltage is marginally 'in the zone' that will cause the transistor to conduct then we have to look why that is so.
Q103 and D103/104 are critical and small differences away from the original design could well alter things enough to cause a problem.
The circuit seems to be finely balanced between the LED lighting and not. By that I mean it will probably be operating at a point very close to needed trigger voltage... a real analogue design that has little leeway for variation.
The first thing though is to see just how close to the trigger point it is running at.
One more thing... hf oscillation could cause something like this to occur. Unlikely in itself but if you have changed any other parts (different output transistors and drivers etc) then its a possible.
Q108 is a constant current sink to optimally load the VAS (voltage amplifier stage) which is Q107. Its a standard practice. The other way would be a bootstrapped resistor chain.
So... how does the LED light?
It does so by there having to be enough current flow in R106 to develop the required turn on voltage for Q104 (the LED driver). At a basic level we can say that Vbe will be around 600mv. That means a current of 0.6/365 has to flow. That's around 1.6 milliamps.
Armed with that information you could remove Q104 and measure the voltage across R106.
Why remove the transistor?
We remove it because the B-E junction will tend to clamp the voltage across the resistor to around 0.6v no matter what the current and we want to see what is there without anything else present that could skew the reading.
If the voltage is marginally 'in the zone' that will cause the transistor to conduct then we have to look why that is so.
Q103 and D103/104 are critical and small differences away from the original design could well alter things enough to cause a problem.
The circuit seems to be finely balanced between the LED lighting and not. By that I mean it will probably be operating at a point very close to needed trigger voltage... a real analogue design that has little leeway for variation.
The first thing though is to see just how close to the trigger point it is running at.
One more thing... hf oscillation could cause something like this to occur. Unlikely in itself but if you have changed any other parts (different output transistors and drivers etc) then its a possible.
Q108 is a constant current sink to optimally load the VAS (voltage amplifier stage) which is Q107. Its a standard practice. The other way would be a bootstrapped resistor chain.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.