Here you go. Thank you.
R103 71.5mV
R104 69.8mV
R146 71.4mV
R147 71.7mV
R3 71.3mV
R4 73.5mV
R146 75.8mV
R147 70.2mV
The complementary input pairs (Q102/103 and Q106/107) I replaced in the right channel were hFE gain matched within 1%, and Vbe were matched around 5%. I used a transistor tester for those measurements. However, the dc servo output(pins 8 and 9) increased from 20mV to to around 100mV after replacing the complementary input transistors.
R103 71.5mV
R104 69.8mV
R146 71.4mV
R147 71.7mV
R3 71.3mV
R4 73.5mV
R146 75.8mV
R147 70.2mV
The complementary input pairs (Q102/103 and Q106/107) I replaced in the right channel were hFE gain matched within 1%, and Vbe were matched around 5%. I used a transistor tester for those measurements. However, the dc servo output(pins 8 and 9) increased from 20mV to to around 100mV after replacing the complementary input transistors.
Thanks. All those look nominal. I'm still suspicious about abnormally high bias current in the right channel, but will have to ponder how to prove that's the case.
One way would be to swap those four devices with their brethren in the other channel, but I'm not suggesting that. Unduly onerous...
One way would be to swap those four devices with their brethren in the other channel, but I'm not suggesting that. Unduly onerous...
Last edited:
Okay, Thanks. If I’m understanding correctly, the voltage spikes should be reduced with the dc servo output?
I'm still struggling with the innards of the amplifier. Am I correct in guessing JP501 (JP503) connects to a panel mounted pot? And that unplugging the cables at JP501(JP503) would leave the path to the mute relays floating? (BTW, what is SW502B?)
A tentative suggestion is to unplug these cables on the theory that leaves no unrecognized paths leading to the power amp inputs. Adding confidence that bias voltage present at R103 reflects bias current at PA input, i.e. Q102/103 and Q106/107.
Perhaps even clearer, is there a pluggable cable at JP101 that could be removed? That would leave nothing remaining. Then momentarily shorting JP101 with an insulated tool should behave exactly like muting relay. Sound right? What do you observe?
A tentative suggestion is to unplug these cables on the theory that leaves no unrecognized paths leading to the power amp inputs. Adding confidence that bias voltage present at R103 reflects bias current at PA input, i.e. Q102/103 and Q106/107.
Perhaps even clearer, is there a pluggable cable at JP101 that could be removed? That would leave nothing remaining. Then momentarily shorting JP101 with an insulated tool should behave exactly like muting relay. Sound right? What do you observe?
Yes, but correction isn't instantaneous. That's the thump you're hearing, I believe.
Correct, JP503/5 connects to the volume pot. I’ve soldered those connections because I had some feedback issues with the pcb connectors. I can re-solder if needed.
SW502 is a push in/out switch that separated or connects the preamp to the amp. I’ve removed that switch and use jumpers for the preamp to amp.
one thing I have checked is the gain relay switch RL507.
I’ll follow up with your other questions later. Thank you.
If it helps, I separated amp and preamp and observed much louder thumping in both channels. Probably shouldn’t do that again.
I’m thinking maybe this a manageable issue. Music at 30% volume is quite loud with my inputs. I typically listen at around 20% and switch between a dac and a turntable. This amp sounds great. I rebuilt it from parts. I can live with turning down the volume before switching inputs/ powering up or down/muting. I’ll have to think about the next troubleshooting steps you recommend. Troubleshooting runs the risk of damaging other components. What I can’t understand is why the dc servo output increased from 20mV to 100mV when replacing the stock complementary pairs with hand matched pairs of the same type -bc846b and bc856b. Do the comp pair resistor voltages reflect how well they are matched, if so it seems the left channel is not as well matched, while the dc servo output is less than 1mV. Is it possible the circuit was designed for some tolerance in matching comp pairs for manufacturing? How is the dc servo output voltage related to the complementary pairs?
Last edited:
I'm not entirely clear what pair resistor voltages you're referring to, but the steady state voltage at the PA output is governed solely by offset errors in the servo amplifer; assuming negligible bias and leakage contributions, in principle the input offset error of U101D will be the error seen at the PA output.
These are all good questions and I'll try to elaborate later.
I think it should be possible to trim away most of the mute thumps by adding an appropriately selected resistor at the PA inputs, no further exploratory intrusions needed. Sorry to leave things hanging...
I appreciate the help, no worries
. Regarding the voltages across resistors R3/4, R46/47 versus R103/104 R146/47. It seems the more consistent voltages in the right channel would indicate those pairs are better matched? If that’s the case then how does that effect (if at all) the dc servo offset?
Last edited:
You may be correct re matching of pairs, but in thumping issue I believe the bigger issue is that the net bias in amp input, the base current in the NPN pair is different than the PNP pair, leading to larger net bias current in the right channel vs. left. I'll offer more detail later, but I'm going to suggest a adding a resistor to trim the bias voltage seen at the input to near 0.I appreciate the help, no worries
Earlier:
At~30% volume
R101= 0.5mV
R1=.02mV (0 volts)
That's plus 0.5mV on R101? Polarity key here. You provided two digits accuracy on R1 voltage. Can report two digits resolution on R101?
Last edited:
I am glad I double checked the polarity. I’ve attached the amp board silkscreen. The R1 and R101 positions are swapped
.I increased the volume a bit more and double checked polarity. It turns out the voltage across R101 is negative.
R101= -1.20mV
R1= 0.02 mV
I am open to adding resistance to reduce the bias voltage.
A couple of ideas:
1. You said the net base current in the NPN vs PNP pairs is different, assuming you’re talking about Q3/4 and Q5/10 and Q103/4 and Q105/10, how about swapping say Q5 with Q105 as a compromise for the sake of improving L/R parity? My thinking is this would reduce the bias current in the right channel and increase it in the left. The dc servo output in the right channel is quite low at 0.8mV.
2. Would bridging (if safe) the left and right inputs at JP501 and JP503 yield any more information? It’s a pain to resolder those connections, but doable.
Last edited:
Last night, I had a brain storm: apply output servo to the PA non-inverting input instead of inverting! I think this will cure the complaint, since the click/thump arises when positive input bias is disturbed by input muting or change of source.
If you trust me,, here are simple mods for the right channel:
1. Disable the servo path from R113. You can either remove R113 or install a short across R137 by contriving a solder-bridge between pins 1 and 2 on the opamp. Either method should work.
2. Install a 2M resistor, one lead at junction of R102, C101, Q101 base, Q107 base and other lead at U101D, pin 14. This takes the servo control from the opposite polarity that's now needed since we've swapped inputs at the PA.
That's it! Be sure to disconnect speakers until proper operation is confirmed.
If you trust me,
, here are simple mods for the right channel:1. Disable the servo path from R113. You can either remove R113 or install a short across R137 by contriving a solder-bridge between pins 1 and 2 on the opamp. Either method should work.
2. Install a 2M resistor, one lead at junction of R102, C101, Q101 base, Q107 base and other lead at U101D, pin 14. This takes the servo control from the opposite polarity that's now needed since we've swapped inputs at the PA.
That's it! Be sure to disconnect speakers until proper operation is confirmed.
ABORT!Last night, I had a brain storm: apply output servo to the PA non-inverting input instead of inverting! I think this will cure the complaint, since the click/thump arises when positive input bias is disturbed by input muting or change of source.
If you trust me,
1. Disable the servo path from R113. You can either remove R113 or install a short across R137 by contriving a solder-bridge between pins 1 and 2 on the opamp. Either method should work.
2. Install a 2M resistor, one lead at junction of R102, C101, Q101 base, Q107 base and other lead at U101D, pin 14. This takes the servo control from the opposite polarity that's now needed since we've swapped inputs at the PA.
That's it! Be sure to disconnect speakers until proper operation is confirmed.
I believe the servo would work, but I no longer believe it will resolve the original problem. So the original configuration remains best. I will have an alternate suggestion shortly.
Sorry for the false suggestion. And sorry for slow response. Too many distractions...
Here's a description of what I believe is happening.
As you've measured, there's -1.20mV bias present at PA input. The PA stage has a gain of 48, about 34dB. When the Mute relay is exercised, the bias is shorted to ground, generating a 1.2mV glitch at the input, which delivers 59mV click at the speaker; the servo nulls this back to 0V with about a half second time constant. When mute is canceled, the reverse glitch is produced.
Apparently, the OEM matching in the left channel is better than what the replacement transistors deliver. But I do believe the bias error could be trimmed out, so that when mute is exercised, there's minimal disturbance generated.
So the right channel presents with -1.2mV bias. This suggests the bias current is 1.2mV/47K = 25.6nA. My initial thought was to trim in 25.6nA from a internal 15V supply from within the amp. But 15V/26nA=5.7E8 ohm, so that's an impractically large resistance for a single conventional resistor.
The attachment suggests a possible implementation, for installation. But perhaps you have bench equipment that you could use for proof of applicability before the tedium of installation. A 10M resistor adjusted by an external bench source could prove/disprove the concept.
Good luck!
Here's a description of what I believe is happening.
As you've measured, there's -1.20mV bias present at PA input. The PA stage has a gain of 48, about 34dB. When the Mute relay is exercised, the bias is shorted to ground, generating a 1.2mV glitch at the input, which delivers 59mV click at the speaker; the servo nulls this back to 0V with about a half second time constant. When mute is canceled, the reverse glitch is produced.
Apparently, the OEM matching in the left channel is better than what the replacement transistors deliver. But I do believe the bias error could be trimmed out, so that when mute is exercised, there's minimal disturbance generated.
So the right channel presents with -1.2mV bias. This suggests the bias current is 1.2mV/47K = 25.6nA. My initial thought was to trim in 25.6nA from a internal 15V supply from within the amp. But 15V/26nA=5.7E8 ohm, so that's an impractically large resistance for a single conventional resistor.
The attachment suggests a possible implementation, for installation. But perhaps you have bench equipment that you could use for proof of applicability before the tedium of installation. A 10M resistor adjusted by an external bench source could prove/disprove the concept.
Good luck!
I really appreciate the help in troubleshooting and providing a solution. I don’t have the equipment to test this theory outside of the amp. Now I know source of the thumping and high dc servo offset. Issues 1 and 2 are manifestations of the same root cause. Thanks for the schematic with your solution. I’m going to replace all the resistors next. Its low cost and low risk. I already triggered fault mode a few times when measuring voltages in the input stage. I figure this is something I can live with. Looking over old notes from when I first started fixing this amp the right side dc servo output was 200mV, before replacing any components in the input stage. Anyhow, if anything now I know there will probably be no damage to the components as long as I don’t initiate the mute relay when the volume is cranked up, which is very rare. Your summary of the issue makes sense. The servo response time you mentioned also explains the timing of dc offset spikes on muting or swings on fast volume adjustments. Very much appreciate your help.
. Thanks for the schematic with your solution. I’m going to replace all the resistors next. Its low cost and low risk. I already triggered fault mode a few times when measuring voltages in the input stage. I figure this is something I can live with. Looking over old notes from when I first started fixing this amp the right side dc servo output was 200mV, before replacing any components in the input stage. Anyhow, if anything now I know there will probably be no damage to the components as long as I don’t initiate the mute relay when the volume is cranked up, which is very rare. Your summary of the issue makes sense. The servo response time you mentioned also explains the timing of dc offset spikes on muting or swings on fast volume adjustments. Very much appreciate your help.
My humble advice would be to leave the left channel alone, as I believe you find its present behavior is acceptable. Swapping transistors (e.g. Q5/Q105) might marginally improve the right channel but might leave both channels unsatisfactory.
The emitter current in each input transistor is about 0.7mA. Data sheets indicate typical beta is about 330, so each base would require about 2.1uA bias current. If emitter currents were perfectly matched and if input NPN and PNP transistors had identical beta, there would be net 0 bias current drawn at the PA input.
In present case, even the problematic right channel shows pretty good matching: the right channel bias voltage implies net bias current is 1.2mV/47K = 25.6nA; but 25.6nA/2.1uA= 0.012--- that is, the NPN and PNP stages are matched/balanced with only 1.2% error. The left channel is even better.
Looking closer at the right PA input, the input bias is -1.2mV, so it could be argued that Q101 is drawing a little too much base current. So I suggest reducing the resistance of R104 about 10% (perhaps by tacking a 1K resistor on top of the existing 100 ohm). This should reduce Q101 current (and its base current, commensurately) and bias voltage should rise. Assuming my thinking is correct, I'd continue to revise the R104 value experimentally until the bias voltage is made acceptably small.
I’ve decided to leave the left channel alone, thanks for the insight and recommendations. I looked at my logs and the right channel complementary pair Betas were matched around .5% according to a transistor tester which does not publish error information. I measured each gain 3 times for accuracy and precision out of 40 NPN and PNP reels. I was careful not to touch the transistors while measuring. The best Vbe match I could find was around 5%, which is very good for NPN/PNP pairs. Yet still the transistor tester does not list any measurement errors. There seems to be something else in the mix here. For one, the input bias is 0 at minimum volume and increase nonlinearly wrt volume. Secondly, I can’t see how Creek Audio spent time in manufacturing to match each complementary pair, rather than designing the circuit to work within some mismatch tolerance. Anyway, I replaced Q101 and Q106 - no change in dc servo output. When I replaced Q102 the dc servo output (now around 400mV) triggered fault mode. I’m going to try and rematch again, this time with a more reliable method. All resistors are within tolerances. Also will check for any shorts or bad connections, and see if resoldering the working q102 clears the fault mode. Is matching most critical for Q102 and Q106?
Last edited:
The output servo addresses input offset voltage error; it injects a correction to drive the PA output to 0V. So I believe it is more challenged by excessive Vbe offset error in each same-sex pair. That is, it's more important that NPN transistors match to themselves and PNP transistors to each other, less important that NPN and PNP match each. That helps input offset voltage. Separate, different issue is how closely NPN bias current matches PNP bias current; the bias servo has no handle on that.
The easiest handle on offset voltage is the servo error error opamp. As you've been noting, the lower the inherent voltage offset error, the lower the correction voltage needed from the servo error amp.
The easiest handle on offset voltage is the servo error error opamp. As you've been noting, the lower the inherent voltage offset error, the lower the correction voltage needed from the servo error amp.
Ok then the only times I have experienced a change in the dc servo output is after replacing Q102/6 and Q101/7 which are the complementary NPN/PNP pairs, right? With the lowest change in Vbe (Vbe (NPN)- Vbe (PNP)) is around 30mV.
Q103/4 and Q105/10 are the same-sex pairs, these were also replaced and very well matched since each pair was in the same reel.