Hi guys,
I've been restoring an MCS 3248 receiver for the past few weeks and I'm getting down into the tuning the performance stages. I've replaced all the old electros (Several were of a breed known to have high failure rates) and replaced all of the signal coupling caps with films of equal value.
Here's the amplifier circuit:
One of the last things left to do is tuning the DC offset and bias. I replaced the bias trimmers with Bourns multi-turns so that part was easy, but the DC offset has me flummoxed. I initially measured ~40mV on both channels, with one slightly lower than the other. I figured the 2SA798 diff pair transistors had drifted (or were never very well matched in the first place) and replaced with matched pairs of KSA992s... which resulted in an offset voltage change of didly. Within 4mV of the original pairs.
So, thinking maybe the current mirror was out of spec and was driving one of the diff pair transistors harder than the other, I matched a pair of KSC1845s and replaced TR117 and TR119, and for good measured zip tied them together with thermal compound between them. I also replaced D101 (Infamous VD1212, blech!) with a pair of 1N4148's in series and replaced R109, R111, and R113 with tight tolerance parts (R111 and R113 are matched within 0.1%). R109 wasn't strictly necessary for the offset but I wanted to equalize the performance between the channels too and resistors are pennies.
I fired it up and... 40mV +-2mV.
At this point I'm wondering if some coupling cap (one of the original films or ceramics) somewhere else in the circuit has gone leaky and is introducing DC or if my matching procedures were crap?
I matched using Ian Fritz's circuit and then did a Vce vs Ic curve trace using my Peak Atlas DCA75 on the pairs that matched up using the test circuit and selected the best ones. Here's an example trace of one of the matched differential sets overlayed:
Vs one of the original 2SA798 pairs:
I'm rebuilding this to learn about electronics and their performance as much as anything else, and I realize that 40mV DC offset isn't world ending, but I'd like to get it into the 20s or below if possible. The fact that all of my tweaking has had nearly zero effect is what is strange to me. I would have expected to at least make it perform differently if not better.
Wanting to learn here, hopefully someone here can enlighten me. Am I tracing the wrong curves? Testing at the wrong current? Is there 40mV of offset built into this design somewhere and I'm just not seeing it?
Cheers,
Nathan
I've been restoring an MCS 3248 receiver for the past few weeks and I'm getting down into the tuning the performance stages. I've replaced all the old electros (Several were of a breed known to have high failure rates) and replaced all of the signal coupling caps with films of equal value.
Here's the amplifier circuit:
One of the last things left to do is tuning the DC offset and bias. I replaced the bias trimmers with Bourns multi-turns so that part was easy, but the DC offset has me flummoxed. I initially measured ~40mV on both channels, with one slightly lower than the other. I figured the 2SA798 diff pair transistors had drifted (or were never very well matched in the first place) and replaced with matched pairs of KSA992s... which resulted in an offset voltage change of didly. Within 4mV of the original pairs.
So, thinking maybe the current mirror was out of spec and was driving one of the diff pair transistors harder than the other, I matched a pair of KSC1845s and replaced TR117 and TR119, and for good measured zip tied them together with thermal compound between them. I also replaced D101 (Infamous VD1212, blech!) with a pair of 1N4148's in series and replaced R109, R111, and R113 with tight tolerance parts (R111 and R113 are matched within 0.1%). R109 wasn't strictly necessary for the offset but I wanted to equalize the performance between the channels too and resistors are pennies.
I fired it up and... 40mV +-2mV.
At this point I'm wondering if some coupling cap (one of the original films or ceramics) somewhere else in the circuit has gone leaky and is introducing DC or if my matching procedures were crap?
I matched using Ian Fritz's circuit and then did a Vce vs Ic curve trace using my Peak Atlas DCA75 on the pairs that matched up using the test circuit and selected the best ones. Here's an example trace of one of the matched differential sets overlayed:
Vs one of the original 2SA798 pairs:
I'm rebuilding this to learn about electronics and their performance as much as anything else, and I realize that 40mV DC offset isn't world ending, but I'd like to get it into the 20s or below if possible. The fact that all of my tweaking has had nearly zero effect is what is strange to me. I would have expected to at least make it perform differently if not better.
Wanting to learn here, hopefully someone here can enlighten me. Am I tracing the wrong curves? Testing at the wrong current? Is there 40mV of offset built into this design somewhere and I'm just not seeing it?
Cheers,
Nathan
Indeed: the tail current is ~1.5mA (0.6V/390), which means 750µA for each member of TR103 or 4.
Assuming a reasonable β of ~300 for these transistors, the base current will be ~2.5µA.
Since R117 is 68K and R105 is 100K, a voltage difference of (100-68)*2.5=80mV will be developed.
That's twice the voltage actually measured, so nothing wrong IMHO: the imperfect null could be designed-in, for E-caps bias for example, or it could be caused by a β lower than typical for TR103.
Normally, the DC resistances of the input paths should be matched, to avoid this kind of (non?)issues, but Japanese designers sometimes follow contorted paths, sometimes for good reasons.
Anyway, there is nothing to worry about: 40mV is a bit high, but will certainly not be problematic
Thanks for all the info guys! I thought that with the current mirror driving the differential pair that there was no way the currents in the pair could be off.
Would there be any risk/harm in lowering the 100k input resistor to 68k? For what reason would it have imbalance designed in? I've eliminated all electrolytic caps from the signal path on the amp board so I should be safe on cap bias.
I was originally going off of info I found in this thread:
Amplifier Distortion, DC-Offset, and You! | Audiokarma Home Audio Stereo Discussion Forums
Which lead me to believe that lowering the offset from ~40mV would result in am improvement in distortion performance and that ~20mV should be the upper end of acceptability.
Would there be any risk/harm in lowering the 100k input resistor to 68k? For what reason would it have imbalance designed in? I've eliminated all electrolytic caps from the signal path on the amp board so I should be safe on cap bias.
I was originally going off of info I found in this thread:
Amplifier Distortion, DC-Offset, and You! | Audiokarma Home Audio Stereo Discussion Forums
Which lead me to believe that lowering the offset from ~40mV would result in am improvement in distortion performance and that ~20mV should be the upper end of acceptability.
Last edited:
One other question I had was, the original diff pair current mirror transistors measured at an hFE of ~480 and the replacements were the highest matched pair of KSC1845s I could manage from my 200 I purchased (hFE ~401). Is the reduced hFE in that part of the circuit likely to cause an issue? I have saved the original pair and I can reinstall if there is a benefit to doing so. I would either swap the other channel out to match the new pair (I have another matched pair at ~399 hFE or so) or put the old ones back into both channels. I'm disinclined to reinstall the 2SA798s as I've read a few different accounts of those going wonky.
Last edited:
Yes, TR103 left base IS somewhat above ground, it MUST be, it´s not an error but a natural phenomenon, because its base current goes through R105, 100k, and of course that drops some voltage.
TR103 right base does exactly the same, in this case its bias current passing gthrough R117.
IF both base currents were the same, both offset voltages the same and both resistors the same, you would have 0 V offset.
But you only corrected one of 3 parameters .
I am quite certain if both resistors match offset will be lower, although not necessarily zero.
As an extra point, most assume current splits evenly between both input differential transistors ... it ain´t necessarily so, by any means ... why would it if the job of both TR103 halves is not the same?
TR103 left supplies whatever´s needed by TR105 base to keep VAS idle current at the desired value, and in this case whatever is supplied by TR107 ...
This alone, which is often ignored, makes TR103 left current vary depending on TR5 Hfe ... just there you have a new variable.
TR103 right will pass the balance of current supplied by TR101 , whatever it is.
Will both halves be the same, split 50/50 all the time?
Think again ... "only by sheer chance".
TR103 right base does exactly the same, in this case its bias current passing gthrough R117.
IF both base currents were the same, both offset voltages the same and both resistors the same, you would have 0 V offset.
But you only corrected one of 3 parameters
. I am quite certain if both resistors match offset will be lower, although not necessarily zero.
As an extra point, most assume current splits evenly between both input differential transistors ... it ain´t necessarily so, by any means ... why would it if the job of both TR103 halves is not the same?
TR103 left supplies whatever´s needed by TR105 base to keep VAS idle current at the desired value, and in this case whatever is supplied by TR107 ...
This alone, which is often ignored, makes TR103 left current vary depending on TR5 Hfe ... just there you have a new variable.
TR103 right will pass the balance of current supplied by TR101 , whatever it is.
Will both halves be the same, split 50/50 all the time?
Think again
... "only by sheer chance".
Last edited:
It was my understanding that the current mirror (TR117, TR119) would force the current collector to emitter on both halves of the pair to be identical (or within the tolerances set by the matching of TR117, TR119, and R111, R113). So if I'm understanding you guys correctly, even though the mirror forces Ice to be identical, having different base currents will result in offset?
Here's a question. I believe these are the resistors in question:
It looks to me like the right hand side is not tied directly to ground, but to some negative voltage value based on the negative rail (Look at the voltages at the emitters of the current mirror on the other channel). Can anyone decipher what negative voltage that's tied to?
Also, how does the tone amp section tied in through C138 affect things (And as an aside, should C138 be an audio grade cap? I thought I identified all of the audio path caps but if audio signals are running through there I think I missed one)
If anyone wants to see the off-board connections, the pdf linked on the last page has the whole schematic.
Thanks,
Nathan
It looks to me like the right hand side is not tied directly to ground, but to some negative voltage value based on the negative rail (Look at the voltages at the emitters of the current mirror on the other channel). Can anyone decipher what negative voltage that's tied to?
Also, how does the tone amp section tied in through C138 affect things (And as an aside, should C138 be an audio grade cap? I thought I identified all of the audio path caps but if audio signals are running through there I think I missed one)
If anyone wants to see the off-board connections, the pdf linked on the last page has the whole schematic.
Thanks,
Nathan
Last edited:
Sorry for multiple posts inline. I've been doing a lot of reading. Is my understanding that the path to ground for the feedback some of the LTP is through the load (speaker) itself correct? At DC C109 and C107 will not pay any current. That leaves the path through R115/117 to the output as the only non-capacitively coupled connection (the EQ network is also capacitively coupled by C138 so should present no path to ground at DC)
With that understanding, is C107 in place to squash high frequency oscillation by allowing it to feedback into the 'inverting' input of the LTP?
With that understanding, is C107 in place to squash high frequency oscillation by allowing it to feedback into the 'inverting' input of the LTP?
You have correctly identified R115-R117 as the only RHS DC path to ground. The load is not relevant because amplifier output impedance is going to be far lower than load impedance, so the load in parallel does not make much of a difference. In small-signal terms, voltage sources are equivalent to ground.
Why they would have made R105 100k but R117 68k is anyone's guess. Choosing the same value for both wouldn't have made much of a difference either way while reducing DC offset, especially since the left leg is likely to be the one with higher current anyway (TR105 base current!). Assuming you've got a 100k or lower log volume pot there, I'd go with 68k for R105, maybe even lower. Maybe you've got a 220k-ish pot floating around to wire up in parallel for testing?
C107 is stability-related compensation business, yes. You often see these in parallel to feedback resistors. They'll compensate for amplifier input capacitance and pull down gain at high frequencies to extend bandwidth. (Input capacitance essentially is in parallel to the ground leg of your feedback network and could accordingly cause some major gain peaking at high frequencies.) Poles and zeroes juggling for control theory geeks.
Yes, C138 should probably be a low leakage (or DIY bipolar) part as it never sees much of any DC.
The tone controls are part of the power amp feedback network here, which makes things a bit hard to understand. It's basically one big non-inverting Baxandall tone control. You may want to simplify it for one given setting and frequency. I couldn't even say what's going on in terms of AC impedance balance, would probably have to drag out the simulator...
Not a major fan of this 1970s output bias circuit BTW (D105 et.al.). Wouldn't be surprised to see a slight positive temperature coefficient. (Is D105 mounted to the heatsink at least?) You'd have to monitor bias current to be sure though. (After the amp has been subjected to some loading, bias should be lower than before.)
Why they would have made R105 100k but R117 68k is anyone's guess. Choosing the same value for both wouldn't have made much of a difference either way while reducing DC offset, especially since the left leg is likely to be the one with higher current anyway (TR105 base current!). Assuming you've got a 100k or lower log volume pot there, I'd go with 68k for R105, maybe even lower. Maybe you've got a 220k-ish pot floating around to wire up in parallel for testing?
C107 is stability-related compensation business, yes. You often see these in parallel to feedback resistors. They'll compensate for amplifier input capacitance and pull down gain at high frequencies to extend bandwidth. (Input capacitance essentially is in parallel to the ground leg of your feedback network and could accordingly cause some major gain peaking at high frequencies.) Poles and zeroes juggling for control theory geeks.
Yes, C138 should probably be a low leakage (or DIY bipolar) part as it never sees much of any DC.
The tone controls are part of the power amp feedback network here, which makes things a bit hard to understand. It's basically one big non-inverting Baxandall tone control. You may want to simplify it for one given setting and frequency. I couldn't even say what's going on in terms of AC impedance balance, would probably have to drag out the simulator...
Not a major fan of this 1970s output bias circuit BTW (D105 et.al.). Wouldn't be surprised to see a slight positive temperature coefficient. (Is D105 mounted to the heatsink at least?) You'd have to monitor bias current to be sure though. (After the amp has been subjected to some loading, bias should be lower than before.)
Last edited:
Great post! Thanks so much for taking the time to write it.
I'll see if I have a trimmer 220k or higher to experiment with. Otherwise I have a grab bag of metal films here I can tack on to the back side of the boards until I find the right value.
Would there be any good reason to want a permanently slightly positive offset? Does introducing imbalance in the diff pair in this way cause any negative affects other than the offset?
I'll see if I have a trimmer 220k or higher to experiment with. Otherwise I have a grab bag of metal films here I can tack on to the back side of the boards until I find the right value.
Would there be any good reason to want a permanently slightly positive offset? Does introducing imbalance in the diff pair in this way cause any negative affects other than the offset?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.