As ZM already suggested: swap things between the good and the bad channel. Once the noise goes with the swap, you know where the noise comes from.
there is just one jumper - that's jumper between input JFet buffer output and autoformer input, placed there as convenient way to set output offset of buffer and then close jumper, thus saving autoformer of possible excessive DC voltage
nothing critical, but anyway
so, if there is noise with jumper removed, it is excluding JFet buffer as possible culprit
autoformer per se can't induce noise
when you short/GND R111 , that means that you grounded output gates in AC domain
if that is killing noise, logic is that culprit is in OS itself
smaller parts ..... those are small Gremlins, but still Gremlins ...... and now everything is suspicious - parts in small CCS ( BD, BC) , IRF510 itself, optocouplers, zener diodes
it will not help in finding exactly what is wrong, but one thing worth mentioning - all parts I'm using are absolutely out of suspicion regarding originality ( no fakes), because I'm getting them from Vendors being looong in business and each part they're selling is traceable all the way to Production
which , of course, doesn't exclude possible sub-quality part - that's happening with every Production I know of
nothing critical, but anyway
so, if there is noise with jumper removed, it is excluding JFet buffer as possible culprit
autoformer per se can't induce noise
when you short/GND R111 , that means that you grounded output gates in AC domain
if that is killing noise, logic is that culprit is in OS itself
smaller parts ..... those are small Gremlins, but still Gremlins ...... and now everything is suspicious - parts in small CCS ( BD, BC) , IRF510 itself, optocouplers, zener diodes
it will not help in finding exactly what is wrong, but one thing worth mentioning - all parts I'm using are absolutely out of suspicion regarding originality ( no fakes), because I'm getting them from Vendors being looong in business and each part they're selling is traceable all the way to Production
which , of course, doesn't exclude possible sub-quality part - that's happening with every Production I know of
Partial success - no noise anymore !
So, I did pull the R channel board (that I have on my left amp channel) off the heatsink, and exchanged a few semis:
- ZD101, 104, 105, all with different makes than before
- M101, IRF510 from the same batch
- T101, BC546 – I have about 200 BC546s, but all from the same source with unknown provenance. So I put in a BC550B instead
One of the old semis must have been the culprit – there is no noise now on that channel anymore. 🙂
Before exchanging the semis, I changed the socketed PC817C (OK101 and OK102) to ones from a different make, also 817Cs. That didn’t change the noise, but it actually did improve offset behaviour – it was tracking the other channel very closely then.
After exchanging the semis, offset instability is still (or again) there. Offset runs away after maybe two hours of operation, but not as far as before. Seems to fluctuate between maybe +80mV and about +150mV after runaway. Will continue to monitor.
Maybe I should try a larger heatsink on M101, mentioned by tombo56 in post #679 in the LuDEF thread:
https://www.diyaudio.com/forums/pass-labs/369990-ludef-68.html#post6828208
On M101, I have the space for a larger heatsink, and have larger heatsinks here as well.
Currently, Q103 has 1nF across B-E, and T101 (BC550B) has no cap across C-E.
All the best, and thanks for the suggestions from you all - that has helped me get rid of the noise ! 🙂
Claas
So, I did pull the R channel board (that I have on my left amp channel) off the heatsink, and exchanged a few semis:
- ZD101, 104, 105, all with different makes than before
- M101, IRF510 from the same batch
- T101, BC546 – I have about 200 BC546s, but all from the same source with unknown provenance. So I put in a BC550B instead
One of the old semis must have been the culprit – there is no noise now on that channel anymore. 🙂
Before exchanging the semis, I changed the socketed PC817C (OK101 and OK102) to ones from a different make, also 817Cs. That didn’t change the noise, but it actually did improve offset behaviour – it was tracking the other channel very closely then.
After exchanging the semis, offset instability is still (or again) there. Offset runs away after maybe two hours of operation, but not as far as before. Seems to fluctuate between maybe +80mV and about +150mV after runaway. Will continue to monitor.
Maybe I should try a larger heatsink on M101, mentioned by tombo56 in post #679 in the LuDEF thread:
https://www.diyaudio.com/forums/pass-labs/369990-ludef-68.html#post6828208
On M101, I have the space for a larger heatsink, and have larger heatsinks here as well.
Currently, Q103 has 1nF across B-E, and T101 (BC550B) has no cap across C-E.
All the best, and thanks for the suggestions from you all - that has helped me get rid of the noise ! 🙂
Claas
IMO bigger heatsinks won’t help with bias instability. On LuDEF, they decrease total span (max – min) of offset change, but otherwise DC offset is with increased temp. going only in one direction on average.
Fluctuation around point, when thermal equilibrium is reached, is not that big.
Yeah, that's the thing that's bothering me. On the right side, the offset fluctuates almost not at at all. Whereas on the left side, after runaway, it fluctuates by maybe 50mV.
This behaviour is exhibited by that channel still after change of T101, M101, SIT, IRFP9140, and the Optocouplers. In other, related news, the IRFP9140 has a gate resistor of 221R and Q103 has 1nF across B-E.
This behaviour is exhibited by that channel still after change of T101, M101, SIT, IRFP9140, and the Optocouplers. In other, related news, the IRFP9140 has a gate resistor of 221R and Q103 has 1nF across B-E.
chede, compare Ugs voltages (up for IRF510, down for IRFP, both ref to out) between two channels
OK, here goes:
Ugs values, or, to be precise, voltages between Gate and Output
Left side (R channel board) – problem side
Ugs 9140 -4.12V
Ug-output 510 +3.16V
Right side – good side
Ugs 9140 -4.16V
Ug-output 510 +3.24V
The values above are a few minutes after turn-on of still warm amp, so in still stable, not-run-away state. Probably because of the small heat sinks of IRF510, voltage between gate and Output goes down faster for IRF510 than Ugs for IRFP9140.
Edit: left side Ug-output for IRF510 starts out at 3.33V, then goes down to 3.19V within 5 minutes after switch-on. /Edit
Regards, Claas
Ugs values, or, to be precise, voltages between Gate and Output
Left side (R channel board) – problem side
Ugs 9140 -4.12V
Ug-output 510 +3.16V
Right side – good side
Ugs 9140 -4.16V
Ug-output 510 +3.24V
The values above are a few minutes after turn-on of still warm amp, so in still stable, not-run-away state. Probably because of the small heat sinks of IRF510, voltage between gate and Output goes down faster for IRF510 than Ugs for IRFP9140.
Edit: left side Ug-output for IRF510 starts out at 3.33V, then goes down to 3.19V within 5 minutes after switch-on. /Edit
Regards, Claas
Last edited:
Offset after run-away fluctuates between, say, +80mV and +160mV. Fluctuates quickly by maybe 20-30 mV, and slower within the broader range - slower movement maybe triggered by variations in rail voltage due to variations in AC line voltage.
Running the amp now with meters connected between gates of 510 and 9140 and amp output, to maybe capture what's happening with Ugs / Ug-output after offset runaway.
I’s a pity that you didn’t exchange Q103 as well. Any CCS instability or variation will change offset.
I’m not sure if a leaking capacitor C106 could make such offset variation. Is quality/state of that part out of question?
I’m not sure if a leaking capacitor C106 could make such offset variation. Is quality/state of that part out of question?
now, when you have noise issue solved, I can't see much for worry
you'll get best offset behavior when impedances from mosfet gates to adjacent rauils are in absolute symmetry
to get that, one need to adjust "level shifter resistor" value, to have symmetrical mosfet voltages as much is possible
now, (absolute value of ) difference between IRF and IRFP is in range of 1V
to elevate gate of IRF for one volt, we need to increase value of said resistor (R115/215) for 1V/Iq of Q103
roughly, that is 1V/(OV65/22R) = 33R
if you increase R115/215 fopr 33R and re-bias amp, Ugs up and down should be so-so equal
now - I'm not loosing my sleep with mild offset fluctuations .... and that's far from anything called runaway ... reason for one channel behaving better can be mix of several factors ..... mains fluctuation, different gains of optocouplers, variation of gate currents of big ones etc.
for amp having output voltage level set in strictly passive manner, you can't expect ideal behavior - slight fluctuation and staying in range is good enough
you'll get best offset behavior when impedances from mosfet gates to adjacent rauils are in absolute symmetry
to get that, one need to adjust "level shifter resistor" value, to have symmetrical mosfet voltages as much is possible
now, (absolute value of ) difference between IRF and IRFP is in range of 1V
to elevate gate of IRF for one volt, we need to increase value of said resistor (R115/215) for 1V/Iq of Q103
roughly, that is 1V/(OV65/22R) = 33R
if you increase R115/215 fopr 33R and re-bias amp, Ugs up and down should be so-so equal
now - I'm not loosing my sleep with mild offset fluctuations .... and that's far from anything called runaway ... reason for one channel behaving better can be mix of several factors ..... mains fluctuation, different gains of optocouplers, variation of gate currents of big ones etc.
for amp having output voltage level set in strictly passive manner, you can't expect ideal behavior - slight fluctuation and staying in range is good enough
Hi tombo56,
thanks for your ideas - C106 I trust. It is a Panasonic FC, bought from Reichelt in Germany, where I have bought Panasonics for 10 years now and never had a dud. Also, bias is always stable. I guess a problematic C106 would show up in some bias instability as well.
As one of the next steps, I would also go to replace Q103. I had monitored the CCS current (voltage drop across R115) on the bench, and always found it stable.
Regards, Claas
thanks for your ideas - C106 I trust. It is a Panasonic FC, bought from Reichelt in Germany, where I have bought Panasonics for 10 years now and never had a dud. Also, bias is always stable. I guess a problematic C106 would show up in some bias instability as well.
As one of the next steps, I would also go to replace Q103. I had monitored the CCS current (voltage drop across R115) on the bench, and always found it stable.
Regards, Claas
Hi Zen Mod,
thanks - in fact, at the beginning, I had the level shifter resistors chosen in a way to provide equal voltages between gates and output for IRF510 and 9140. But I hadn't been happy with offset swings or stability then (but that could have been different causes ...), and changed the level shifter resistors to provide just the Ugs I measured for my SITs at 24V/2A.
But upon further reflection, as you laid out in your post #472, the level shifter resistor better should cover Ugs of SIT plus the Delta in Ugs between IRF510 and 9140. That way, the gate potentials with regard to the amp output would be symmetrical around C105 // C106.
That's an easy fix to try - will solder in a different resistor tomorrow in my R115 string to add approx. 33R to string.
Interestingly, the right side has stable offset with the 1V difference 😛. Movement between 30 min and 2.5 hours run-time maybe from -10mV to +5mV. I have a 5U/400 chassis, so it takes a really long time to reach thermal equlibrium.
Regards, Claas
thanks - in fact, at the beginning, I had the level shifter resistors chosen in a way to provide equal voltages between gates and output for IRF510 and 9140. But I hadn't been happy with offset swings or stability then (but that could have been different causes ...), and changed the level shifter resistors to provide just the Ugs I measured for my SITs at 24V/2A.
But upon further reflection, as you laid out in your post #472, the level shifter resistor better should cover Ugs of SIT plus the Delta in Ugs between IRF510 and 9140. That way, the gate potentials with regard to the amp output would be symmetrical around C105 // C106.
That's an easy fix to try - will solder in a different resistor tomorrow in my R115 string to add approx. 33R to string.
Interestingly, the right side has stable offset with the 1V difference 😛. Movement between 30 min and 2.5 hours run-time maybe from -10mV to +5mV. I have a 5U/400 chassis, so it takes a really long time to reach thermal equlibrium.
Regards, Claas
Last edited:
dunno
I had zero (old so easily forgetting) problems while breadboarding and finally proofing pcbs, with parts I'm regularly buying in quantity
I have ton of pulls, but lazy to even search for them, always easier to take new part from little drawer...... that's why I'm sure that I'm using freshly bought ones
which component exactly was cause for noise issues, and what can be culprit for different behavior of two channels , can't say with certainty ...... you know where you bought parts for your pcbs .......... least known source is always most questionable
I have luck with my guys - they are dealing parts for decades, with great experience regarding quality....... where we need to differentiate absolute quality and guaranteed quality
I had zero (old so easily forgetting) problems while breadboarding and finally proofing pcbs, with parts I'm regularly buying in quantity
I have ton of pulls, but lazy to even search for them, always easier to take new part from little drawer...... that's why I'm sure that I'm using freshly bought ones
which component exactly was cause for noise issues, and what can be culprit for different behavior of two channels , can't say with certainty ...... you know where you bought parts for your pcbs .......... least known source is always most questionable
I have luck with my guys - they are dealing parts for decades, with great experience regarding quality....... where we need to differentiate absolute quality and guaranteed quality
Usually I buy parts either at Reichelt, or, when I can't get it there, Mouser.
Can't remember problems with the Reichelt parts in the past.
When hunting for the noise, I changed several parts at once.
The IRF510 and the Zeners (ZD101, 104, 105) have all been from Reichelt.The BC546 I took from a tape of two hundred I got from a Polish distributor that was selling on eBay - provenance of the BC546 unknown. That's why I replaced it not with another BC546 from the same batch, but with a BC550B.
Regards, Claas
Can't remember problems with the Reichelt parts in the past.
When hunting for the noise, I changed several parts at once.
The IRF510 and the Zeners (ZD101, 104, 105) have all been from Reichelt.The BC546 I took from a tape of two hundred I got from a Polish distributor that was selling on eBay - provenance of the BC546 unknown. That's why I replaced it not with another BC546 from the same batch, but with a BC550B.
Regards, Claas
By the way, offset just took off after 2.5 hours ...
Voltages given are measured between gates and output of amp.
1 hour:
9140 -4.09V
IRF510 +3.07V
Offset -17mV
1.5 hours:
9140 -4.08V
IRF510 +3.06V
Offset -13mV
2 hours:
9140 -4.08V
IRF510 +3.05V
Offset -10mV
~ 2.5 hours:
9140 -4.08V
IRF510 +3.06V
Offset +4mV
-> Taking off
~ 2.5 hours plus 2 minutes more:
9140 -4.08V
IRF510 +3.06V
Offset +150 - +220mV
Voltages given are measured between gates and output of amp.
1 hour:
9140 -4.09V
IRF510 +3.07V
Offset -17mV
1.5 hours:
9140 -4.08V
IRF510 +3.06V
Offset -13mV
2 hours:
9140 -4.08V
IRF510 +3.05V
Offset -10mV
~ 2.5 hours:
9140 -4.08V
IRF510 +3.06V
Offset +4mV
-> Taking off
~ 2.5 hours plus 2 minutes more:
9140 -4.08V
IRF510 +3.06V
Offset +150 - +220mV
offset seems stable now
So, here's an update:
Two days ago, I added 32R to the level shifter resistor string on the left side, so that we have about equal Ugs of 9140 compared to voltage between gate of IRF510 and output. The latter consists of Ugs of IRF510 plus the voltage over the level shifter resistor, minus Ugs of SIT.
I had the amp running the last two days, with some occasional breaks to let it cool down and start again cold.
After the first half hour, the offset on the left side now only shows a little drift with temperature (+/- 10mV), and more importantly, not a runaway event.
I'm about ready to close the cover on my SissySIT R.3, and check offset only occasionally (should have a look next summer, when temperature of room and heatsink will be warmer 😉).
Interestingly, the right side is even more stable. I have a SIT with 1.9V Ugs there, and the level shifter with 62R only compensates for SIT Ugs, not the difference in Ugs of 9140 and IRF510. However, that side is even more stable, with offset variation after the first half hour in the order of +/- 1mV.
Thanks for all your suggestions and help ! 🙂
All the best, Claas
So, here's an update:
Two days ago, I added 32R to the level shifter resistor string on the left side, so that we have about equal Ugs of 9140 compared to voltage between gate of IRF510 and output. The latter consists of Ugs of IRF510 plus the voltage over the level shifter resistor, minus Ugs of SIT.
I had the amp running the last two days, with some occasional breaks to let it cool down and start again cold.
After the first half hour, the offset on the left side now only shows a little drift with temperature (+/- 10mV), and more importantly, not a runaway event.
I'm about ready to close the cover on my SissySIT R.3, and check offset only occasionally (should have a look next summer, when temperature of room and heatsink will be warmer 😉).
Interestingly, the right side is even more stable. I have a SIT with 1.9V Ugs there, and the level shifter with 62R only compensates for SIT Ugs, not the difference in Ugs of 9140 and IRF510. However, that side is even more stable, with offset variation after the first half hour in the order of +/- 1mV.
Thanks for all your suggestions and help ! 🙂
All the best, Claas
Last edited: