I have an early model Musical Fidelity B1 whose sound I really enjoy, but has problems with thermal runaway in one channel - always in the same pair of transistors.
(Schematic: http://www.vintage-radio.net/forum/attachment.php?attachmentid=21781&d=1228922797)
After several such events and consequent rebuilds, decided to look more closely at the components and see if there are any issues which could be the cause. The most likely possibilities seem to be oscillation (perhaps due to the transition frequency of modern 2N3055's being higher), or a damaged / wrongly selected / fitted component earlier in the signal path.
Since one channel has always worked, it seemed unlikely that this was a design problem, but while reading up on parallel power transistors, it stuck me that there appears to be a possible flaw which fits the issue.
T19 and T21 are always the power transistors which blow and they often take the driver T17 with them. The B1 has an unusual thermal design, as the power transistors have individual heat-sinks, despite being in parallel pairs, which is not the best arrangement it seems, but can be mitigated by careful matching of transistors and in particular by fitting low value emitter series resistors. This helps prevent variations in characteristics between pairs of transistors causing one to carry more current than the other and to become hotter, leading to higher BE voltage and further heating - leading to thermal runaway. The emitter resistor will drop more volts if carrying more current, lowering the BE voltage and thereby current carried relative to its partner - resulting in balance due to negative feedback.
It would seem from the schematic that the emitter resistors have been put in series with the collector in this design, which won't fulfill the needed negative feedback function.
The channel which has always worked likely has a well matched pair of transistors in the lower section...
I'm about to rebuild the amp again and considering modifying the layout by moving R63 and R65 to between the 3055 emitters and -ve rail. Any thoughts / advice appreciated - may well be missing something!
(Schematic: http://www.vintage-radio.net/forum/attachment.php?attachmentid=21781&d=1228922797)
After several such events and consequent rebuilds, decided to look more closely at the components and see if there are any issues which could be the cause. The most likely possibilities seem to be oscillation (perhaps due to the transition frequency of modern 2N3055's being higher), or a damaged / wrongly selected / fitted component earlier in the signal path.
Since one channel has always worked, it seemed unlikely that this was a design problem, but while reading up on parallel power transistors, it stuck me that there appears to be a possible flaw which fits the issue.
T19 and T21 are always the power transistors which blow and they often take the driver T17 with them. The B1 has an unusual thermal design, as the power transistors have individual heat-sinks, despite being in parallel pairs, which is not the best arrangement it seems, but can be mitigated by careful matching of transistors and in particular by fitting low value emitter series resistors. This helps prevent variations in characteristics between pairs of transistors causing one to carry more current than the other and to become hotter, leading to higher BE voltage and further heating - leading to thermal runaway. The emitter resistor will drop more volts if carrying more current, lowering the BE voltage and thereby current carried relative to its partner - resulting in balance due to negative feedback.
It would seem from the schematic that the emitter resistors have been put in series with the collector in this design, which won't fulfill the needed negative feedback function.
The channel which has always worked likely has a well matched pair of transistors in the lower section...
I'm about to rebuild the amp again and considering modifying the layout by moving R63 and R65 to between the 3055 emitters and -ve rail. Any thoughts / advice appreciated - may well be missing something!
Last edited:
Hi Everdine, this is a class A design with rather high quiescent current and DC feedback mechanism, using R63, R64 as current sensors.
I studied this configuration in detail in 2014 and even designed my own version, based on the same principle, but fully complementary and having some other specialties - you can see it here: My class A
It works perfectly - with moderate size heatsinks I had, I ran my prototype at 0.5A per output pair. Bias, once it's set on a warm amplifier, is very stable - I have never had runaway issues. I used emitter resistors for better current distribution between the pairs, but in this particular case they are not as vitally important, as they would be in a classic class B/AB design.
In this kind of design, the top half and the bottom half are actually two separate amplifiers, running in parallel with each of the halves pulling its side. It looks like in certain conditions the bottom side of the problematic channel oscillates, killing the output section. Rather difficult to diagnose without an oscilloscope though.
So... can you arrange an oscilloscope and likely a function generator?
I studied this configuration in detail in 2014 and even designed my own version, based on the same principle, but fully complementary and having some other specialties - you can see it here: My class A
It works perfectly - with moderate size heatsinks I had, I ran my prototype at 0.5A per output pair. Bias, once it's set on a warm amplifier, is very stable - I have never had runaway issues. I used emitter resistors for better current distribution between the pairs, but in this particular case they are not as vitally important, as they would be in a classic class B/AB design.
In this kind of design, the top half and the bottom half are actually two separate amplifiers, running in parallel with each of the halves pulling its side. It looks like in certain conditions the bottom side of the problematic channel oscillates, killing the output section. Rather difficult to diagnose without an oscilloscope though.
So... can you arrange an oscilloscope and likely a function generator?
Hi Valery,
thanks for taking the time to have a look at this - much appreciated...
As I understand it, the B1 is generally considered a class AB amp, which is usually biased at around 25mA per output device (see: http://www.vintage-radio.net/forum/showthread.php?t=35190). It is possible to run further into the class A region, although this is know to cause heating problems due to power transistors being individually mounted on rather small heatsinks.
Amp layout:
http://www.vintage-radio.net/forum/attachment.php?attachmentid=21873&d=1229428635
I see your design has a common heatsink, which is clearly a much better arrangement thermally. Perhaps you were thinking of the B1's close relative, the A1, which runs mostly in class A? (Musical Fidelity A1 › Technical)
Good point concerning the role of R63 and 64 - seems at least in the case of the upper section, the resistors have a dual role, being both used for current sensing and as emitter resistors. However, the negative feedback control relates to the current flowing through T19 and 20 only, so it appears there's a design assumption that at least T19 and 21 are closely matched to in performance. (T20 and T22 have emitter resistors for protection, so matching not so important).
In the case of T19 and 21, if the pair are not closely matched then thermal runaway must be a potential issue, as there is no direct NFB control for T21, no thermal coupling with T19 (different heatsinks) and no emitter resistor.
The role of R63 in current sensing rules out moving it (missed that!), but could perhaps just move R65 to T21's emitter, as it doesn't have a current sensing role and is linked to the most likely device to cause problems.
Concerning oscillation, it's certainly a possibility and would naturally make any thermal issues worse. I've sourced a set of 10 Multicomp 3055's which have a transition frequency of 800Khz. The ON-semis fitted previously transition at 2.5MHz, so hopefully the new set will help with stability. Plan to build a simple test rig to match the transistors as closely as possible.
Have recently fixed my scope and have access to signal sources on CD (don't have a function generator currently), so hopefully that will be good enough to test for oscillation...
All the best,
Martin
thanks for taking the time to have a look at this - much appreciated...
As I understand it, the B1 is generally considered a class AB amp, which is usually biased at around 25mA per output device (see: http://www.vintage-radio.net/forum/showthread.php?t=35190). It is possible to run further into the class A region, although this is know to cause heating problems due to power transistors being individually mounted on rather small heatsinks.
Amp layout:
http://www.vintage-radio.net/forum/attachment.php?attachmentid=21873&d=1229428635
I see your design has a common heatsink, which is clearly a much better arrangement thermally. Perhaps you were thinking of the B1's close relative, the A1, which runs mostly in class A? (Musical Fidelity A1 › Technical)
Good point concerning the role of R63 and 64 - seems at least in the case of the upper section, the resistors have a dual role, being both used for current sensing and as emitter resistors. However, the negative feedback control relates to the current flowing through T19 and 20 only, so it appears there's a design assumption that at least T19 and 21 are closely matched to in performance. (T20 and T22 have emitter resistors for protection, so matching not so important).
In the case of T19 and 21, if the pair are not closely matched then thermal runaway must be a potential issue, as there is no direct NFB control for T21, no thermal coupling with T19 (different heatsinks) and no emitter resistor.
The role of R63 in current sensing rules out moving it (missed that!), but could perhaps just move R65 to T21's emitter, as it doesn't have a current sensing role and is linked to the most likely device to cause problems.
Concerning oscillation, it's certainly a possibility and would naturally make any thermal issues worse. I've sourced a set of 10 Multicomp 3055's which have a transition frequency of 800Khz. The ON-semis fitted previously transition at 2.5MHz, so hopefully the new set will help with stability. Plan to build a simple test rig to match the transistors as closely as possible.
Have recently fixed my scope and have access to signal sources on CD (don't have a function generator currently), so hopefully that will be good enough to test for oscillation...
All the best,
Martin
Last edited:
A quick progress update - matched a set of driver and Multicomp 2N3055 transistors as closely as possible using a DVM and fitted them to the problem channel, after checking all other components in the driver stage onward.
Added a pair of 0.22R 5W wire-wound emitter resistors to T19 and 21 on the reverse side of the board.
Tested using a software signal generator + 10R 10W wire-wound load, after setting the bias to approx 25mA per section and adjusting DC offset with no load and no signal.
Ran the amp up to 17.5v peak-to-peak (12.25 RMS, approx 15W) with sine wave @1KHz and left it there for 15 minutes. The load resistors got pretty hot, so discontinued, but no sign of thermal runaway, cross-over distortion or oscillation. Let them cool and repeated at 10KHz with the same result. Attempted to induce oscillation in both cases by dropping the input signal in and out repeatedly, but nothing visible on the scope...
Initial audio tests with a pair of Heybrook Point 5's sounded good and bias was stable, so moved over to my main system and a pair of AE100's which also sound fine, in fact more controlled particularly in the bass region.
Have run the amp at high volume for a few minutes at a time (the conditions which caused thermal run-away repeatedly previously) without problems, so hopefully the issue is solved. Will report back if it reoccurs...
Added a pair of 0.22R 5W wire-wound emitter resistors to T19 and 21 on the reverse side of the board.
Tested using a software signal generator + 10R 10W wire-wound load, after setting the bias to approx 25mA per section and adjusting DC offset with no load and no signal.
Ran the amp up to 17.5v peak-to-peak (12.25 RMS, approx 15W) with sine wave @1KHz and left it there for 15 minutes. The load resistors got pretty hot, so discontinued, but no sign of thermal runaway, cross-over distortion or oscillation. Let them cool and repeated at 10KHz with the same result. Attempted to induce oscillation in both cases by dropping the input signal in and out repeatedly, but nothing visible on the scope...
Initial audio tests with a pair of Heybrook Point 5's sounded good and bias was stable, so moved over to my main system and a pair of AE100's which also sound fine, in fact more controlled particularly in the bass region.
Have run the amp at high volume for a few minutes at a time (the conditions which caused thermal run-away repeatedly previously) without problems, so hopefully the issue is solved. Will report back if it reoccurs...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.