I've got this Sansui AU-X1 in the workshop in order to find a solution to prevent the issue with output stage oscillation, that had caused some AU-X1 to burn out the output stage.
I've been reading some threads on several fora e.g. audiogon etc. regarding this issue, and it seems like this issue is only present when no speakers/load is connected - even only a set of connected headphones should prevent this oscillation. I did not find any add-on/mod solutions though!
Then in a short "brain fart" it came to me, that if one build in a separate seudo headphone load (resistors, inductor and cap only), the problem should be solved. This would only put very little load on the output stage and would not affect the performance in any way......
What do you guys and girls think??
Thanks for any input to come
I've been reading some threads on several fora e.g. audiogon etc. regarding this issue, and it seems like this issue is only present when no speakers/load is connected - even only a set of connected headphones should prevent this oscillation. I did not find any add-on/mod solutions though!
Then in a short "brain fart" it came to me, that if one build in a separate seudo headphone load (resistors, inductor and cap only), the problem should be solved. This would only put very little load on the output stage and would not affect the performance in any way......
What do you guys and girls think??
Thanks for any input to come
Last edited:
I think you are very courageous to work on such a beached whale, it is one of the worst amps I have worked on in terms of access and structure.
If you publish the schematic, I think there will be people more serious than me to answer you but I would be surprised if such a simple answer was given to such a complex design.
good luck .
If you publish the schematic, I think there will be people more serious than me to answer you but I would be surprised if such a simple answer was given to such a complex design.
good luck .
I'm just looking at the power amp schematic - thanks for sharing.
1. The Zobel is placed after the output coupling inductor and that's a pretty big no-no in Self and Cordell's books. You do run the risk of HF OPS oscillation with this type of scheme. By placing the Zobel after the output L, you are isolating the OPS from the Zobel. The Zobel is also not placed on the main amplifier PCB (where it should be to keep HF circulating current loops as small as possible), so any wiring L only exacerbates the problem.
2. The OPS is an EF3 if I'm reading the schematic correctly. These are prone to instability unless treated with care - again, Self and Cordell discuss this in-depth (see especially Cordell).
3. There are no values shown for the OP transistor base stoppers, so I assume they are linked out. Again, with an EF3, this is not recommended practice nowadays - typically you would use 2.2 - 10 Ohms. JC used 10 Ohms, I use 4.7 Ohms and both Self and Cordell discuss a range of values in their excellent books.
4. There is a base stopper network in the pre-driver transistors. Normally, you would place these base stoppers and associated base load capacitors in the base circuit of the driver devices. Cordell shows IIRC 10 Ohms with a 470 pF to the rail, but I have seen and used 22 Ohms and 470 pF. An isolation network consisting of 5.6-10 Ohms with a decoupling cap (20-50uF) to ground would be placed in the collectors of the pre-driver transistors. This latter newrok is to ensure there can be no coupling between the driver and pre-driver via the supply rail - see Cordell.
5. There are capacitors placed in parallel with the pre-driver base stoppers. Normally, you would not do this as it defeats the purpose of the base stoppers. What you want to do is dampen any oscillatory tendencies at HF with the resistors - the caps are bypassing them.
6. The VAS devices have 3p Miller caps that work in conjunction with 2 caps (can't quite make the value out) that go from the VAS collectors to the JFET device inverting input. You can use this type of compensation, but you have to be very careful about creating an HF zero formed by these caps and the gain setting resistor as this can cause HF gain peaking in the loop gain.
7. The slew rate is specified as 260V/us and rise/fall time of 0.5us. Given the vintage of this amp design and the obvious compensation design anomalies compared to modern designs, the alarm bells are ringing. The 1W 8 Ohm bandwidth is stated as 'DC~500 kHz' which opens the amp up to all sorts of RFI, and should it go into oscillation for whatever reason, these types of BW's don't help.
Since 1979, when this amp was launched, a lot of progress has been made in amp design regarding stability and compensation and its vintage in this respect is pretty evident looking at the schematic.
I suspect some of the things on this amp were done in the interest of prioritising distortion and slew rate over stability.
However, this is a beautiful-looking amp, so it deserves to be fixed/tweaked and take pride of place in someone's system!
I'd start by adding a 10 Ohm (5W) + 0.047uF (250VDC) Zobel on the amplifier side of the output coil (leave the original one in place). Place it on the main amplifier PCB, keep the leads short and return the 0V side of the Zobel to the 0V connection coming onto the PCB - don't run long connections back to the PSU 0V. The second thing I'd do is limit the input BW to 120-150 kHz. I can't read the component values (68o Ohms and 100 pF? If so, the -3 dB BW is 2.3 MHz) but 1k and 1nF would set the -3 dB BW to 160 kHz and clean things up IMV.
Specs:
Some internal pics
1. The Zobel is placed after the output coupling inductor and that's a pretty big no-no in Self and Cordell's books. You do run the risk of HF OPS oscillation with this type of scheme. By placing the Zobel after the output L, you are isolating the OPS from the Zobel. The Zobel is also not placed on the main amplifier PCB (where it should be to keep HF circulating current loops as small as possible), so any wiring L only exacerbates the problem.
2. The OPS is an EF3 if I'm reading the schematic correctly. These are prone to instability unless treated with care - again, Self and Cordell discuss this in-depth (see especially Cordell).
3. There are no values shown for the OP transistor base stoppers, so I assume they are linked out. Again, with an EF3, this is not recommended practice nowadays - typically you would use 2.2 - 10 Ohms. JC used 10 Ohms, I use 4.7 Ohms and both Self and Cordell discuss a range of values in their excellent books.
4. There is a base stopper network in the pre-driver transistors. Normally, you would place these base stoppers and associated base load capacitors in the base circuit of the driver devices. Cordell shows IIRC 10 Ohms with a 470 pF to the rail, but I have seen and used 22 Ohms and 470 pF. An isolation network consisting of 5.6-10 Ohms with a decoupling cap (20-50uF) to ground would be placed in the collectors of the pre-driver transistors. This latter newrok is to ensure there can be no coupling between the driver and pre-driver via the supply rail - see Cordell.
5. There are capacitors placed in parallel with the pre-driver base stoppers. Normally, you would not do this as it defeats the purpose of the base stoppers. What you want to do is dampen any oscillatory tendencies at HF with the resistors - the caps are bypassing them.
6. The VAS devices have 3p Miller caps that work in conjunction with 2 caps (can't quite make the value out) that go from the VAS collectors to the JFET device inverting input. You can use this type of compensation, but you have to be very careful about creating an HF zero formed by these caps and the gain setting resistor as this can cause HF gain peaking in the loop gain.
7. The slew rate is specified as 260V/us and rise/fall time of 0.5us. Given the vintage of this amp design and the obvious compensation design anomalies compared to modern designs, the alarm bells are ringing. The 1W 8 Ohm bandwidth is stated as 'DC~500 kHz' which opens the amp up to all sorts of RFI, and should it go into oscillation for whatever reason, these types of BW's don't help.
Since 1979, when this amp was launched, a lot of progress has been made in amp design regarding stability and compensation and its vintage in this respect is pretty evident looking at the schematic.
I suspect some of the things on this amp were done in the interest of prioritising distortion and slew rate over stability.
However, this is a beautiful-looking amp, so it deserves to be fixed/tweaked and take pride of place in someone's system!
I'd start by adding a 10 Ohm (5W) + 0.047uF (250VDC) Zobel on the amplifier side of the output coil (leave the original one in place). Place it on the main amplifier PCB, keep the leads short and return the 0V side of the Zobel to the 0V connection coming onto the PCB - don't run long connections back to the PSU 0V. The second thing I'd do is limit the input BW to 120-150 kHz. I can't read the component values (68o Ohms and 100 pF? If so, the -3 dB BW is 2.3 MHz) but 1k and 1nF would set the -3 dB BW to 160 kHz and clean things up IMV.
Specs:
Some internal pics
Last edited:
You have to be careful with that. What that does if chosen correctly (and after all other compensation design is completed), is get you a few degrees of extra phase margin. However, if you overcook it, you get loop gain peaking and you can cause problems
I wrote an article on my website talking about how amplifier design progressed between the mud 1960’s when solid state started replacing tubes through to the mid 1990’s when solid state designers got a grip on feedback and compensation. The first solid state amp that showed some knowledge of how to deal with the high levels of feedback solid state amps provided (tube amp loop gains were much, much lower in comparison so never had the problems that became exposed with solid trade amps) was Bart Locanthi’s SA600 when he was at JBL. But, he came from an analog computing background so knew about how to deal with [high] feedback - most transistor amp designers did not.
You can see some of these issues in this Sansui Amp. Huge loop gains, very marginal compensation to try to get high slew rates and low distortion, omitted base stoppers and bypassed base stoppers to try to maximise bandwidth. Compare this design to a modern VFA like the Wolverine to see how much things have changed!
Most signal sources today are bandwidth limited to a shade over 20 kHz. Slew rates need not be higher than 1V/us per peak output volt (see Cordell) and some practioners recommend even less than this. Amplifier bandwidths on modern amps are typically half the Sansui figure.
You can see some of these issues in this Sansui Amp. Huge loop gains, very marginal compensation to try to get high slew rates and low distortion, omitted base stoppers and bypassed base stoppers to try to maximise bandwidth. Compare this design to a modern VFA like the Wolverine to see how much things have changed!
Most signal sources today are bandwidth limited to a shade over 20 kHz. Slew rates need not be higher than 1V/us per peak output volt (see Cordell) and some practioners recommend even less than this. Amplifier bandwidths on modern amps are typically half the Sansui figure.
I believe the diodes form a Baker clamp: as TR13 approaches saturation, D05 diverts base current into the collector and prevents the transistor from saturating and "sticking". Makes for more graceful clipping behavior.