I have a suspicion that the Sanken output devices may be dodgy, as from all accounts none of the Chinese ebay ones are original parts, those being obsolete. I've ordered some from Profusion as they have new old stock of the real thing.
I had the same problem once years ago with a couple of danish LC-Audio modules "The End MK2". On each channel, a 470nF from signal ground to heatsink solved the problem 🙂
You might try it from either signal ground or power ground.
You might try it from either signal ground or power ground.
Last edited:
round numbers:
assume that the area of the collector pad is 0.5 square inch.
assume a mica insulator, with relative permeability of 3
assume the insulator is 0.002 inches thick.
capacitance would be 169 pF from the collector to the heatsink.
That looks like -j473 Ohms at 2 MHz.
It's a pretty high Q cap I'd imagine...I wonder if that's enough to encourage oscillation when it goes to ground?
interesting experiment---what if you ground the heatsink thru a few values of C...
150 pF - kind of cuts total in half...
1 nF - lets most of it be there in all its glory
10 nF - equivalent of a short...
all these might be hints about root cause...
assume that the area of the collector pad is 0.5 square inch.
assume a mica insulator, with relative permeability of 3
assume the insulator is 0.002 inches thick.
capacitance would be 169 pF from the collector to the heatsink.
That looks like -j473 Ohms at 2 MHz.
It's a pretty high Q cap I'd imagine...I wonder if that's enough to encourage oscillation when it goes to ground?
interesting experiment---what if you ground the heatsink thru a few values of C...
150 pF - kind of cuts total in half...
1 nF - lets most of it be there in all its glory
10 nF - equivalent of a short...
all these might be hints about root cause...
These would be series elements for a phase correction network drawn in the wrong place. They should link the collector of Q13 to the negative supply.
I can only add that I have a few of those amps, but did not have oscillation problems. They measure better with the large Sankens too, compared to other outputs I tried, so I suggest you try to fix it without swapping outputs.
djoffe, interesting experiment! If that is actually an issue, why is it not widespread and well known to cause problems? Which takes me to Rallyfinnen's comment that he's built the same amps with no issues. But Rally, did you ground the heatsinks? Also there are at least two, possibly more versions of the pcb.
I'll continue to tinker with it and report back when I have success! Thanks for all the help folks.
I'll continue to tinker with it and report back when I have success! Thanks for all the help folks.
Always a chance chinese parts are fake. However the schematic you posted shows CFP (Sziklai pair) output stage. These can have stability issues. If you look at designs like ESP P3A there is an extra capacitor (100pf) on the negative rail drive transistor between base/collector to aid stability.
This isn't the first LJM design that has been intentionally confused to make it difficult for pirate copying, which is rife at Chinese on-line selling platforms. Take a look at the L12-2 thread for a really crazy effort at confusion.
I would have thought it was clear from earlier comments that this isn't one of Doug's designs - just the general topology is similar. It's still a very successful and probably the most popular of LJM's designs but just inspired by the model designs featured in D. Self's articles, web pages and books.
I would have thought it was clear from earlier comments that this isn't one of Doug's designs - just the general topology is similar. It's still a very successful and probably the most popular of LJM's designs but just inspired by the model designs featured in D. Self's articles, web pages and books.
The heat sinks should be be in electrical contact with the case. So should the safety earth and the zero volts connection of the power supply to keep these at the same potential.
There could be an anomaly in the way you have wired the input and earth connections and it would be helpful to see images of the respective layouts.
mjona, I've tried various grounding scenarios, none solve the issue. My usual system is to use a 10r resistor in parallel with a 0.1uF cap between mains earthed chassis and signal ground and a bridge rectifier as a fault path, with the lifted signal ground and chassis earth point at the PSU capacitor common point. At present I have the heat sink floating and everything works fine. If I connect to any grounding point directly or through a capacitor or resistor the problem starts up. If I pull the output devices away from the heat sink the oscillation stops. The harder I press the devices to the grounded heat sink the worse the oscillation gets. I've used mica washers and paste. So either the Sanken devices are not suited to this circuit design (some versions are sold with different devices) or the Chinese Sankens are dodgy. I will try some small capacitors in strategic points as suggested elsewhere to see if that helps. Cheers. Matt
With the small capacitors in strategic places, if you look at the Cob rating for the devices you are using, the 2SA1295 has double the capacitance (Cob) compared with the 2SC3264, I do not understand why (I am lazy with this stuff if it just works, why reinvent the wheel), but the CFP output stage seems sensitive to this imbalance and so adding some additional capacitance can help.
If the devices are fake you will be chasing a red herring, but useful to look at....
If the devices are fake you will be chasing a red herring, but useful to look at....
I said in my post that it could be a capacitance effect and your comment on "pressing the devices hard to the heat-sink makes it worse " just backs up what I said it would mean faulty devices .
Adding small capacitance's is just using the distributed compensation method it might work there again it might not but watch out you don't change the frequency response .
The answer really is to find the actual cause not "sticking plaster it ".
Adding small capacitance's is just using the distributed compensation method it might work there again it might not but watch out you don't change the frequency response .
The answer really is to find the actual cause not "sticking plaster it ".
I had a similar problem with a high voltage amp long ago. Going to a thicker ceramic insulator under the devices, that lowered the capacitance, solved it. IMO, it's still a design flaw of some sort.
2SA and 2SC output transistors have higher fT than 2SB and 2SD types used in the second circuit. That implies an increase in gain at high frequencies in the region where out phase shift in the return signal from a reactive load is likely. I don't see any output coil in that circuit.
Few years back I did play with "improving" the P3A amp mentioned earlier (simulations only). Changing bootstrapped VAS to active current source, and using a current mirror in the LTP - ended up very similar to this design.
In simulations the original P3A is stable, but with those two changes the design seemed inherently unstable. You might find changing the current mirrors in the LTP back to resistors (whilst technically inferior) helps with stability by reducing gain in the LTP a bit.
Have found that when playing with other amplifier designs, "improving" with a current mirror - which will give higher gain in the LTP, brings instability with it.
But some people design like that, as one high end manufacturer will claim, things get more interesting on the edge of stability...
In simulations the original P3A is stable, but with those two changes the design seemed inherently unstable. You might find changing the current mirrors in the LTP back to resistors (whilst technically inferior) helps with stability by reducing gain in the LTP a bit.
Have found that when playing with other amplifier designs, "improving" with a current mirror - which will give higher gain in the LTP, brings instability with it.
But some people design like that, as one high end manufacturer will claim, things get more interesting on the edge of stability...
That circuit is 90% a DS design I have seen it before and that includes the current mirror that's why I said I was surprised it was unstable as DS went to great lengths to prove otherwise and actually challenged other audio designers to fault it .
He would be a "might unhappy " if his designs were attacked and went to very great lengths to show how wrong they were.
Its also a case of anybody modifying a well know designers basic circuit changes the stability of the design by not keeping to the recommended PCB layout ,as I use variations on JLH designs it took me a while to get stability ---without a load of compensation capacitors ( something JLH wasn't keen on ).
I achieved this by changing the layout of the components without extra small value capacitors and sticking with JLH,s version of compensation.
D.Self uses his well known 100pf capacitor which, in his circuits proved to be more than enough to stop any oscillation , others use different methods but when he brought out his "Blameless " Design and a kit for it ,it was quickly bought by other design engineers --does that not tell you something ?-- high praise indeed !
As Mjona says where is the output/coil parallel resistor ? -D.self designs incorporate those components, does the person modifying the D.Self design have enough confidence his design stability exceeds D.Self,s ?
He would be a "might unhappy " if his designs were attacked and went to very great lengths to show how wrong they were.
Its also a case of anybody modifying a well know designers basic circuit changes the stability of the design by not keeping to the recommended PCB layout ,as I use variations on JLH designs it took me a while to get stability ---without a load of compensation capacitors ( something JLH wasn't keen on ).
I achieved this by changing the layout of the components without extra small value capacitors and sticking with JLH,s version of compensation.
D.Self uses his well known 100pf capacitor which, in his circuits proved to be more than enough to stop any oscillation , others use different methods but when he brought out his "Blameless " Design and a kit for it ,it was quickly bought by other design engineers --does that not tell you something ?-- high praise indeed !
As Mjona says where is the output/coil parallel resistor ? -D.self designs incorporate those components, does the person modifying the D.Self design have enough confidence his design stability exceeds D.Self,s ?
Hi . Some idea for test ... Try to ground heatsink through inductor ,made from solid wire , like output inductors in other amps. Will amp oscilate or not ?
Many years ago I had similar strange problem that manifested only when heatsink was grounded. It came out a bad insulation of the temperature sensing transistor.
Last edited:
I connected a 100pf cap on neg driver cb, that cured that half of the cycle, so I did the same to the other rail driver and the problem is solved! I looked at the 1KHz square wave performance and that looked pretty much the same as before. However when I connect a 1uF cap across the load there is a lot of overshoot and ringing. It does damp out and does not cause instability though. I also tried putting an inductor//resistor in series with the output, this doesn't make any visible difference on the scope. I have a feeling that, as suggested this design was finessed for slower output devices than the ones I've got. Thanks for all the input folks.
You really need to include an output inductor with a damping resistor - and especially with a CFP stage.
The reason your amp oscillated when you grounded the h/sink is that you treated a capacitance to ground from the VAS output. That’s why when you pressed on it it got worse as the capacitance increased.
Probable that the amp was marginally stable and the added capacitance pushed it over the edge. Are you sure the Chomsky cap marked 101 Is not 10 pF instead of 100 pF?
The reason your amp oscillated when you grounded the h/sink is that you treated a capacitance to ground from the VAS output. That’s why when you pressed on it it got worse as the capacitance increased.
Probable that the amp was marginally stable and the added capacitance pushed it over the edge. Are you sure the Chomsky cap marked 101 Is not 10 pF instead of 100 pF?