SA-8800 Oscillation?

Yay, the skeleton has been pushed back in the closet and the door is closed. :thumbsup: I entered all the transistors, played around with the simulation to get rid of the schematic errors I made, figured out the value for VR5 to set dc offset in the simulation, adjusted values for the other VRs and It now behaves like an amplifier, though imperfectly. A 1kHz 10 mV input produces a ~500mV p-p output. I think I have reached the limit of my current capabilities with LTSpice. More to learn, but need to pass it on for further examination. Who gets it next? :)
 
I think it is only you and me :) Good to hear you are up and going, pls email me the SA-8800 .asc file and I'll run with it.
 
thanks to Merlyn doing the hard work of entering the schematic into ltspice, we have a working SA-8800 PA in simulation.
In order to get it to run and bias up per the procedure, I had to do some mods.
Using 2 of 1N4148 diodes as a sub for D9(STV2), produced too much bias in the ops, so I shorted one out. In practice not sure what to do in this case.
Ran a DC bias sim, added the bias voltages, some important op currents to use as a reference. They seem to match what Pioneer has on the schematic.
Based on this sim and the parts used, it seems to behave properly. I did bias,THD,freq response/bode plots,square wave tests.
I am still doing some work to determine phase margins, loop gain stuff etc.
It is a finicky design, change one device type and you have to fine tune it again.

In this design, is D9(STV2) attached to the heatsink for the op bjts?

As far as bias procedure goes, they say
With VR9=1,adj VR7 for 56mV across R133,135 (59mA)
With VR7 adjusted, adj VR9 for 70mV across R133,135 (75mA)
That is a lot of bias, it will result in the op devices dissipating around 3W static each, does this make sense, have others experienced these hot devices?
 

Attachments

  • Pioneer SA-8800 NSA PA(ltspice).pdf
    43.6 KB · Views: 14
I looked through this thread, it does not indicate what the bias was measured as.

I figured out that the STV-2(D9,10) attaches to the heatsink.

In the sim, using the components as shown you can not use the STV-2, since using only one 1N4148 type will make it bias up correctly. You can use any bjt as a diode (C,B shorted) if it is easier to mount on the heasink, a TO-126, BD139 should work nicely, I'll try that in sim. It works fine.
I read some threads on diyaudio, others have had the same problem I see in sim, when you change any of the bjt types in the driver,ops, you have to fine tune the bias generator to match. It is a fine tuned circuit, but does have stellar performance once it is in tune.
Using 82pf for C45 gives the best square wave response. this is actually a benefit, since this provides more stability.
So to the thread owner, Qsilver or anyone else in this position, I can suggest to build one channel up as I show in sim and try it out.
 
Last edited:
I'm sorry I haven't been more attentive towards this, my computer went down over the weekend and I've just now got it back roughly to where it was. Though not installed half the programs again yet. As I said before I have four STV-2H's. Two are mounted on the heat sink and two are just on the circuit board in place of two resistors in later designs. My SA-8800 goes back to 1979 I think so one of the first ones. Just wanted to say that in case that alters how the circuit is fine tuned.

I'm happy to go ahead with working balancing the channel. Just let me know what I have to do and I'll get to it.

In the sim, using the components as shown you can not use the STV-2, since using only one 1N4148 type will make it bias up correctly. You can use any bjt as a diode (C,B shorted) if it is easier to mount on the heasink, a TO-126, BD139 should work nicely, I'll try that in sim. It works fine.

Will we need to watch for thermal runaway in this regard?

I looked through this thread, it does not indicate what the bias was measured as.

Originally, the bias was as you mentioned, 56mV, then 70mV with the second adjustment. With these new transistors in place, the idle current at minimum is around the 56-70mV mark without any adjustment IIRC.

So the oscillation's cause is due to changing the Drivers and Outputs? How does dissipation on the drivers look? As these transistors handle less current than the originals. Again, IIRC.
 
Sorry to hear about my comp problems.
Do you have a couple of pics of the ops bjt mounting to the heatsink?
Will we need to watch for thermal runaway in this regard?
yes you need thermal compensation using either the STV-2 or a BD139, mounted on the heatsink, what ever one works to get the bias range adjustments.
They decided to remove the two STV-2 and replace with a 220 ohm R for some reason. I could not get it to bias up correctly with the STV-2, so I used the 220 ohm instead.
I was also looking at the Sa-9900 which is very close to the same design, it uses 430 ohms
.
So the oscillation's cause is due to changing the Drivers and Outputs?
An assumption.
Are your original drivers defective?
You are using what for new ops bjts?
 
The whole reason I started this was because I recapped it and ran it for a few days and then it blew out one of the outputs on the left channel and smoked a resistor.

I removed the outputs, found the dead one and decided to replace all of them at once. After that I had more problems and I discovered I had some big problems with the drivers. They looked very black and when powered off one of the 2SA913's would start to oscillate and breakdown, discharging the main capacitors into the circuit, so I decided they would have to be changed as I thought that might have been the original cause of the outputs being blown.

I've managed to fit the new outputs in the same orientation as the originals with new tapped holes in the heatsink. Though they are not connected at the moment. I used MJL3281A and MJL1302A for the outputs and KSA1381 and KSC3503 for the driver transistors.

I could not get it to bias up correctly with the STV-2, so I used the 220 ohm instead.

Should I replace them with 220 Ohm resistors?

For temperature compensation, you can see the current STV-2H attached to the heatsink on the second picture towards the bottom.
 

Attachments

  • 20180416_182810.jpg
    20180416_182810.jpg
    88.1 KB · Views: 28
  • 20180416_182817.jpg
    20180416_182817.jpg
    72.9 KB · Views: 26
  • 20180416_182832.jpg
    20180416_182832.jpg
    87.3 KB · Views: 29
  • 20180416_182926.jpg
    20180416_182926.jpg
    127.5 KB · Views: 25
  • 20180416_182916.jpg
    20180416_182916.jpg
    114.1 KB · Views: 26
Thx
Should I replace them with 220 Ohm resistors?
That is what I am thinking.
I will sim again with KSA1381 and KSC3503 for the drivers and report back. I do not think that they should be used, only 50mA of Ic. Okay for MOSFET ops but no good for 2EF bjt ops.
 
I think the other alternative I have seen is KSA1220 and KSC2690. They look to be much better in terms of Ic. 1.2A? What have you been simming with up until now? I have those if they're better. Looks like they will be.

EDIT: Just loaded up the PDF you attached on your post. You used MJE15033/2 for drivers? If those check out in terms of being fast enough for the non-switching aspect I have those too?
 
Last edited:
That is correct, some have used KSA1220 and KSC2690 for drivers. I did read someone even using 4MHz devices ksa940/ksc2073.
I have not heard of many being able to bias this amp up properly with these subs.
The original devices are 2sa913,2sc1913, 150V/1A/15W devices in a TO-200 case. I would try MJE15033/2 for drivers, they are fast at 30MHz devices. I did read someone even using 4MHz devices
There are/were faster drivers 2sa1930/2sc5171, but they are now obsolete by Toshiba.
There is also Sanken 2SA1859/2SC4883 which another one to try, Digi-key have them.
The sim runs okay with KSA1381 and KSC3503 for the drivers but they are no good under a heavy load, as I think that they do not have enough Ic rating for this purpose.
 
That's what I thought. I had read that the 2690/3503 and their counterparts were required for the high speed, 150-155MHz. I ruled out the MJE15033/2 due to their lower fT of 30MHz. Does the THD check out when using these? I tried getting some of the Toshiba's but saw they were obsolete and I didn't want to run the risk of counterfeits in an amp like this.

I've got the MJE15033/2's if they're good, I can put those in now along side the other adjustments I need to make?
 
Last edited:
Does the THD check out when using these?
Yes it does, measures about the same with MJE15033/2 & KSA1381/KSC3503
Give MJE15033/2 a try and replace the two STV-2 diodes mounted on the pcb with 220 ohm R's. measure, adjust bias, test for operation using scope on o/p. See how your measurements on the actual compare with the sims.

edit: When using MJE15033/2, change R121 from 470 to 75 ohms, to get more current in the driver stage. Might need a small HS on them if they get too hot to touch.
measure the bias to start with using both VR7, VR9 set to minimum
 
Last edited:
I don't think I've got 75 Ohm to hand. I have 82. Will that do for a test? Actually I'll have to see what I do have a 1/2W rating. 100R @ 1/2W might be the closest I have... Or two 150R 1/4 in paralell?
 
Ok, I tried that. However, I was working on the right channel instead of the left. I used two 150 ohm 1/4W resistors in parallel, replaced the two STV-2H's with 220 Ohm resistors and fitted the MJE15033/2's and reconnected the output transistors.

Powered up, and its still not stable. The Idle current started at 50mV and continued to rise at around 2mV per second until the 100W bulb in the DBT resisted enough to stop it rising. Without it it continued to rise over 70mV at which point I powered it back down.

The left channel is still oscillating much more than the right so perhaps I should make the same changes to that channel too, just to try and calm things down? The Drivers didn't seem to get too hot just yet but were noticeably hotter when I did a quick test before re connecting the outputs to test idle current.

At first, the scope showed very high frequency pulses. My scope was struggling to capture them (faint trace) on the 20kHz-100MHz scale they looked like square waves with a slight ramp down along the top. As I was about to photograph it they disappeared and left what I have photographed. This signal was captured much lower, on the 50-200Hz scale so its certainly slower.

Also worth noting that I removed the 200pF capacitors that I had piggybacked onto the feedback loop.
 

Attachments

  • 20180416_235650.jpg
    20180416_235650.jpg
    47 KB · Views: 7
For now I suggest to work on 1 ch at a time until we sort it out hopefully.
As predicted in the sim, using the STV-2 diode as the thermal FB element, attached to the heatsink, produces to much bias, so it has to be replaced by a single diode when using these bjts as replacements. I used a BD139 in the sim (B/C shorted, using the B/E junction as your diode). You can use a KSC2690 or KSC3503 instead. If you attach it to the place where the STV2 was mounted, you have to insulate/thermal comp. the tab from the heatsink ground. I guess you could just use the opposite of the metal tab side to start with, so no insulation is necessary.
Doug Self in his design books suggested to mount the thermal element on top of one of the o/p bjts to have a fast thermal feed back response. Try to keep these leads as short as possible.
Good luck
Rick
 
I also simulated the SA-9900, I will post the schematic here so you can see the small differences that they made.
 

Attachments

  • Pioneer SA-9800 NSA PA (ltspice).pdf
    45.1 KB · Views: 15
Yes, I agree, I'll make the same changes to Channel 1 (Left) and then concentrate on that channel for now. Channel 1 is the worst of both of the channels and the one that originally blew out.

I think I'll mount the KSC3503 to the same spiot as the STV-2H for now but possibly move it over to an output if I can manage it. As far as I can see, both the KSC3503 and the KSC2690 are fully insulated transistors so no insulator/thermal comp is needed? Correct me if I'm wrong. So I connect B & C together and pass current through B to E?

E would be the cathode end of the STV-2H?
 
As far as I can see, both the KSC3503 and the KSC2690 are fully insulated transistors so no insulator/thermal comp is needed?
Good no insulator req, only a small amount of paste. I mixed up with TO-225 that has the tab. Play with re-location after if we get this sorted out.
I think I'll mount the KSC3503 to the same spot as the STV-2H
Easiest way for now.
Correct me if I'm wrong. So I connect B & C together and pass current through B to E?
That is correct, make a diode out of the BE junction. B/C is the anode and E is the cathode. Follow the ltspice schematic as drawn. base goes to Q19,23, emitter goes to VR7 pot.
 
Ok, I've redone channel 1, Drivers changed to MJE15033/32, 470 Ohm 1/2W resistor changed to 75 Ohm 1/2W (2x 150 Ohm 1/4W for testing purposes), All STV-2H's removed. Two onboard changed to 220 Ohm resistors, temperature compensation changed to the KSC3503 with B/C Short.

It's stable! No oscillation that I can see at the moment and idle current is sitting at 0.05mV and very slowly rising to 1mV.

I replicated this to channel 2 and it matches channel 1's behaviour. So what's next? Should I try to bring up the idle current on channel 1?
 
Last edited:
Back
Top Bottom