SA-8800 Oscillation?

yes a ceramic disc cap is fine. instability is usually caused by too much phase delay in the response. hard to say how stabile the design is without measuring the performance. changing comps can cause this instability. have to experiment/simulate the design in order to see whats going on. I doubt that it is the power supplies if they are stable and noise free.
 
I got around to looking at this again. I put in the extra capacitance on both amplifier channels at the same time. I tried putting in 10pF extra with no change. Added another 10pF and still no change. I then removed these and, maybe went a little over the top, but I then added 200pF just to see if this would make any difference. It did. I can see clear wave forms on the scope now, although they do look quite strange... the left channel has quite a large waveform, where the right channel is much smaller.

I've uploaded a photo of the waveform at the output terminal of the left channel and one of the fluroscan which is showing the left VU illuminated.

The right channel waveform looks to be similar, just smaller. Any ideas? I'm starting to wonder if I've put something in backwards in the NSA circuit...?
 

Attachments

  • 20180324_191156.jpg
    20180324_191156.jpg
    56.2 KB · Views: 31
  • 20180324_191222.jpg
    20180324_191222.jpg
    34.3 KB · Views: 30
I think we need to simulate this design to see how it works. Sometimes these designs are on the verge of instability, one small change throws them off into oscillations. Sims can help uncover the phase margins involved.
If you are so inclined, try to input the schematic into ltspice. I can help you run the sims. The effort is to input the schematic. I have never sim'd a NSA design, now is a good time to do so.
Using the sub components that you have used, we have good models for them. We can determine what is going on. Other than that we are just guessing. There are number of compensation methods going on in this design.
I think it is just you and me, everyone else it seems to have lost interest or can not offer any further advise.
 
Last edited:
I've got ltspice but I've not used it too much. I will be able to put the design in. I guess there is no quick way, I'll just have to "draw" it in? I can do that and I can get hold of the transistor model data etc but when it comes to simulating I'm lost. Should I try and put in the whole NSA Amp board or just one amp channel? My gut says the whole thing...

On a side note, I've seen a few people, such as Leesonic re-do these amps when they blow with the same components that I've used and the amp just works.... that's what threw me.

I also have a spreadsheet somewhere with all the ceramic caps listed... I think I used conditional formatting to show when they were over their 20% margin and a number of the ones I tested seem to be getting dangerously close.. I'll try to dig it back out...
 
If you are so inclined, try to input the schematic into ltspice. I can help you run the sims. The effort is to input the schematic. I have never sim'd a NSA design, now is a good time to do so.
I am working on learning ltspice, but cannot find a potentiometer symbol. What do you use?
Thanks!

EDIT: also what do you use for the STV-xxx diodes?
 
Last edited:
That is right, no quick way to enter a schematic, just enter it, slow and surely. Once you get the hang of the quirky way it works, then you get better at it.
Only one channel necessary, it is good to match the ref designators as shown on one of the channels, does not matter which one, if they are identical. R?,C?,Q?,D? etc to match the one channel so we can refer them. Do not worry about polarized ecaps either, since first runs we use perfect caps with no losses, you can refine the caps later if necessary.
This NSA amp is different, it is much faster, doing so, it leads to stability problems when you alternate comps and do not re-adjust (tune) the compensation. We shall see how it behaves in sim.
I do not know if 20% margins on caps will make or break this design, we can play with that in the sim.
I do not use a pot but make up the equivalent of the pot value used, by using a voltage divider made of 2 fixed resistors. You run a ".op" sim to calculate the bias levels, adjust the voltage divider ratio to get the bias you need, through a few trial runs.
In the case of the pot in the ops bias tree, I just use one value, since the pot is usually 100R, I just chose a value between 1-100 that sets the bias in the ops to where I want it to be.
I use some 1n4148 diodes in a series string for a STV diode array, it is all you need, since there is no thermal feedback modeling involved. If someone wants to get advanced when we could model the STV diodes based on the data sheet, but it really is not needed for these simple sims at this time.
Good luck guys, I'll fire the starters pistil and see who wins the race, :) take your time, it ain't a race. Once you are done I can look it over and we can work together on how to run the sims. BTW I got better at this when I read Bob Cordells PA design book and he had a great tutorial. Also from members at diyaudio. Lookup the group thread devoted to ltspice on diyaudio, it is excellent, with folks knowing way more than I know there to help us out.
Basically build off what I have supplied and modify it to match the NSA design.

Cheers
 
That is right, no quick way to enter a schematic, just enter it, slow and surely. Once you get the hang of the quirky way it works, then you get better at it.
Only one channel necessary, it is good to match the ref designators as shown on one of the channels, does not matter which one, if they are identical. R?,C?,Q?,D? etc to match the one channel so we can refer them. Do not worry about polarized ecaps either, since first runs we use perfect caps with no losses, you can refine the caps later if necessary.
This NSA amp is different, it is much faster, doing so, it leads to stability problems when you alternate comps and do not re-adjust (tune) the compensation. We shall see how it behaves in sim.
I do not know if 20% margins on caps will make or break this design, we can play with that in the sim.
I do not use a pot but make up the equivalent of the pot value used, by using a voltage divider made of 2 fixed resistors. You run a ".op" sim to calculate the bias levels, adjust the voltage divider ratio to get the bias you need, through a few trial runs.
In the case of the pot in the ops bias tree, I just use one value, since the pot is usually 100R, I just chose a value between 1-100 that sets the bias in the ops to where I want it to be.
I use some 1n4148 diodes in a series string for a STV diode array, it is all you need, since there is no thermal feedback modeling involved. If someone wants to get advanced when we could model the STV diodes based on the data sheet, but it really is not needed for these simple sims at this time.
Good luck guys, I'll fire the starters pistil and see who wins the race, :) take your time, it ain't a race. Once you are done I can look it over and we can work together on how to run the sims. BTW I got better at this when I read Bob Cordells PA design book and he had a great tutorial. Also from members at diyaudio. Lookup the group thread devoted to ltspice on diyaudio, it is excellent, with folks knowing way more than I know there to help us out.
Basically build off what I have supplied and modify it to match the NSA design.

Cheers
I have the component symbols entered for the power amp, GWH-115 (left channel only) and am in the process of entering designators and values. This part of the process is much slower than entering the symbols. I am not adding any of the "protection" circuitry components as they do not contribute to the function of the amplifier.
 
Last edited:
Good stuff, the wiring is also time consuming.
Wiring was done concurrent with the component symbols. I found that doing the designators and values as a second pass helped me locate errors I had made, mostly npn/pnp mix ups. It has also been a great way to review and learn more about the circuit's functions :D.
 
Wow I'm stunned. I'm still trying to find time to start it and you sound like you near the end of it! Are you plotting it out as it is in the schematic? My version of the SA-8800 has a couple notable difference from the schematic. It has an extra pair of STV-2H diodes in place of two resistors... I'll double check where they are and get back to you.
 
Wow I'm stunned. I'm still trying to find time to start it and you sound like you near the end of it! Are you plotting it out as it is in the schematic? My version of the SA-8800 has a couple notable difference from the schematic. It has an extra pair of STV-2H diodes in place of two resistors... I'll double check where they are and get back to you.
I would say about half way through the process, as entering the designators and values is MUCH slower than placing components and wiring. But it is a learning experience I am enjoying :). I am laying it out much like the schematic, except where that makes no sense or is impossible. I may find some of those skeletons referred to in your signature :rolleyes:.
 
OK, the skeleton has popped out of the closet :( . None of the(substituted) transistors listed in Post 12 are in LTSpice.
What do I use and/or where do I get the data models?:dunno:
I have the simulation schematic done except for transistors :confused:. What's next?
 
OK, the skeleton has popped out of the closet :( . None of the(substituted) transistors listed in Post 12 are in LTSpice.
What do I use and/or where do I get the data models?:dunno:
I have the simulation schematic done except for transistors :confused:. What's next?

You should be able to find simulation models for many of the replacement transistors on the Onsemi web site.

Les.
 
Ok, I think these are the models - I've attached them to the post.

Also, I mentioned the STV-2H variation in my amp earlier on...
My board has what I think are FOUR STV-2H's. Where R83, 84, 85 and 86 are located, I have STV-2H - its also marked on the silkscreen as being the STV.

So that's where the difference is.
 

Attachments

  • Transistors.zip
    3.7 KB · Views: 13
I have never sim'd a NSA design, now is a good time to do so.

Excellent thread and keep us up to date. I'm fighting with LTspice as we speak and would be very interested to see how this all pans out. Once NSA went into a PAxxxx chip it all became a big mystery.
 
I provided you with the models, maybe you have either not put them in the correct folder in the lib or that you are having a difficult finding them, do to the naming of them. Use the same models as used in the SX-780 and CA-800 designs.
You should be able to place the mouse over the comp symbol,bjt, RMB, pick a new transistor, the bjt lib show appear with the models I provided. if you do not see the models I provided like ksa992, then the models are not installed correctly, you have to recursively change permission to be owned by you in order to edit the libs. You can access the libs from within ltspice, file,open, navigate to lib folder, select file viewing as discretes, which include bjt, ... etc
If you are having trouble defining the models, do not worry, I can do that portion as a last step, but we have to be using the same models for the sim runs.

As for John, explain what you are having issues with? maybe I can help out.
 
Last edited:
I provided you with the models, maybe you have either not put them in the correct folder in the lib or that you are having a difficult finding them, do to the naming of them. Use the same models as used in the SX-780 and CA-800 designs.
Will the "standard" files you sent work with LTSpiceXVII, since you are using an older version? I did find the location in my version . .
 
Back
Top Bottom