Hi, this is my first post here so go easy!
I was hoping someone might be able to assist me with a biasing issue I'm having with my Sony TA-FB920r integrated amp. The left channel died a few years ago so it went into storage as I don't like binning things that could be fixed. I read a lot about common issues with these amps and seemingly bias drift can lead to the output stages failing. Sure enough, both output MOSFETS dead short, amp stuck in Protect mode. I removed both dead devices which cleared the protect mode so I then removed and tested transistors and other components in the various stages prior to the power amp, replaced the o/c gate resistors and finally the pair of MOSFETS.
The channel is now working but I cannot get any bias voltage across the bias test points, it just sits at zero Vdc at any point on the trimmer. I don't want to leave it as is and blow another set of (expensive!) output devices so I must have missed something? Any help would be greatly appreciated 👍
I was hoping someone might be able to assist me with a biasing issue I'm having with my Sony TA-FB920r integrated amp. The left channel died a few years ago so it went into storage as I don't like binning things that could be fixed. I read a lot about common issues with these amps and seemingly bias drift can lead to the output stages failing. Sure enough, both output MOSFETS dead short, amp stuck in Protect mode. I removed both dead devices which cleared the protect mode so I then removed and tested transistors and other components in the various stages prior to the power amp, replaced the o/c gate resistors and finally the pair of MOSFETS.
The channel is now working but I cannot get any bias voltage across the bias test points, it just sits at zero Vdc at any point on the trimmer. I don't want to leave it as is and blow another set of (expensive!) output devices so I must have missed something? Any help would be greatly appreciated 👍
Hi Terry,
Welcome to the forum!
I assume the amp output re ground is about 0V? I understand the bias at the test points is 0mv. As you vary the bias trimmer, do you observe any voltage across C508? Across R519? Please let us know what find. You replaced Q5111 and Q512? Did you replace with the same types? Where did you acquire them?
Good luck!
Welcome to the forum!
I assume the amp output re ground is about 0V? I understand the bias at the test points is 0mv. As you vary the bias trimmer, do you observe any voltage across C508? Across R519? Please let us know what find. You replaced Q5111 and Q512? Did you replace with the same types? Where did you acquire them?
Good luck!
Thank you so much for your reply; I carried out measurements as you've described: the voltage across R519 was 2.5Vdc trimmer at zero, rising to about 3Vdc with the trimmer at midway. For C508, the voltage across it was 3.9Vdc trimmer at zero rising to about 4.4Vdc trimmer in the mid-position.
I actually purchased the output devices some years ago but life got in the way of ever getting around to doing anything! I believe I bought them from a company called 'littlediode' here in the UK.
I couldn't measure anything above a few millivolts of DC on the speaker output terminals if that's what you were getting at?
Many thanks again!
I actually purchased the output devices some years ago but life got in the way of ever getting around to doing anything! I believe I bought them from a company called 'littlediode' here in the UK.
I couldn't measure anything above a few millivolts of DC on the speaker output terminals if that's what you were getting at?
Many thanks again!
You thinking what I'm thinking...
I wonder if these are not not genuine devices and are in fact regular HEXFET's remarked.
Ah, ok, the plot thickens! I'd better pull them and do a bit of testing.
If they do turn out to be fakes then I must have been wrong about buying them from littlediode, they're a proper bona fide component supplier I've used many times before.
Do the measurements more or less confirm to you the o/p devices are probably the culprit?
If they do turn out to be fakes then I must have been wrong about buying them from littlediode, they're a proper bona fide component supplier I've used many times before.
Do the measurements more or less confirm to you the o/p devices are probably the culprit?
The fact you say it is working (playing OK) does suggest an issue with the FET's. The common and cheaper HEXFET needs about 4 volts between gate and source to conduct That is per FET so you would need around 8 volts in total across R519 to achieve the point where they begin to conduct.
The other big problem with the HEXFET is that you would suffer thermal runaway without a temperature compensated bias generator (Q508 in contact with the heatsink) if you did bias them to a suitable value.
As it stands and if this is what has happened then thermal runaway won't be an issue at this point because you are nowhere near the point where they would conduct. So the amp is safe in that regard. One unknown of course (again assuming this is the cause) is not knowing what they actually are.
The other big problem with the HEXFET is that you would suffer thermal runaway without a temperature compensated bias generator (Q508 in contact with the heatsink) if you did bias them to a suitable value.
As it stands and if this is what has happened then thermal runaway won't be an issue at this point because you are nowhere near the point where they would conduct. So the amp is safe in that regard. One unknown of course (again assuming this is the cause) is not knowing what they actually are.
I'll do my best to see if I can identify the devices, my best bet might be to remove the devices on the working channel and use my basic automatic component tester (surprisingly good!) to see what it comes up with. It would give me the opportunity to renew the thermal paste anyway.
I've done a bit of research and the original FETs are quite rare now and by no means cheap!
I also looked at issues with Sony's FA-30ES model as it's essentially identical to mine but adds an MC phono stage. I found a German webpage describing a repair using more easily available IRFP240 and IRFP9240 devices in place of the original FETs, the author describes only having to adjust the values of a couple of resistors in the bias circuit. Sadly he doesn't specify which and the new values and it's a bit beyond my knowledge. Do you think this is something worth pursuing or should I bite the bullet and buy original output devices? I'm a bit worried I might miss something, cook them and end up £30 lighter!
Here's a link to the FA-30ES repair site:
https://old-fidelity-forum.de/thread-23268.html
I've done a bit of research and the original FETs are quite rare now and by no means cheap!
I also looked at issues with Sony's FA-30ES model as it's essentially identical to mine but adds an MC phono stage. I found a German webpage describing a repair using more easily available IRFP240 and IRFP9240 devices in place of the original FETs, the author describes only having to adjust the values of a couple of resistors in the bias circuit. Sadly he doesn't specify which and the new values and it's a bit beyond my knowledge. Do you think this is something worth pursuing or should I bite the bullet and buy original output devices? I'm a bit worried I might miss something, cook them and end up £30 lighter!
Here's a link to the FA-30ES repair site:
https://old-fidelity-forum.de/thread-23268.html
It is easy to tweak the bias generator but my worry would be thermal stability. Now if you deliberately run at a low bias then that might not be an issue.
I'll take a look at the circuit and its bias generator a little later.
The big giveaway would be a high Vgs (gate source voltage) needed to get the FET's to conduct compared to the originals.
I'll take a look at the circuit and its bias generator a little later.
To get more bias voltage across R519 you would need to increase R512.
2k2 may be high enough or possibly 2k7. That will allow the voltage across R519 to reach the levels needed to bias common vertical FET's. 10mv across the test points is 100 milliamp bias current.
This is where it gets tricky because the bias current will tend to rise as the transistors warm and what you absolutely do not want is a situation where it starts to runaway drawing ever more current. Normally the bias generator transistor Q508 would be in thermal contact with the heatsink to compensate and reduce the bias voltage as the temperature rises. If you wanted to try with no thermal compensation then I would recommend setting the current lower so you have say just 2 millivolts across the test points. Then monitor the amp and get it hot (play loud) and check the current again.
Always always start with the preset turned to give minimum bias current.
I would also always recommend using a dbt (dim bulb tester) while initially experimenting to save any major disasters happening.
2k2 may be high enough or possibly 2k7. That will allow the voltage across R519 to reach the levels needed to bias common vertical FET's. 10mv across the test points is 100 milliamp bias current.
This is where it gets tricky because the bias current will tend to rise as the transistors warm and what you absolutely do not want is a situation where it starts to runaway drawing ever more current. Normally the bias generator transistor Q508 would be in thermal contact with the heatsink to compensate and reduce the bias voltage as the temperature rises. If you wanted to try with no thermal compensation then I would recommend setting the current lower so you have say just 2 millivolts across the test points. Then monitor the amp and get it hot (play loud) and check the current again.
Always always start with the preset turned to give minimum bias current.
I would also always recommend using a dbt (dim bulb tester) while initially experimenting to save any major disasters happening.
That's brilliant, thank you for your reply and time.
I have been using a dbt throughout the process, it made sense to make one as it's so simple but effective.
That's very interesting what you have said re. the bias generator transistor, as you point out, it's on the main board rather than on the heatsink. Is this a design oversight then? Would mounting it on the heatsink with flying leads and a spring clamp be beneficial? Drilling and tapping an M3 hole in the heatsink would be straightforward and there's plenty of room.
I'm torn between modding and just buying original devices as I have to accept it has very little intrinsic value, I am enjoying the tinkering aspect and learning a bit more about audio repairs, so thank you for that!
I have been using a dbt throughout the process, it made sense to make one as it's so simple but effective.
That's very interesting what you have said re. the bias generator transistor, as you point out, it's on the main board rather than on the heatsink. Is this a design oversight then? Would mounting it on the heatsink with flying leads and a spring clamp be beneficial? Drilling and tapping an M3 hole in the heatsink would be straightforward and there's plenty of room.
I'm torn between modding and just buying original devices as I have to accept it has very little intrinsic value, I am enjoying the tinkering aspect and learning a bit more about audio repairs, so thank you for that!
Q508 in itself is not critical and if you are happy drilling holes then you could use a small flatpak such a TO126 outline. Something like a BD131 insulated from the heatsink of course. Lots of options.
I think I'd be inclined first of all to just get it working at low bias and see how it fares. Why not just try increasing that resistor first and see at what point the FET's you fitted start to conduct. The reason the transistor is not already on the heatsink is because the original FET's are a different technology and intrinsically have a negative temperature coefficient meaning the current does not increase as they get hot, the opposite of the common HEXFET's
As it stands now there is zero bias current but the big question is how does it sound? There will be some crossover distortion because of the zero bias but is it audible?
I think I'd be inclined first of all to just get it working at low bias and see how it fares. Why not just try increasing that resistor first and see at what point the FET's you fitted start to conduct. The reason the transistor is not already on the heatsink is because the original FET's are a different technology and intrinsically have a negative temperature coefficient meaning the current does not increase as they get hot, the opposite of the common HEXFET's
As it stands now there is zero bias current but the big question is how does it sound? There will be some crossover distortion because of the zero bias but is it audible?
Thank you once again for all of your help, I took your advice and just had a listen and to my ears at least, it's pretty decent. I can't hear any channel imbalance or obvious distortion.
I thought I'd do something else before adjusting the value of R512 as you've suggested, I've hooked up an 8 ohm 100 watt resistor to the speaker output and I'm feeding in a 1kHz sine wave, I can't see any obvious distortion on the trace but it's only a very basic 'scope. The undamaged channel produces very slightly more voltage but the traces look the same to me.
The temperature of the unknown output devices is about 65C at the onset of clipping (around 17V p-p, still supplied through the dbt)
The voltage at the bias test points climbs to about 85mV at pre-clipping onset, so 850mA as measured across a 0.1ohm resistor in the circuit if I'm understanding right?
Shall I leave as-is do you think or is it worth carrying out the bias mod?
I thought I'd do something else before adjusting the value of R512 as you've suggested, I've hooked up an 8 ohm 100 watt resistor to the speaker output and I'm feeding in a 1kHz sine wave, I can't see any obvious distortion on the trace but it's only a very basic 'scope. The undamaged channel produces very slightly more voltage but the traces look the same to me.
The temperature of the unknown output devices is about 65C at the onset of clipping (around 17V p-p, still supplied through the dbt)
The voltage at the bias test points climbs to about 85mV at pre-clipping onset, so 850mA as measured across a 0.1ohm resistor in the circuit if I'm understanding right?
Shall I leave as-is do you think or is it worth carrying out the bias mod?
The voltage across the test points under load is just a reflection of the load current and so does not tell anything about the quiescent conditions.
It is worth doing the test with the 1kHz sine again but this time set the amplitude really really low, so just millivolts of signal across the load resistor. That is where the distortion would show.
This is just a simulation running zero bias. Look how low the signal is across the load. The distortion is always at the zero crossing point. Higher frequencies will show it better as well such as 5 or 10kHz.
You can also here surprisingly low levels of distortion on a sine (it must be a pure sine though, a function generator sine isn't good enough) and you will hear it as a rough edge to the tone. Again listen at the lowest possible levels to hear it.
Its up to you on the mod 🙂 If it were mine I would be wondering what the replacement parts might actually be and also testing them a bit more to be 100% sure they behave like HEXFET's. If they are fake then they could be anything, all bets are off.
Small differences in the two channels are more likely due to differences in volume and balance control matching as much as anything else.
It is worth doing the test with the 1kHz sine again but this time set the amplitude really really low, so just millivolts of signal across the load resistor. That is where the distortion would show.
This is just a simulation running zero bias. Look how low the signal is across the load. The distortion is always at the zero crossing point. Higher frequencies will show it better as well such as 5 or 10kHz.
You can also here surprisingly low levels of distortion on a sine (it must be a pure sine though, a function generator sine isn't good enough) and you will hear it as a rough edge to the tone. Again listen at the lowest possible levels to hear it.
Its up to you on the mod 🙂 If it were mine I would be wondering what the replacement parts might actually be and also testing them a bit more to be 100% sure they behave like HEXFET's. If they are fake then they could be anything, all bets are off.
Small differences in the two channels are more likely due to differences in volume and balance control matching as much as anything else.
Hmmmm, excellent food for thought! I've had to use the sine wave tracks on an Alan Parsons test cd for my source playing through my Fiio dap. I don't have any other way to produce a sine wave except an android app! I did try to monitor the wave on my 'scope at very low amplitude but it's not the best and doesn't trigger reliably until the amplitude is increased (same for both channels).
I'm actually erring on the side of getting the IRP240 and IRFP9240 pair as I can find them for less than £10 for both. At least that way I'll have known devices as I've not been able to get anything meaningful from my admittedly rudimentary testing of the existing pair. Do you think this might be the most sensible approach? It looks like the chap on the German forum had some success so it's certainly an option. Just for clarity; the IRP/FP devices need a higher gate voltage to conduct which I can achieve by changing the value of R512 - I think I'll do this anyway as I'm 99.9% sure my existing replacement devices are indeed fake so at least I can carry out some careful experiments whilst waiting on the new known devices to arrive.
Just out of interest, please could you explain the little bias circuit to me as I can't see how it can ever produce more than +/- 2.6V on the bases of Q509 & Q510 as that's the limit of the supply voltage isn't it?
Apologies for all the questions and possibly some stupid ones and many thanks once again
I'm actually erring on the side of getting the IRP240 and IRFP9240 pair as I can find them for less than £10 for both. At least that way I'll have known devices as I've not been able to get anything meaningful from my admittedly rudimentary testing of the existing pair. Do you think this might be the most sensible approach? It looks like the chap on the German forum had some success so it's certainly an option. Just for clarity; the IRP/FP devices need a higher gate voltage to conduct which I can achieve by changing the value of R512 - I think I'll do this anyway as I'm 99.9% sure my existing replacement devices are indeed fake so at least I can carry out some careful experiments whilst waiting on the new known devices to arrive.
Just out of interest, please could you explain the little bias circuit to me as I can't see how it can ever produce more than +/- 2.6V on the bases of Q509 & Q510 as that's the limit of the supply voltage isn't it?
Apologies for all the questions and possibly some stupid ones and many thanks once again
The -/+2.6 volts are not a supply, added together (so 5.2 volts in total) it is the voltage dropped across the bias transistor. That voltage is determined by hard the transistor conducts and that is determined by the two resistors to the base. The curent flowing in Q508 is a constant and determined by Q507 which is a constant current generator. Fully conducting and the bias voltage (across Q508) is close to zero and then as we turn it off the voltage will get ever higher. We could go nearly as high as the main rails (so -/+44 v across Q508 ) with it fully off. That would let the magic smoke out 😉
You could also for example replace all that bit of circuitry with a single preset (although there would be no temperature compensation at all and the current is a little bit to high to put through a presets delicate wiper)
The LED gives a minimum bias voltage (say 2 volts) because it is in series with the transistor. It may also serve some form of temperature compensation for the original FET's but it should be nothing to worry over.
https://en.wikipedia.org/wiki/Rubber_diode
The IRFP devices are the obvious one one's to use if going the HEXFET route.
That's a great explanation and many thanks for the link, I think it's slowly sinking in!
I tried the R512 change and fitting a 2k7 allows the bias 'voltage' to start coming up right near the top of the pot's range. Would a 3k3 be better here? I set it at 2mV initially (voltage across R519 now just over 7 Vdc) and then left the amp idling for an hour or so, still sat at 2mV, so ran a sine wave through it into the 8ohm resistor at about 15V RMS. Again no issues so checked around with my Flir (I'm a sparky so we do thermography on MCC's etc) and the only thing that stood out was the temperature of Q509 & Q510 was about 30C above the unmodified channel (85C vs 55C).
I presume this is an unavoidable side effect of modifying the bias circuit to increase the gate voltage? Is it worth fitting some small TO220 heatsinks or even moving these transistors onto the main heatsink?
I'm starting to question if I should worry about going to the trouble of swapping over to the IRFP devices now!
I tried the R512 change and fitting a 2k7 allows the bias 'voltage' to start coming up right near the top of the pot's range. Would a 3k3 be better here? I set it at 2mV initially (voltage across R519 now just over 7 Vdc) and then left the amp idling for an hour or so, still sat at 2mV, so ran a sine wave through it into the 8ohm resistor at about 15V RMS. Again no issues so checked around with my Flir (I'm a sparky so we do thermography on MCC's etc) and the only thing that stood out was the temperature of Q509 & Q510 was about 30C above the unmodified channel (85C vs 55C).
I presume this is an unavoidable side effect of modifying the bias circuit to increase the gate voltage? Is it worth fitting some small TO220 heatsinks or even moving these transistors onto the main heatsink?
I'm starting to question if I should worry about going to the trouble of swapping over to the IRFP devices now!
I suppose ideally you would aim for the correct bias with the preset around the midpoint but it doesn't matter as long as you can achieve the value you want. Remember the actual bias current is the current flowing in the 0.1 ohm resistors. That is the value to check.
To get 7 volts across R519 takes I=V/R which is 7/330 which is 21 milliamps. Lets say Q509/510 have about 40 volts across them. That gives a power dissipation W=I*V which is 40*0.021 giving 0.84 watts each. So yes, a little bit of heat generation.
4 volts across R519 works out to a shade under 0.5 watt per driver transistor. So it might possibly be worth a very small heatsink. I wouldn't move the transistors off the board, that could bring other issues with stability due to adding wires. Other options (now we are down to real tweaking and modifying) would be to also raise the value R519. Higher value means less current for the same volt drop and that lowers the dissipation in the transistors.
Very small heatsinks are maybe the answer tbh.
To get 7 volts across R519 takes I=V/R which is 7/330 which is 21 milliamps. Lets say Q509/510 have about 40 volts across them. That gives a power dissipation W=I*V which is 40*0.021 giving 0.84 watts each. So yes, a little bit of heat generation.
4 volts across R519 works out to a shade under 0.5 watt per driver transistor. So it might possibly be worth a very small heatsink. I wouldn't move the transistors off the board, that could bring other issues with stability due to adding wires. Other options (now we are down to real tweaking and modifying) would be to also raise the value R519. Higher value means less current for the same volt drop and that lowers the dissipation in the transistors.
Very small heatsinks are maybe the answer tbh.
That's great, thanks again for your assistance. Space is a bit tight on the board where the transistors are located so I'll fit a couple of 'clip-on' heatsinks so there's no chance of them touching eachother or anything else! Also reduces any mechanical stress on the device legs as it would be very tricky to solidly mount larger 'sinks.
It's a shame I don't have the kit to clearly see any distortion at the zero crossing point, earlier you wrote that initially setting the bias to 2mV across the 0.1ohm resistor (so 20mA current) would be a good place to start. What are the benefits of a higher bias setting or is it better to be cautious given the unknowns and leave it set quite low?
The only exposure to the effects of bias I've had anything to do with is on valve guitar amps where running the valves hot or cool affects the tone.
It's a shame I don't have the kit to clearly see any distortion at the zero crossing point, earlier you wrote that initially setting the bias to 2mV across the 0.1ohm resistor (so 20mA current) would be a good place to start. What are the benefits of a higher bias setting or is it better to be cautious given the unknowns and leave it set quite low?
The only exposure to the effects of bias I've had anything to do with is on valve guitar amps where running the valves hot or cool affects the tone.
Definitely better to be cautious. What I would say is see how it looks for thermal stability by hooking up your meter (semi permanently with wire) to the test points and then experiment with bias currents. The manual says 10 mv which is 100 ma current. If you try that then watch the current. Does it drift ever higher as the FET's warm. If so then set a lower initial value such that the current never exceeds say 130 ma when you have the amp hot hot hot 😉 Play music and then mute the sound and check the current is not excessive.
Always make the final adjustments on full mains (without bulb)
If you have a proper test CD with 1kHz sine then that should be a super low distortion source. If you listen at really low volume it should sound very pure with no harmonics.
The zipped folder has an MP3 track of 1kHz that also has increasing crossover distortion added. By the time you get to even a few seconds in the 'pure' tone has already been lost and you can hear the harsher harmonic in the background. Even playing on laptop speakers and the effect is clear. Play it and keep jumping back to the beginning as a reference of 'pure'.
With your speakers you do not ever want to hear that harsher tone (test with your CD playing). In practice even very low bias current totally remove any audible sound like that.
The image shows the base 1kHz tone at the top and the ever increasing lower level 3kHz tone at the bottom.
