sam9
I use the same method as Satji, I drill a hole of the diameter of the TO-92 case and I glue it here
This schematic uses a buffered Vbe multiplier :
Note that in this circuit the Vas is biased at 30mA and this is quite troublesome since most unbuffered Vbe multipliers wouldn't produce enough negative feedback at such a high Ic
MikeB :
This may seem stupid, but try placing a 10 to 47 ohms resistor in series with the BD139 emitter. The bias current would cause an almost constant voltage drop of a few mV on this resistor and this will add a constant voltage to the apparent Vbe of the BD139. Then you will have to readjust the bias potentiometer to a lower setting and thus there will be less negative feedback
I use the same method as Satji, I drill a hole of the diameter of the TO-92 case and I glue it here
This schematic uses a buffered Vbe multiplier :
An externally hosted image should be here but it was not working when we last tested it.
Note that in this circuit the Vas is biased at 30mA and this is quite troublesome since most unbuffered Vbe multipliers wouldn't produce enough negative feedback at such a high Ic
MikeB :
This may seem stupid, but try placing a 10 to 47 ohms resistor in series with the BD139 emitter. The bias current would cause an almost constant voltage drop of a few mV on this resistor and this will add a constant voltage to the apparent Vbe of the BD139. Then you will have to readjust the bias potentiometer to a lower setting and thus there will be less negative feedback
Eva / Sajti,
I'm laughing (at myself) regarding mounting the TO-92. The solution is so simple and obvious it is invisible to someone looking for something more complex.
Eva,
I recognize your buffered Vbe-multiplier as a version of one described briefly by Self. You use much larger resistor values which I assumes is rlated to achieving a sutable distribution of current. I've passed it by because of the somewhat understated warning that oscillation is a possability. I take it that oscillation has not been an issue for you in this case. The idea looked good but I had other things to pursue first before taking on something that might leave me fighting oscillations. It is encouraging to see someone successfully use it.
I'm laughing (at myself) regarding mounting the TO-92. The solution is so simple and obvious it is invisible to someone looking for something more complex.
Eva,
I recognize your buffered Vbe-multiplier as a version of one described briefly by Self. You use much larger resistor values which I assumes is rlated to achieving a sutable distribution of current. I've passed it by because of the somewhat understated warning that oscillation is a possability. I take it that oscillation has not been an issue for you in this case. The idea looked good but I had other things to pursue first before taking on something that might leave me fighting oscillations. It is encouraging to see someone successfully use it.
Eva,
I just wired in the additional transistor and resistor to the existing PCB to test. Different values but the effect seems good improves stability. Initial setting is easiest if I deliberatly over-biased first, letting things get too hot, them adjust back down. Starting with and underbias and cautiously increasing the adjustment is very tedious.
I also note that the correspondence between the current through the RE resistors and the voltage across the Vbe multiplier does not match the tables in Self. It appears that they hold only for the "blameless" configuration or perhaps that there is a different relationship in the case of mirror-image input sand VAS sections. I tried to set the voltage per Self to 2.89V to see what would happen and the RE current was way too high andthe temperature of the heatsink was enough to fry an egg. When I set the RE current where it should br the Vbe voltage measured ~2.3V. Whether is is optimat or not, I can say but was good enough that THD was below my ability to measure with my ancient ebay bought gear.
I just wired in the additional transistor and resistor to the existing PCB to test. Different values but the effect seems good improves stability. Initial setting is easiest if I deliberatly over-biased first, letting things get too hot, them adjust back down. Starting with and underbias and cautiously increasing the adjustment is very tedious.
I also note that the correspondence between the current through the RE resistors and the voltage across the Vbe multiplier does not match the tables in Self. It appears that they hold only for the "blameless" configuration or perhaps that there is a different relationship in the case of mirror-image input sand VAS sections. I tried to set the voltage per Self to 2.89V to see what would happen and the RE current was way too high andthe temperature of the heatsink was enough to fry an egg. When I set the RE current where it should br the Vbe voltage measured ~2.3V. Whether is is optimat or not, I can say but was good enough that THD was below my ability to measure with my ancient ebay bought gear.
What resistor values are you using?
I hadn't any oscillation problem, but I think the key for stability is to operate the buffer transistor at a much higher Ft than the sense transistor
BC560C datasheet states Ft=200Mhz for Ic=30mA and the transistor I used has approx. Hfe=425 [Ib=70uA]
P2N2222A datasheet states Ft=90Mhz for Ic=1mA, but I'm operating it at Ic=100uA so extrapolating the Ft graph it may have less than 20Mhz Ft and this produces a dominant pole on the system. It's Hfe is about 100 for that Ic
The 22k resistor demands another additional 30uA of Ic from the sense transistor. Increasing this current reduces the negative thermal feedback effect and also the buffer stability margin, since the Ft of the PN2222A quickly increases as its Ic is increased and oscillation may arise if both Ft are too close
With proper resistor values, the idle bias current is inmediately stabilished as the circuit is powered up, there is no need to wait until the heatsink gets hot to have proper biasing as usual
I hadn't any oscillation problem, but I think the key for stability is to operate the buffer transistor at a much higher Ft than the sense transistor
BC560C datasheet states Ft=200Mhz for Ic=30mA and the transistor I used has approx. Hfe=425 [Ib=70uA]
P2N2222A datasheet states Ft=90Mhz for Ic=1mA, but I'm operating it at Ic=100uA so extrapolating the Ft graph it may have less than 20Mhz Ft and this produces a dominant pole on the system. It's Hfe is about 100 for that Ic
The 22k resistor demands another additional 30uA of Ic from the sense transistor. Increasing this current reduces the negative thermal feedback effect and also the buffer stability margin, since the Ft of the PN2222A quickly increases as its Ic is increased and oscillation may arise if both Ft are too close
With proper resistor values, the idle bias current is inmediately stabilished as the circuit is powered up, there is no need to wait until the heatsink gets hot to have proper biasing as usual
May I ask your attention for an article of Malcolm Hawksford: "OPTIMIZATION OF THE AMPLIFIED-DIODE BIAS CIRCUIT FOR AUDIO AMPLIFIERS, JAES, vol.32, no.1/2, pp.31-33, Jan/Feb 1984" (See:http://www.essex.ac.uk/ese/research/audio_lab/malcolms_publications.html. There you can also find his article on distortion correction circuits for audio amplifiers)
Marc.
Marc.
The link doesn't work correctly. Try this one: http://www.essex.ac.uk/ese/research...cs/J8 Optimization of the amplified diode.pdf
Marc.
Marc.
Vbe multipliers on bipolar transistor power amplifiers should track any increase in case temperature at the output devices. They should react either so that the OP stage quiescent remains the same as temperature rises, OR even reduces somewhat with temperature rise. This is not just to avoid thermal runaway; you want to keep output stage quiescent current as constant as you can because the sonic characteristics of a Class AB amplifier change considerably with fluctuations in quiescent current.
If you have a particularly heavy session of music, then when it stops, the OP devices will have heated appreciably and their base/emitter voltage will have dropped considerably (roughly 2mV/degree C). This drives up the quiescent if the Vbe reacts slowly, as it will do if placed (remotely) in the heatsink. This problem is exacerbated by use of small emitter resistors, and for this reason I often choose 0.47 ohm, since it is included in the global feedback loop anyway and the full effect does not manifest as an appreciably higher output impedance.
A better idea I have found is to mount the Vbe device (BD139 is fine) on top of one of the output devices. This means it tracks the case temperature with a quick response, rather than the slower temperature rise at the heatsink. You then suitably choose collector/base and base/emitter resistors/pots in the Vbe multiplier to thermally track predictably, and if things are difficult you can add 10-22 ohms of emitter resistor on the Vbe device, though this is normally not necessary.
Because you are dealing with a thermal feedback system the math is complex, and the phenomenon is disguised by marked hysteresis. This application really needs to be 'tuned' empirically for best results.
Cheers,
Hugh
If you have a particularly heavy session of music, then when it stops, the OP devices will have heated appreciably and their base/emitter voltage will have dropped considerably (roughly 2mV/degree C). This drives up the quiescent if the Vbe reacts slowly, as it will do if placed (remotely) in the heatsink. This problem is exacerbated by use of small emitter resistors, and for this reason I often choose 0.47 ohm, since it is included in the global feedback loop anyway and the full effect does not manifest as an appreciably higher output impedance.
A better idea I have found is to mount the Vbe device (BD139 is fine) on top of one of the output devices. This means it tracks the case temperature with a quick response, rather than the slower temperature rise at the heatsink. You then suitably choose collector/base and base/emitter resistors/pots in the Vbe multiplier to thermally track predictably, and if things are difficult you can add 10-22 ohms of emitter resistor on the Vbe device, though this is normally not necessary.
Because you are dealing with a thermal feedback system the math is complex, and the phenomenon is disguised by marked hysteresis. This application really needs to be 'tuned' empirically for best results.
Cheers,
Hugh
But this is only valid for devices whose metal tab is still visible after they are mounted on the heatsink, ie : TO-3 metal can, TO-218 and TO-220
TO-247 and TO-268 [TO-3P] devices are not suitable for that purpose since it has already been demonstrated that the plastic resin case does not track the temperature of the die at all, it tracks ambient temperature instead. It's normal for these cases to be 20ºC cooler than the die and to respond with more than 30 seconds of lag to die temperature changes
An alternative for these full isolated devices may be to drill a small hole of the diameter of a TO-92 case in the heatsink, just in a corner of the area where the power device is mounted, and insert here a TO-92 sense transistor from the other side, coupling it directly to the copper tab of the power device with thermal grease
TO-247 and TO-268 [TO-3P] devices are not suitable for that purpose since it has already been demonstrated that the plastic resin case does not track the temperature of the die at all, it tracks ambient temperature instead. It's normal for these cases to be 20ºC cooler than the die and to respond with more than 30 seconds of lag to die temperature changes
An alternative for these full isolated devices may be to drill a small hole of the diameter of a TO-92 case in the heatsink, just in a corner of the area where the power device is mounted, and insert here a TO-92 sense transistor from the other side, coupling it directly to the copper tab of the power device with thermal grease
Cortez said:
Just use the shotest wire, and use very small diameter, because the wire can cool down the transistor. TO-92 dissipates the most of the heat, on the legs.
I found, that if You use at least .56, or .68ohm as emitter resistor there will be no problem with the BD139 mounted on the heatsink.
sajti
Almost everything is stable with such high emitter resistor values as 0.56 or 0.68 ohms
But I did my experimentation with 0.1 ohm emitter resistors and 100V Vce devices with +-41V rails, and I got stability only with a buffered Vbe multiplier mounted in a hole on the heatsink just near one of the power devices. Placing the Vbe multiplier over the plastic case of one of the output devices produced only a bit better feedback effect than placing it hanging it in the air near the heatsink
I've seen some commercial amplifiers whose Vbe multipliers were mounted on the PCB up to an inch away from the heatsink and they still didn't show thermal runaway. High voltage devices working with low voltage rails, low bias and high value emitter resistors are allways very easy to stabilise
sam9 :
Is R3 330k or 330ohms?
Isn't Q1 drawn upside down?
But I did my experimentation with 0.1 ohm emitter resistors and 100V Vce devices with +-41V rails, and I got stability only with a buffered Vbe multiplier mounted in a hole on the heatsink just near one of the power devices. Placing the Vbe multiplier over the plastic case of one of the output devices produced only a bit better feedback effect than placing it hanging it in the air near the heatsink
I've seen some commercial amplifiers whose Vbe multipliers were mounted on the PCB up to an inch away from the heatsink and they still didn't show thermal runaway. High voltage devices working with low voltage rails, low bias and high value emitter resistors are allways very easy to stabilise
sam9 :
Is R3 330k or 330ohms?
Isn't Q1 drawn upside down?
Figure 2 in the Hawksford paper seems to be pretty widlley adopted.
Figure 3 I've not seen before but is something I vaguely wondered about from another viewpoint - namely using two transistors to double the area of contact and/or in the case where the drivers are being tracked/coupled (CFB topology) provides a way to couple to both the pos and neg (top 7 bottom?) sides.
This thread has been helpful but has suggested a number of ideas I have to set aside for now or I will get too distracted.
Figure 3 I've not seen before but is something I vaguely wondered about from another viewpoint - namely using two transistors to double the area of contact and/or in the case where the drivers are being tracked/coupled (CFB topology) provides a way to couple to both the pos and neg (top 7 bottom?) sides.
This thread has been helpful but has suggested a number of ideas I have to set aside for now or I will get too distracted.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.