Hi Eva,
Thank you for your thoughts on the matter.
Your view is purist and correct within its limits. A good compromise emitter resistor for a double emitter follower output stage is 0.33R to 0.47R. 0.1R, as you found, increases criticality alarmingly and is not strictly 'safe' in commercial use. 0R22 is probably the lowest value in wide commercial use.
In the commercial environment for Vbe multipliers a TO126 like the BD139 mounted atop one of the outputs is very effective, giving slightly quicker thermal feedback than when mounted in a hole in the heatsink. This approach confines you to a TO92 which has less effective thermal transfer than a TO126; swings and roundabouts.....
The top of an output device, despite the packaging, runs hotter than the heatsink, and has far less thermal inertia than the insulating washer/heatsink to which it is attached. Thus, when the device warms in service the temperature at the top of the package is hotter than at the heatsink, and hysteresis is less. Thus it is a valid point for thermal feedback and works very well.
Cheers,
Hugh
Thank you for your thoughts on the matter.
Your view is purist and correct within its limits. A good compromise emitter resistor for a double emitter follower output stage is 0.33R to 0.47R. 0.1R, as you found, increases criticality alarmingly and is not strictly 'safe' in commercial use. 0R22 is probably the lowest value in wide commercial use.
In the commercial environment for Vbe multipliers a TO126 like the BD139 mounted atop one of the outputs is very effective, giving slightly quicker thermal feedback than when mounted in a hole in the heatsink. This approach confines you to a TO92 which has less effective thermal transfer than a TO126; swings and roundabouts.....
The top of an output device, despite the packaging, runs hotter than the heatsink, and has far less thermal inertia than the insulating washer/heatsink to which it is attached. Thus, when the device warms in service the temperature at the top of the package is hotter than at the heatsink, and hysteresis is less. Thus it is a valid point for thermal feedback and works very well.
Cheers,
Hugh
Hate to sound like I'm bikering, but two days ago I used a thermocouple accessory to my DMM to compare the two. The top of the TO-247 (2sc5200) read ~20 degC cooler than the heatsink. There had been 30-60 minutes for everything to stabilize. I considered that I might not have done an adequate job of sticking the themocouple to the surface of each, so I repeated using Kapton tape to avoid the clamping mechanising pulling off any heat. As a crude check I also used the "finger test" to confirm that the heatsink was much hotter.
Note that I'm not relying on something I read in a book or the internet -- it's my own measurements.
One possible reconciliation of the differeng findings is that based on several months of actual experience, my heatsink is adequate for normal but vigorous (i.e., loud) use in a home adio environment. A larger or more effective heatsink which would let you drive a contiuous sine wave into a dummy load for hours or even days mught well disapate heat fast enough to remain cooler than the top of a TO-247 device.
Hugh is clearly more experienced in this matter than I am, yet I do not believe my eyes (and index finger) are deceiving me. I'm concluding for the moment that there are unidentified parameters at work.
How about using a thermistor in the Vbe multiplier? By using one with appropriate characteristics and/or putting it in series with an appropriate value of fixed resistance, it could be made to give any amount of feedback you want.
Furthermore, since thermistors can be very small, it could be placed right at the interface between transistor and heatsink, or inside the heatsink under the transistor, or even inside the metal tab of the transistor itself.
Like sam9, I've measured the temperature of various plastic packaged transistors at various places and the case is definitely much cooler than the heatsink. I haven't done the same test with metal cases.
Furthermore, since thermistors can be very small, it could be placed right at the interface between transistor and heatsink, or inside the heatsink under the transistor, or even inside the metal tab of the transistor itself.
Like sam9, I've measured the temperature of various plastic packaged transistors at various places and the case is definitely much cooler than the heatsink. I haven't done the same test with metal cases.
I just had to order some aluminum sheet for a non-audio project and decided to add a couple of small, thin (~0.5mm) copper sheets that I should be able to sandwitch between the output devices and the heatsink. I'll bend them 90 deg, mount a Vbe-multiplier on the flange and trim off any excess copper. This seems like it may be the most direct way to thermally communicate with the output divices in an EF arrangement. The material won't be here for a week or so it will be while be fore the experiment, but i'll definately post any useful (or disapointing) results.
Andy_C,
TO3-P, standard plastic pack, on 2SC5200/2SA1943.
Sam9,
Naturally I won't refute your results, nor indulge an IUC (intergalactic urinary contest) at this stage, except to say it might be that with a large, 0.3C/watt heatsink for four devices on two 55W AKSA modules the high thermal inertia is ensuring case temperature is higher than sink temperature!
I will say that your idea about using a flanged copper seat beneath each device to abut the heatsink with the sensing Vbe multiplier on the flange is a good one. However, Cu and Al are very different on the electrochemical series, and the junction is quite likely to corrode badly, particularly in the presence of heat.
Cheers,
Hugh
TO3-P, standard plastic pack, on 2SC5200/2SA1943.
Sam9,
Naturally I won't refute your results, nor indulge an IUC (intergalactic urinary contest) at this stage, except to say it might be that with a large, 0.3C/watt heatsink for four devices on two 55W AKSA modules the high thermal inertia is ensuring case temperature is higher than sink temperature!
I will say that your idea about using a flanged copper seat beneath each device to abut the heatsink with the sensing Vbe multiplier on the flange is a good one. However, Cu and Al are very different on the electrochemical series, and the junction is quite likely to corrode badly, particularly in the presence of heat.
Cheers,
Hugh
Well, yes, and I notice from your site you appear to be using nice Conrad heatsinks. The cost of shipping a pair of those up here is more than a whole box full of so-so units off e-bay. Meaning, I'm using the e-bay jobs. There are more than a few mysteries from my stand point with regard to heatsinks and one is the radiating surface. I presume that in the case of two heatsinks with equal surface area and charateristics, the one that can loose heat to the air quicker will also respond quicker to temperature changes at the device.
Hopefully, they will last long enough to satisfy my curiosity. I admit I associate this problem more with metals imersed in salt water (a couple of nautical types in the family). However, from time to time someone posts here about mounting output devices "bareback" usually in connection with a Class-A heater. I presume it is the plating on the copper that prevents disaster. I lso presume the plating is nickelbut wonder if tin would do as well. (I can do some crude tin plating.)
sam9 said:
It can be good idea. I will try it! But for better thermal coupling I will use the copper between the case, and the insulator. With this solution it has no contact to the aluminium heatsink, and the temperature will higher. OK, I have to insulate the Vbe-multiplier transistor, but it's not so complicated in case of TO-92 device...
sajti
My friend had an idea: Keep the heatsink on constant temperature. To do it, use larger hetasink than necessary, and put temperature controlled heaters on it! If the amplifier heats up the sink to -say- 50degree, let's keep it on 60, for every time.
There will be no big changes on the junction of the output devices, and this means less problem with bias, and less problem with other parameters, depended by temperature.
I never tried, but it can work...
sajti
There will be no big changes on the junction of the output devices, and this means less problem with bias, and less problem with other parameters, depended by temperature.
I never tried, but it can work...
sajti
Hi,
Doug Self carried out experiments for his EF topology output stages and his published results confirm AKSA.
Put your To126 or To220 Vbe on top of the To264/To3P output.
But for a CFP (shiklai) the Vbe must monitor the driver temp or in a 3 deep then the predriver.
I like that idea of gluing a To92 onto the back of the output.
Anyone tried this and comparing Vbe voltage to To92 glued to the heatsink?
regards Andrew T.
Doug Self carried out experiments for his EF topology output stages and his published results confirm AKSA.
Put your To126 or To220 Vbe on top of the To264/To3P output.
But for a CFP (shiklai) the Vbe must monitor the driver temp or in a 3 deep then the predriver.
I like that idea of gluing a To92 onto the back of the output.
Anyone tried this and comparing Vbe voltage to To92 glued to the heatsink?
regards Andrew T.
AndrewT,
To backtrack a bit, a number of us are familiar with Self's experiments but got different results when attempting to repeat them. At this point it looks like heatsinking is an important variable. Hugh is using a much bigger heatsink than I am and that allow the heatsink to run cooler than the topside of the plastic case. The heatsink I'm using is adequate for ordinary to vigorous listening levels but I know that it is too small to run large test signals in to a 4-ohm dummy load for long periods. The result is that the heatsink is about 20 deg C hotter than the topside of the device.
Rather than a conflict of opinion, what I think has happened is we found out something we didn't know before. Namely that in selecting where to mount the Vbe multiplier the choice of heatsinking plays a role. Obviously bigger is better, but bigger may not always possible given other physical constraints.
To backtrack a bit, a number of us are familiar with Self's experiments but got different results when attempting to repeat them. At this point it looks like heatsinking is an important variable. Hugh is using a much bigger heatsink than I am and that allow the heatsink to run cooler than the topside of the plastic case. The heatsink I'm using is adequate for ordinary to vigorous listening levels but I know that it is too small to run large test signals in to a 4-ohm dummy load for long periods. The result is that the heatsink is about 20 deg C hotter than the topside of the device.
Rather than a conflict of opinion, what I think has happened is we found out something we didn't know before. Namely that in selecting where to mount the Vbe multiplier the choice of heatsinking plays a role. Obviously bigger is better, but bigger may not always possible given other physical constraints.
sam9 said:
As I imagine the heatsink must be cooler, than the top of the transistor. There is thermal resistance between the case and the heatsink. The dissipated power makes temperature rising on this resistance. So if You have 1K/W between the case, and the heatsink (I think this is common value for TO-247), and You dissipate 20W, the temperature on the case (and the top of the transistor) must be 20 degree higher than the heatsink. With few W dissipation low idle current, the dirrence is very small..
sajti
satji:
The plastic resin material of the case is a very bad heat conductor but it radiates heat to the air quite well since it's surface is black and textured, and there there is allways natural convection air flow near the case caused by the big hot surface of the heatsink
This makes the thermal resistance die-to-case equal or higher than the thermal resistance case-to-air and explains why our measurements show that the case is at a lower temperature that the heatsink just near the power device [or at the opposite side]
Note that the thermal resistance die-to-case may be as high as 100ºK/W while the thermal resistance from the die to the heatsink just near the power device may be only 2ºK/w
Imagination does not work well for these purposes, altough I'm aware that people uses it a lot, particularly in audio
Anyway, the evidence is right here, in each plastic-case transistor mounted in a heatsink over the world, and is waiting for anybody with temperature measurement facilities to discover it...
The most beautiful thing about scientific findings is repeatability
The plastic resin material of the case is a very bad heat conductor but it radiates heat to the air quite well since it's surface is black and textured, and there there is allways natural convection air flow near the case caused by the big hot surface of the heatsink
This makes the thermal resistance die-to-case equal or higher than the thermal resistance case-to-air and explains why our measurements show that the case is at a lower temperature that the heatsink just near the power device [or at the opposite side]
Note that the thermal resistance die-to-case may be as high as 100ºK/W while the thermal resistance from the die to the heatsink just near the power device may be only 2ºK/w
Imagination does not work well for these purposes, altough I'm aware that people uses it a lot, particularly in audio
Anyway, the evidence is right here, in each plastic-case transistor mounted in a heatsink over the world, and is waiting for anybody with temperature measurement facilities to discover it...
The most beautiful thing about scientific findings is repeatability
Eva said:
I know that the plastic case is not the best heat conductor. But I think it's thin to lose lot of heat. To lose lot of heat You have to put some heatsink to the top of the case. And If You mount TO-126 device, it will hide some surface, and there will be no radiation.
But due the bad conducting plastic, You have to install the sensor very carefully, to avoid the cooling caused by the screw, the cables, You use. I think that the distance between the junction, and the top of the case is no more than 2mm, so bad heat conducting is not serious problem.
sajti
satji :
I suggest you to take a plastic device and a heatsink and do some temperature measurements. This is the best thing one can do when in doubt
Note that there may be strong temperature gradients across the heatsink and that the place just near [or in the oppsite side] of the power device responds almost instantaneously to dissipation changes while the rest of the heatsink shows increasing lag with distance
I suggest you to take a plastic device and a heatsink and do some temperature measurements. This is the best thing one can do when in doubt
Note that there may be strong temperature gradients across the heatsink and that the place just near [or in the oppsite side] of the power device responds almost instantaneously to dissipation changes while the rest of the heatsink shows increasing lag with distance
Eva said:
I think it's very difficult predict the thermal condition. I have two different type of heatsink with same surface area but one of them has 5mm base, the other has 19mm. This two heatsink works absolutely differently with same transistors, and same dissipation...
My only problem is, that I have ceramic insulators without holes. I tried, but not possible to drill any holes. So I use massive aluminimum bridges to fix the TO-247 devices. But this bridge cool down the top of the case...
I use higher value emitter resistors, and the bias remains fix over wide temperature range. This emitter resistors protects the output devices against reactive loads, and reduce the short circuit current, so simple fuse is good enough to protect...
sajti
Pending arrival of the thin copper pieces mentioned in an earier post, I found a piece of "thin" scrap aluminum that would serve as a sensor between the plate of the cases and the heatsink. A hour of work with Dremel tool, drill and sandpaper and it was assembled. With Vbe transistor attached to the sheet of Al bias tracks somewhat better. I also notices that the temp differential between the heatsink and the back of the plastic case is much smaller. Under these circumstances this is NOT a good thing as I have obiously simply increased the thermal resistance between devive and heatsink to a considerable degree. However, it does reinforce the observation that the effectiveness of the heatsink is a factor.
If you are unconcerned about the survival of the output devices,optimal Vbias tracking could perhaps be achieved by dumoinf the heatsinks and just atacth the Vbias transistor plate-to-plate with one of the output devices. (Caution: For readers whose first language is not English the preceding sentence is intededed as irony, not a serious suggestion.)
If you are unconcerned about the survival of the output devices,optimal Vbias tracking could perhaps be achieved by dumoinf the heatsinks and just atacth the Vbias transistor plate-to-plate with one of the output devices. (Caution: For readers whose first language is not English the preceding sentence is intededed as irony, not a serious suggestion.)
Just a reminder regarding the origin of this thread, when Self wrote about attaching the Vbe-multiplier to the top of the output device, he was specifically talking about a TO-3 metal can. Some of us thought this sounded good and we should do the same with a plastic flat-pack. It did not work out for me and some others, but Hugh Dean has had success with this approach.
An aside, early in the thread somone posted not too helpfully that we should dusp bump BJT-EF amps and go MOSFET. As a matter of fact it has been my intention for some time to make my next ground-up project an L-MOSFET amp. Biasing hassles are one but not the onlt reason.
An aside, early in the thread somone posted not too helpfully that we should dusp bump BJT-EF amps and go MOSFET. As a matter of fact it has been my intention for some time to make my next ground-up project an L-MOSFET amp. Biasing hassles are one but not the onlt reason.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.