Hi,
I am in progress of building an amp with TDA7250 and TIP142/147 transistors. After putting everything in the enclosure I've discovered the heatsink on the transistors reaches 55C on quiescent current alone (which appears to be about 130mA per channel and rail). The heatsink is aluminum 120x100x25mm and finned (transistors are contacting the sink via a thermally conductive silicone pad + thermal paste on both sides). The amp is powered with a 300VA 2x29V transformer (gives about 41 V DC per rail).
The question is: is this sort of temperature normal or is it overheating?
If this beast is indeed overheating, do you have any recommendations considering:
- I can't fit much more aluminum inside the enclosure
- a fan is acceptable as a last resort only
- I don't want to swap the tranformer for a lower sec. voltage unit (due to additional cost)
Thanks for any comments or suggestions.
Live long and prosper!
I am in progress of building an amp with TDA7250 and TIP142/147 transistors. After putting everything in the enclosure I've discovered the heatsink on the transistors reaches 55C on quiescent current alone (which appears to be about 130mA per channel and rail). The heatsink is aluminum 120x100x25mm and finned (transistors are contacting the sink via a thermally conductive silicone pad + thermal paste on both sides). The amp is powered with a 300VA 2x29V transformer (gives about 41 V DC per rail).
The question is: is this sort of temperature normal or is it overheating?
If this beast is indeed overheating, do you have any recommendations considering:
- I can't fit much more aluminum inside the enclosure
- a fan is acceptable as a last resort only
- I don't want to swap the tranformer for a lower sec. voltage unit (due to additional cost)
Thanks for any comments or suggestions.
Live long and prosper!
That is a radical temperature. Even my stupidly heat sinked dynakit ST120 only reaches this temperature if I play a church gig for a couple of hours without a fan. That heat sink is 1/8" thick and has no fins.
130 ma quiescent current seems to be too high. 20 ma is about right on TO3 cased parts. Try putting a potentiometer across one of the voltage spreader diodes so you can partially short it out. or something. +- 42 v (1.4 x 29v) rails is pretty mild, IMHO.
130 ma quiescent current seems to be too high. 20 ma is about right on TO3 cased parts. Try putting a potentiometer across one of the voltage spreader diodes so you can partially short it out. or something. +- 42 v (1.4 x 29v) rails is pretty mild, IMHO.
130mA bias current is certainly high but not unheard of bias current for a single complementary pair in class AB. The heatsink seems about right for the task, unless you are talking of more than a single pair, stereo channels etc. Actually, I can't tell what its features, metal thickess etc. are, let alone know its rating in deg.C/W. It could be around 1 degree per Watt which would be plenty for a pair of TO3s on 40V rails with say, 25 mA bias for about 60W output. Darlingtons probably need slightly higher bias to allow for the additional driver bias current.
The dissipation of 130mA bias current is only about 10W or so and that means a temp. rise of perhaps 10C above ambient - nowhere near enough to get to 55C unless your ambient is well over 40C, assuming of course, that the heatsink is vertically aligned and has free circulation of air.
Different diode types also have different voltage drops. If the design specified 1N4148 and you happened to use 1N4004 or similar, you certainly would get higher bias and heatsink temperature. Also, I think 4 diodes are often too much for a typical EF2 output stage and you often see 3 diodes with or without a resistor or pot to increase the voltage drop just a little more and provide adjustment, somewhat as indianajo suggests.
I think there must be more than just bias current thermal dissipation here. It could well be oscillating to get that hot and you'll need a scope or other suitable piece test gear to identify and fix it. If you have a rat's nest of wires in a temporary setup though, you may have your answer right there.
The dissipation of 130mA bias current is only about 10W or so and that means a temp. rise of perhaps 10C above ambient - nowhere near enough to get to 55C unless your ambient is well over 40C, assuming of course, that the heatsink is vertically aligned and has free circulation of air.
Different diode types also have different voltage drops. If the design specified 1N4148 and you happened to use 1N4004 or similar, you certainly would get higher bias and heatsink temperature. Also, I think 4 diodes are often too much for a typical EF2 output stage and you often see 3 diodes with or without a resistor or pot to increase the voltage drop just a little more and provide adjustment, somewhat as indianajo suggests.
I think there must be more than just bias current thermal dissipation here. It could well be oscillating to get that hot and you'll need a scope or other suitable piece test gear to identify and fix it. If you have a rat's nest of wires in a temporary setup though, you may have your answer right there.
Hi,
thanks both of you for your thoughts.
Ian Finch: the transistors are in TO-218 package and I have both channels on the heatsink (4 darlingtons total), so there should be a little over 20W going through that heatsink. The schematic is the same as in page 6 of [wiki=http://www.datasheetcatalog.org/datasheet/SGSThomsonMicroelectronics/mXvrvuw.pdf ]%[/wiki]. The heatsink is positioned horizontally, but there are vents just above it on the upper chassis wall.
However, what I wanted to ask is if it is normal to let the heatsink go that hot and how much more heat I should be comfortable pumping into it (in terms of maximum temperature) before I should take further measures to cool it (like using a fan).
thanks both of you for your thoughts.
Ian Finch: the transistors are in TO-218 package and I have both channels on the heatsink (4 darlingtons total), so there should be a little over 20W going through that heatsink. The schematic is the same as in page 6 of [wiki=http://www.datasheetcatalog.org/datasheet/SGSThomsonMicroelectronics/mXvrvuw.pdf ]%[/wiki]. The heatsink is positioned horizontally, but there are vents just above it on the upper chassis wall.
However, what I wanted to ask is if it is normal to let the heatsink go that hot and how much more heat I should be comfortable pumping into it (in terms of maximum temperature) before I should take further measures to cool it (like using a fan).
As Ian stated, the heatsink seems appropriate for one pair of output devices at the given operating conditions. I did no calculations yet, but it sounds rather undersized for two channels resp. four output devices.
This mounting position turns the heatsink into a rather useless chunk of metal, vents in the chassis or not. It will get significantly hotter than its specifications would suggest, as normal convection cooling is drastically reduced in the horizontal position.
Finned heatsinks have to be mounted vertically and the top and bottom plates of the enclosure need vent openings above and below the heatsink.
Rundmaus
Never forget that a 'heatsink' is a small sink plus a big radiator.
If the radiator part is not allowed to work, the sink part does not work either.
Heatsinks are finned to create a large surface area from where to exchange heat with the surrounding air. This air has to continuously move: Cool air in, warm air out.
If you rely on natural draft, the fins have to be vertical! And the spacing of the fins has to be more than ~10 mm. Air has to move freely through those fins.
For this reason you will also need holes below the heatsink, as well as above.
If the fins are horizontal, or denser than 10 mm, you need a fan with a well defined air path through the enclosure and the heatsink.
It does not take much air, you can slow the fan down to inaudible speed.
EDIT: Also see post 5
Last edited:
Radical in midwestern english means too much. That is, 55 deg C is too much. Pay attention to the vertical (good) versus horizontal (bad ) comments above. There are no spreader diodes in figure 6 of the schematic, so what I said about shorting them doesn't apply. I suppose you are measuring the 130 ma idle current across the 0.15 ohm resistors?
TO-218 parts have even less capability to withstand idle current and heat than TO3 parts.
I would check for oscillation like Mr. Finch said. Unfortunately, working scopes are a bit beyond my means, I purchased a second one last year and it now need a complete electrolytic cap replacement with the boards glued in. !@#$ You can detect ultrasonic oscillations with a 200kohm/volt analog VOM with 2 VAC and 20 vAC scales. Not a DVM, none of those I have purchased have given reliable readings for any frequency above 120 hz. You put a blocking capacitor (0.47 uf, 600v) in series with the plus probe, and look for a reading at the output with the amp input shorted. If you have a reading, that is an oscillation. Then change the blocking capacitor to 470 pf or smaller. If you still have a similar reading with the small cap, the oscillation is ultrasonic.
Oscillations can be poor layout, or bad solder joints at the anti-oscillation caps (the picofarad ones.) Or after a meltdown due to shorted output transistors, burned out anti-oscillation caps, particularly those rated less than the rail voltage.
Good luck. The TDA7250 is interesting, but since it comes without a heatsink tab I think I would prefer an LM49810 from TI/National Semi.
TO-218 parts have even less capability to withstand idle current and heat than TO3 parts.
I would check for oscillation like Mr. Finch said. Unfortunately, working scopes are a bit beyond my means, I purchased a second one last year and it now need a complete electrolytic cap replacement with the boards glued in. !@#$ You can detect ultrasonic oscillations with a 200kohm/volt analog VOM with 2 VAC and 20 vAC scales. Not a DVM, none of those I have purchased have given reliable readings for any frequency above 120 hz. You put a blocking capacitor (0.47 uf, 600v) in series with the plus probe, and look for a reading at the output with the amp input shorted. If you have a reading, that is an oscillation. Then change the blocking capacitor to 470 pf or smaller. If you still have a similar reading with the small cap, the oscillation is ultrasonic.
Oscillations can be poor layout, or bad solder joints at the anti-oscillation caps (the picofarad ones.) Or after a meltdown due to shorted output transistors, burned out anti-oscillation caps, particularly those rated less than the rail voltage.
Good luck. The TDA7250 is interesting, but since it comes without a heatsink tab I think I would prefer an LM49810 from TI/National Semi.
Last edited:
So, I've flipped the heatsink to a vertical position, there are now large vent holes beneath it (8mm + 2x 6mm holes between each fin pair) and holes just above it. It still gets to 53C (my previous measurement of 55C may have been inaccurate, I terminated after about 30 minutes, it may not have reached its full temperature).
Flipping the heatsink seems to have helped a little (I can feel the warm air rising from it through the vent holes. However, rearranging the guts freed up some space inside the chassis, so I am going to switch the existing heatsink (120x100x35) for a larger one (190x100x50). Being of almost twice the surface area, it will hopefully suffice. If you're interested, the current setup look like this: [wiki=https://plus.google.com/photos/114041417468443473448/albums/5862559745780688705/5863653583955342450]%[/wiki]
As indianajo and Ian Finch have suggested, I've checked for oscillations with a scope, there seem to be none (output is quiet when input is short-circuited). Which also seems to be supported by another measurement (using one of those power consumption meters that you plug into the wall socket and connect your device through it) - the consumption reads 24W (2-3W go to the other transformer I have installed), so the amp doesn't appear to eat too much power.
Fins on the heat sink are really a very good idea for naturally cooled amps. I don't know how the salvage condition is in your country, but I found a couple of really nice 7 cm tall x 4 cm wide x 2 cm deep heatsinks with 12 fins in a Toshiba projection TV sitting out for the garbage last month. Plenty of flat area for two TO3's or TO-247's. I had to take a screwdriver down the block to pull that and the 20 cm speakers. I didn't even have to bring the TV home, it only took a couple of minutes. I don't really see anything similar in the Mouser catalog, but equivalent sized sinks look about $20 apiece. In a flat heatsink, my Wurlitzer 4500 organ with about a 20 W power amp has 15 cm x 25 cm for one TO3 pair.
Other possible donor heatsink systems are dead switching power supplies, particularly the ones for desktop PCAT's, and variable speed motor drives thrown out by factories.In a different economy, one may have to pay the salvage guy to get these, they may not sit around at the pickup point waiting for the trash truck like here. Everybody here is so desparate to import something new from ***** that they won't walk a block down the street and pick one up off the ground.
Other possible donor heatsink systems are dead switching power supplies, particularly the ones for desktop PCAT's, and variable speed motor drives thrown out by factories.In a different economy, one may have to pay the salvage guy to get these, they may not sit around at the pickup point waiting for the trash truck like here. Everybody here is so desparate to import something new from ***** that they won't walk a block down the street and pick one up off the ground.
Last edited:
Not so good, I'm afraid. There are warehouses that collect old electronics, but they generally don't like people putting stuff apart in there
. It is good to have an electronics shop owner as a friend if they are willing to let you browse through the old stuff .The new heatsink is not that expensive (170 CZK, equiv. to about 9 USD) plus shipping of course
.
Hi
'Good that it's working OK. That new heatsink should be plenty but I was looking at ST Micro application note and was working out how the 4.7 uF integrator caps determine quiescent current. I thought it was ingenious because it's the output current that charges them. I'm not certain how much these can be adjusted to set bias but I suggest first adding 1uF in parallel to see how much current change there is and reducing say to 3.3uF if the change was to higher current and so on.
I would think 80-100 mA would be plenty for any darlington o/p stage and you could even use a much more appropriate and cool running 40mA with less risk.
'Good that it's working OK. That new heatsink should be plenty but I was looking at ST Micro application note and was working out how the 4.7 uF integrator caps determine quiescent current. I thought it was ingenious because it's the output current that charges them. I'm not certain how much these can be adjusted to set bias but I suggest first adding 1uF in parallel to see how much current change there is and reducing say to 3.3uF if the change was to higher current and so on.
I would think 80-100 mA would be plenty for any darlington o/p stage and you could even use a much more appropriate and cool running 40mA with less risk.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.