FWIW I use a high Hfe TO92 (BC556c) glued to the heatsink, in the middle of the power transistors.
Probably a metallic TO3 case is a good sample of the chip temperature; not so sure (at all) that the top of a *plastic* case is reliable as a sampling point, there must be some reason nobody uses that.
Probably a metallic TO3 case is a good sample of the chip temperature; not so sure (at all) that the top of a *plastic* case is reliable as a sampling point, there must be some reason nobody uses that.
The Thermal Resistance between junction and top of plastic package is far too high, despite the short distance the heat needs to travel.
The few builders that do this, seem not to realise they are not getting a good sample of chip temperature (Tj).
Thermals as like to call them is not a simple thing and beyond all useful comments above there a few more to know ..
--Bias will vary depending on the heatsink temperature in a classic EFP amplifier
--Using different heat sinks for P and N transistors generally is a bad practice
--Stacking P on the one side and N on the other is also not the 100% best way to go is to make one N one P one N one P and so on and on
--In thermals one should also calculate room temperature ...far more in countries that have wide temp variations ,
--Finally while many use many of the above methods for attaching the transistor the job is only complete if more tests are done to avoid overcompensation where bias will drop to a noticeable level
Kind regards
Sakis
--Bias will vary depending on the heatsink temperature in a classic EFP amplifier
--Using different heat sinks for P and N transistors generally is a bad practice
--Stacking P on the one side and N on the other is also not the 100% best way to go is to make one N one P one N one P and so on and on
--In thermals one should also calculate room temperature ...far more in countries that have wide temp variations ,
--Finally while many use many of the above methods for attaching the transistor the job is only complete if more tests are done to avoid overcompensation where bias will drop to a noticeable level
Kind regards
Sakis
^^^^^^^^^^^^
Fully agree, of course.
Our friend Sakis is showing his actual experience with real World products.
Simulation is great, excellent tool to save time and money and to pre-check something before building it, but the ultimate test lies in actually building the thing , firing it up and using it to play music at realistic levels, with proper speakers.
One problem I have with Simulation parts libraries, is that all include quite accurate **electronic parts** models, but none that I know of includes models of actual speakers, or speaker cabinets, or their internal crossovers , not even, say, 10 Ft of speaker cable
How can a simulation of a Power Amp mean much if it does not include such VERY important parameters?
Who cares about a 5 degree shift in signal phase in some transistor collector at 200KHz when the load it's driving has 60 degree shift at, say, 90 Hz?
Please think about it.
Fully agree, of course.
Our friend Sakis is showing his actual experience with real World products.
Simulation is great, excellent tool to save time and money and to pre-check something before building it, but the ultimate test lies in actually building the thing , firing it up and using it to play music at realistic levels, with proper speakers.
One problem I have with Simulation parts libraries, is that all include quite accurate **electronic parts** models, but none that I know of includes models of actual speakers, or speaker cabinets, or their internal crossovers , not even, say, 10 Ft of speaker cable
How can a simulation of a Power Amp mean much if it does not include such VERY important parameters?
Who cares about a 5 degree shift in signal phase in some transistor collector at 200KHz when the load it's driving has 60 degree shift at, say, 90 Hz?
Please think about it.
Because it's a pretty good bet that the amp has more than 60 degrees of phase margin for stability at 90 Hz. Whether it has even 5 degrees at 200 KHz can't be assumed without analysis. More often than not, however, it's all the little things that you DON'T model (but really could if you dug deep enough) that determines the ultimate SQ that can be achieved.
When I've got an output stage that's distributed all over the place - for a big PA amp - I use more than one thermal compensator diode. One directly mounted to one of the NPN outputs, one on the NPN driver, one on a PNP output, and one on their driver. Take up the slack with the Vbe multiplier on the heat sink. For each diode, use an overmolded TO-220 transistor with C-B shorted. It will stay stable even with separate heat sinks.
When I've got an output stage that's distributed all over the place - for a big PA amp - I use more than one thermal compensator diode. One directly mounted to one of the NPN outputs, one on the NPN driver, one on a PNP output, and one on their driver. Take up the slack with the Vbe multiplier on the heat sink. For each diode, use an overmolded TO-220 transistor with C-B shorted. It will stay stable even with separate heat sinks.
Regarding temperature errors with the output transistor case as reference point; don't forget the sense transistor is not a significant sink and plastic case thermal resistance is less relevant when the sinking effect is small compared to that of the main heatsink, where the error will remain high, probably more than 20 C. less than case temp. at the contact surfaces, even. It has been stated here that temp. equillibrium can take 20 mins using typically positioned 'sink mounted sensors. IOW, any precision is a pointless exercise with the traditional line up, placed between the drivers, (of all the stupid places to sense output transistors!). This is likely just the result of manufacturing convenience or perhaps a relic of days when few manufacturers thought there were was any need for precise bias.
It soon becomes clear when experimenting with using this approach, that a different (lower) thermal coefficient is necessary to standard assemblies so you can take advantage of the faster, more accurate bias without overcompensating.
I refer to "Audio Power Amplifier Design Handbook" 5th ed. chapter 15 "modelling the EF output stage" pp 390-398. 'A good read, even if the practicalities of commercial product assembly would probably preclude the refinements. 'Not sure what the the new edition will retain of this text, though.
If you want to be really tricky and use Bonsai's radiated heat sensing of the sensor's collector lead, you still have an error of temperature difference but it could well be a more "manufacturer friendly" assembly. I think that is the main reason we don't see fancy heat sensing mounts although I do have flying lead sensors on my ROTEL 1060. At the price, they should have used audiophile grade silver wire for them, too!
It soon becomes clear when experimenting with using this approach, that a different (lower) thermal coefficient is necessary to standard assemblies so you can take advantage of the faster, more accurate bias without overcompensating.
I refer to "Audio Power Amplifier Design Handbook" 5th ed. chapter 15 "modelling the EF output stage" pp 390-398. 'A good read, even if the practicalities of commercial product assembly would probably preclude the refinements. 'Not sure what the the new edition will retain of this text, though.
If you want to be really tricky and use Bonsai's radiated heat sensing of the sensor's collector lead, you still have an error of temperature difference but it could well be a more "manufacturer friendly" assembly. I think that is the main reason we don't see fancy heat sensing mounts although I do have flying lead sensors on my ROTEL 1060. At the price, they should have used audiophile grade silver wire for them, too!
nice ...
please check tips from the repair world :
AS JM Fahey said and WG_SKI in pro amps it should be more than one sense points i have seen some of Hand MAde In China pro amps mistakes like that , call them convenience mistakes where boards of L and R are identical and sense point is one fair enough and OK for the one board but on the other the sense point is next to the ventilator with catastrophic results
Then again in home amplifiers some users to gain some safety and make a party added a 12V computer vent on the top of the amplifier and still managed to blow the amp ...Then come to my shop wondering why ?
Pretty obvious if you blow cold air on the Vbe this one will think that the amp is cold and act accordingly .Obviously if you need to cool something that will be ONLY the outer area of the heat sink .
An one even more funny
At one of the horrible Elektor designs if memory serves i think the ""Gigante" or something else even more stupid called on the boards there was accommodation for circuits for controlled ventilation but if some of the air was leaking to the front end and blow air on the LTP scheme that temp variation was enough to vary the offset in the output ...
It is well known that if you fart next to one of these amplifiers oscillation may occur together with some offset and bias variations !!!!
Happy regards
Sakis
please check tips from the repair world :
AS JM Fahey said and WG_SKI in pro amps it should be more than one sense points i have seen some of Hand MAde In China pro amps mistakes like that , call them convenience mistakes where boards of L and R are identical and sense point is one fair enough and OK for the one board but on the other the sense point is next to the ventilator with catastrophic results
Then again in home amplifiers some users to gain some safety and make a party added a 12V computer vent on the top of the amplifier and still managed to blow the amp ...Then come to my shop wondering why ?
Pretty obvious if you blow cold air on the Vbe this one will think that the amp is cold and act accordingly .Obviously if you need to cool something that will be ONLY the outer area of the heat sink .
An one even more funny
At one of the horrible Elektor designs if memory serves i think the ""Gigante" or something else even more stupid called on the boards there was accommodation for circuits for controlled ventilation but if some of the air was leaking to the front end and blow air on the LTP scheme that temp variation was enough to vary the offset in the output ...
It is well known that if you fart next to one of these amplifiers oscillation may occur together with some offset and bias variations !!!!
Happy regards
Sakis
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.