Hiraga 20W - thermal issues

[Mikie]

I was going to ask Geoff this question, but it may be better to post it here, for the next victim, I mean enthusiast, that decides to build this.

The driver transistors are supposed to be in thermal contact with one another, so the quiescent current drift is minimised. Am I right so far?
Now, what about making them in contact with the output transistors as well, as in all on a single heatsink? Is this going to improve things or mess it all up?

[GKU]

Hi Mikie,

I have read your post some times a go but did not find time to test your proposal with my Hiraga 20W amps.
Did you do some experiences in the meantime?
I have just build a test arrangement on a 200x200x40mm heatthink, but did not yet finish the new regulated PSU.
I would appreciate if you could report any experiences made in the meantime. As soon as I have finished the new PSU (hope in about 2 weeks) I will report my findings to.

Rgds

GKU

[cunningham]

From a standpoint of design, I would think that minimizing the temperature change of the drivers would minimize bias current change through them. If the drivers were mounted to the outputs, the temperature change would be larger unless of course, your heatsink is large enough to not have a large temperature change. In a recent circuit I use seperate heat sink for drivers(much smaller obviously). However, it is sort of oversized for the application. This seems to give me more stable results since the outputs are biased by the drivers in my circuit and change in temperature of the drivers is small. Your circuit may be different, but temperature coefficient conditions still apply. Maybe the servo should be mounted to the outputs????? A BJT's Vbe turn-on voltage decreases with temperature increase, according to the design of the device, & can lead to thermal runaway.

[EWorkshop1708]

Originally posted by cunningham
From a standpoint of design, I would think that minimizing the temperature change of the drivers would minimize bias current change through them. If the drivers were mounted to the outputs, the temperature change would be larger unless of course, your heatsink is large enough to not have a large temperature change. In a recent circuit I use seperate heat sink for drivers(much smaller obviously). However, it is sort of oversized for the application. This seems to give me more stable results since the outputs are biased by the drivers in my circuit and change in temperature of the drivers is small. Your circuit may be different, but temperature coefficient conditions still apply. Maybe the servo should be mounted to the outputs????? A BJT's Vbe turn-on voltage decreases with temperature increase, according to the design of the device, & can lead to thermal runaway.

So do you recommend mounting the biasing transistor/diodes to the heatsink for the outputs? If the drivers are cool to the touch, but the outputs are hot, is thermal runaway still an issue?

[AndrewT]

Hi,
please excuse me butting in & only knowing half the story.
What kind of output stage is hanging on the end of the Hiraga?
regards Andrew T.

[GKU]

Hi Andrew,

to answer your question:

I have two Hiraga Class A, one is the Super 30W stereo version driving 2x2 FOSTEX FS41RP magnetostats used as dipole and connected in series in my active LS-system, covering the frequency range from 200 - 1500 Hz.
The second one is a 20W mono version driving an ORIS 150 Hornsystem from BD-Design using AER MD3 driver units.
The stereo version I do currently redesign to mono versions due to positiv experiences with the 20W version.

The JLH version of Fig. 2, sometimes is used in the ORIS system to for comparison. Both amps perform similar. Subjectively the Hiraga amp sound a bit more smooth in the heights compared with the JLH, but this can be caused from differend PSU's used. Have not yet tried to power the JLH with the PSU from the Hiraga, which is a regulated version origin developed for the Kaneda 50W amp.

Rgds

GKU

[jazz]

Hi there,

I'm not sure if this contributes to you're question but....
I've build the hiraga 'le classe A' 20 Watt a few times (I think this is the design you're talking about) but have never before read or heard that the drivers should be thermally coupled. This could be my fault but i've never experienced problems without thermal coupling.

If you want to eliminate the dc drift at the output there is another way (at least it worked for me). there is a 1k8 resistor coming from the opposite psu line to the base of the output transistors. Divide this resistor up in to 1k2 / 600R or something like that and put a cap to ground in between the resistors. This solved the dc stability prob for me and had no sonic effects in my situation. That is just my situation though and it might not work for you. Anyway if that's what you're after you might give it a try.

regards
Joris

[GKU]

Joris,

Many thanks for your input.
Recently I have talked with Jean Hiraga about this issue. Jean did not recommend a thermal coupling of the drivers due to problems to get tightly matched pairs. In the case of tightly matched pairs, there should be no problem to thermally couple them. Today I have chanced one of my amps which contains very tight matched drivers by mounting the drivers electrically isolated on one heatsink with ~3.5K/W. After the whole day running and observing the DC-drift I can say that the drift was minimized to ~ +/- 10 mV around a value of 20mV. The temperature measured at the heatsink after 2 hours of running has been stabilized at 38°C over the whole observation time.
The drivers w.o. heatsink achieves about 60-65°C.
But in addition I will try your proposal to.

Rgds

GKU

[cunningham]

Originally posted by EWorkshop1708


So do you recommend mounting the biasing transistor/diodes to the heatsink for the outputs? If the drivers are cool to the touch, but the outputs are hot, is thermal runaway still an issue?

The BIASING transistor (or diodes) biases less to the outputs when it is heated so it should be the same temperature as the outputs(in theory) to control thermal runaway. The drivers should not get very warm so they shouldn't affect it much.

[GKU]

Hi there,

finally I have managed in a pragmatic approach to test DC-thermal drift on my Hiraga 20/30W systems by using thermally coupled driver transistors as well as driver transistors assembled together with the power transistors on one big heatsink in comparison with the original design from the journal L'Audiophile.

1. Original design
This design use for each power trans one single heatsink with ~ 1K/W and no thermal coupling of the driver trans.
Used PSU: original passive version with 1.4 F caps.

Result:
After switching on at RT, the average DC-offset per channel starts at about 90-150 mV offset. After a stabilization time of around 20 minutes the offset vary between +50 and - 10mV. which according to Jean Hiraga is normal due to thermal sensitivity of single heatsinks and no thermal coupling of the driver trans.
Power heatsink temperature after stabilization: 82-85°C.

1.1 Original design but thermally coupled driver trans
The drivers has been assembled electrically isolated but thermally coupled on one small heatsink with ~3.5W/K .

Result:
The initial DC-offset was the same, but the stabilization time could be reduced to ~15minutes. The final offset vary similar between +50 and 0mVm, thus no major improvement achieved.
Power heatsink temperature after stabilization: 82-85°C.
Driver heatsink temperature: 36C

2. Modified mono version
2.1 driver and power trans assembled together thermally coupled on one big heatsink.
In this case, the same cirquit as used for tests in the original version with the same driver and power trans were assembled together on one single big heatsink (200x200x40mm, 0.4K/W).
Used PSU: original passive version with 1.4 F caps.

Result:
Due to higher heatcapacity of the heatsink, the thermal stabilization take longer ~20-25 minutes. But the offset could not be stabilized and vary between +150 and -65mV.
Power heatsink temperature after stabilization: 72-75°C.
Conclusion: Not applicable

2.2 Same as 2.1 but driver trans thermally coupled on one separate small heatsink like 1.1
Used PSU: original passive version with 1.4 F caps.

Result:
Time till thermal stability same as 2.1 but DC-offset slightly reduced and more stable at +40 to 0mV.
Power headsink temperature: 72-75°C
Driver headsink temperature: 36C
Conclusion: Not as bad

2.3 Same as 2.2, but PSU regulated with additional caps as described in the post "Hiraga 20W layot".

Result:
Time till thermal stability same as 2.1/2.2, but DC-offset again reduced and much more stable (+30 to +10mV)
Conclusion: Best result!

Remark:
All transistors used were tightly matched. Driver trans gain 115 and power trans gain 130.


Wrap up:
It seems that a more stable PSU will positively impact offset stabilization.
Driver and power transistors assembled thermally coupled on one heatsink can not be recommended.
Thus, my two mono versions using already the big headsink have been finally modified by thermally coupled driver trans on one small 3.5 K/W heatsink.
The original 20/30W stereo amp will be transferred soon into two mono versions according to the 2.3 configuration.

One additional and important remark i have to make:

At Class A amps, the heatsinks must be thermally isolated from the case to allow the heatsink to thermal stabilize and do not heat up the case influencing the parameters of the other transistors on the main circuit, which also influence DC-thermal drift.

I hope this tests can help somebody to fine tune their equipment.

Rgds

GKU