Vbe Thermal Coupling Issue

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am trying to use D.Selfs suggested method of thermal coupling the Vbe Multiplier by attaching it directly to the top of a TO-247 output (his example is TO-3) transistor. It is an EF output topology. Setting quiescent current is cery "fiddly" it is very hard to find a stable point. Any distrubance sends the measured current (volts across the RE resistors, actually) off in one direct or the other. It is kind of like sitting in the center of the saddle surface that is often used to describe chaos theory!

Setting quiescent current is not new to me, but this behavior is. The heatsink does not seem to be an issue as the situation persists even with a "monster" sink. However, in the past I've always bolted the Vbe multiplier to the main heatsink. Is it possible that the Vbe multiplier to to sensative and over-compensates in one direction or the other? Is TO-247 just not ameanable to this - my temperature probe says the heatsink gets hotter than the top of the package so that is another suspiicion I have.

I'm going to experiment with a thermal bar and with going back to main heatsink attachment. But while I'm doing the metalwork, I would be interested if anyone has experienced something similar.
 
This fact is something I have already commented altough nobody seemed to trust me. I also did the same experiment in the past and got the same frustrating results as you

The worst place to mount a Vbe multiplier is over the plastic case of a power device. This plastic resin usually has very high thermal resistance and just does not follow the die temperature at all, actually it's allways sustantially cooler than the heatsink and much cooler than the die itself. The temperature of the plastic surface is mostly dependent on ambient temperature only
 
Eva,

Thank you for your prompt reply. I made a some searches on the forum but didn't come up with your post. Anyway, it is reassuring to find confirmation of my results.
----------
Anatech,

I had considered a thermal bar to address the multiple device issue, but Eva's comment makes that moot.
----------------
All,

I noticed some time back that ESP's P3a doesn't couple the Vbe multiplier but doesn't suffer for it. I'm wondering if the thermal coupling business isn't as critical as claimed.
 
Administrator
Joined 2004
Paid Member
Hi sam9,
All bipolar stages need thermal compensation of some sort unless they are not biased on. Whether they need a vbe multiplier or not is another question. Your thermal bar would be the heatsink in this case.

I am just prototyping an amplifier right now that does not appear to need a vbe multiplier. I designed for it, but left it unstuffed on the board. Bias seems stable right now. However, the circuit has built in compensation, the vbe multiplier was / is in case I need it at higher power levels and temperatures. I won't know until I beat it up. That is some time away yet.

So depending on your circuit and thermal layout, you may or may not require the vbe multiplier (bias control circuit).

-Chris
 
I'm quite sure mine does need it.

I've seen cicuits by Hagerman and Self which are claimed improve the tracking and stability but sill require some kind of thermal coupling.

Just as interesting was a comment that the need for thermal tracking could be avoided by using output triples. The example cited was the Quad 303. I look at the schematic and it is quasi-complementary. I think the reason you can avoid the coupling requirement is that it's the pre-drivers which which determine the bias of the rest And they can be kept at a low stable temperature if the current demants are small enough. I assume thiwould not apply to an EF tripple but would to a full CFB with predrivers.

But these are all ideas for the future. Right now I have to go drill and tap a heatsink!
 
My explosive experience;

I had an experience once where my Vbe transistor was mounted directly to a very large heatsink.

The FETS ran away thermally for no apparent reason very quickly after the bias seemed to settle . This happend in a few seconds and resulted in destroyed FETS. The FETS were warm and the heatsink was cold.

My diagnosis was the heatsink was too big (wierd huh?) and / or the sensor location on the heatsink was thermally too far from the output devices.

I fixed this by attaching the output devices and Vbe transistor to a seperate mounting bar and then attaching this to the large heatsink. The Vbe now senses the temperature of the mounting bar rather than the heatsink and this improved the tracking no end.

This solution was specific to that amp. On smaller heatsinks I mount the Vbe directly on the heatsink within 1-2 centimetres of the output devices. This is helps track the output devices closely. For large heatsinks I always try to use a seperate mounting bar. I never mount the Vbe directly to an output device. Even with matched devices one could cruising while the one next to it is screaming, but the Vbe multiplier thinks everything is fine beacuse it is mounted on the cruiser.

A friend of mine has (and continues to) mount the Vbe on the FET directly. He has not had any failures but his bias does swing a lot more than mine. I think he is lucky so far.

My $0.02 worth.


Cheers
 
In my experience not all bipolar devices show the same thermal behavior in all circumstances :

- Devices with lower Vce rating [<=100V] start turning on at a lower Vbe [0.5V] and show a much sharper knee of current versus Vbe turn on characteristic. These devices are harder to thermally stabilise and his Vbe-on drops quicker as temperature increases. Devices with higher Vce rating [>=150V] show higher Vbe-on and more progressive turn-on, being easier to thermally stabilise and showing also less impact of temperature over Vbe

- Vbe-on and Vbe temperature coefficients are not constant. They depend mostly on Vce and Ic. For the same Ic : The higher the Vce the lower the Vbe-on and the higher the temperature coefficient of Vbe, so the same transistor model is much harder to thermally stabilise at higher Vce than at lower Vce

- In the low Ic range [<500mA approx.] and given fixed Vbe and Vce values : The higher the Ic, the quicker Ic increases as temperature is increased. This means that the same amplifier may be thermally stable with 50mA bias but may suffer thermal runaway if bias is increased to 100mA


To understand all this you may try to thermally stabilise a pair of BD911/BD912 or TIP35C/TIP36C devices [rated at 100V Vce] with +-45V rails and 50mA bias. Then try to stabilise a pair of Sanken 2SA1295/2SC3264 devices [rated at 230V Vce] with the same +-45V rails and 25mA bias. You may also repeat the experiment with +-15V rails for the 100V Vce devices and +-85V rails for the 230V Vce devices

Huge differences in bipolar transistor behavior will show during these experiments

Now I wonder why Rod Elliot suggests 250V and 350V Vce devices for +-35V supply rails. This is an overkill, there are better things to do with these devices, I think

Also note that the only way I've found to thermally stabilise 100V Vce devices with up to 100mA bias when used near its Vce and dissipation limits is to use a buffered Vbe-multiplier with a low-vbe transistor as the sense element [PN2222A], operated at an Ic around 100uA for maximum temperature coefficient [these small signal devices show their higher temperature coefficients at very small Ic]

Actually the temperature coefficient of any Vbe multiplier may be adjusted by increasing or decreasing sense transistor Ic to match the output devices behavior, and also to compensate for the temperature difference between heatsink and transistor dies, whose value is allways proportional to heatsink temperature
 
sam9 said:
I noticed some time back that ESP's P3a doesn't couple the Vbe multiplier but doesn't suffer for it. I'm wondering if the thermal coupling business isn't as critical as claimed.

That's because the P3a output stage is Complementary Feedback Pair where the Vbe of the output devices (and hence their thermal instability) is servoed out by the drivers. Thus you only need to sense the driver thermal situation. As the drivers run cool/don't see much thermal cycling, it's safe to run the Vbe multiplier in free air.
 
Hmm, i have the reverse problem...
I use a bd139 as vbemultiplier mounted between outputdevices on
heatsink. When amp is powered on, after ~10secs the quiscent
current reaches it's maximum, and then drops below the value
at the beginning, when amp was completely cold. The problem is,
the amp can drift into classb if getting very hot...
It starts with 100ma, climbs to 120ma, then drops to 50ma.
Outputdevices are mjl3281/1302. What am i doing wrong ?

Mike
 
MikeB

I think you are experiencing too much negative feedback
Try to run the BD139 at a higher Ic, this will reduce the amount of negative feedback. Try also a sense transistor with higher Vbe like BC54x/55x

Also, I'm curious about :

- What is the Vbe of the output devices just when the heatsink is at ambient temperature and the amplifier is turned on?

- What is the bias current measured *inmediately* after playing at full output swing for 5 to 10 minutes?

Remember that the transistors are allways substantially hotter than the heatsink when playing loud and that they quickly cool down to the temperature of the heatsink when you stop playing. This has a great impact over the bias current, so its actual magnitude is only measurable inmediately after playing, idle measurements are useless. I recommend keeping the multimeter connected and switching to the right scale just at the same time you pause the music

Your amplifier may be actually biased at 200mA at high temperatures, altough you may only see 20mA at idle
 
MikeB,

I had same problem with one of my amplifier. The solution is to keep the Vbe multiplier colder. You can increase the current of the VAS, but I think for this action You have to redesign the whole amplifier. This is not good idea.
So the Vbe multiplier contains the transistor itself, and the resistors which set the voltage. My experience is that the current sharing between the resistors and the transistor makes fine adjustment. More current on the transistor makes the compensation harder, and more current on the resistors results less compensation. So very low current on the resistors can result problem, so 10% is the minimum I recommend.
Another solution to mount the sensor transistor far away from the output devices. If it's hard to do, just use more insulators between the heatsink, and the Vbe multiplier.

sajti
 
Eva said:
MikeB

I think you are experiencing too much negative feedback
Try to run the BD139 at a higher Ic, this will reduce the amount of negative feedback. Try also a sense transistor with higher Vbe like BC54x/55x

Also, I'm curious about :

- What is the Vbe of the output devices just when the heatsink is at ambient temperature and the amplifier is turned on?

- What is the bias current measured *inmediately* after playing at full output swing for 5 to 10 minutes?

Remember that the transistors are allways substantially hotter than the heatsink when playing loud and that they quickly cool down to the temperature of the heatsink when you stop playing. This has a great impact over the bias current, so its actual magnitude is only measurable inmediately after playing, idle measurements are useless. I recommend keeping the multimeter connected and switching to the right scale just at the same time you pause the music

Your amplifier may be actually biased at 200mA at high temperatures, altough you may only see 20mA at idle


Eva,

my experience is that few milliamper change of the current on the Vbe multiplier makes not too much change. But extreme low currents (less than 1mA) makes the compensation better. So I found that some darlington works well on this situation, because the very small current on the first transistor.
I used BC517, with very good result for 5mA Ic. I guess that the first transistor runs with 10-20uA only.
For better compensation use the Leach type multiplier, with diodes. I found that it makes the compensation more sensitive.

sajti
 
Eva said:

- What is the bias current measured *inmediately* after playing at full output swing for 5 to 10 minutes?


I haven't played at full swing for a long period, it's too loud...
But i played it at least to increased temp, i guess 40°, and the
current was down to ~6ma. After about a minute the current
returned to 50ma. It seems to me that the circuit is overcompensated.

The Ic through the bd139 is very low, below 2ma, maybe the bd139
has too big thermal drift at these low currents ?
I chose bd139 because of its to126 case. I use bd139 from onsemi.

I have not done exact measurements on the vbe's of outputtransistors yet...

to sajti, the current through the resistors is quite high, i use a 5k
plus a 1kpot+500ohm, makes about 6k for 3.6v. Vas current is 2ma.

Mike
 
Ok, I've moved the Vbe multiplier to the heatsink. It behaves much better - not rock steady but +/- 2mV (or mA equivalent) around the desired point. It takes about 10-20 minutes just to get the temperature up to a stable point, then after that a series of small adjustments eachfollowed by a waiting period of a couple of minutes until it settles down to the new value.

Now I want to go back to some of the info in these posts about different transistors for this (I'm using an MJE340 because that's what I have on hand) and forcing more (or less) current through the muliplier by changing the resistors. Different behavior of with different transistors had ocurred to me but the division of current between rersistors and the bias transistor had not.
 
Also note that the only way I've found to thermally stabilise 100V Vce devices with up to 100mA bias when used near its Vce and dissipation limits is to use a buffered Vbe-multiplier with a low-vbe transistor as the sense element [PN2222A], operated at an Ic around 100uA for maximum temperature coefficient [these small signal devices show their higher temperature coefficients at very small Ic]

Eva,

Two questions:

Could you post an example schematic of the buffered Vbe-multiplier?

As a practical matter how do you mount a TO-92 device to a heatsink? (Themal adhesives like are used on some Pentium heatsinks? A hand-made spring clip?)
 
MikeB said:


I haven't played at full swing for a long period, it's too loud...
But i played it at least to increased temp, i guess 40°, and the
current was down to ~6ma. After about a minute the current
returned to 50ma. It seems to me that the circuit is overcompensated.

The Ic through the bd139 is very low, below 2ma, maybe the bd139
has too big thermal drift at these low currents ?
I chose bd139 because of its to126 case. I use bd139 from onsemi.

I have not done exact measurements on the vbe's of outputtransistors yet...

to sajti, the current through the resistors is quite high, i use a 5k
plus a 1kpot+500ohm, makes about 6k for 3.6v. Vas current is 2ma.

Mike


Mike,

just try to use half of all the resistors in the Vbe multiplier. So, 2.2k+470ohms pot+220ohms looks OK for my point of view.
To test Your system is not necessary to drive the amp with full power for long time. Just put some clothes over the heatsink, and let your amp to heating up. Keep the monitoring of the bias, during this period. Just check if the bias remains same even with different temperatures.

sajti
 
sam9 said:


Eva,

Two questions:

Could you post an example schematic of the buffered Vbe-multiplier?

As a practical matter how do you mount a TO-92 device to a heatsink? (Themal adhesives like are used on some Pentium heatsinks? A hand-made spring clip?)

I made 5mm diameter, and 5mm deep hole into the heatsink, and fixed the TO-92 device, with some glue. It works well and the failure of this transistor is not common.

sajti
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.