Biasing/thermal compensation of Thermal Trak transistors

[Bob Cordell]

The OnSemi ThermalTrak transistors are in the nice big T0-264 plastic package. This package appears to have significantly more metal heatsink contact area, which should provide a much better Theta C-S than the TO3P or TO-247 packages. This is nice.

But does anybody here know of a decent source for TO-264 mica or other type insulators?

Thanks!
Bob

[AndrewT]

Hi,
the To247 insulators I use are just wide enough for To264.
They have plenty height.

The sticky back versions can be attached accurately to the back of the To247.
I would not be happy using the small mica insulators for the bigger package.

[Bob Cordell]

Originally posted by AndrewT
Hi,
the To247 insulators I use are just wide enough for To264.
They have plenty height.

The sticky back versions can be attached accurately to the back of the To247.
I would not be happy using the small mica insulators for the bigger package.

Thanks, Andrew.

What material are your sticky-back versions made of (e.g., aluminum oxide, silicon rubber, etc.) and how does its thermal resistance compare to mica with grease?

Regards,
Bob

BTW, I was thinking that the ThermalTrak diode would allow for an excellent opportunity for doing an experiment for comparing thermal resistance of various insulation approaches (including no insulation), by heating up the ThermalTrak BJT with bias and measuring the junction temperature by measuring the change in the junction drop of the ThermalTrak diode.

[AndrewT]

Hi,
I am using Bergquist Sil Pad 900S.
1.6W/mK,
0.2C/W (but this one omits the contact area, so is kind of close to useless as a spec).

I think this is slightly better than 0.002inch thick Mica.

Sil Pad 2000 is twice as conductive, but is 65% thicker and 300% dearer.

[pooge]

GROUP BUY!!!

[anatech]

I will very happily stick to - I mean with - mica and grease.

-Chris

[GK]

Originally posted by Bob Cordell
Looks like it worked. For convenience, here is the complete post.


This is a ThermalTrak bias circuit that I came up with that I have been working with. It shows a couple of examples of how the ThermalTrak diodes can be used.

The circuit is in the form of a Class-AB 100 watt output stage in simulation form. The circuit is essentially a triple Darlinton “T” circuit that employs two pairs of NJL3281D – NJL1302D OnSemi devices.

Q1-4 just make up a pair of 10 mA current sources that, in combination with the input voltage source, simulate a typical VAS.

Q5-6 and D1-2 make up a feedback-based bias spreader. D1-2 are the key ThermalTrak diodes that work to establish a stable, temperature-tracking bias. The use of the feedback-based bias spreader, as opposed to a conventional Vbe multiplier, eliminates the influence, drift and uncertainty of the pre-driver and driver Vbe’s in establishing the output stage operating point.

Q7-8 and Q9-10 are the pre-drivers and drivers. D5-6 provide a diode drop of voltage offset to make up for the Vbe of Q5-6. D5-6 and Q5-6 are NOT on the heatsink. The relationship of voltage drop for a given current for D5-6 as compared to Q5-6 is involved in setting the bias current.

The voltage drop across R5-6, caused by R7, sets the voltage that will ultimately appear across the emitter resistors R13-16 (albeit modified slightly by some Vbe differences). In practice, R7 would include the bias-setting pot.

The d.c. feedback from the emitters of Q9-10 to the bases of Q5-6 sets the output stage operating point, and forces a tracking relationship between D1-2 and the Vbe’s of the output transistors.

If D1-2 had the same junction drop as Q13-14, and Q5-6 had the same junction drop as D5-6, you can see that the voltage drop across R5-6 would have to equal the voltage drop across R13-14. Corresponding junction drops are at essentially the same temperature. Note that power dissipation, and thus self-heating, in Q5-6 and D5-6 is quite low. This is how the feedback-based bias spreader works. Although in practice the above-mentioned equalities do not hold perfectly, they are close enough, and net differences are made up by trimming R7 to set the idle current.

The two “extra” ThermalTrak diodes, D3-4, are used between the emitters of the driver transistors to establish the idle current of the drivers while keeping the impedance between the emitters very small, so that the drivers can operate in push-pull to provide turn-on and turn-off current to the output transistors. Keeping the impedance between the driver emitters very low at high frequencies is essentially what is often done with the speedup capacitor. The ThermalTrak diodes make it possible to do this in a d.c. fashion with the necessary precision because they track the output transistor Vbe’s.

Cheers,
Bob

Just out of curiosity, have to tried running a sim with lower value bypass/coupling caps for the "base spreader" circuit transistors.
You've got a pair of nfb loops there enclosing the driver and pre driver BJT's, which you might be able to make use of at higher frequencies by using smaller caps. That could provide a lower drive impedance for the output transistors.

Cheers,
Glen

[AndrewT]

Originally posted by anatech
I will very happily stick to - I mean with - mica and grease.

I use the technique of rotating the greased interface as the bolt is tightened up. The rotation is limited to about +-20degrees in a oscillatory motion to expell the excess componund and more importantly squeeze out any trapped air bubbles.

If this is done with To247 mica on a To264 package there is a high risk of misalignment of the insulator since it is "hidden" from view.

Has anyone found a supplier of To264 sized mica in the UK?

[Bob Cordell]

Originally posted by G.Kleinschmidt



Just out of curiosity, have to tried running a sim with lower value bypass/coupling caps for the "base spreader" circuit transistors.
You've got a pair of nfb loops there enclosing the driver and pre driver BJT's, which you might be able to make use of at higher frequencies by using smaller caps. That could provide a lower drive impedance for the output transistors.

Cheers,
Glen

Hi Glen,

You have a very good point here. The circuit I showed was mainly for simulation illustrative purposes to keep things simple. There are actually a number of different ways that the bypassing/local compensation can be done. The brute-force way I chose to illustrate is particularly convenient because of the fact that for simulation I feed the "center point" from a voltage source. In a real amplifier, the signal feed comes from both ends of the spreader with the upper and lower current source transistors replaced with the VAS transistors.

I did a couple of simulations with hard drive at high frequencies where I actually saw some difference (improvement) when I used LARGER capacitors. In a real amplifier, with this brute-force arrangement, I'd probably use 10 uF electrolytics, possibly bypassed by 0.1 uF caps. However, there are also more elegant approaches to compensating/bypassing this arrangement, such as putting collector-base capacitors around the spreader transistors while deliberately increasing somewhat the source impedance to their bases. This pretty much amounts to local Miller effect compensation.

In any of the bypassing/compensation approaches for this scheme, where we are considering the local closed-loop effect on impedance at the output of the driver transistors, we need to distinguish between common-mode impedance and differential mode impedance. In this context, I would more clearly define it as signal-mode impedance and bias-spreading-mode impedance. The local feedback in this bias scheme will, I believe, primarily affect the bias-spreading-mode impedance. However, I must admit that I was not consciously trying to manipulate those impedances here and have not given it a whole lot of thought.

Cheers,
Bob

[pooge]

Originally posted by Bob Cordell



I think the sensing diode is so large (same junction drop as the BJT Vbe at 1/4 the current) so as to make things about right at a typical value of bias current. For example, if the bipolar is biased at 100 mA Class-AB idle current, the junction drop of the thermal diode will be the same at a reasonable current of 25 mA. One could argue, I suppose, that a "reasonable" current might be more like a 10 mA value flowing in the VAS. There is also the possibility that OnSemi was deliberately playing some games with delta-Vbe temperature characteristics of junctions operating at different current densities (as is the practice employed in bandgap voltage references).

Anyway, I was a little surprized at the size as well.

Cheers,
Bob

Can't this be said about any other temperature sensor scheme, in that the Vbe transitor or sensing diode(s), used to sense temperature, are operated at a meager current wrt the current in the output transistor(s). In other words, this issue with the TermalTrak diodes does not appear to differ from any other scheme.