wrote:
It provides negative thermal feedback.

The voltage across the transistor will decrease as the transistor
temperature increases. This prevents thermal runaway.

That's what it's supposed to do... I claim it can't work as shown. It's
not quite that simple as tacking a few diodes on the bias circuit and
laying them on top of the transistors... LOL

www.telstar-electronics.com

Tacking the diodes on to the case of the transistor works better than
attaching them to the heat sink due to the relatively slow thermal response,
the heat sink changes temperature slower that the cap on the transistor. If
you want as fast a response as possible then the diodes are placed about as
well as you can get. I've also seen small diodes clipped on to the flange of
TO-3 style cases for the same reason. I've seen some high power transistors
used for motor drive and high frequency inverter use that has the
temperature sensing diode, or thermistor, built inside of the package.

http://pdf1.alldatasheet.co.kr/datas...50HM75STG.html

As the temperature of the base-emitter junction goes up the required forward
voltage for a given base current decreases. If nothing was done, the
base-emitter bias voltage held fixed, the base current for increases and
thus the collector current too. Of course increased collector current
results in more power dissipation, thus more heat and higher resulting
temperature, i.e. thermal runaway.

The diode used for bias temperature compensation has to very closely match
the Ib-Vbe curve of the transistor to work effectively . You can't use just
any diode and expect it to work very well. The problem with mounting the
diode on the heat sink is the large thermal mass. If the transistor starts
to go in to thermal runaway the heat sink will take a significant time to
heat up, thus reducing the bias to the base-emitter junction, long after the
runaway has started. You want to get the compensation right as soon as
possible, reduce the base-emitter bias, before the junction gets too hot.
The small cap on the transistor will heat up much faster that anything else
giving the designer a chance to design a fast responding bias circuit which
could avoid transistor destruction.

Diode bias circuits are on their way out in favor of more advance bias
circuits. The chief problem with diode bias compensation is either too much,
the transistor gets starved for needed bias current, or not enough, tendency
to go in to thermal runaway.

The reason the transistor can get starved for bias current is simple. The
Beta, current gain, of a bipolar transistor is not constant over the
operating range, it varies based on the instantaneous operating point. At
some point the gain is so low that the current flowing through the bias
compensation diode may drop to a low value, or go to zero, limiting the bias
current forcing the operation to go non-linear over part of the cycle.
That's why some of the simple diode bias circuits draw so much standing
current. It has to supply the worst case bias current. Newer active bias
circuits use a voltage source type design. The bias voltage tracks the
transistor temperature and since it is a "voltage source" there is no
practical limitation on the bias current. The transistor will draw what it
needs without being limited by the current limiting resistors in the diode
bias type of circuit.

I'm sure somebody is going to nit-pick this post. They're welcomed.

--
Regards,
Leland C. Scott
KC8LDO

"Telstar Electronics" wrote in message
ups.com...
wrote:
It provides negative thermal feedback.

The voltage across the transistor will decrease as the transistor
temperature increases. This prevents thermal runaway.

That's what it's supposed to do... I claim it can't work as shown. It's
not quite that simple as tacking a few diodes on the bias circuit and
laying them on top of the transistors... LOL

www.telstar-electronics.com

On Sun, 24 Sep 2006 23:30:24 -0400, "LeIand C. Scot"
wrote in :

snip
I'm sure somebody is going to nit-pick this post. They're welcomed.


Not picking nits, just taking a different approach.....

Thermal bias compensation works to a point but neglects one important
issue: it takes time for heat to travel from the junction to the
outside of the package, and thermal runaway can happen faster than a
-thermal- compensation circuit can respond to it. Since the heat
buildup is due to excessive EC current, it makes more sense to control
the EC current directly.

There are two solutions that use this approach. One is to include a
resistor on the emitter with a TC opposite of the transistor. Not
precision but much faster response. The other is to monitor the EC
voltage and the base current; send both measurements to a differential
OP amp and use the output as feedback for the bias regulator.

I tend to favor the first choice because it has the added benefit of
improving linearity. I would only use the second choice in a high
profit, 'stick-it-to-the-consumer' product.

On Sun, 24 Sep 2006 21:59:32 -0700, Frank Gilliland wrote:

On Sun, 24 Sep 2006 23:30:24 -0400, "LeIand C. Scot"
wrote in :

snip
I'm sure somebody is going to nit-pick this post. They're welcomed.


Not picking nits, just taking a different approach.....

Thermal bias compensation works to a point but neglects one important
issue: it takes time for heat to travel from the junction to the
outside of the package, and thermal runaway can happen faster than a
-thermal- compensation circuit can respond to it.

Exactly. That's why those diodes are place on the ceramic cap of the
device and not on the heat sink.

Since the heat
buildup is due to excessive EC current, it makes more sense to control
the EC current directly.

There are two solutions that use this approach. One is to include a
resistor on the emitter with a TC opposite of the transistor. Not
precision but much faster response.

This is done in many audio amps. The chief problem is the negative
feedback introduced by the emitter resistor. At auto frequencies this
resistor is bypassed by a rather large electrolytic capacitor sized such
that at the lowest frequency of interest the reactance is much smaller
that the emitter resistor value. Thus the "AC" gain isn't affected much by
the emitter resistor.

Believe it or not I've seen many of the old Motorola RF devices use
internal emitter resistors. Those took the form of many small tungsten
bonding wires from different areas of the emitter structure to the emitter
terminal. The main idea there was the many wires, resistors, in parallel
resulted in a very small overall emitter resistor. Also they found that a
problem called "second break down" would occur if they didn't do this.
What it amounted too was local hot spots, thermal runaway, in tiny areas
of the transistor's emitter structure. I think the term they used for RF
devices built this way was "emitter ballasting".

The other is to monitor the EC
voltage and the base current; send both measurements to a differential
OP amp and use the output as feedback for the bias regulator.

You would have to look at the "DC" emitter current minus the "AC"
component, which I don't think is going to be so easy to do.

Regards,
Leland C. Scott
KC8LDO

Leland C. Scott wrote:
Exactly. That's why those diodes are place on the ceramic cap of the
device and not on the heat sink.

Having the sensing on top of the transistors is a poor location.
The internal die is in intimate contact with the heat sink... not the
top! The heat sink... preferably near the device is the proper location

for any tracking device.

http://auctions.yahoo.com/i:SkyWave%...fier:117239910

"Telstar Electronics" wrote in message
ups.com...
Leland C. Scott wrote:
Exactly. That's why those diodes are place on the ceramic cap of the
device and not on the heat sink.

Having the sensing on top of the transistors is a poor location.
The internal die is in intimate contact with the heat sink... not the
top! The heat sink... preferably near the device is the proper location

for any tracking device.

Are you really that dense?

U-Know-Who wrote:
Are you really that dense?

Thanks for that great post... you really add a lot to any conversation.

www.telstar-electronics.com

On 25 Sep 2006 16:27:30 -0700, "Telstar Electronics"
wrote in
. com:

Leland C. Scott wrote:
Exactly. That's why those diodes are place on the ceramic cap of the
device and not on the heat sink.

Having the sensing on top of the transistors is a poor location.


It's better than the heat sink.


The internal die is in intimate contact with the heat sink... not the
top!


Wrong. The ceramic package isn't hollow; on the contrary, it contacts
more of the junction's surface area than the heat sink flange (which,
BTW, doesn't make "intimate contact" with any part of the junction
because it is insulated from the die by the Be Oxide substrate). The
fact is that a transient pulse can heat and blow the junction before
it can dissipate into the -anything-, which it's more likely to happen
when the transistor is already hot from normal operation.


The heat sink... preferably near the device is the proper location

for any tracking device.


What part of "heat sink" don't you understand?

Frank Gilliland wrote:
Wrong. The ceramic package isn't hollow; on the contrary, it contacts
more of the junction's surface area than the heat sink flange (which,
BTW, doesn't make "intimate contact" with any part of the junction
because it is insulated from the die by the Be Oxide substrate).


If the ceramic cap is tied so well to the heat source internal to the
transistor... then why don't you just attach the heat sink to the
ceramic caps on your new design?... LOL

www.telstar-electronics.com

On Mon, 25 Sep 2006 19:06:49 -0400, "Leland C. Scott"
wrote:

+++Believe it or not I've seen many of the old Motorola RF devices use
+++internal emitter resistors. Those took the form of many small tungsten
+++bonding wires from different areas of the emitter structure to the emitter
+++terminal. The main idea there was the many wires, resistors, in parallel
+++resulted in a very small overall emitter resistor. Also they found that a
+++problem called "second break down" would occur if they didn't do this.
+++What it amounted too was local hot spots, thermal runaway, in tiny areas
+++of the transistor's emitter structure. I think the term they used for RF
+++devices built this way was "emitter ballasting".
*********

I have never seen tungstun bonding wire. All I have ever seen is gold.
You need a soft malible metal to bond to the die pads on any
semicondcutor. The bond wire is sonic heated to the aluminum metal die
pad. This forms the nice ball on the die pad that is a weld of the
aluminum and gold. Tungstun is far to hard a metal for bonding.

Emmitter ballasting is done on the die within the emmitter matrix.
There are several metods of fabricating an RF transistor. Major
factors are power, frequency and device operating point. For most
transistors operating below 50 MHz use an interdigitated emmitter
geometry. Incorparated within are current balancing resistors in the
emmitter matrix. This does increase die size and reduces gain. It does
spread heat and current more evenly through the die. Interdigitated
emmitters will have multiple bond wires.