diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Solid State (http://www.diyaudio.com/forums/solid-state/)
-   -   Moving the Multiplier to the heatsink, or To3, or Collector lead (http://www.diyaudio.com/forums/solid-state/221349-moving-multiplier-heatsink-to3-collector-lead.html)

AndrewT 11th October 2012 11:15 AM

Moving the Multiplier to the heatsink, or To3, or Collector lead
 
The plastic package transistors have a metal backplate that is directly connected to the Collector of the BJT transistor (Drain in a vertFET and Source in a latFET).
The contact face of the backplate is at Tc.
The inner face of the backplate is at a slightly higher temperature than Tc (Tc + a bit). This is due to thermal resistance to heat flow across the thickness of the backplate.
The plastic package in contact with the front surface of the backplate will be at the same slightly elevated temperature of Tc + a bit.
The outer surface of the plastic package will be at a very considerably lower temperature than Tc + a bit. Again this is due to the thermal resistance of the plastic to the heat flow from the backplate through to Ta on the inside face of the amplifier. It is likely that the plastic is >10C lower than Tc and possibly >20C lower. The heatsink Ts adjacent to the backplate will be Tc - a bit, again due to thermal resistances and heat flows.
The collector lead is directly attached to the backplate with a fairly thick copper lead. Some are welded on, but many (collectors) are part of the backplate copper. The temperature of the lead next to the plastic package is very close to Tc in temperature. This is because there is little heat flow out through the collector lead.
Nowhere, external to the transistor package, is there any temperature that monitors Tj.
All locations are a compromise. That compromise determines what ratio of temp change the Vbe multiplier has to apply to approximate the tempco required to maintain near optimum ClassAB bias conditions for minimising
I reckon that the collector lead is the closest in Tc for a monitoring temperature outside of the package.
We could use D.Self's or Cordell's simulation methods to arrive at a model for this collector lead monitoring.
Any volunteers?

Quotes from Previous thread for context below:

Quote:

Originally Posted by madtecchy (Post 3196671)
Hi Nico.
Forget i mentioned the changing the 0.25 for 0.1. I had another little read of Selfs book and it mentions a lower resistance lowers the distortion But that is regarding CFP outputstage and not EF2. Also i noticed another problem maybe? the bias transistor is close to the heatsink but not quite touching it.. after reading Self i understand the thermal issues with an EF2 stage need etra care. i propose to stick the bias transistor to one of the output devices to hopefully improve this area..

Kind Regards Mark

Quote:

Originally Posted by Nico Ras (Post 3197367)
A lot of people (on DIY) stick this transistor on top of one of the output devices. I have seen this a lot in the threads here.

I don't subscribe to this and believe it is best to tie this transistor onto the heat sink so it responds slowly to the temperature variation a form of hysteresis.

Tying onto one of the output transistors can introduce thermally generated distortion in that the thermal compensation reacts quickly to one transistor but slowly to the other thereby distorting or compressing the output signal.

This is measurable, but maybe not audible at lower temperatures.

Quote:

Originally Posted by AndrewT (Post 3197393)
buy some sot23 transistors.
Glue one sot23 with the legs sticking out/up, to the collector leg of an output BJT.
Use the legs of the sot23 for your Vbe multiplier.

This gives a much quicker response to Tj changes and helps maintain the optimum bias level. Not as good as the NJLs from ONsemi, but far better than using the heatsink coupled multiplier.

Nico,
you'll gather from my reply, that I disagree with your philosophy.

Quote:

Originally Posted by madtecchy (Post 3197409)
I Nico .
I did think about hysteresis. I also read that the Upper surface of output devices is the hottest part opposed to the heatsink. I also read that EF2 is more critical of the bias point .. Now to do some more reading as here is a compromise . Wich is going to give less distortion..

1: The way it is now with very loose thermal coupling ( poor bias tracking ?)
2: Affix bias transistor to one output device ( only sensing heat of one output device )
3: Use NLJ,s insted of existing output devices ( very good thermal tracking of bias )
4: Maybe it is possible to use a bias transistor on both output devices and get an average of both just like on a heatsink with the bias transistor fitted between the devices directly to the heatsink.

Kind Regards Mark

Quote:

Originally Posted by madtecchy (Post 3197431)
Hi Andrew..
Looks like i have opened a can of worms. . im begining to come to the conclusion that most every component change in an amplifier raises lots of other conciderations and has an affect on other parts of the circuit .. The amplifier in question was gained very cheaply and sounds very pleasing. as this is a working amplifier it helps me relate the theory to the practical element i find this the best way for me to learn .. I think i need to make a list and prioritise the modifications i would like to make .

The effect on lowering distortion by using current mirrors " low cost transistors "
seems like one of good value . however this will alter the gain of the front end.
and therfore the overall gain of the complete amplifier.. Is it possible to alter the gain of the front end only, to bring it to a level the next stage can deal with .. Maybe i have this all wrong . please bear with me as the learning curve is starting to get steeper the more i start to learn..

Kind Regards Mark

Quote:

Originally Posted by madtecchy (Post 3197434)
Hi again andrew .
Can i not just use the existing bias transistor wich is a to92 package and glue that to the output device????

Quote:

Originally Posted by AndrewT (Post 3197452)
A To92 is massive compared to a sot23.
The high mass of the To92 will reduce the response speed and increase the temperature error from the Tc you are trying to measure.
We really need Tj, but even the NJL and Sanken can only get us closer to Tc. I know of no device that allows us to monitor Tj.

Adding a current mirror changes the open loop gain (OLG) of the amplifier. That increase in OLG may make the amplifier oscillate and destroy it's self.
You can reduce the OLG by adding in or increasing the LTP degeneration, but you would still need to adjust the compensation to make the amp stable into all likely loads.

Quote:

Originally Posted by Jay (Post 3197567)
Quote:

Originally Posted by Nico Ras (Post 3197367)
A lot of people (on DIY) stick this transistor on top of one of the output devices. I have seen this a lot in the threads here.

My "school of thought" is to make the VAS (including temperature compensation) free of stray capacitance. I like to put both drivers/VAS transistors very close with each other. As the consequence of this, the compensation transistor will be put in the middle of NPN and PNP output transistors, and consequently, on the heatsink.

Quote:

Originally Posted by Nico Ras (Post 3197367)
I don't subscribe to this and believe it is best to tie this transistor onto the heat sink so it responds slowly to the temperature variation a form of hysteresis.

The question is, is the compensation transistor used to linearize gain and to lower distortion or SIMPLY as protection against thermal runaway.

Quote:

Originally Posted by Nico Ras (Post 3197367)
Tying onto one of the output transistors can introduce thermally generated distortion in that the thermal compensation reacts quickly to one transistor but slowly to the other thereby distorting or compressing the output signal.

Logically, for protection purposes we need to control only one device: the most disturbing one. For most power BJT, it should/usually be the NPN.

Quote:

Originally Posted by Nico Ras (Post 3197367)
This is measurable, but maybe not audible at lower temperatures.

This "fact" is more important for practical less theoretical minds. Theoretically, to get the perfect tracking is very possible, but there are too many variables to be considered (properties of transistors used, heatsink size, bias used related to TR properties, circuit used, etc etc), leading to question: is it necessary? May be not for most circuits.

Easier is to change the level of temperature tracking/coupling physically or mechanically without "changing" the circuit (just an idea). I guess using TL431 in combination with diodes it is possible (by shorting some of the diodes based on measurement). I'm not sure tho...

But primitively changing the contact level between the TC transistor and tracked transistor, physically, is possible (no detailed idea how the mechanics should work, yet).


Bonsai 11th October 2012 12:10 PM

That's exactly the thinking I used in both my Ovation 250 and the e-Amp.

I noted in my write up that the ON parts with integrated sense diodes will give a better result on sensing the output device die temp, but this is still only a half (EF2) or one third (EF3) f the overall tcomp challenge.

So, in the e-Amp I used a BC847 to sense the collector lead temp on one of the output devices and an NTD to shape the compensation curve. I mount all three of the output devices on th same heatsink (pre-driver, driver and OP device). The result, after set-up, is a very stable bias point.

Macleod 11th October 2012 05:39 PM

I also assume that temperature reading is affected by less innertia when done on the lead. This mays help to track faster power change and avoid overbiasing after loud music play..?

Bonsai... BC847 smd package? How do you put that on a lead? Glue? small pcb? on the same pcb close to the lead..?

Bonsai 11th October 2012 11:28 PM

No, I mount it on the underneath side of the PCB very close to the collector lead and then couple with a very small amount of silicon grease.

Macleod 12th October 2012 09:30 AM

I assume it would be better to have a low inertia measurement of the temperature with some offset than a low offset measurement with some inertia..?

Bonsai 12th October 2012 11:54 AM

Quote:

Originally Posted by Macleod (Post 3198589)
I assume it would be better to have a low inertia measurement of the temperature with some offset than a low offset measurement with some inertia..?

Yes

wahab 12th October 2012 01:13 PM

Quote:

Originally Posted by AndrewT (Post 3197470)

The outer surface of the plastic package will be at a very considerably lower temperature than Tc + a bit. Again this is due to the thermal resistance of the plastic to the heat flow from the backplate through to Ta on the inside face of the amplifier. It is likely that the plastic is >10C lower than Tc and possibly >20C lower. The heatsink Ts adjacent to the backplate will be Tc - a bit, again due to thermal resistances and heat flows.
The collector lead is directly attached to the backplate with a fairly thick copper lead. Some are welded on, but many (collectors) are part of the backplate copper. The temperature of the lead next to the plastic package is very close to Tc in temperature. This is because there is little heat flow out through the collector lead.

Just use a thermal insulation in the plastic side of the VBM
using hot-melts that cover completely the device.

For minimal influence of the leads , cut them , just let 3 mm
and connect to the board using very low diameter plastic insulated
wires.

On another note , using a strip of two diodes pairs , each pair
being sticked to a TOP3 power device with thermal contact grease
appear as working well , i had 20% current variation using
this combination with a pair of Sanken darlingtons ,for heatsink
temperature variation of about 55 C.


All times are GMT. The time now is 11:49 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2