Current project: 1kW

[kASD]

Re: Re: Re: Re: Commentable Thoughts

I have often thought (but never tried it) that paralleled mosfets would be better off with separate heatsinks rather than thermally tied to their neighbours. If a mosfet gets hotter because of hogging the current, it's temperature increase will be greater if it is thermally isolated from the others. This greater temperature increase will result in a greater reduction of current back toward what it's neighbours are carrying.

Of course, this all presumes that the Rds *increase* with temperature has a greater effect than the Vgs threshold *decrease* with temperature =at the same time=.

In a saturated switching circuit, e.g. a SMPS it certainly will. In a linear amp cct... not so certain.

Mr CIRCLOTRON
IRF technical team recommended me to use good thermal coupling between paralleled MOSFETS in linear operation.

So i am only implementing their recommendations.

Regards
Ampman

 

[Eva]

Let's make some calculations about THD

Depending of output current and voltage, a worst case at full power may be :

Gain of output devices may vary between 10 and 100
Gain of drivers may vary between 25 and 100
Gain of VAS buffers may vary between 50 and 100

So 4 ohms on the output [of each channel, 8 ohms total due to bridge mode] may appear to the VAS as 125k to 1Meg ohms

The VAS is loaded with 1K ohms so the gain variation of the triple darington of the output stage would represent less than 1% THD in the worst case

Assuming a worst case, we may have about 500mV of VBE variation across the triple darlington. Those 500mV should have a resistive component and a non-linear component so this may account for less than 0,5% extra THD asuming +-75V output swing

The remaining stages have local feedback and in the worst case should have THD lower than 1%

Since VAS is loaded with such a low impedance, output impedance should not be much higher than te resulting value of the emitter resistors of the output devices in paralell plus the drop in the VBE junctions, lets say about 0,2 ohms, so damping factor may be 20 or better

So in the end, we may be talking about 1000W into 8 ohms with no more than 2% THD and DF > 20 [maybe 0,5% THD at low frequencies assuming less gain variations]

At half power THD may be as low as 0,1% due to much lower gain variation in output devices

Supply rejection may be good as long as the supply of the first stage is regulated and quiet

I don't like exotic no-global-feedback full-discrete designs at all, but these numbers doesn't look bad for that kind of exotic amplifier

Most exotic no-global-feedback desings [including tubes, mosfets, transformers, etc..] don't go better than 50W with lots of THD and less damping

 

[The Saint]

Perhaps I can help here

I would agree to a point with ampman, IRF devices do require good heatsinking with fan cooling, (In High powered applications exceeding 200 Watts into 8 Ohms) the pairs do need to be in close proximity with each other and in very good thermal and mechanical contact with the heatsink, but the use of source resistors is very important with these devices. I am guessing that ampman only uses very light biasing in the o/p stage, possibly around 40 to 50ma per devices perhaps less which would account for the lack of thermal biasing control or source resistors.
Below is a quote from APT application notes, it aptly explains the condition that araises in all Vertical type MOSFETs when high voltage is applies to these devices, approaching 100 volts.

Quote:
Like all MOSFETs, the
gate threshold voltage, Vth, has a negative temperature
coefficient. This makes operation as a linear amplifier
difficult.
When forward biased with a constant gate voltage, the
quiescent drain current will rise as the temperature of the
die increases. Operating at the typical drain voltage for
these parts, about one third of the rated BVdss, the power
dissipation due to the increasing Idq results in “hot spotting”
and subsequent thermal runaway. This is an unstable
system. The dissipation increases so rapidly that the outside
surface of the case does not follow the internal junction
temperature. As a result, a bias compensation scheme that
uses temperature sensing cannot keep up with the Vth shift
and the device is destroyed.
The power dissipation within the die is a direct function of
the operating voltage. By lowering the operating voltage
the thermal loop gain can be reduced to a point where the
gate threshold shift can be compensated for. Thermal
stability can be achieved by sensing the case temperature.
Linear operation thus becomes practical at 100V and below.
While this is less than 25% of the rated BVdss and results
in less gain, a very rugged and useful linear amplifier results.
End quote:

The fact is that these device love lots of heatsinking both top and the bottom of the devices, most of my amplifiers in the 500 watt to 1kw range have a small amount of heatsinking on the top of the devices, this aids the Vgs multiplier in controlling the Gate threshold as it can sence all of the devices at once, instead of just one device. and I have 105 volts +- on the o/p stage.
With not one amplifier blowing up.

 

[IVX]

km,
don't worry, the truth is only one and you are right. I had simulation "current feedback amp_***" at 15 minute after publication the post, it was pure class B with huge THD (but claims 0.001 or maybe 0.0000000001%- actually is no matter as 2/3 of 58v =27VAC RMS (isn't 38VAC)).
DIYer can't lie, but can have mistake, expert is contrary seems third type does exist.
Sorry for some offtopic.

 

[millwood]

thanks anthony for the reference.

Operating at the typical drain voltage for
these parts, about one third of the rated BVdss, the power
dissipation due to the increasing Idq results in “hot spotting”
and subsequent thermal runaway.

Well, in the "Can Class B amp have thermal runaway", I mentioned that thermal runaway is due to hot spotting and thus can happen in any amps. People disagreed and thought thermal run away can only happen with bias current.

Seems the above from IRF disagree with that notion.

Any thought?

 

[Eva]

Wouldn't be a good idea to distinguish between thermal runaway due to idle current runaway and thermal runaway due to bad current sharing when multiple devices are used in paralell?

The second case should not be as quick or as destructive as the first one

 

[kASD]

Commentable Thoughts

Perhaps I can help here

I would agree to a point with ampman, IRF devices do require good heatsinking with fan cooling, (In High powered applications exceeding 200 Watts into 8 Ohms) the pairs do need to be in close proximity with each other and in very good thermal and mechanical contact with the heatsink, but the use of source resistors is very important with these devices. I am guessing that ampman only uses very light biasing in the o/p stage, possibly around 40 to 50ma per devices perhaps less which would account for the lack of thermal biasing control or source resistors.

Anthony seems to be right in certain way.
The output device biasing in my amp is around 50-70mA per devices. The fact is that when the amplifier i.e. is used in professional field is never used for low power output instead of it is used in hard driving conditions while operating at full power, So this biasing doesn't pose a problem at this condition. Though the heating of device self compensates its thermal tracking.

Best Regards To Mr. Anthony Holton for Clarifying the statement to all of u.
AmpMaN

 

[Ultima Thule]

Wouldn't be a good idea to distinguish between thermal runaway due to idle current runaway and thermal runaway due to bad current sharing when multiple devices are used in paralell?

The second case should not be as quick or as destructive as the first one

I don't!

Actually the thing is, at least from a pure electronic design engineers view, that I would prefer to say just "thermal runaway" which is known as the phenomenom with some semiconductors, else there would be "millions" of name for thermal runaway which degrade the understanding between engineers. Engineers should use simple and universal name so engineers can communicate all ower the world and just be aware what can cause it.

I must admit when I enterd this forum in the beginning (and I'm still quite new here) that even if I am an electronic design engineer since many years I had difficulties to interpret what people was talking about, I am used to the "universal" engineering language but here in the audio community has an own langauage developed for the same electronics related to audio(probably also because some people here are selfteached in electronics and have focused only in audio electronics).

The audio issue, I think it is just fine if audio people and especially those who are not really electronic design engineer in their profession would adopt the more universal engineering laguage model which makes it easier for all wethere he or she is an engineer or not, I think it's just good to know that their are diffrent cases that can cause thermal runaway(because there are several, not only in audio eelctronics) and people can reffer to them. But please spare the fantasy what it should be named and just say Thermal Runaway(the discussion in the other thread about thermal runaway was good).

Just my view and I hope people would share this view.

Cheers

 

[ChocoHolic]

Hi Eva!
I love your distorsion estimation.
But there is one point were I would expect higher THD.
It looks like T5+T6 together with R7 & R8 are doing the gain.
I hope that the change of the base emitter voltage of both BJTs together is less than 0.5V, because from my understanding we must
see this change in relation to the input signal not in relation to the output signal level....
These simple gain stages tend to have high 3rd harmonics.
If I remember right (damn long back ago...!) then "Tietze/Schenk" is analysing the distorsion of such single BJT amp (not darlington) very nice and results in acceptable distorsion for small signals.
Especially 3rd harmonic was poor (I think) when this topology is running at high output swing...
OK, here the designer went for a darlington configuration, but he
is making a hell of gain out of this stage. R8 is 12 Ohms, which will
not bring much linearization......

Also I think the damping factor will be less than in your estmation, because you simplification would be pretty fine in a class A design,
but here we should also consider the change in the base-emitter voltages of the output stage.

Well even if this Amp would not be "my favourite design" it will have some good properties:
-It is quite fast. And it should be able to provide this feature
without tendency to instable HF-ringing!!!!
Also a property which should welcomed for beginners.
(But be carfefull with continuos large HF-signals at the input,
the output stage may burn. Because of the storage time
of the heavy BJTs the quiescient current will jump up under
this condition..... normal audio signals should be fine)
-Thermal compensation should work ..., (somehow, but acceptable)... as D5, T12, T13, D6 will compesate the drift of T15...T22.
I tried this compensation in an Amp 11 years ago and it was OK.
(And it still works for my subwoofer.....
8x150W@4 Ohms ... now paralleled (4times) and bridged to deliver 1200Wrms into 2 Ohms, well infact I am loading it now with less than 1 Ohm and it pushes up to 2kW ;-).... OK, I know a modern
switch mode solution should replace this thing sometimes....)



Cheers to all
Markus