John Curl's Blowtorch preamplifier part II

[anatech]

Hi John,
All I did was to support something that Bob mentioned, an easy concept. I have no idea why you felt it was necessary to take issue, then insult me.

Chris, if you are not an engineer or a physicist, you are not in a position to judge.

That, is insulting.

So, what letters do you have behind your name? PEng, CET?
Does it matter? No, it really doesn't. I have spent my life doing work in this field, just as you have (I guess). I am not about to insult you, and haven't.

Please adjust, I am in too much pain.

I'm really very sorry to hear you are uncomfortable. It's not pleasant and makes everything that much more difficult.

But, guess what John? I know several other members who are just as, if not more uncomfortable than you are. Personally, on a good day, I have a couple good hours to function somewhat. On bad days, I can't even think straight. Those are the days I don't post here at all. As it is, I had to take a break from this post. I'm on 80 mg of Oxycontin per day, plus other stuff. Increasing doses because they can't get my situation under control. Plus A.R.T., Chiro, and nerve blocks every week. There are people worse off than I am who are members here as well.

Stop whining and sniping at people. You aren't as bad off as you may think, although I'll accept that you may be in the worst pain you have felt so far. It can get worse. It's still not going to cut it as an excuse.

I do wish you well along with a speedy recovery. At least you will recover, I will not.

Try to treat others around you with some respect. now, if you have a problem with me, get over it. Instead, try letting me know in realistic terms what the problem is.

-Chris

 

[PMA]

3 different idle currents, 2 x 50V supply, push-pull output stage, class AB.

 

[PMA]

Same for smaller output voltage. One can see that at higher idle the power dissipation approach to a constant, or at least fluctuates much less than in a class B.

 

[AndrewT]

Hi,
it's even better than PMA is posting.
All music has transients. We all listen at much lower levels than maximum output.

Transients by their definition do not last long.

The power devices and the heatsink they are attached to have thermal inertia.

The junction temperature will fluctuate very little for low amplitude power changes.
Due to the thermal inertia, the case temperature will not fluctuate much for high amplitude transient power changes.

How do we estimate/measure/calculate the changes in junction temperature when we can feel/measure that the device case temperatures vary little in normal music reproduction?

 

[Bob Cordell]

Easy is better than complex. Chris, you are making it complex. This happens when techs try to do what engineers are trained to do. It is easier, in this case, to use the engineering equations.
Someone: Dimitri, Cordell, Wurcer, please bail out this situation. It is on a page of a favorite book owned by most of you.

Hi John,

I tend to agree with Chris here. As an engineer, I know equations are very important and use them every day. However, equations are not always the best way to express concepts. When one has a good grasp of the concepts and trends, then the equations become even more valuable and less subject to error or mis-interpretation. I've seen far too many people blindly plug into equations and draw wrong conclusions.

On the other hand, we must always guard against error in intuitive arguments and generalizations. This is where equations are especially important. Linked to this is simulation, which is not unlike a more convenient and detailed application of equations. There have been many times when I have discovered that my intuition has been less than perfectly accurate when subjected to the scrutiny of analysis by equations or simulation.

Cheers,
Bob

 

[Bob Cordell]

Same for smaller output voltage. One can see that at higher idle the power dissipation approach to a constant, or at least fluctuates much less than in a class B.

Hi PMA,

Thanks for showing this. This is the point I was making in Post 452.

BTW, an interesting side note is that this tends to be synergistic with MOSFETs, where gm doubling basically cannot happen. Bias them nice and toasty and they will have very little transconductance droop, a gairly large class A range, and rather small deviation of power dissipation as a function of output power. The price, of course, is higher quiescent dissipation.

Cheers,
Bob

 

[Bob Cordell]

Hi,
it's even better than PMA is posting.
All music has transients. We all listen at much lower levels than maximum output.

Transients by their definition do not last long.

The power devices and the heatsink they are attached to have thermal inertia.

The junction temperature will fluctuate very little for low amplitude power changes.
Due to the thermal inertia, the case temperature will not fluctuate much for high amplitude transient power changes.

How do we estimate/measure/calculate the changes in junction temperature when we can feel/measure that the device case temperatures vary little in normal music reproduction?

Hi Andrew,

There was a lot of good discussion of this over on the ThermalTrak thread. The thermal time constant of the heat sink is quite long, on the order of tens of minutes. The thermal time constant of the power transistor case (heat spreader header) is on the order of seconds (maybe 5). This is important because the case is separated from the heat sink by the thermal resistance of the insulator, perhaps on the order of 0.5 C/W.

The thermal time constant of the die/junction is much shorter, sometimes on the order of 20 ms. It is separated from the case by a thermal resistance typically between 0.5 and 0.9 C/W. Short-term transistor power dissipation changes of, say, 20W can cause a transient temperature difference that would like to asymptote to on the order of 10-18C, enough to upset short-term bias, given a junction TC on the order of 2 mV/C. That theoretical 26 mV number for voltage across RE can quickly be seriously compromized on a short term basis as a result of program-dependent power dissipation.

Cheers,
Bob

 

[scott wurcer]

It is an important distinction when talking about "time constants" that heat obeys the diffusion equations and can never be modeled exactly in SPICE by R's and C's.

BTW there is a beautiful paper by (I think) Alberto Billoti (sp?) that derives the behavior of temperature vs. arbitrary device geometry from first principles, rectangular coordinates only IIRC. Such maths would make a Russian jealous .

 

[dimitri]

Alberto A. Bilotti, “Static Temperature Distribution in IC Chips with Isothermal Heat Sources,” IEEE Trans.
Electron. Devices, vol. ED-21, No. 3, pp. 217-226, March 1974.

 

[scott wurcer]

Thanks Dimitri, That's one journal that I don't have on my DVD archive.

 

[Bob Cordell]

It is an important distinction when talking about "time constants" that heat obeys the diffusion equations and can never be modeled exactly in SPICE by R's and C's.

BTW there is a beautiful paper by (I think) Alberto Billoti (sp?) that derives the behavior of temperature vs. arbitrary device geometry from first principles, rectangular coordinates only IIRC. Such maths would make a Russian jealous .

Very good point, Scott. Using lumped time constants and RC representations is just a starting point, but it does lend valuable insight.

We use FlowTherm a lot at work for these kinds of things. The thing cranks for quite a long time when modeling the temperature of all of the components on a board in a shelf.

Cheers,
Bob

 

[AndrewT]

Bob,
I'll try again.
Using your 20W as a medium term output from a 200W amplifier, we can expect many seconds even minutes operation at the 2W (-20dB) level.
This output will fall to tens of milliseconds at the 20W (-10dB) output level and for a 200W amplifier that has a 2pair output stage is just 5W per device. I can't tell you the device dissipation since that will depend on the load reactance, but I could estimate it will be of the same order as the output power. The devices will dissipate on alternate halves of the overall cycle and thus have cool down periods between half transient.

The short term transients should last just a millisecond or so and hopefully they will be short of clipping.

It's on this basis that I suggested that PMA's posting show a power dissipation that is likely to indicate temperature variations that are worse than what happens in an amplifier reproducing music.

I am not specifically referring to change in bias conditions nor how the bias compensation should work. Just variations in output device junction temperature.

I do however support PMA's contention that the higher the bias, whether optimum or higher, helps swamp out deltaTj with variations in output power that follow normal music waveforms.

 

[jan.didden]

[snip]@Jan: I think there is no way to get constant power with real world complex loads and EF's/SF's unless one uses cascodes which generate an (approximated) optimum load line for the instanteous current, regardless of voltage (as they aren't corelated).

Other way would be not to use plain EF's/SF's, then...

- Klaus

Klaus,

I was wondering for resistive loads, because intuitively you can image that there is some condition where a doubling of output current leads to exactly halving of Vce which would mean no change in dissipation. Worth looking at I thought.

The math however seems to indicate that there isn't such a condition except for one specific Vout (which depends on Ibias, Vcc and Rl), because the Vout is in both Ic (Vout/RL) as well as in Vce (VCC-Vout) which leads to a quadratic term in Pc which cannot be cancelled in the general case.

Ohh well.

jd

 

[Wavebourn]

It is an important distinction when talking about "time constants" that heat obeys the diffusion equations and can never be modeled exactly in SPICE by R's and C's.

BTW there is a beautiful paper by (I think) Alberto Billoti (sp?) that derives the behavior of temperature vs. arbitrary device geometry from first principles, rectangular coordinates only IIRC. Such maths would make a Russian jealous .

We Russians when learned how to deal with dissipated heat studied as well works of Grashof Prandtl. Were they Italians as well?

 

[john curl]

Thanks, PMA.

 

[Nelson Pass]

I was wondering for resistive loads, because intuitively you can image that there is some condition where a doubling of output current leads to exactly halving of Vce which would mean no change in dissipation.

You can approximate that with cascoding, an example being the ZV9
with slightly different resistor values. It has the advantage of being
current dependent, so that the value and phase of the load are
not of primary importance.

 

[KSTR]

Yes, I learned this trick from you, Nelson

 

[john curl]

Now, we are cooking.

 

[Bob Cordell]

Bob,
I'll try again.
Using your 20W as a medium term output from a 200W amplifier, we can expect many seconds even minutes operation at the 2W (-20dB) level.
This output will fall to tens of milliseconds at the 20W (-10dB) output level and for a 200W amplifier that has a 2pair output stage is just 5W per device. I can't tell you the device dissipation since that will depend on the load reactance, but I could estimate it will be of the same order as the output power. The devices will dissipate on alternate halves of the overall cycle and thus have cool down periods between half transient.

The short term transients should last just a millisecond or so and hopefully they will be short of clipping.

It's on this basis that I suggested that PMA's posting show a power dissipation that is likely to indicate temperature variations that are worse than what happens in an amplifier reproducing music.

I am not specifically referring to change in bias conditions nor how the bias compensation should work. Just variations in output device junction temperature.

I do however support PMA's contention that the higher the bias, whether optimum or higher, helps swamp out deltaTj with variations in output power that follow normal music waveforms.

Hi Andrew,

I'm not sure we're really disagreeing here. My main point was that we should not underestimate the amount of variation that can occur in junction temperature as a result of program power variations. It is certainly true that the heat sink has a long time constant, but such program-dependent variations in junction temperature will occur even if you bolt the output transistors to the Brooklyn Bridge.

It is certainly true that the extent of such variations depends strongly on the nature of the program material and how loud it is being played. Hopefully, our amplifiers will sound good with all kinds of program material. Consider the 1812 Overture, with extended periods of high program levels followed by very soft levels and vice versa. Consider also other material that might have bass notes that drive an amplifier to 1/3 power for over 100 ms, where output stage dissipation may be near maximum.

A simple 100W/8-ohm amplifier with 45V rails might have output stage dissipation of 10 watts quiescent and nearly 50W at 1/3 power. That's quite a swing. Such an amplifier might typically employ two output pairs.

Junction temperature variations do have consequences, and those include possible intervals of over-bias or under-bias.

Cheers,

Bob

 

[john curl]

Try an 800W amp into 4 ohms.