Derating, am I missing something?

I have built many amplifiers, and I have found that I prefer TIP41C and TIP42C transistors for my output. It's just a preference. And they're cheap and very easy to come by.

For a 130W amplifier, I used two TIP35/6C in parallel for the output to provide sufficient rating. I think these are TO-3P. The TIP41/2 are TO-220's. I like to mount the transistors directly onto anodized aluminium for best heat transfer.

So I'm doing a little calculating to find out if I can use the TIP41/2's for such an amplifier. The maximum average power for one would be about 30W. The heatsink is 0.6K/W, so by my calculations, and assuming an 8ohm load, one TIP41C and one TIP42C should theoretically be enough? Is this correct? For load variations, I would use 2 of each, but in a theoretical sense, would one do it?

P = 30W
Ts = 30(2)(0.6) + Ta = 66C [Assume Ta=30C, the entire output on one heatsink]
Tc-s = 30(1) [Assume 1K/W for the interface]
Tc = 96C

At this temperature, 30W should be fine.

[IMGDEAD]http://s33.postimg.org/imbn3jx5b/TIP41_Derating.png[/IMGDEAD]

Well, cutting it close, but for music it should be fine, right? Or do you look at the derating in terms of peak power?

Still, I would use a parallel output to play it safe, and to deal with load variations.
 
Last edited:
If you maintain a good heat transfer using a quality compound, the heat generated will be 0.6 X power = 18 degrees above ambient. So, say ambient is 25 then 25 + 18 is 43 degrees. Look at the power expected for 43 degrees. Almost cold so low losses.
This is the average power as peak power will take a while to heat the sink up. The peak power is down to the transistor specification.
 
Ok, thanks. So small transistors do fine for this sort of power then (provided the transistor peak ratings are well within limits).

When I read about the many projects here and elsewhere online, why does everyone use massive transistors? If the heat transfer is fine, then aren't these huge transistors overkill?
 
It seems fashionable, on this site, for people to build without taking into account design properties. Yes, a 50W transistor will run a 30W amplifier but as headroom may be required for reliability. I would double the power required. The heating we spoke about is the same for one or 12 transistors, so if you use 12 transistors they will generate the same amount of heat as one because you are still dissipating the same power differential.
I think the same people go for huge motor start capacitors in the belief that this is an improvement on the original designers design. Never mind about most designer, myself included, gained many qualifications and degrees in electronics and the sciences.
Tip 41 and Top 35 are nice transistors. Cheap and plentifully but an MJ (any big number you can think of), looks nicer. Ha! Ha!
 
I have built many amplifiers, and I have found that I prefer TIP41C and TIP42C transistors for my output. It's just a preference. And they're cheap and very easy to come by.

For a 130W amplifier, I used two TIP35/6C in parallel for the output to provide sufficient rating. I think these are TO-3P. The TIP41/2 are TO-220's. I like to mount the transistors directly onto anodized aluminium for best heat transfer.

So I'm doing a little calculating to find out if I can use the TIP41/2's for such an amplifier. The maximum average power for one would be about 30W. The heatsink is 0.6K/W, so by my calculations, and assuming an 8ohm load, one TIP41C and one TIP42C should theoretically be enough? Is this correct? For load variations, I would use 2 of each, but in a theoretical sense, would one do it?

P = 30W
Ts = 30(2)(0.6) + Ta = 66C [Assume Ta=30C, the entire output on one heatsink]
Tc-s = 30(1) [Assume 1K/W for the interface]
Tc = 96C

At this temperature, 30W should be fine.

[IMGDEAD]http://s33.postimg.org/imbn3jx5b/TIP41_Derating.png[/IMGDEAD]

Well, cutting it close, but for music it should be fine, right? Or do you look at the derating in terms of peak power?

Still, I would use a parallel output to play it safe, and to deal with load variations.
You have not finished.

You now need to look at the SOAR curve plotted in the datasheet and see what maximum current can pass at your Vce. This current @ Vce is what you need to derate.
Driving a resistive load where current and voltage are in phase your will find that when Vce = rail voltage that the current is approximately zero to the load. = zero power dissipation.
and when Vce = zero the current to the load is at maximum and power = zero.

In between these you find that the output dissipates power and that should not exceed the derated SOAR curve. i.e apply that 30W of reserve you had from the first calculation.

But here's comes the difficult bit:
when driving a reactive load, voltage and current are not in phase.
When Vce= rail voltage the current is no longer zero. Power is dissipated in the device.
And when Vce= ~half rail voltage, the current can be similar to the maximum passed to a resistive load.

Read ESP for a little bit more, but the real one worth reading is David Eather.
Phase Angle Vs. Transistor Dissipation
Semiconductor Safe Operating Area
 

Attachments

  • Prac1.pdf
    76.1 KB · Views: 33
  • Prac2a.pdf
    99.9 KB · Views: 26
Last edited:
Thanks, Andrew. Yes, it's well within the SOA, and yes, I'm talking about purely resistive. I know there's the consideration of a reactive load, which is most important, but the first step is to design for a resistive load. I have read the article on ESP, but didn't gain too much from it. It's well known stuff. The major thing there is about reflected power. I'll do some reading on David Eather's articles.
 
Another point is the output transistor is only on for half of the time.

Also music is no where near a sine wave so it can have quite high transients.

I found the output power is highest in output transistor at 2/3 of output voltage with a sine wave.

Yes, absolutely. That's where class ab is at its least efficient. The average power at this point is the power I'm talking about when I talk about average power.
 
you calculated the near30W using Tc for the temperature de-rating.
You have not taken accounr of SOAR yet.
That is the stage you completely missed.

How much current can the device pass when Vce is high?
That is the part you have missed !

Hi Andrew. You are correct that this is a most important consideration. At the moment I'm looking at a resistive load, 8 ohms. The example is a 130W amplifier where the Vcc = |Vee| = 50V. The SOA is well within limits, I'm trying to determine whether I understand the derating correctly. As far as I know, the derating shown is for DC, or average current (and I believe others have confirmed this now). The SOA curve I looked at was for DC.

This is Vce vs Ie for the example. The intersection of these two give the maximum Ic and Vce possible. Green is Vce, red is Ie (approx. Ic):

[IMGDEAD]http://s33.postimg.org/ljnd6iqjz/Vce_Ie.png[/IMGDEAD]

This is the SOA curve provided:

[IMGDEAD]http://s33.postimg.org/5v8udvnqn/SOA.png[/IMGDEAD]

My original question also isn't considering the SOA, but only the derating (I know they are related, but I want to understand the limits indicated by the derating curve).
 
The SOAR you have attached is for 25°C
You have to de-rate the current to take account of your Tc.
You do that by moving the whole set of plots DOWN by the factor of your de-rating.

For Tc=96°C and Tcmax =150°C the de-rating factor is:
{150-Tc}/{150-25} = 150-96/150-25 = 0.432
The current shown as a straight horizontal line at 6Adc becomes 2.59aAdc
10A becomes 4.32Apulsed or peak.
The first kink moves down from [email protected] to [email protected]
The second kink moves down from 2.2Adc @ 33Vce to 0.95Adc @ 33Vce
Draw those new de-rated lines into your graph. See how close your red dot gets to the 5ms line !!!!!!!!!
And your red dot is in the wrong place for 1.91Apk @ 33.76Vce This is a log.log scaled graph.

It has been discussed here that DC rating are too conservative for an audio amplifier driving a resistive dummy load. Instead there seems to be a consensus that the 100ms SOAR is a reasonable limit for LF audio signals.

When driving a reactive load, the load curves are completely different. READ Eather !!!!!!!

My simulations have indicated that for BJT devices driving a severe reactance load when kept reasonably warm (not hot) that you can get approximately total Maximum device power / 5 as reliable maximum output power for robust devices and divide by 6 for less robust devices that have a poor SOAR at higher voltage. (ex for 1943/5200 divide by 6)

Are your devices rated @ 60W?
Total maximum device dissipation is 60W+60W.
Divide by 5 and you get a reliable maximum output power of 24W from 1pair in domestic listening conditions when driving severe reactance speakers. A reasonable maximum for mild reactance speakers where the output devices are virtually cold would be roughly double that in a domestic environment.
 
Last edited:
For Tc=96°C and Tcmax =150°C the de-rating factor is:
{150-Tc}/{150-25} = 150-96/150-25 = 0.432
The current shown as a straight horizontal line at 6Adc becomes 2.59aAdc
10A becomes 4.32Apulsed or peak.
The first kink moves down from [email protected] to [email protected]
The second kink moves down from 2.2Adc @ 33Vce to 0.95Adc @ 33Vce

Thanks, this is the derating debacle I'm trying to find out about. This means the derating curve doesn't show average / DC power, but peak power.

From what I've read, this doesn't make sense. The SOA remains as is without derating. For the derating curve, the junction temperature doesn't change with peak power.

I will read more. I haven't had a chance to read Eather yet, but I intend to soon.
 
Hi,

A simple rule of thumb for speakers is to design for a
resistive load half of the speakers nominal impedance.

That covers most reactance issues, for most speakers.

Don't imagine for a second that an amplifier designed
near its SOA limits into 8R is good for 8 ohm impedance.

It isn't, and any design that takes devices to their limits
is inherently very unreliable, and will inevitably expire.

There is nothing clever about using devices near their
limits, clever is a robust and reliable implementation.

rgds, sreten.
 
Hi,

A simple rule of thumb for speakers is to design for a
resistive load half of the speakers nominal impedance.

That covers most reactance issues, for most speakers.

Don't imagine for a second that an amplifier designed
near its SOA limits into 8R is good for 8 ohm impedance.

It isn't, and any design that takes devices to their limits
is inherently very unreliable, and will inevitably expire.

There is nothing clever about using devices near their
limits, clever is a robust and reliable implementation.

rgds, sreten.

Hi sreten. I couldn't agree more. In this case, the calculations indicate the transistors are at or near their limits, right? Meaning, I can parallel the output with a second set of transistors safely?

I mean, the SOA doesn't need to be derated?
 
And that's why essentially no current amplifier designs use TO-220 case power transistors any more. You can barely get an Rth,jc of about 1 K/W on these with a bit of a following wind, as opposed to more like 0.75 K/W for TO-3P or around 0.4 K/W for TO-264 or TO-247 (see e.g. this appnote). Practical TO-220s seem to vary a lot, with some sources giving worst-case values of 4 or even 8 K/W, but since TIP41s are rated 65 W at 25°C Ta and Tj,max = 150°C, "official" (maximum) Rth,jc can't be higher than about 1.9 K/W. It clearly is not a deal-breaker on a medium-power amp, but when you're trying to squeeze >100 W / 4 ohms from a single pair of outputs on the edge of SOA like a lot of mass-market home theater receivers and the like do, that just wouldn't fly.

Parallel output devices are always a good way of increasing SOA, as obviously dynamic power dissipation and current are now shared. They'll also add capacitance to be driven and idle power consumption though, so keep that in mind.