Power Amplifier for 100kHz.

It's old, but discrete and can take some pain. Too many IC's have thermal limiting. Fig 1b? dunno, I just Googled the circuit diagram as an example of the architecture.

It's a few watts, coupled capacitively to various very lossy lines, maybe 30R. I just want to know if there are any bright ideas for improving it within the constraints.

No, PNP's are pink and NPN's are blue. I'll have none of this LGBTQ nonsense.

A few watts, the heatsink isn't huge, there's no room. Tj on the output pair is probably around 210'C at 180'C ambient. It doesn't come with a lifetime guarantee.

I guess that this will be driving a few km of cable from a very deep hole in the ground.

John

Maintaining class AB bias over that temperature range is going to be difficult. I'd look instead at having no Vbias so the power stage operates pure class B and then have a fast small class A stage fill in the cross-over distortion. In other words the Quad feed-forward aka current dumping idea of the 1970s.

piglet

One trick is to have complementary class-B followers and add one resistor from the bases/gates to the output. That makes the output transfer curve continuous, granted nonlinear but the feedback mostly fixes that.

No quiescent bias current, no thermal runaway hazard.

Or use a TCA0372 power opamp. The thermal limit doesn't work!

Some local feedback around the output stage would get my vote. The Sziklai pairs have their own local feedback, but that doesn't fix the crossover problem.

Another approach would be to turn TR4 into a diff pair. TR3's collector swing is going to waste, and that would let you keep the open-loop gain the same, while stabilizing the tail current.

Cheers

Phil Hobbs

Thanks, that second one particularly sounds like a good idea. I'll see what the sim says. Local feedback on the output stage sounds trickier, I'll have to think.

One thing I thought of was to use multiple smaller output pairs in parallel, and have a DC offset for each one. Imagine replacing the Vbias diode with a string of a few series diodes, and connecting the bases of one output pair across D1, the next pair across D2 etc. Would need resistors from each pair of emitters to the output. That should give lower crossover in more places. Maybe.

To reduce the tempco of the first stage's tail source, one approach would be to use a LED or a SiC diode instead of two silicon ones in series.

I've sometimes used a LED with an emitter follower to make a temperature-compendated low noise 1-V reference. You have to futz around with choosing LEDs to make that work well, but some orange Avago LEDs got down to a couple of hundred microvolts/K.

I don't know whether there are LEDs that survive long enough in your conditions. Alternatively, the GB01SLT06-214 SiC rectifier seems to have about the same -2 mV dV/dT as a BJT, while dropping more voltage, so that might work.

Cheers

Phil Hobbs

(Who still wants to make a thermoacoustic fridge to make downhole instruments' lives easier.)

Some years ago Jim Thompson posted an audio amplifier design which used current mirrors to provide bias to the output transistors for the express purpose of keeping crossover distortion low over a large temperature range. He claimed it was the bees knees, but a quick search failed to turn it up. Perhaps someone else saved it or remembers the thread?

Glen

Yeah, the circuit linked is the basis of my amplifier rather than the actual thing, and I do use a red LED for this - one used elsewhere which works at 180'C (and very few do). It actually does very well at tracking the transistor Vbe plus about 1.2V, I think more by luck than judgement. Upshot is that Iconst changes from 13mA to 14mA from 20'C to

Is it ONLY 100 kHz you want to amplify ?

If so, you might incorporate a filter on the output.

There are other ways to make 100 kHz too if you think about it for a while.

boB

Doug Self has a version of this in his audio-amplifier book (not sure if he mentions it in his online materials). He refers to it as "output- inclusive compensation".

Briefly, you split the Cdom capacitor into a series pair (each being twice the nominal Cdom value), and connect the junction of the two to the output node via a resistor. Over an intermediate range of frequencies, the Vas feedback is partly local (collector-to-base via the two halves of Cdom) and partially from the output stage. Self claims that this can knock down residual crossover distortion to negligible (even unmeasurable) levels.

I used this technique in an audio amplifier I finished building this year. It seems to work, and I've been unable to detect any adverse effect on stability (either in SPICE simulation, or in measurement of the finished amp).

The 'cable' is horrible. There are many different types, absolutely and completely non-negotiable, but for the longest, getting 100kHz through is often not achievable. The signal looks like white noise. Many sinusoids from 5kHz to 100kHz.

No you haven't. Not using conventional components.

RL

Bugger! I could have sworn it was working at 180'C (along with all the other parts of the system), but it seems you know better. I must have a faulty oven.

I'd better warn all the other downhole instrumentation companies too!

But yes, selected conventional components, analog and digital. And yes, I know that if you extrapolate the graphs, most of the parts de-rate to negative power dissipation.

I used to build giant NMR gradient amplifiers. I used one opamp per mosfet to make a sloppy fet into a nearly perfect device.

SiC is good too.

When I first started working in this area, I was very surprised on my first day to see a colleague doing a crude temperature test using a hot air gun and a thermocouple, just checking before a long term test in an oven.

I'm not giving anything away which isn't well known to those in the business by saying the part in question was an ordinary 8-bit PIC. Operating at 180'C. An 85'C part IIRC, though I suspect the only difference between that and a 125'C part is the part number. I've had PICs running at 200'C, though 180'C is the usual benchmark.

Those involved have lists of components which they've tested. That takes considerable time and money so they don't readily reveal that information.

How about reducing the _amplifier_ temperature range with a Peltier element ?

This will:

- increase component life time

- simplify biasing

However, there are several issues with Peltiers:

- it doesn't tolerate strong vibration

- sufficient extra power must be available to drive the element, often more than the power that you want to transfer out from the amplifier

- the hot side can be well above 200 C if the ambient is hot, remember to dissipate both the amplifier losses and in addition the power for the Peltier

The Peltier can be used to cool the amplifier in a hot environment. Reversing the Peltier current and it can be used to warm up the amplifier in a cold environment. Thus the bias design is simplified.

Not practical, I'm afraid. These instruments are in a tube (a pressure housing) to cope with over 1000 Bar, and the inside diameter is maybe

Peltier modules are sometimes used for an individual small part, a camera chip for example, though I haven't used one.