Greetings All,

I'm currently working on learning about sine generation techniques and 3 level PWM for driving an H-bridge. I came across this "Precision Tri-Wave Generator Circuit" from National Semiconductor and a differential pair circuit to convert the triangle wave to a sine wave with fairly high accuracy. My issue is that I was told I need to temperature compensate the signal controlling the base of the BJT differential pair if I want to have good accuracy but I'm not sure how to do this.


TI has an example log shaper circuit with clamping diodes as temperature compensation but I cant tell how this compensates for temp in any way. I think the diodes are for standard reference and the double diodes are there for a 1.4V reference of some kind instead of 0.7V with one, but again, don't know how they compensate for temp.


Any help, or literature, in understanding what is actually going on in the diode section of TI circuit or how to modify a circuit for temperature compensation is what I'm really after.

Thanks in advance

 

Who told you that you need temperature compensation? What temperature range does he want and how accurate does he want it?

 

It was an engineer at one of Harley's test facilities (we use their dyno). I was there for a design team test and we got to talking about my pet project and he told me he thought it needed temp compensation. We had to start testing again so he didn't tell me anything beyond that. Your guess is as good as mine.

 

It was an engineer at one of Harley's test facilities (we use their dyno). I was there for a design team test and we got to talking about my pet project and he told me he thought it needed temp compensation. We had to start testing again so he didn't tell me anything beyond that. Your guess is as good as mine.

I think you are okay if you use the transistors shown as they are matched and in the same package.
Edit:
What will the triangle wave circuit look like?

 

Great to hear! I'll go ahead and test what I have once I get components and tell you if it works.

 

I think you are okay if you use the transistors shown as they are matched and in the same package.

The LM394 is a matched pair on one piece of silicon, so the two transistors not only match each other, they simply must be at the same temperature as each other. I would bet you can get the PNPs in monolythic pairs, too. If that is inconvenient, you can glue the PNPs together.

 

Remember that the differential pair in the first circuit (TI Example) makes the triangle wave look like a sine wave by acting as a logarithmic clipper. As temperature increases, the voltage at which visible compression takes place will increase and the result will be a change in harmonic content and a change in amplitude.

This is explained in the National Semiconductor Applications Note 222 Super Marched Bipolar Transistor Pair.
users.ece.gatech.edu/~lanterma/sdiy/datasheets/transistors/AN-222.pdf


The important part is that the base-emitter voltage is proportional to temperature (T in the numerator).

Compensation for this change requires the amplitude of the applied waveform to be changed by +3300 ppm/°C. You can probably do a good job of it with a combination of a thermistor and a fixed resistor.

 

What will the triangle wave circuit look like?

Just saw your edit, here you go: (the specs can be found in the original pdf attached to my first post)