Not for the output stage.
Why don't you go with one of the discrete transistor designs on this site? Those have already taken device variation into account.
The class D chip amp that you linked requires great care in layout. It is not DIY-friendly.
Ed
Why don't you go with one of the discrete transistor designs on this site? Those have already taken device variation into account.
The class D chip amp that you linked requires great care in layout. It is not DIY-friendly.
Ed
I've been taught that there are several error reduction techniques, and they fall in a few categories:
1. Negative feedback
This is used all over the place
2. Error feedforward
This is almost never used in audio, but there are exceptions. You can see QUAD's current dumping as a clever combination of negative feedback and error feedforward, and @jan.didden has also used error feedforward.
3. Compensation
A well-known compensation technique is balancing, like in a differential pair. When you use a single transistor as an amplifier input stage, you get an input offset voltage equal to its VBE, which spreads with process and spreads over temperature. Use two matched transistors connected as a differential pair and you compensate for most of the offset.
4. Modulation
This is a quite broad category. You can think of chopper techniques to get really small offsets, dynamic element matching in DACs, pulse-width modulation in class-D amplifiers so the precise characteristics of the output transistors don't matter much as long as they turn on and off well, pulse code modulation, that is, digitizing audio signals.
Two reasons:
1) the aforementioned engineering challenges with imprecise semiconductors
2) I just finished designing a circuit around the tpa lol
Although nobody says I can't make another one in spice and then see which I like better
Funny you should say that. That's actually one of the reasons I picked the TPA3255. Two of the channels I was going to use for that.
What about if you're using more than just a pair? I guess the better question would be to what extent do I need to compensate for variance? If it's just a matter of running it in differential that's not too terrible.
Interesting. I would have thought you needed to account for variances in semiconductors for modulation (like if one varied more than another then the modulus would look messed up). So would a polar amplifier be an effective means to modulate a signal as well?
I’d have to say that modern semiconductor processes make device variations a lot less than in the past. For Many devices hfe grading or ranking is not offered anymore since the devices fall in a much narrower range. As an example the ksa992 (fjv992) is only sold in the “f” ranking.