Forums

clever huh?   first time i have seen this.  what it really is, is some local feedback around the current mirror.  the lower impedance load forces  the current source to supply more current to the drive of the output transistors. all those output transistors in parallel really form an extremely stiff drive current without global feedback. in single ended form, perfect for low impedance ribbons.   and 30v/10a switchers are cheap.  lot of heat however, 1.5 amp bias current. (total) i have a single ended gain stage to go with it.  basically half of the balanced version with a cfa capable input section. input section can have local feedback as shown. or remove that resistor. there is a later version with the input stage cascode just like the balanced board. will publish that one in a bit. pioneera09pre.pdf

Funny:)  I think for the existing  AIs it is easiest to be funny than design a working circuit. There are a lots of funny people in the internet for source. How long?  

i have not think it will design the circuit this AI is a chatting machine learn from  all the chats from the internet. (my keyboard is bad so there will be mistakes) I read about AI circuit design it exists. But it is f*cking good from this standpoint. I think I found its main source for the circuit but the result was not the same. I was awed but I like it because I can use it not for this day to day.

AI is not yet at the stage where you should trust the answers given you.  For example, I used an AI to translate an algorithm (Reverse Bits in Word by Lookup Table on this "Bit-Twiddling Hacks" page) to Rust, and the values in the lookup table that it came up with -- it had trouble converting the macros to values, plus it didn't use macros -- were wrong.  Fortunately, I know Rust well enough to have vetted and corrected the answer, but I feel the same needs to be done for everything -- let it suggest to you, but don't copy and paste the answers, make sure you understand everything it's doing with its solution. That said, it can be a very powerful tool -- I used the entire translation (fraction of a second) and vetting process (an hour or two at the most) as a learning experience. I would recommend you do the same with your efforts.  E.G. don't assume that the fault was yours, make sure you understand exactly what the fault was, and learn what it would take to fix it.  Or if it was the AI's fault, learn what it would take to have caught it before building the circuit.