Has anybody tried. . . tossing the 22uf?
Hey, I was wondering about replacing the 22uf cap (zobel for the gainclone loop) with a plain ordinary wire jumper, so that only the 10k R is in use.
That does two things:
It grounds DC out of the loop.
It makes slightly (very slightly) higher DC offset.
For the larger gainclones, this warms up the sound and cools off the chip. Has anybody tried it with the baby?
Hey, I was wondering about replacing the 22uf cap (zobel for the gainclone loop) with a plain ordinary wire jumper, so that only the 10k R is in use.
That does two things:
It grounds DC out of the loop.
It makes slightly (very slightly) higher DC offset.
For the larger gainclones, this warms up the sound and cools off the chip. Has anybody tried it with the baby?
AndrewT said:
I'm not referring to blocking DC.
There is already DC blocking via input filter decoupling cap, and there is already AC shunting via the zobel on the speaker output. Not talking about either one.
Aren't we taking about LM1875? From pins 2 to 4, there's a fixed l-pad structure of resistors, pass of 100k, shunt of 10k.
However, my schematic has a 22uf cap between the 10k and the ground. It is shunting AC.
This is no longer a pure signal because it is modified many, many times by the presence of that capacitor shunting AC (music) at specific frequencies during amplification. That's a noise.
Furthermore, the value of the cap doesn't go in proportion to the value of the l-pad structure between different models of Overture chip amp.
So, not only is it a curious thing to do (make a noise during each pass of a copy, of a copy, of a copy sort of situation), but the recommended capacitor isn't consistent in proportion between models. So, the recommended bandwidth of this noise varies.
That's altogether puzzling.
richie00boy said:
Of course. Either way it is minute.
I'm referring to BrianGT's www.chipamp.com lm3886tf with specific directions that this specifc cap should be left off because it makes an awful sound.
One possible difference is that the fixed l-pad structure on lm3886 is 22k against 680 ohms wheras the lm1875 has 100k against 10k. The 22k version has a much firmer grip than the loosey-goosey hum-prone 100k structures. So, there's a possible difference.
paulb said:
Hi Paul,
I'm really glad to see someone mentioning composite designs for chipamps. I've been playing with them in LTspice for a year or so, in my spare time (but had to resort to using a TI OPA541 spice model (E-version), since there don't seem to be any good models for the National chipamps). My real work has gotten in the way of contructing any of them, for quite a while now. But I'm hoping to finally be able to try building some more of my designs, within the next month or two.
The first thing I noticed, after wrapping an opamp's feedback loop around a chipamp, was that the THD could be made almost arbitrarily low. I have many designs that have THD @ 20 kHz that's less than .0001% (and much lower, for all lower frequencies). It's also possible to simultaneously be able to drive large capacitive loads, with no overshoot or ringing, for squarewave inputs, using output slew rates that are the maximum the chipamp can provide (but not try to push it any faster). It should be interesting to find out how they sound.
I eventually was able to simulate them with fairly-detailed models, including models of speakers, models of speaker cables, models of real power supplies that use models of measured transformers and even models of the mains supply, including parasitics everywhere, even down to including the parasitic inductances and resistances of wires and PCB traces. (With LTspice, I was also able to use WAV files of music for the simulated circuits' inputs, and create WAV files from the outputs.
When simulating composite chipamp circuits, I found it fascinating to look at the signal being sent to the chipamp, with, for example, a squarewave at the main input. The outer feedback loop "knows" to add sharp peaks at the squarewave's rise and fall points, to improve the chipamp's output squarewave, and everything can be fine-tuned to get the best-possible edges out of the chipamp, but with no overshoot or ringing.
I've also done some chipamp designs that use Hawksford-style "error correction" topologies, where, basically, opamps in an outer feedback loop subtract the input from the output to get an "error" signal, which is then subtracted from the input to the chipamp. Some of the best of those even phase-shifted (delayed) the input signal (for the feedback loop) to try to get it better time-aligned with the output feedback before the error signal was derived, and used super-fast opamps in the feedback loop, to minimize time-alignment errors.
I'm also a fan of DC Servos, and have done quite a bit of work on different designs, for them, mainly in trying to get the AC response down to almost nothing, at very low frequency, while still maintaining a reasonably-fast response time. There is a thread, here, somewhere, with some of the initial ones, which includes links to the LTspice files for them.
There's some good stuff about composite amplifiers, in general, in Walt Jung's "Op Amp Applications Handbook", which is available as a free download from the Analog Devices website, at:
http://www.analog.com/library/analogDialogue/archives/39-05/op_amp_applications_handbook.html
Sorry to have blathered-on, for so long, about all of that. Carry on.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.