LM3886 feedback network

[BarryV]

Hi guys,

Can an LM3886 (or other chipamp) use a feedback network that is tpically used for a low-pass network instead of (or in addition to) the typical resistive network? See this page for examples of what I am asking: essentially replace the op-amp in the network with a chipamp.

Op-amp Filter - The Active Low Pass Filter

My hunch is that there may be stability issues, but I'd like the experts to weigh in!

Thanks for your time,


BarryV

[GloBug]

I might have read something like that on Linkwitz site, I'm not positive though.

Have a look around. Linkwitz Lab - Loudspeaker Design

[jcx]

audio power chip amps typically have a minimum closed loop gain spec >10 for stable operation

the OPA541 is one power chip amp exception that is unity gain stable - though not specifically designed for audio


the min gain requirement makes typical audio chip amps unusable as a low pass filter with a feedback C - the C increases the feedback at higher frequency where the low pass filter response is rolling off - violating the minimum gain spec and causing the circuit to oscillate at 100s of kHz

shelving low pass that flattens out at high frequency can be made if the ultimate gain is >10 or you can learn about "noise gain compensation" to make the chip stable at lower but still "flat" gain at high frequency

the Sallen-Key with gain filter could work but the gain has to be >10 - and then the sensitivity of the frequency response to small component changes (tolerance, drift) becomes large


releative to "good" small signal audio op amps, power chip amps are slow, have low gain - doing any filtering/signal processing with small signal op amps will typically give more accurate filter response

[BarryV]

jcx,

Thank you very much!!!

What about using 100% passive filtering ahead of the input stage of a chipamp? There are HP networks at the inputs of some power amps, usually in the form of a series cap ... why not put in a large shunt cap to perform a low pass function (with a series R).

I am asking because I am interested in doing a point-to-point LM3886 amp that contains a low-pass network. The addition of an opamp pretty much is out of the question.


Thanks again jcx!

BarryV

[AndrewT]

read the jFET passive filtering threads.

[KSTR]

Sallen-Key highpass filters can easily be done with ChipAmps in the ususal gain=20x config, just divide down the resistive feedback (to the SK-Cell) to get the desired overall gain for the SK-Cell (typically 1x...2x ref. Vin) with the same effective impedance as given by the design formula.

SK-Lowpass is harder but doable, too. To avoid a capacitive divider (can't be used, effectively), divide down the output with a low impedance resistor divider to the desired level and tap off to the SK feedback capacitor from there. There is some error from this series impedance but that can be kept low enough for practical work.

LTspice is your design friend for this, to check if the transfer functions look like the should.

Remember that design frequency Fp of any SK-Cell is alwas sqrt(C1*C2*R1*R2), while the system's Q is determined by the ratio of C1/C2 and R1/R2

[godfrey]

I simmed one with a Butterworth low pass at about 100Hz (and a 1'st order subsonic roll-off). My main concern was that without an input buffer, you're at the mercy of whatever source impedance the filter's being driven from.

It's behavior turns out to be surprisingly civilized. The red curve is for the circuit as shown, and the green curve is with it driven from a 5K source impedance. With any normal source impedance (say up to a few hundred ohms), the error should be negligible.

With a strategically placed pot, one could make a nice sub-amp, with cutoff frequency adjustable over an octave or so. Another pot could be added for adjustable Q. I'm starting to like this.

Originally Posted by BarryV

j
What about using 100% passive filtering ahead of the input stage of a chipamp?

That can work too, but then you're limited to very low Q filters. For similar complexity, an active filter gives more flexibility.

[GloBug]

Looks like you have a handle on it.

You should continue forward with this project.