[<<][staapl][>>][..]Mon Nov 12 15:29:42 CET 2007
i changed the circuit back to 1K input impedance, 100K feedback in
first stage. the second stage has 1K + 100nF, and 100K feedback, and I
have no idea why this works: less noise, and it seems to have a good
response in the intended range..
maybe because most sounds have a 1/f response? i don't know... it
responds well to whistles, which is nice.
this is a 10 kHz pole... so it's basicly set up as a differentiator?
maybe because i have GBW rolloff this works? i'm puzzled.
i tried with a LM358N and it gives a lot more noise.
i tried TL072CN and it gives too much bandwidth! so, i use a
compensated integrator with 1K/100n in the source and 220K/4.7n in the
feedback section. looks like this is final enough.. maybe beef up the
amp just a tiny bit more..
2 x 220K
1 x 100K
2 x 10K
3 x 1K
2 x 15pF
1 x 4.7nF
2 x 100nF
2 x 10uF
1 x 10MHz
1 x 18F2620
1 x TL072CN
1 x LED(red)
2 x 6 PIN HEADER
SPK Rg 220K | R2 220K |
o /-/\/\/\--\ o--/\/\/\---o
| | __ | | __ |
| Rs 1K | | \ | R1 1K | | \ |
o--/\/\/\--o-| - \___o--||--/\/\/\--o-| - \_____o LINE
|\ @ ____| + / C1 100nF __| + /
| ||@ | |__/ | |__/
|/ @ o---------------------------/
=== Cs |
| 10uF |
| Cn |
| 10uF |
GND 2.5V 5V
First stage gives 220 x amplification.
TL072 (TI version, i'm using ST version) has a GBW of 3 MHz, with 220
x amplification this gives rolloff at 13 kHz. so for the first stage
Second stage is a band pass filter with 22 x amplification:
G . . . . ._________
/ . . \
/ . . \
t1 = R1 C1 = 100us -> 10kHz
t2 = R2 C2 = 1000us -> 1kHz
because f1 > f2 the gain is not R2/R1 but R2/R1 * f1/f2.
a bit quirky, but it works.. maybe i should try with exchanging the
time constants so f2 > f1.
looks like these changes keep the transfer function the same, with a =
R2 -> 1/a R2
C2 -> 1/a C2
R1 -> a R1
C1 -> a C1
so, there's a reason to do it like i did! the capacitors are
smaller. so where's the trade-off? maybe noise due to large resistors?
however, when f2 > f1 the gain is independent of the capacitors.
let's make this a bit more intuitive. what happens when C1 is made 10x
larger, so f1 = 1kHz, and C2 is made 10x smallr, so f2 = 10kHz? the
gain is now 10x more, so then the gain can be reduced by making R2
10x smaller, which again requires C2 to be 10x larger. so the net
R1 -> R1
C1 -> 10 C1
C2 -> C2
R2 -> 1/10 R2
this gives a 1uF capacitor. so alternatively R1 can be made 10x
larger, which requires C2 to be made 10x smaller. giving 10K and 470pF
respectively. (EDIT: this is what i did. works fine).
makes more sense now. so is there a ciruit that has independent
frequency and gain?
hmm.. i just tried the LM358N again, and it gives good results
also. guess the TL022 was just too low bandwidth? yup. 0.5 Mhz. hmm
the LM358N is only 1MHz ?