# [sdiy] Odp: Re: Single supply op-amp filter biasing question

Michael E Caloroso mec.forumreader at gmail.com
Fri May 5 20:03:24 CEST 2017

```You have to be real cautious with decreased dynamic range in single
rail active circuits - reactive elements can potentially exceed the
output swing.  Overshoot due to Q can be sneaky and can make the opamp
go into phase inversion.  Noise can be a problem too - I just replaced
all the 4558s in my ARP 16 voice piano with OPA2134s and it did
improve the S/N ratio.

MC

On 5/5/17, Roman <modular at go2.pl> wrote:
> 2V on R1, leaves (9-2)=7V on RX. So RX is 7/2 times bigger.  or in another
> words  Ubias = 9V * (R1/(R1+RX))   You will loose a bit of amplitude too, as
> the divider R1/RX attenuate by 2dB. This can be compensatted by adding some
> gain to Sallen-Key.   How about blocking the DC at filter output with 22uF,
> so then you can set the bias at 4.5V with some considerably bigger resistor
> like 300k, that will not influence neither filter characteristics, nor
> amplitude in any significant way. It will need quite a few seconds after
> power up to settle though.   Roman  Dnia 5 maja 2017 17:02 Tom Wiltshire
> <tom at electricdruid.net> napisał(a):  Thank you very much, Roman.
> Where does 3.5 times greater come from? I could see 4.5 times greater, since
> 9/2=4.5, but how does 3.5 work?   Thanks,  Tom    On 5 May 2017, at 15:25,
> Roman Sowa <modular at go2.pl> wrote:   yes, you could, but I would just
> add the "X" resistor straight to 9V and skip 100k trimmer and 22u.
> 56K in parallel with X is your new value of the first resistor for filter
> calculations so keep that in mind.   To rise bias by about 2V, you need X to
> be 3.5 times greater than R1 (let's call that what you have there as
> 56K). And R1 in parallel with X should be 56k.  Now do your math.   Roman
> W dniu 2017-05-05 o 16:03, Tom Wiltshire pisze:   capacitor. But since the
> signal that I'm trying to retrieve with this  filter is a DC level,
> using a capacitor (AKA highpass filter) to  connect the input doesn't
> seem feasible. So could I do something like  this, and how would I calculate
> the value of X?   electricdruid.net electricdruid.net   Thanks, Tom

>

```