rough LPF design

[Omega_Void]

Hi folks,

I plan to build an opamp LPF circuit. It's intended to process a line-level signal from a mic preamp going to a DI box. I determined the response I need using VST in the computer, but I need something which will work live.

I've drawn up a schematic circuit, it's kind of cobbled together from bits of other circuits I've seen. I don't know much about designing circuits, and I'm worried I may have overlooked something important.

So before I build it and perhaps find problems I won't know how to fix - can anyone recommend changes to the input/output/bias stages or the PSU section? Your advice would be deeply appreciated!

Thanks in advance!

[Nigel Goodwin]

What are you trying to do?, that looks a completely weird circuit!.

[sidiy]

A few notes, it's not necessary to follow all my ideas as the circuit seems generally correct:
- 10k in series at the input seems a bit big; are you planning to use a jfet input op-amps?
- 10k at the output may be increased to 47k and connected to ground instead of vb. If you still want to connect it to midvoltage (to avoid clikcs and pops) you may add a second vb network separated, only for the outputs. You may also need a coupling cap at the output before or after this resistor...
- is there a resistor between the last GIC section of the filter and the 10n capacitor connected to vb? At first view the filter seems a 9th order elliptic, i'm not sure what it is supposed to do...
- why the 470ohm in series with 1N diode in the power supply section?
- you may use a diode in series with the battery pack (or after the switch) to protect against reverse connection...

Well now I'm really out of advices...

[fdeck]

Some comments...

1. Do you really need 54 dB per octave cutoff? That seems pretty steep. Maybe you could get away with fewer poles.

2. A nice website, http://www.beis.de/Elektronik/Filter/Filter.html has tools for computing active filters using fewer op amps. Even the 3-pole might be sufficient.

3. About the power supply, I assume you want to stick with a single secondary, such as an AC wall xfmr. A possible alternative to a single supply design would be a split supply using a half wave voltage doubler circuit with grounded midpoint. There is more ripple, but big caps are cheap, and the regulators eat the ripple anyway. I have always had good results with this arrangement, and as a bonus, 12-V wall xfmrs are more common and possibly cheaper. You would then use two 9-V batteries with a shared ground, which should be sufficient for line level signals.

4. For low currents, I like the 78Lxx regulators because they don't need a heat sink, and have built-in current limiting. The 78xx need a heat sink, even at very low output current.

[Omega_Void]

Thanks for your advice guys - I really appreciate your taking time to post and help me out!

The filter's a 10kHz 8th order elliptical. The design is basically lifted from an application note for the FilterShop 3.0 program, but I nudged the resistor values to the nearest E24.

@sidiy -

The 10k in series at the input - maybe I should drop it to 100R, or could I just leave it out? Maybe I'm just thick, but I don't see what a resistor there would achieve except a drop in voltage?

Regarding the opamps, I'm thinking in the end to use maybe OPA2134 or OP275 which I'll have to mail-order. I have some TL072 right now which I'll try for starters.

I didn't include a coupling cap on the output 'cos using a passive DI the transformer would not pass DC, and with an active DI the DI input would necessarily have a coupling cap if DC blocking was required. Am I missing something?

The 470ohm in the PSU came straight out of a Boss pedal schematic. No idea what it's for, maybe I'll just leave it out?

What I don't get is why the output needs a separate midvoltage reference. I don't doubt you, but I just don't get it

Thanks a lot for your recommendations!

@fdeck -

I guess I'll use the uA78L24 regulator - I kinda figured the higher-current reg would run cooler, thanks for putting me straight there!

I decided against using a shared ground for the batteries 'cos I'll be using rechargeables, and I worry that uneven drain might decimate my headroom. Maybe I'm being alarmist, but this way if one battery were to suddenly fail to a short in the middle of a gig... I really don't know if that's likely but I'm heeding Murphy's law here. With that decided and a biasing network therefore required, a split supply just didn't seem necessary. But perhaps when I look at the relative costs of the transformers I'll change my mind and use your voltage-doubler scheme!

Thanks again for posting!

It's now looking like this:

[sreten]

The filter's a 10kHz 8th order elliptical.

Hi, And it is of some practical use ? /sreten.

[Omega_Void]

I'll be using it to tidy up an eeevil distortion patch. It sounds cool (to me anyway) but gets real fizzy and hissy up top.

Only VST plug which fixed it was the brickwall filter in Wave Arts TrackPlug, which is 10th order elliptical. 8th order should be enough though.

All-pole filters just didn't tidy it up enough. By the time the fizz was gone the top end of th sound was losing too much bite. Even 48dB per octave didn't hack it. Just didn't have steep enough roll off.

I figure it'll be good for patching in the loop on mega-dirty guitar sounds also. Anything that sounds cool but gets scrappy up top.

[unclejed613]

you could try adding a small integrating cap to the fuzz circuit. not sure what you're using for fuzz (nothing i'm familiar with at least), but most fuzz circuits use an amplifier and a nonlinearity, the most common these days being an op amp and diodes in the feedback loop. if you put a 200pf cap across the diodes in the feedback loop, it gets rid of the "fizz" quite well.

[Omega_Void]

Sadly 'tis a digital patch!
But thanks for the tip, it may well come in handy sooner or later...
:-)

[unclejed613]

if it is a digital patch, then IIRC you get a true "brick wall" at half of your sampling rate. can you run at a 20khz sampling rate?