VC Low Pass Filter improvements?

[Hotstuff]

Hey all,

I have been working for a little while on a VC-LFP using the LM13700. It sounds really great with the Electric Druid TapLFO on the external CV.


The design was based partially on the 'Sub-Commander' filter from musicfromouterspace.com, the BBlender from http://seanm.ca/stomp/bblender.html and I'm using a LMC7660 voltage pump to get ±9V to power the IC's.
Using the internal buffers from the LM13700 cause a loud "thump" when bringing up the CutOff. Replacing these with simple opamp buffers fixed the problem totally.

I'm stuck with a couple questions though:
- Could I simplify the design and reduce parts? Perhaps the blend section is a bit much. How could I improve this?
- There is a little white noise coming from the filter. Searching the forums, I've seen other people mention OTA's being naturally noisy. How could I try to reduce the noise?
- The unused buffers from the LM13700 seems like a bit of a waste, has anyone dealt with the 'thump' problem and how can it be fixed?

Thanks!

[frequencycentral]

Do you have the achematic that thumps? It could be related to the CV input conditioning circuit rather than the buffers.

[slacker]

Looks good, great idea combining it with the TapLFO.

You can simplify it quite a bit. You don't need IC1A or the cap and resistor following it. You can feed the filter straight from the output of IC1B into the left hand side of the 39k resistor. I don't think you need the 4k7 resistors from the output of IC3, looks like they're left over from using the LM13700 buffers. You don't need IC5A, you can connect the blend pot to the output of IC3B. You can probably get rid of the caps connected to the blend pot, you're using a bipolar power supply so the outputs of the opamps should be 0 volts DC, so the caps aren't needed.
I'm not sure what the 100k resistors from the outputs IC2A and B are for.

You could use one of the unused buffers in place of IC1B, just connect it like you would a normal transistor buffer, and maybe the other one in place of IC5B. The opamps probably do a better job though.

Hope that helps.

[Hotstuff]

If I replace the IC3 buffer stages simply with the internal buffers from LM13700 I get a 'thump' as I sweep the CV from low to high. Putting 100K resistors in series between the OTA and buffer stage seems to reduce the thump a little. After the thump the audible lower sound frequencies kick in. Thanks for the tips so far!

[Hotstuff]

Here's an updated schematic!

It's all sounding very good, although still a little noise coming from the filter section. I hope someone might have some tips to deal with this?

[PRR]

> OTA's being naturally noisy. How could I try to reduce the noise?

Use a better gain-control block.

However they all hiss. And you can spend a LOT more for a little less hiss.

This is the 21st century. You can drop your signal and CV into a 20-bit 20-buck DSP, twist it in unimaginable ways, and come out pretty clean.

But you got these OTAs. Use the diodes, that's what they are for. You can bring signal level up going in, giving effectivly lower hiss.



Try it, but don't nail things down. You *may* not like it. In the no-diode OTA, distortion rises gradually, and you may be used to playing in the bent-zone. With the diodes, distortion is low to a point and then CLIP!! You may lose your sweet-spot.

Guide:

Diode current should be near 1mA more-or-less. The exact current is not important. 10K to +5V to +12V is fine.

Changing "39K" to 15K is 2.5:1 (8dB) more signal (and thus better signal/hiss ratio).

In abstract, the original "47K" should be 39K for unity-gain. However the filter-action chops-off a big part of the sound. Musically, we like a bit more than unity-gain for what is left, so the abstract 39K ideal is rounded-up to 47K. Now that the input is 15K, we make the feedback a bit more than 15K, and 18K is a handy value.

You can try this without yanking the 39K and 47K. Tack in the new 10K resistors to pins 2 and 15 to get distortion down. Tack a 24K across the 39K, tack a 29K across the 47K to bring up level. (22K and 27K may be handier values.) If that isn't better, four snips and the mod is un-done.

[Hotstuff]

Thanks for the tips, optimising the sound/noise ratio does seem like a good idea. Sadly, adding the 10K resistors on the diodes causes the filter section to go silent!

edit: Bringing the input resistor down to 10k gives much more of a volume boost than 15k. Doesn't seem to overdrive yet. What would be the best practice to bring the volume back down again before the blend pot?

Thanks!

[PRR]

> bring the volume back down again

If the ratio of the "39K" to the "47K" is nearly unity, the output level should be fine.

If you can use 10K in front in the "39K" place, try 12K or 15K at the "47K" place.

[Hotstuff]

It worked! Using the 10k & 12k combo reduced the hiss pretty significantly. I have an a/b setup using a breadboard so the difference is quite clear. How far do you think I can push it?
The resonance has increased with the new combination so I'll need to bring that down a little.

Still, I've not been able to use the diodes... I tried increasing the resistor values and eventually the sound returns but I don't thinks it's helping much compared to leaving the diodes unconnected.
Cheers!

[PRR]

> The resonance has increased

Ah, yes. Reduce that 22K by about the same ratio as you changed the other resistors.