Using 4-pole VCFs as 2-pole

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: Kent, luketeaford, Joe.

Post Reply
Borogove
Common Wiggler
Posts: 56
Joined: Sun Feb 25, 2018 10:56 pm

Using 4-pole VCFs as 2-pole

Post by Borogove » Sun Oct 10, 2021 3:55 pm

I was discussing the CEM3397/PA397 voice IC (dual DCO shaper, mixer, and 4-pole VCF) with someone, and we got onto the question of how the DSI/Sequential synths such as the Prophet Rev2 use it to get a 2-pole filter mode, given that there are no taps between the filter stages. Our hypothesis is that two of the external capacitors which contribute to the filter tuning are switched out to much smaller values (or disconnected entirely?) in 2-pole mode, pushing the cutoff frequency of those stages much higher, so the most significant part of the filter response is 12dB/octave even though it may go to 24dB/octave eventually. Since the resonance control is designed to yield self-oscillation in the 4-pole mode, it doesn't get anywhere near oscillation in two-pole, since the resonance is split between two different frequencies.

I have a few questions:

1) Can someone confirm this hypothesis?

2) The datasheet values for the CEM3396 (almost identical to the 3397) show 33nF caps for the first three stages, and 470pF for the fourth. (A note on the fourth gain cell in the filter says "70x", which presumably relates.) Which caps would be switched out for the 2-pole mode? The first pair of 33nF, or the 3rd and 4th?

3) I'm interested in applying a similar 2/4-pole switch for the SSI2144, which similarly uses 4 external caps, with the datasheet suggesting 3x 6.8nF and 1x 560pF. Would the same trick work there?
Last edited by Borogove on Sat Oct 16, 2021 11:04 pm, edited 1 time in total.

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Wed Oct 13, 2021 8:19 pm

Strange, I thought the CEM3396 was a cascaded topology. That capacitor configuration makes me think ladder now. Interesting!

I’ve never seen one (personally) configured as a 2 pole, except if the 3396/PA397 is ladder, in the DSI/Sequential synths. Maybe they exist, but I’ve always been more interested in State Variables and cascaded OTA/VCA types.

That might explain how the resonance is almost non-existent in two pole mode.

I’ve been working with the SSI2144 a lot lately. I’ll see if I can figure out how that would work.

Now I’m intrigued. It would be simpler in a cascaded topology. In which case a 2164 or 2140 would be the clear choice.
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Wed Oct 13, 2021 9:24 pm

the 3396 looks ike a cascaded topology. maybe they just buffer the second cap and take it as an output? could also just bypass the feedback path at the same time.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Wed Oct 13, 2021 11:33 pm

guest wrote:
Wed Oct 13, 2021 9:24 pm
the 3396 looks ike a cascaded topology. maybe they just buffer the second cap and take it as an output? could also just bypass the feedback path at the same time.
True. It does have “some” resonance though, just not much. Maybe it leaks back in though somehow. I really don’t know enough about the chip. It actually sounds like a cascaded filter to me, and the block diagram does look that way too. (which is why I originally thought that)

Having worked a bit with ladders lately though, they seem to have that nF range on the later poles with a pF range on the first section. I haven’t seen that on a cascaded filter before. Usually each section is identical.

I’m not the world’s most knowledgeable filter guy though. Maybe the math works out in a way that I’m not thinking of. My experience is pretty much in the SVF and more typical 4 pole cascade areas. (MOTM 440, 2040/2140, Ripples, some Roland, and my own DownLow (13700 based) all have identical cells for example)

Could a ladder be drawn that way in cells like the datasheet diagram? I wasn’t confused when reading it before, and I thought I knew the sound, playing my Prophet Rev2, but now I have to admit, I’m slightly confused :lol:

They don’t really say much about it in the datasheet.

Figure 22 in the SSI AN701 doc is what I’m used to seeing:

http://www.soundsemiconductor.com/downloads/AN701.pdf
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Thu Oct 14, 2021 2:57 am

for a ladder, youd have to have connections for either side of the caps, or it would be a single-sided drive like a steiner parker filter, which has worse common mode. im guessing the larger cap in the last stage is due to a higher control current being used in the last stage OTA to allow for a higher output drive in the last stage. the previous stages only need to drive the next OTA, but the last stage might need to feed a bunch of amps, etc. also, filter noise might be dominated by that output stage, so higher current helps there?
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Thu Oct 14, 2021 10:03 am

True, hadn’t thought of that. Makes sense for the final drive too. Probably saves a bit of extra buffering, able to feed the output, feedback, etc. Interesting.

I originally thought it was a cascaded filter, but never really paid attention to the cap config, and that’s definitely what I found confusing now. Thanks for the input!

I always thought this chip was cool. I like how the oscillators are clock driven for example.

Anyway, as to getting a 2 and 4 pole response from the same filter, I would still recommend the 2164 or 2140. They’re basically made for this task. Fully configurable cells.
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Thu Oct 14, 2021 12:59 pm

yeah, definitely cool chip. the 3109 has high output drive on the 2 and 4 OTA buffers, as a way of selecting 2 or 4 pole.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Thu Oct 14, 2021 2:02 pm

guest wrote:
Thu Oct 14, 2021 12:59 pm
yeah, definitely cool chip. the 3109 has high output drive on the 2 and 4 OTA buffers, as a way of selecting 2 or 4 pole.
Interesting! I’ll check it out! Now I’m intrigued a bit. It’s just funny that I’ve looked at tons of four pole cascade schematics from VCA to discrete to various OTA, and custom chips, and never saw this before. Completely threw me in the case of the 3396 block diagram :lol:

Pretty sure all the Haible, Sowa, Osamu, Gillet ones Ive seen and their derivatives all had identical cells, and those were usually the places I’d go first to look at how to do these things. Now with the SSI app note, I’ve been sticking to that and their datasheets.

One thing I would like to do is either a 2164 or 2140 based Xpander style filter with logic or software (if I can get the help) mode switching. I made an all analog fully adjustable mix filter like this once, but 80%+ of the settings were useless. Using specific configurations and ratios is pretty important I think. :hihi:
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Thu Oct 14, 2021 5:51 pm

on the 3109 the OTA cells are the same, but the buffer cells are not.

the xpander sounds like a cool project. would you replicate what they did, or go through and design all your own settings? it looks like they swap out one of the capacitors to reduce the max poles by one.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Thu Oct 14, 2021 9:00 pm

I’m not 100% sure yet on the approach actually. I’m not sure how many modes it really needs, but I’d love to do several, including something more than the usual four modes of SVFs.
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Thu Oct 14, 2021 10:58 pm

i have a design on paper that ive been procrastinating on that is 2x SVF, which i want to actually build up and listen to at some point, as i think it will be interesting to have 2 seperate cutoff frequencies. i suppose the traveller filter did this. i wonder if there is a not-so-complicated way to get more cutoff filters, sort of a mini vocoder.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Fri Oct 15, 2021 2:26 pm

If you haven't laid out a PCB yet, I have a triple SVF PCB. I'd be happy to send it your way. It's LM13700 based. It has an input and output mixer. The mixers are for parallel routings, but there are also jack connections for patching in series configurations. After the Zyrinqz (Syrinx clone-ish) filter I did, I decided to ditch the four pole section, and put one more state variable section in its place. (more flexible) Attaching the schematic. If you'd like a PCB let me know.

Here's video of some of the sounds the old one made (with two bandpass only SVFs and one four pole lowpass):



Attachments
FlexFilter.PDF
(333 KiB) Downloaded 13 times
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Fri Oct 15, 2021 3:02 pm

thats awesome! those are exactly the sort of sounds i was imagining. the schematic should be enough for me, ill start digging in.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 7006
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Re: Using 4-pole VCFs as 2-pole

Post by J3RK » Fri Oct 15, 2021 3:19 pm

Nice! :tu:

The SVFs themselves are based roughly on Scott Bernardi's OG synth SVF. I wanted to use inverting integrators, and I also liked that the linearization diodes were used. He didn't mind me chopping it up, so that's where the basic filter cores are from. The matched pairs could be chopped out if desired. I use LS358s there. They're quite inexpensive for a good log conformant pair.

The individual outputs could probably use buffers, but I've never had an issue with overloading the integrators, so I just left them as-is.

The buffered precision shunt probably also isn't needed. I was going to do something different with it, where it might have actually helped more.

R63, R9, R27 could be on the -10 reference, and adjusted slightly for it. The linearization diode supplies could have been on the reference voltage as well. Right now, just the pots are using them, in which case they could have been unregulated, or use cheaper LM types.

Otherwise it should be pretty straight-forward.

Oh, one last item. I think I picked 15K resistor values for the linearization diodes to keep the overall current input to each OTA at a certain level. I also laid this out when I was using 15V supplies, and didn't decrease those when I converted to 12V. So they may need a small downward tweak.
Are you lemon? Does your head come to a nub?

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Fri Oct 15, 2021 6:04 pm

cool, thanks for the tips, thats super helpful.
openmusiclabs.com

Borogove
Common Wiggler
Posts: 56
Joined: Sun Feb 25, 2018 10:56 pm

Re: Using 4-pole VCFs as 2-pole

Post by Borogove » Sat Oct 16, 2021 11:06 pm

guest wrote:
Wed Oct 13, 2021 9:24 pm
the 3396 looks ike a cascaded topology. maybe they just buffer the second cap and take it as an output? could also just bypass the feedback path at the same time.
The filter goes directly into the internal amplitude VCA, and there's no way to re-inject the signal bypassing the later stages of the filter, so that won't work.

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Sat Oct 16, 2021 11:57 pm

shoot, hadnt noticed that. your idea of bypassing two of the stages sound good. but, i think id use a resistor to replace the caps in those stages. or perhaps you can buffer the cap on the first stage and inject it into the cap node on the 3rd stage (disconnecting the cap on the third stage, and shorting the second stage to ground)
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 6650
Joined: Mon Aug 19, 2013 11:49 am

Re: Using 4-pole VCFs as 2-pole

Post by guest » Sun Oct 17, 2021 12:00 am

another way might be to buffer the 2nd stage output and feed it into the cap on the gain setting RC network (disconnecting the cap from ground and feeding there). so you get a mix of 2 pole and 4 pole, which might add up to 2 pole. the advantage of this method is that the resonance will still work similarly.
openmusiclabs.com

Post Reply

Return to “Music Tech DIY”