MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

Coming up with a Filter for "Pinging"
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Author Coming up with a Filter for "Pinging"
robin87
Hi everyone!

Just to clear things up, because there has been some confusion in related discussions:
„Pinging“ a filter refers to a technique where you feed a short trigger or envelope (somewhere between 1 and 5ms works well) into the audio input of a filter on the edge of self oscillation, which should produce a percussive sound. Pitch is set by the frequency of the filter, while the decay can be fine tuned with the resonance knob. Velocity depends on the shape and amplitude of the actual trigger, a 5v envelope with a bit of decay will sound different than a 10v trigger with sharp edges. This makes for all kinds of great percussive sounds, lots of classic analogue drum machines work like this.
It is not related to plucking a lowpass gate for example, or sending short envelopes to the cutoff cv of a self oscillating filter. Both techniques make for good percussion sounds, though.

I guess every filter capable of self oscillation should be able to be pinged, but some seem to be better suited to the task than others. Filters I’ve tried:

Yusynth Steiner VCF:
Great filter, not so much for pinging though. Doesn’t track very well, just very hard to tame in general. The Resonance is really hard to dial in and it sounds rather nasty.

Minimoog VCFs of all kind:
Minimoog VCF is known for not self oscillating in the lower range. Pinging kind of works, but is inconsistent throughout its range.

Serge VCFQ (both CGS and R*S ):
Really good for pinging, produces a nice and clean sine. It even comes with a trigger input specifically for this task. (I still prefer to use dynamic triggers into the audio input.) Since the VCFQ will not self oscillate without feedback patching, the resonance is very tweakable and can be tuned for really long decay in the lower range. Sounds kind of weak and clicks in the upper ranges, though.

Mutable Instruments Ripples:
Not quite as clean and loud as the VCFQ, but pings very well throughout the entire audio range. Tuning the decay is a bit harder, the usable range for pinging is between one and two o’clock on the resonance knob. Will start to oscillate earlier in the very high ranges.


So far it’s a tie between the VCFQ and Ripples, but both have their shortcomings. The cwejman res-4 seems to be the gold standard when it comes to pinged filters, unfortunately that’s way out of my price range for now.
I’d like to experiment with a couple of filters and try to come up with a circuit that produces clean, consistent pings throughout its entire range and serve double duty as a nice vcf as well.

I’m looking for some advice where to start, what filter topology is best suited to producing clean sine wave pings, how to tweak the resonance path for easy fine tuning of the decay, etc.

Here are some of my criteria:

Bandpass Resonse:
Seems to produce the best pings in my experience. Would also be the most useful for filtering duties in my case, I tend not to use lp/hp as much.


Clean sound:
I’m not really looking for „character“ here. Don’t get me wrong, I appreciate gritty filters, but I already have that covered. Since I plan to eventually hook up a couple of filters in parallel and use them as a filter bank as well, I guess clean sound and low distortion are preferable.

Good tracking:
Not looking for 10+ octaves here, but it should be able to track V/Oct reasonably well.

Self Oscillation:
Clean sine waves with the resonance fully CW would be nice to have, but I’d rather have precise control over the resonance. Don’t want a filter that starts to go crazy once you get past 12 o’clock on the resonance. Would also be fine with having to use external feedback for oscillation, like in the VCFQ. Maybe use a switch to get it into full self oscillation.

My first intuition is to start with a simple state variable design. Dr. Sketch-n-Etch shared a basic 2164-based schematic and some great insight in this particular thread:
https://www.muffwiggler.com/forum/viewtopic.php?t=65976&start=0&postda ys=0&postorder=asc&highlight=

Seems like a great place to start, so I’ve ordered a couple of 2164s for prototyping.

A AS3320 based VCF has also crossed my mind, since they are (kind of) cheap and readily available now. I’ve never tried a 3320 filter, but apparently they sound pretty clean. Should be a pretty straight forward build, especially since I’ll probably be fine with bandpass response only. Ordered one of those as well.

A THAT2180 based approach could be worthwhile, but since I could not find a ready-to-prototype schematic and my understanding of EE and circuit design is very limited, I’d probably save that for later.

Anyway, I'd love to hear some of your thoughts on this. Any advice is appreciated!

Thanks! thumbs up
clorax hurd
What happens when adding resonance to LP filter with 4 poles in the same place (=in synth world, 99% of 4-pole LP filters are of that kind) is analysed here http://www.timstinchcombe.co.uk/index.php?pge=poles
I guess that another part of the story can be the VC element adding some non-linearity to the signal. I guess it can't do much to the original impulse (has all the harmonics already anyway), but can do thing to the wave it creates.

If you want BP then in my opinion the best approach would be trying to design such BP filter, that it's delay time at center freqency would be equal to period of that (center) frequency, or some of it's subharmonic.
So that the frequencies which can pass thru the filter the most, will also be the ones which can resonate when feedback ("resonance") is added.

either way, the phase/group-delay aspect of filter is very important in my opinion, because filter pinging is essentially a karplus strong synthesis (which is based on delay).

(so i'd also experiment with polarity of the feedback/resonance part)

filters have different delay times for different frequencies. that delay vs frequency graph is defined by the constellation of poles and zeros. most synth LP filters have just 2 or 4 poles in one place... like more 1-pole filters tunes to same frequency stacked... probably because it's "musical" that way.
but for your purpose i'd optimize the filter design.
not sure how difficult would that be and if possible at all.

with BP filters it seems much less predictable to me.
with LP or HP you get higher chance that the passed frequencies will also hit the resonating frequencies will hit the passband again...i think that's why the pinging can sound nicer than regular accurate digital karplus strong...

and also why more difficult to tune.. having just good tracking VCF is not always the whole story. (but you can get lucky)... so i'd first make sure it tracks ok, before spending money on 2180 to make it perfect...

(not sure if all thing i written are correct. just thinking aloud after thinking about these things a lot during the last 3-4 years)
wsy
Clorax hit it on the head, but it's a sad head.

Resonance pinging can only give you sine waves; all of the interesting fun is in the envelope.

So, while it is fun (and I don't deny it!!! NOT ONE BIT) it's also something you would get better control
of just by using a VCA and an EG (i.e. a MATHS). Or even use the VCA as an LPG.

Then you also get a lot more flexibility in the waveform (i.e. harmonic or inharmonic structure). Just
put a wavefolder after the VCA (or, after the pinged filter) and have at it.

Sorry to be a downer...

- Bill
robin87
Thank you both for sharing your thoughts!

clorax hurd wrote:


Very interesting, admittedly most of that went straight over my head, I'm afraid. I’m very much a novice when it comes to analog circuit design. Still, the comparison to karplus strong is especially intriguing, since that is something I’m very interested in right now.
(Just ordered some AD9833s last week, planning to put my pair of Doepfer A-188s under tight digital control, but that’s a topic for another thread... )


wsy wrote:
Clorax hit it on the head, but it's a sad head.

Resonance pinging can only give you sine waves;
[...]
Sorry to be a downer...

- Bill


Not sad at all, in fact that’s exactly what I’m after: a nice percussive sine for further processing. Or better yet, multiple filters that produce a couple of sines with independent pitches, levels and decay times. (See where this is going? I mentioned a certain expensive and hard to get quad bandpass in my first post, didn’t I? grin )

What I’m looking for is a way to fine tune a filter to produce these pings consistently over it’s entire range.
Like I said, the VCFQ is great for pinging, but I find the sounds get weaker/more „clicky“ the higher you go in frequency. The MI Ripples gives me better pings in the higher range, but it’s more prone to distortion and self oscillation.

My ideal filter would produce pings with the same „strength“ (I guess you could say amplitude/decay time, although it would probably sound more natural if the decay time gets shorter the higher you go) at all frequencies. But a lot of the filters don’t do this, for example they will go from nice, boomy pings in the lower range to wild self oscillation in higher registers. (Or hardly ping at all past a certain frequency).

For me it’s not about reinventing the wheel, I guess it’s more a question of finding a filter topology suitable for pinging and then tweaking it to my liking. In other words, fine tune the range and behaviour of the resonance.

Sounds simple, but as I'm a bloody beginner in everything circuit design, I'm grateful for any advice i can get!
clorax hurd
robin87 wrote:
What I’m looking for is a way to fine tune a filter to produce these pings consistently over it’s entire range.


I am not sure why they normally don't. I feel like ideal VCFs in theory should. So I would suspect the implementation is the problem. Using VCA chip as VC element sounds like a good idea. (that2180 may be overkill, 2164 should do)

btw CGS version of VCFQ schematic is available. https://www.cgs.synth.net/modules/cgs112_vcfq.html
That one is using serge gain cells made of discreet parts instead of VCA chips.

I'm just thinking that having it tidy and predictable is perhaps easier for LP filters, as they cause constant delay to the low frequencies which are not attenuated.
For HP and BP filters the delay time is variable in the passband which can (but maybe won't) cause issues. (even if it doesn't seem like an issue for ideal models fed by uniform noise bursts.)

robin87 wrote:
Like I said, the VCFQ is great for pinging, but I find the sounds get weaker/more „clicky“ the higher you go in frequency.


The ping response should be shorter in high frequencies. Resonance = the delayed ping is fed back into filter input, just slightly attenuated.
Lets say that the resonance is attenuating to 99% of the original amplitude. On output the first period will be the impulse with 100% volume, 2nd period is delayed&filtered impulse of 99% volume, 3rd period is 2*delayed and 2*filtered impulse of 0.99*0.99 volume... ... and it takes 69 !periods! until you get to 50% volume = -6dB.
The decay length with karplus strong is about periods and not milliseconds.
For 100Hz and 0.99 resonance it will be 69 periods = 0.69 seconds until the decay gets to 50%,
while for 1000Hz and 0.99 resonance it will be still 69 periods to get, but for 1000Hz it's now just 0.069 seconds, so for 10x frequency sound also the decay is 10x faster...

and sure it's more clicky for high frequencies - it's a filter. if you filter impulse with lower cutoff LP (or BP) filter, you're removing the clickyness. when moving the cutoff higher you're not filtering the clickyness anymore.

so more clicky and faster decay for higher frequencies is just natural.

if you want
robin87 wrote:
pings with the same „strength“ (I guess you could say amplitude/decay time,


then you'd probably need to have frequency CV affect also the resonance in a very accurate way... very possible in digital, but too much hassle in analog... especially when you consider how sensitive the resonance amount is... it's a very small range we care about when you want audible decay but not self-oscillation...

I guess it will be quite difficult to make it better than the R*S VCFQ.
Nice mod for the VCFQ could be limiting the resonance range to where the interesting resonant decays happen. normally it's just few percent of the range...streching that to whole pot range / whole 0-5V CV range could give you more control
Grumble
In my synth I use several programmable switched capacitor filters like the MAX26x series.
They are not just programmable in frequency, but also in Q-factor (resonance) and you can even get them to self oscillate.
cygmu
If you want the ping duration to be roughly equal at all frequencies, you will need to raise the Q (resonance) as the frequency rises, I think. Theory follows -- I don't necessarily understand all this properly but here goes anyway. One source for it is https://ccrma.stanford.edu/~jos/fp/Decay_Time_Q_Periods.html

For an ideal second order low-pass, the impulse response (ping!) is a decaying sine whose decay time constant is Q cycles. What this means in practice is that a 400Hz ping decays twice as fast as a 200Hz ping if the Q is fixed.

A regular state-variable topology gives you independent control over centre frequency and Q. That is what is used in the Serge VCFQ for instance. It means that unless you change something deliberately, you have the same Q throughout the frequency range which is what gives the shorter clicks at high frequencies. So maybe something like a VCFQ with a circuit that automatically adjusts Q with frequency, to partially compenstate for this effect, would do what you want?
robin87
clorax hurd wrote:

and sure it's more clicky for high frequencies - it's a filter. if you filter impulse with lower cutoff LP (or BP) filter, you're removing the clickyness. when moving the cutoff higher you're not filtering the clickyness anymore.

so more clicky and faster decay for higher frequencies is just natural.


Of course, this makes perfect sense. Thinking about it, it's even desirable for natural sounding percussions.

clorax hurd wrote:

I guess it will be quite difficult to make it better than the R*S VCFQ.
Nice mod for the VCFQ could be limiting the resonance range to where the interesting resonant decays happen. normally it's just few percent of the range...streching that to whole pot range / whole 0-5V CV range could give you more control


As for the range of the resonance control: I think it works very well as is on the VCFQ. In my limited understanding, the lack of fine control over resonance on a lot of VCF designs is due to the fact that they:

a) Often use some sort of linear control in the feedback path, either a linear pot or an OTA based VCA. A logarithmic response of the resonance control would be preferable for fine tuning, especially when on the edge of self oscillation. Both the so called „blackmer gain cells“ used in the VCFQ or the fancy VCA ICs are exponential. As I’ve learned rom Dr. Sketch-n-Etchs posts in the thread I’ve linked to in the first post, the resonance works by reducing the negative feedback of the bandpass stage, so an exponential VCA in the feedback path would result in a sort logarithmic response of the resonance control, right? Sorry if I’m mixing things up here or use the wrong terms, I’m just starting to grasp the basics. Please correct me if I’m wrong!

b) most VCFs just have way to much resonance on tap. Seriously, I don’t get why so many filters start to tip into crazy self oscillation once you go past noon. Even using a reverse audio tapered pot won’t help, because the once you reach the more controllable part of the pots travel, you’re already in „scare your dog“ territories. I wish more filters came with a trimpot to set the maximum resonance.




clorax hurd wrote:

then you'd probably need to have frequency CV affect also the resonance in a very accurate way... very possible in digital, but too much hassle in analog... especially when you consider how sensitive the resonance amount is... it's a very small range we care about when you want audible decay but not self-oscillation...


cygmu wrote:
So maybe something like a VCFQ with a circuit that automatically adjusts Q with frequency, to partially compenstate for this effect, would do what you want?


Interesting, this has crossed my mind as well.. turns out the idea is anything but new. Looking at the VCFQ again, that seems to be exactly what’s going on with R12. Both Ken Stone and Random*Source included it, but basically recommended against placing it by saying it’s hardly ever used on the actual serge units. It’s referred to as „Automatic Resonance Control“ and feeds the output of the CV stage back to the summing node of the Resonance CV Input via a 100k resistor.

I'll try to install it and see how it works out, will report back here.


After all, I think the VCFQ makes a for good starting point.
Yet if you take away the more esoteric features like the AGC and trigger inputs, it’s really just a textbook state variable filter, isn’t it?

Random*Source replaced the gain cells with that2180s and there are a number of 2164 based SVF designs similar to Dr. Sketch-n-Etchs schematic out there, even discrete designs have been done.

I guess I'll just have to start breadboarding and see how they all differ sonically and performance wise.
cygmu
I had forgotten about the ARC resistor -- well remembered/spotted. That is definitely something to experiment with. Try different values from 100k down to 15k or so to see what effect they have.

The VCFQ's resonance control does indeed have an exponential response because it is controlling the Blackmer-style gain cell, both in the 1970s design and in the R*S version with 2180s. The 2180 is pretty much a drop in replacement for the original gain cell, though you have to tweak the control circuitry a tiny bit -- just a few resistor changes I believe.

I'm looking forward to hearing how your experiments go!
wsy
That's an interesting point though - one should be able to just patch resonance compensation into the system
by multing off of the 1V/octave pitch thru an attenuator into the resonance CV in.

Worst case, if resonance in uses one of those "it's level until you plug something in, then it's attenuator" use a signal
processor to offset + gain it.

THAT should give you a nice even pluck...

- Bill
robin87
cygmu wrote:
I had forgotten about the ARC resistor -- well remembered/spotted. That is definitely something to experiment with. Try different values from 100k down to 15k or so to see what effect they have.


Yes, might even wire in a pot temporarily for testing.

cygmu wrote:
The VCFQ's resonance control does indeed have an exponential response […]


Please correct me if I’m wrong, but shouldn’t it be just the opposite? The gain cell / 2180 is attenuating the inverted feedback from the bandpass out and the resonance is increasing as the gain approaches zero.. So a exponential response in the feedback path would result a logarithmic response of the resonance pot, wouldn’t it?


cygmu wrote:
The 2180 is pretty much a drop in replacement for the original gain cell, though you have to tweak the control circuitry a tiny bit -- just a few resistor changes I believe.


I’ve been wondering about this. Couldn’t find any reference schematic for a 2180-based SVF and at my current skill level I’m not confident with taking stabs in the dark on handful of 6€ ICs.. wink That’s why I decided to start with the 2164 for now.

Maybe even mess around with blackmer gain cells after that. That specific serge circuit is proprietary but the topic seems to be well documented. With surface mount matched transistor pairs readily available, it could be fun to give it a try. In fact, I’ve stumbled upon pictures of a naked res-4 once and while it had 2180s on it (no surprise, I think it gets used a lot in his designs) there were only two of them per voice, along with a group of four surface mount devices that might have been matched transistor pairs. (At least that’s what they looked like.) So maybe even a mixed approach might be worthwhile as well?

Right now I'm working on a schematic based on the 2164 SVF design I've mentioned before, lifting the CV Input stages from the MI Instruments Ripples/Shelves. That should give me a workable schematic from which to start, allowing me to mess with things like "Automatic Resonance Control" later on, i guess.

I even think about going straight for a manufactured PCB, as I don't really enjoy breadboarding. Perfboard would be an option, but changing parts later on can end up pretty messy. I used to etch my own pcbs in the past, so whenever I wanted to try something, I made layout and went for a full prototype right away.
Probably going for surface mount this time, since i find it easier to swap smd components in and out using a hot air desoldering station. I'd make sure add test points and provide pads/jumpers for additional components, trimmers, etc. wherever i think they could come in handy, of course.
cygmu
robin87 wrote:

cygmu wrote:
The VCFQ's resonance control does indeed have an exponential response […]


Please correct me if I’m wrong, but shouldn’t it be just the opposite? The gain cell / 2180 is attenuating the inverted feedback from the bandpass out and the resonance is increasing as the gain approaches zero.. So a exponential response in the feedback path would result a logarithmic response of the resonance pot, wouldn’t it?


I can see why you might think so -- putting an exponential control in the feedback path of an op amp is a good way to generate logarithmic response -- but I don't think that's how it works out here.

The resonance in a state variable filter is obtained by feeding back the BP output into the input with a gain of -1/Q. The gain cell we're discussing controls the magnitude of this gain (the sign is dealt with by the summing arrangement). The circuit takes the CV, inverts it, and feeds it to the gain cell, so the gain is controlled in a exp(-CV) relationship. So the Q is controlled in a 1/exp(-CV) = exp(CV) relationship.

I hope that makes sense! I might have become confused along the way but I think this is right.

Edited to add: I always get confused about the structure of the SVF and go hunting for the diagram at the top of p21 of http://electronotes.netfirms.com/EN194.pdf which helps. I am linking it here so that I can find it again when I need it!
robin87
cygmu wrote:


I can see why you might think so -- putting an exponential control in the feedback path of an op amp is a good way to generate logarithmic response -- but I don't think that's how it works out here.

The resonance in a state variable filter is obtained by feeding back the BP output into the input with a gain of -1/Q. The gain cell we're discussing controls the magnitude of this gain (the sign is dealt with by the summing arrangement). The circuit takes the CV, inverts it, and feeds it to the gain cell, so the gain is controlled in a exp(-CV) relationship. So the Q is controlled in a 1/exp(-CV) = exp(CV) relationship.

I hope that makes sense! I might have become confused along the way but I think this is right.

Edited to add: I always get confused about the structure of the SVF and go hunting for the diagram at the top of p21 of http://electronotes.netfirms.com/EN194.pdf which helps. I am linking it here so that I can find it again when I need it!


Thanks very much for this right away, thinking about it more helped me find an error in my schematic! thumbs up

I'm still confused though. I thought decreasing the gain of the negative bandpass feedback actually increases the resonance, with a gain of zero resulting in maximum resonance? otherwise, why would you invert the cv at all?

Again, I don't want to suggest that you're wrong, I know close to nothing about these things and I'll have a lot of reading and experimenting to do, that's why I keep asking questions. The fact that I'm looking at two different schematics at once (2164-based SVF/VCFQ) just adds to the confusion, as i begin to wonder whether they are really as similar as i think. (this is the schematic I'm talking about, btw: https://www.muffwiggler.com/forum/viewtopic.php?p=904155#904155 )


Another question: Wouldn't a exponential response of the Q control be the opposite of what i need, assuming my main goal is to dial in the ringing of the filter precisely?

My VCFQ becomes very resonant with the Q knob past 10 o'clock, after that you basically control the decay of the ringing, which makes it great for finely tuned pinging. When turning the knob and listening to the resonance, it doesn't feel exponential to me. (In the way a exponential VCA reacts, for example.)

Thank you very much for your input, really appreciated!


- UPDATE! -

I just finished putting together a schematic of the filter I'd like to start with. "putting together" is the right term here, as it is a combination of different parts of circuits shared by other people. Aside from adding some optional components, (which might even turn out to be useless,) I contributed nothing original to this. Do you guys think it would be ok to post the schematic here anyway?

It's basically a filter core by Dr. Sketch-n-Etch, combined with CV input stages copied straight from the Mutable Instruments Ripples/Shelves.

I should probably rename this thread first, as people might assume that I'm designing an original filter here, which is not the case. For now, I just want to build something that I can mess with and tweak to my liking.
cygmu
robin87 wrote:

I'm still confused though. I thought decreasing the gain of the negative bandpass feedback actually increases the resonance, with a gain of zero resulting in maximum resonance? otherwise, why would you invert the cv at all?


That's right. The resonance is determined by Q, and the bandpass feedback gain is -1/Q, so as you reduce the magnitude of the gain you increase the Q.

From a quick look, the 2164 schematic seems to follow the same block structure as the VCFQ one, which is as shown in the Electronotes article I linked. I always get in a muddle with the signs of the various stages and summers but any working schematic will have them right I guess smile

One thing to bear in mind when comparing the schematics is that the 2164's control voltage input is an attenuation input, i.e. more CV means more attenuation, while the 2180 has both + and - CV inputs, and the Serge gain cell has a gain input, i.e. more CV = more gain. The Random Source VCFQ uses the + input on the 2180 I believe.

Quote:

Wouldn't a exponential response of the Q control be the opposite of what i need, assuming my main goal is to dial in the ringing of the filter precisely?


I'm not sure. For keeping the ring time constant across the frequency range you certainly want exponential response: each 1V extra frequency CV doubles the frequency, so you need to double the number of cycles the ping takes to decay, so you need to double the Q; which is to say, you halve the magnitude of the feedback gain.

I just had a play with my VCFQ, trying to see what change in Q CV would double the length of the decay of a ping, just by eyeballing it on the scope. The response did indeed seem to be exponential, with the decay doubling every 0.4V. I managed to smoke a prototype circuit that was lying about on the worktop while doing so, too, so my focus might not be the best -- take the measurements with a pinch of salt!

I agree that the VCFQ Q control does feel quite usable for dialling in decay, so perhaps the exponential response isn't a problem for that, or perhaps I have just got it all wrong!

(I sent you a PM by the way.)
cygmu
robin87 wrote:

I just finished putting together a schematic of the filter I'd like to start with. "putting together" is the right term here, as it is a combination of different parts of circuits shared by other people. Aside from adding some optional components, (which might even turn out to be useless,) I contributed nothing original to this. Do you guys think it would be ok to post the schematic here anyway?


I'm not the arbiter of fairness or anything but I think that if you have assembled it from other schematics that have been made publicly available and you are asking for help refining the circuit you are trying to build, then that is exactly in the spirit of a DIY forum and you should feel free to post it.
robin87
cygmu wrote:

I'm not the arbiter of fairness or anything but I think that if you have assembled it from other schematics that have been made publicly available and you are asking for help refining the circuit you are trying to build, then that is exactly in the spirit of a DIY forum and you should feel free to post it.


EDIT: Never mind the schematic below, there are some mistakes in there. I made some progress and will post an updated one soon.

I guess you're right. If anybody might object to this, I'll remove it on the spot. So here we go:



As mentioned before, this is basically a SVF core posted here on Muffwiggler by Dr. Sketch-n-Etch (David Dixon) and CV input stages by Oliver Gillet.

I added some comments in the schematic. Does anybody see some potential issues with this right away? I only included BP and LP outputs for now. Is buffering the Outputs recommended in this case?


cygmu wrote:

I'm not sure. For keeping the ring time constant across the frequency range you certainly want exponential response: each 1V extra frequency CV doubles the frequency, so you need to double the number of cycles the ping takes to decay, so you need to double the Q; which is to say, you halve the magnitude of the feedback gain.


That makes sense. However, I think a uniform decay throughout the audio range might not even be desirable. I think the shorter decay on higher frequencies will feel more natural, as that's how acoustic instruments usually work. Just not quite as weak as I've heard on some filters. I think a potentiometer to dial in the amount of Freq CV fed to the resonance input could turn out to be a nice feature.

edit: sorry for the bad quality, should have split the schematic into two pages.
cygmu
I don't think I understand the CV section at top left. You're inverting the CV there but then clamping everything with diodes -- is it correct?

[Edit: aha, thanks, you pointed out in a PM that it's a rectifier.]

I think you want an input resistor from the Resonance pot to its op amp.
robin87
cygmu wrote:
I don't think I understand the CV section at top left. You're inverting the CV there but then clamping everything with diodes -- is it correct?

I think you want an input resistor from the Resonance pot to its op amp.


You're right of course, I actually made a mistake while redrawing the Resonance CV stage from the MI Shelves. The missing resistor (200k) actually goes from the negative reference voltage to the wiper of the pot, the cw lug to ground and the ccw lug to the opamp input.

To be honest, I don't really know why he chose this configuration. Same goes for the diodes in the frequency CV stage. I guess i just assumed this to be standard pratice for the 2164, since it's the same in both shelves and ripples (and in another circuit I've seen, can't remember right now.)
I've seen this diode configuration before, referred to as a "improved precision rectifier"

Should probably stop for now and get back at it tomorrow with a fresh pair of eyes, trying to understand the cv stages a bit better first.
wsy
I just tested the idea of feeding a little of the frequency CV forward as resonance CV as well and it works. Not perfectly-perfectly, but
works well enough.

I used MATHS with channel 2 being the frequency CV and channel 3 being a DC offset (necessary because the
RESONANCE knob becomes an attenuator on the two filters I tried it with - an MMF and an APERTURE. A RENE made a nice
testing staircase

It's VERY twitchy on those filters though- it takes very little feedforward CV to kick it into resonance at the higher frequencies.

Then I tried it on my Synchrodyne, and WHOA - beauty ensued!!! Of course, the Synchrodyne takes a while to
shift (it's basically a VCO banging a PLL multiplier banging a switched capacitor filter) so all that resynch turned into wonderful
slews and glitches.

VERY cool. Never would have tried it if it wasn't for you folks.

- Bill

- Bill
Grumble
Just a little off topic: how is a synchrodyne filtering at low frequencies? I am asking this because I use max26x puscaf for filters and since the clock frequency is 200x the filter frequency it becomes audible at low frequencies.
robin87
Just sat down with the two pingable filters in my rack, to try out some ideas that came up yesterday:

I think a stock VCFQ with ARC is not the solution here. The issue here is that I get almost no decay at all in the high range. Everything past one o'clock on the frequency pot gives me nothing but tuned clicks that don't really ring, even with resonance cranked.

With Ripples, it's pretty much the opposite. You can get nice pings at high frequencies if you tune the resonance just right.( Difficult to do, since the range is pretty narrow.) Bnd the higher you go, the more it tends toward self oscillation. I can tune in a beautiful ping at a given frequency, but if I jump up an octave (without readjusting the resonance) I get screaming feedback. Escpecially annoying with sequenced pings.

Any idea what sort of response i could expect from a 2164 based state variable filter? there are a couple of those around, both commercial and DIY, so I guess somebody has tried pinging them?
wsy
Grumble wrote:
Just a little off topic: how is a synchrodyne filtering at low frequencies? I am asking this because I use max26x puscaf for filters and since the clock frequency is 200x the filter frequency it becomes audible at low frequencies.


To be honest, I never noticed it.... but I suspect that you might be able to tease
the switch frequency out if you tried hard enough.

I think I've also gotten it into comb filter territory which is cool too, but I didn't explore it deeply
enough (nor put it on a spectrum analyzer enough) to know for sure.

The Synchrodyne is definitely a wierd piece. Definitely not a one trick pony... but definitely not
your typical filter either. But it's pushed other, newer modules out of my rack more than once.

Note- because of the PLL middle, how the PLL responds to rapid 1V/octave changes is
NOT obvious - well, it's obviously not acting like a regular filter, but controlling takes some
trial and error even if you understand PLLs and how they can be used to multiply square
wave frequencies by integer fractions and get clean results. On the other side, it's not necessary to
know how that works in order to coax some really cool sounds out of the Synchrodyne as you change pitch.

And of course, the PLL fun only happens when you are changing pitch or somehow make the PLL unstable....

- Bill
Grumble
I use an Arduino to program an AD9833 which supplies the puscaf its clock signal.

cygmu
Grumble wrote:
I use an Arduino to program an AD9833 which supplies the puscaf its clock signal.


I have enjoyed poking around in your collection of youtube videos since you posted this one -- thanks! How many arduinos are there in your synth? smile
Grumble
I haven’t counted them, but one in every module exept the multies and mixers.
So about 20 Mr. Green
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Page 1 of 2
Powered by phpBB © phpBB Group