Cytomic "The Drop" Resonant Filter

[@midnight]

but Synth Squad doesn't report any latency, regardless of oversampling settings, iirc...

[andy-cytomic]

but Synth Squad doesn't report any latency, regardless of oversampling settings, iirc...

That is right, Synth Squad does not report it's latency. The latency is a different fractional sample amount for every frequency, which is impossible to report to hosts that take a single integer sample delay.

[ryandfl]

but Synth Squad doesn't report any latency, regardless of oversampling settings, iirc...

That is right, Synth Squad does not report it's latency. The latency is a different fractional sample amount for every frequency, which is impossible to report to hosts that take a single integer sample delay.

[valhallasound]

but Synth Squad doesn't report any latency, regardless of oversampling settings, iirc...

That is right, Synth Squad does not report it's latency. The latency is a different fractional sample amount for every frequency, which is impossible to report to hosts that take a single integer sample delay.

The same "fractional sample amount" of latency is true for every IIR filter, which includes EQs and VCF emulations. I presume that Andy is not reporting any latency, as the closest integer to the average latency at lower frequencies is zero. This is the case for Hilbert networks, which half-band filters are closely related to.

Sean Costello

[ftech]

but Synth Squad doesn't report any latency, regardless of oversampling settings, iirc...

That is right, Synth Squad does not report it's latency. The latency is a different fractional sample amount for every frequency, which is impossible to report to hosts that take a single integer sample delay.

My thoughts exactly.............this is making my head hurt

[andy-cytomic]

The same "fractional sample amount" of latency is true for every IIR filter, which includes EQs and VCF emulations. I presume that Andy is not reporting any latency, as the closest integer to the average latency at lower frequencies is zero. This is the case for Hilbert networks, which half-band filters are closely related to.

Sean Costello

Hi Sean,

This isn't correct. The group delay does not approach zero samples at low frequencies. For the twin path polyphase halfband iir filters in Synth Squad it is around 4.2 samples at DC. Here are the full gory details http://cytomic.com/files/dsp/synthsquad ... oup-x2.png

Now, if you don't mind I would like to get back on topic, please ask questions about The Drop, not Alchemy or Synth Squad!!!!

[Bronto Scorpio]

Now, if you don't mind I would like to get back on topic, please ask questions about The Drop, not Alchemy or Synth Squad!!!!

OK
A simple (and probely annoying) one to get you started
When will the drop be released? Still in december or maybe a bit earlier?
I can't wait to try it!

Cheers
Dennis

[valhallasound]

The same "fractional sample amount" of latency is true for every IIR filter, which includes EQs and VCF emulations. I presume that Andy is not reporting any latency, as the closest integer to the average latency at lower frequencies is zero. This is the case for Hilbert networks, which half-band filters are closely related to.

This isn't correct. The group delay does not approach zero samples at low frequencies. For the twin path polyphase halfband iir filters in Synth Squad it is around 4.2 samples at DC. Here are the full gory details http://cytomic.com/files/dsp/synthsquad ... oup-x2.png

Interesting. To me, at least. So it seems like the best solution would be to run things at 240kHz, so you could accurately report a latency of 21 samples. I haven't double checked my math on this, but I am sure that it is correct.

Now, if you don't mind I would like to get back on topic, please ask questions about The Drop, not Alchemy or Synth Squad!!!!

So, what filter models do you plan on having in there? SVF seems like a good choice - are you going to incorporate the diode clipping as found in the Oberheim SEM feedback path (and what the heck are those diodes doing in there)? MS20, Moog, diode ladder, Steiner, Polivox, what are we talking about here?

Sean Costello

[andy-cytomic]

Interesting. To me, at least. So it seems like the best solution would be to run things at 240kHz, so you could accurately report a latency of 21 samples. I haven't double checked my math on this, but I am sure that it is correct.

So, what filter models do you plan on having in there? SVF seems like a good choice - are you going to incorporate the diode clipping as found in the Oberheim SEM feedback path (and what the heck are those diodes doing in there)? MS20, Moog, diode ladder, Steiner, Polivox, what are we talking about here?

Sean Costello

I apologize for saying "That isn't correct" without further clarification, I was only referring to the DC group delay of the resampler filter being not zero, your other statements about iir filters in general having non-linear phase response is true (most of the time), and the group delay does go to almost zero at DC (most of the time).

You can actually change the transition bandwidth of the particular hiir filter in synth squad to get a very close to integer number of samples at DC, without needing to change the sample rate. Perhaps for a mastering processor this would be needed, but it doesn't really matter too much for a synth.

Ok, now back on topic for me too:

I'm releasing The Drop with SVF (yes with the diodes) and Sallen-Key filters only, and I will add a Cascade OTA, and Cascade ladder later on. Please email me to discuss what the diodes are doing in the SEM-1A filter, but I feel this isn't the place to go into details and I want to stay on topic

[andy-cytomic]

OK
A simple (and probely annoying) one to get you started
When will the drop be released? Still in december or maybe a bit earlier?
I can't wait to try it!

Cheers
Dennis

December looks like a realistic release date still. I was hoping to get it out before then, but unfortunately I have spent the past two months working on something to be announced at the upcoming AES show. I can't go into details but it is a big deal and a very positive thing for Cytomic.

[mystran]

So, what filter models do you plan on having in there? SVF seems like a good choice - are you going to incorporate the diode clipping as found in the Oberheim SEM feedback path (and what the heck are those diodes doing in there)?

I can answer this; it's pretty simple.

If we look at the first opamp, we have a basic inverting mixer. The operational principle of this is that positive input being connected to ground, the opamp will do it's best to maintain the negative input at ground potential as well.

Now, we have three voltages (Vin, Vlp, Vbp) connected to the negative input through three resistors, and a negative feedback resistor. If (for simplicity) we assume all the resistors equal, we have currents from the inputs as (Vin + Vlp + Vbp) / R. In order to keep the negative terminal at ground, we get Vout/R=-(Vin+Vlp+Vbp)/R or Vout = -(Vin+Vlp+Vbp). So we have an elementary inverting unity gain mixer (can't remember if the SEM-1A is unity gain but it's not really important here).

Now, when we put the diodes in parallel with the resistor connecting the Vbp voltage, there will be exponentially growing current through the diodes as Vbp increases (this is how diode clippers works as well). One can think of the effective resistance in the Vbp path as going down. What happens, is that the summing opamp needs to compensate this increased current by putting out equal (but opposite) increased current through the feedback path, so the gain of the Vbp component increases: we get more band-pass feedback.

Now, in a SVF design as band-pass feedback increases, the resonance goes down. So what the diodes do is allow more band-pass feedback when the bandpass amplitude increases. This decreases the resonance, limiting it to some level dependent on the diode voltage drop and the additional resistors around the diodes (which we ignored above; the circuit works without them as well, but limiting the current through the diodes can make things a bit smoother).

Quite simpe, isn't it?

[Bronto Scorpio]

OK
A simple (and probely annoying) one to get you started
When will the drop be released? Still in december or maybe a bit earlier?
I can't wait to try it!

Cheers
Dennis

December looks like a realistic release date still. I was hoping to get it out before then, but unfortunately I have spent the past two months working on something to be announced at the upcoming AES show. I can't go into details but it is a big deal and a very positive thing for Cytomic.

Nice
What's good for Cytomic can only be good for us

Cheers
Dennis

[ryandfl]

December looks like a realistic release date still. I was hoping to get it out before then, but unfortunately I have spent the past two months working on something to be announced at the upcoming AES show. I can't go into details but it is a big deal and a very positive thing for Cytomic.

Congrats...on whatever it is! It looks like there is an AES even later this month. Is that the show where you will announce it or are you referring to a show further down the road?

December is going to be a big month between The Drop and Dive.

I know you have a stance against discounts, but will there be a package price for The Drop + The Glue + possibly the to be announced thing if it is a plug?

[andy-cytomic]

So, what filter models do you plan on having in there? SVF seems like a good choice - are you going to incorporate the diode clipping as found in the Oberheim SEM feedback path (and what the heck are those diodes doing in there)?

...but I feel this isn't the place to go into details and I want to stay on topic...

I can answer this; it's pretty simple...

And seemed involve loads of off topic waffle when all you needed to say was "the diodes limit resonance". I'm all for general discussion, but can we please try to stay a little more on topic? General details of circuits are all that is required so everyone can remain included in discussions.

[andy-cytomic]

Congrats...on whatever it is! It looks like there is an AES even later this month. Is that the show where you will announce it or are you referring to a show further down the road? ...

I know you have a stance against discounts, but will there be a package price for The Drop + The Glue + possibly the to be announced thing if it is a plug?

Yep, the announcement will be made on 20th October.

There will be a discount for all existing customers as well as a discount for new customers that buy both at once. I haven't decided the final amounts, but I will announce them before release since it will hopefully aid in peoples purchasing plans.