I'm really intrigued by the Undertone Audio console where the Q control affects the shape of the shelf EQ and introduces a bump at the top and bottom of the knee. A lot of plug-in EQ's have this ability. Does anyone know how this is done in an analog circuit? Something based on a state variable circuit would be useful.
looking for a resonant shelf EQ schematic
- Thread starter Fenris
- Start date
Help Support GroupDIY Audio Forum:
The first approach would be a LC filter, and then you mix a positive or negative version of this signal to the original. The inductor could be changed for a gyrator in some cases, the high pass is quite straight forward because the inductor goes to ground. Then you can go with any filter that has a low pass or high pass, as SVF and change the Q as you can do with SVF and work on that, once you have the filtered frequency then you add it to the original and that should be it. Little labs makes a resonant 4th order LPF and sum it with the signal, only boost LF, I don't know what topology they use... http://www.littlelabs.com/vog.html
JS
JS
Using a SVF (state variable filter) topology, similar to most parametric EQ, you can vary Q. To make a shelving EQ you sum both the BP and LP (or HP) section outputs. I did this inside a recording console (back in the late '70s) with parametric EQ in the input channels, I made the top and bottom EQ sections switchable between peak or shelf.
I don't know how useable the Q adjustment was for shelf EQ but it was easier for me to keep it than disable it.
JR
I don't know how useable the Q adjustment was for shelf EQ but it was easier for me to keep it than disable it.
JR
Passive EQs that are not properly terminated have a tendency to resonate at the cutoff point. The Pultec HLF-3C is designed for 600 ohm operation but when run into 10k+ loads will exhibit a peak at the corner frequency.
Fenris said:
Are you sure it's the shelf that has the bump and not an HPF or LPF? The bump usually is an under damped filter. A shelf is basically half a bell. I've never seen a bump in a shelving EQ response. Some people refer to the Q of a shelf as the "slope". I find this incredibly misleading as it conflates an shelving EQ with a filter. They are very different animals.
Gold said:
The thing is that you do a filter SVF in this case for example, which does have Q and slope, and then add or subtract that signal from original, when summing 18dB for example as the shelving point you just see the characteristics of the filter, the interaction happens when both signals are equal, as they are -18dB part of the filtered section on a 2nd order filter, you have cancelation, plus the resonance you already has at the peak, so what you get is a shelving but with a peak at the end off the boosting freq and a cut at the beginning of the slope, when boosting, when cutting everything is backwards. The curves would be something like this when varying the Q of the filter (not the eq per se)
JS
No, you just sum and under dumped LPF or HPF with the original signal, it already has the peak because it underdamped and as 2nd order is reversed in phase at the end, it generates the dip because cancelation when both signals are equal in amplitude (filtered and original).
I think this could be made in passive, and quoting mj
JS
PS: I don't know what happens with the group delay, which is something that really affects the signal, more than phase itself, they are correlated but not the same thing, of course, a filter with a constant group delay (just a delay) would be almost in phase for low freq and starts to rise out of phase as the frequency increases, but this usually like to hear, when different freq components of the same signal arrive to us in different moments our ears aren't so happy, it even could be all in phase but HF much less delayed than LF for example, all the freq one period late (ideal case) all would be in phase, so we see nothing in the phase plot but group delay is way out.
I think this could be made in passive, and quoting mj
mjrippe said:
JS
PS: I don't know what happens with the group delay, which is something that really affects the signal, more than phase itself, they are correlated but not the same thing, of course, a filter with a constant group delay (just a delay) would be almost in phase for low freq and starts to rise out of phase as the frequency increases, but this usually like to hear, when different freq components of the same signal arrive to us in different moments our ears aren't so happy, it even could be all in phase but HF much less delayed than LF for example, all the freq one period late (ideal case) all would be in phase, so we see nothing in the phase plot but group delay is way out.
Thanks Joaquins, it sounds like I can start with a resonant HPF and create the kind of resonant shelf I want. My board has 3 band sweep EQ (state variable) and sweep hi and lo pass (Sallen-Key). It would be cool to utilize the circuitry in the console instead of building something from scratch. I was able to add a resonance control to the Sallen-Key lo pass, but the Q increases with frequency and the overall level is boosted. I attached a schematic. Apparently this is known as a Steiner Resonant Filter and was used in the Synthacon in a discrete version.
Apparently, a state variable circuit will do a resonant bandpass, but not a resonant hi pass or lo pass.
Apparently, a state variable circuit will do a resonant bandpass, but not a resonant hi pass or lo pass.
Attachments
L´Andratté
Well-known member
A state variable will give resonance effect in all modes hp/bp/bs/lp, by choking off some of the feedback of the first integrator, it will also not change gain with q when done right.
look up this e.g.
http://sound.westhost.com/articles/state-variable.htm#s30
I have build a voltage controlled version of the synthacon filter with 6 pole lowpass, driven by 3x vtl5c3/2s and one silonex for q control, sounds so beautiful!
look up this e.g.
http://sound.westhost.com/articles/state-variable.htm#s30
I have build a voltage controlled version of the synthacon filter with 6 pole lowpass, driven by 3x vtl5c3/2s and one silonex for q control, sounds so beautiful!
I found the schematic I was looking for! This Calrec schematic is similar enough to the circuit in my board, and by comparing the 4 bands I can see exactly what they did to add a bell/shelf switch to a state variable circuit . Looking at the hi band, I see a 4-pole switch. The first section switches between the output of the 2cd opamp (BP) and the output of the 1st opamp (HP). The second section raises the input impedance of the 1st opamp, presumably to lower the volume. The third section disables the Q switch. The fourth section does something funny with the 3rd opamp (LP), it engages two polarized capacitors in parallel with the feedback loop and lowers the resistance in the ground path. There's an extra opamp that inverts the polarity of the HP output, so the cut/boost knob isn't reversed.
It looks like they didn't have to change the frequency, or combine the BP and HP.
What is the purpose of the 33 uf tantalum caps that were added to the hi band? Are they DC blocking caps, power supply decoupling caps, or both?
Did I interpret the schematic correctly? What's the purpose of the fourth section? Does it lower the LP frequency? It looks like they cheated on the lo shelf, the shelf switch simply lowers the frequency of the HPF so it's below the audio band.
The shelf doesn't need to be switchable, which simplifies things.
It looks like they didn't have to change the frequency, or combine the BP and HP.
What is the purpose of the 33 uf tantalum caps that were added to the hi band? Are they DC blocking caps, power supply decoupling caps, or both?
Did I interpret the schematic correctly? What's the purpose of the fourth section? Does it lower the LP frequency? It looks like they cheated on the lo shelf, the shelf switch simply lowers the frequency of the HPF so it's below the audio band.
The shelf doesn't need to be switchable, which simplifies things.
Attachments
The capacitor between output and inverting input inhibits the opamp. The 2nd integrator's output (LP) is nil. That makes the circuit a 1st-order SVF. Indeed, a 1st-order filter does not have a Q control.Fenris said:
Although it is located near the HI band, it really blocks DC for the whole string of potentiometers. Can be thought as a pendant to the caps at the outputs of IC6, 8, 10 & 12.
saint gillis
Well-known member
It's not exactly the point, but could be inspiring
http://www.yusynth.net/Modular/EN/BANK/index.html
http://www.yusynth.net/Modular/EN/BANK/index.html
Or downward. Or change the amplitude. Anyway it wouldn't do anything useful.Fenris said:
Using the two existing integrators and retaining the Q control allows creating 2nd-order shelves, however there are a lot of details to look after. The polarity of the summed LP or HP signal is one. Stability is another.
You have to consider the fact that, at low amounts of boost or cut, there cannot be a significant difference between a 1st-order shelf and a higher-order one.
2nd-order shelf make a difference when they're used at rather high amounts of B/C, particularly when used as an intermediate between a shelf and a hi-pass (which they actually are).
Any leads on an existing model with a 2cd order shelf? Designing something from scratch is beyond me. I was able to add useful resonance controls to the Sallen-Key filters in the separate filter section, if necessary I'll just turn the hi and lo EQ bands into Sallen-Key circuits.
I suggest you look for the K+H (Klein & Hummel) UE1000, as well as the older valve-based UE100. They are the only analog products I know that offer(ed) choice of 1st or 2nd order shelves. Indeed they are Sallen & Key based.Fenris said:
These schemos are not for the faint of heart. I find the UE100 easier to read.
Mr. Gyraf has a schemo on his site. What a beast! I once had an idea of building an LC EQ where L and C are on separate switches so you can change the bandwidth ... I don't speak German but it looks like they did the same thing in the UE-100. Wild!
Sallen-Key is a lot simper than state variable and I already got the resonance control working. That's probably what Undertone used. Thanks for the tips. Will post schemos & response curves when I get it working.
Sallen-Key is a lot simper than state variable and I already got the resonance control working. That's probably what Undertone used. Thanks for the tips. Will post schemos & response curves when I get it working.
Not a single inductor in sight in the EQ sections of the UE100, although there are several chokes in the PSU and one in the output amp.Fenris said:
Someone there did really understand the virtues of active RC filtering. Very advanced for the times.
Oh yes, indeed.
And with near-obsessive attention to detail: E.g. the frequency-determining capacitors in there are all composed of two parallel (styrene'ish?) capacitors in parallel - one with positive and one with negative temperature coefficient.. nearly no frequency drift vs. temperature
Jakob E.
