SSLtech

Constant Q/Constant Bandwidth
« on: June 28, 2005, 04:51:49 PM »
I've redrawn a well-known midrange (And I don't mean Demi Moore's bellybutton!) here:



The switch on the upper right marked "E/G" switches between Constant Q (G-series) and Constant Bandwidth (E-series) assuming that I haven't made any mistakes with the re-draw.

I just don't really see how it does what it is supposed to do... can anyone explain it to me???

Oh, and please don't ask me to draw and post the rest of the EQ schematic... It's a current product, and I really don't want to get everyone into trouble. This isn't for cloning purposes...

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.


PRR

Constant Q/Constant Bandwidth
« Reply #1 on: June 29, 2005, 01:38:53 AM »
The difference is subtle and I can't math it out.

But for a grok: move the cut/boost wiper to 1/4 or 3/4, the half-cut or half-boost positions. The impedance seen looking into the wiper rises above 5K ohms. At mid-position it is ~5K. At the ends it is zero.

With the switch open, that's what the R-C-R-C Wein bride sees as source.

With the switch closed, the Wein bridge sees 1K or less, at all positions.

Since the Wein bridge starts with 5K11, which is roughly its impedance, it acts different when fed from <1K or from around 5K.

At the ends it makes no difference. In the middle it makes no difference (no EQ). But the half-boost/cut areas will give different curves.

bcarso

Constant Q/Constant Bandwidth
« Reply #2 on: June 29, 2005, 01:56:53 PM »
I don't have the time to analyze it either, but it reminds me of the difference between the Kerwin-Hueslman and Tow-Thomas state variable filters.  The KH has one sum/diff amp and two ideal integrators, whereas the TT has a lossy integator, sum/diff, and an ideal integrator.  KH is constant Q and TT is constant bandwidth iirc.

There's a description of those and several more in the gigantic tome from Chen and CRC, the Circuits and Filters Handbook.  Many many typos and the crudest ~machine schematic drafting I have ever seen, but a lot of material in 2861 pages.

SSLtech

Constant Q/Constant Bandwidth
« Reply #3 on: June 29, 2005, 02:21:55 PM »
Quote from: "buttachunk"
the bandwidth and freq pots are dual ??

Yep. Frequency control is altering the wein bridge, controlling two legs similarly. (each leg has a 5.11k padding resistor, a capacitor (15nF or 4.7nF depending on HMF or LMF) and a pot gang, so they remain the same as each other at all settings.

The Q is split. One half governs feedback around the bridge, but it also affects overall gain at the same time, so the second gang is used to track the output level. The trimpot is used to match the gain of the bandwidth (close enough) for a constant level across it's bandwidth range.

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.

SSLtech

Constant Q/Constant Bandwidth
« Reply #4 on: June 29, 2005, 02:32:33 PM »
By the way, that oddly-drawn cut/boost pot is a center-tapped element, so that wire doesn't connect through to the wiper (unless the wiper happens to be at dead center of course!)

-Just in case that's how it's being read.

PRR, that helps a little, BCarso, I see what you're getting at... I'm working on a better understanding.

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.

Svart

Constant Q/Constant Bandwidth
« Reply #5 on: June 29, 2005, 02:37:55 PM »
"but a lot of material in 2861 pages."

I would hope so!  :shock:
Welcome to the GroupDIY leper colony! when something falls off, we just replace it with a tube!
occupation: General Electron Mayhem

Alesis X2 information repository:
http://www.theopiumdenproductions.

gyraf

Constant Q/Constant Bandwidth
« Reply #6 on: June 30, 2005, 04:02:49 AM »
Quote
that would make cloning pretty difficult..


..Omeg will make centre-tapped..
..note to self: don't let Harman run your company..

SSLtech

Constant Q/Constant Bandwidth
« Reply #7 on: June 30, 2005, 08:31:26 AM »
...you think that's a custom pot, wait until you hear about the top & Bottom...

Instead of trying to convert the filter from bell to shelving (as most manufacturers do) This particular manufacturer uses two conpletely seperate tracking filters for bell & shelf. That makes the HF and LF Frequency pots 4-gang reverse log... plus a linear gang for recall... then the cut & boost pots have to be dual-gang center-tapped, plus of course the obligatory linear gang for recall...

Cloning this would be outrageous in terms of the cost of the pots; it's okay for the manufacturer since they sell this as a console EQ and an outboard EQ, so they buy thousands upon thousands of every value of pot at a time.

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.

Svart

Constant Q/Constant Bandwidth
« Reply #8 on: June 30, 2005, 09:07:41 AM »
hmm still sounds like we should investigate cloning..
Welcome to the GroupDIY leper colony! when something falls off, we just replace it with a tube!
occupation: General Electron Mayhem

Alesis X2 information repository:
http://www.theopiumdenproductions.

SSLtech

Constant Q/Constant Bandwidth
« Reply #9 on: June 30, 2005, 11:36:54 AM »
Each channel would need:

2 x 22k linear single gang center tapped with detent at center

2 x 22k linear dual gang center tapped with detent at center

2 x 47k reverse log dual gang

2 x 47k reverse log quad gang

2 x 10k log dual gang

1 x (? value) reverse log quad gang with switch at the clockwise end

1 x (? value) linear three gang with switch at the clockwise end

See how much that lot would set you back, then bear in mind that when people work on these consoles, the first thing that they usually do is patch in an outboard EQ... The manufacturer isn't selling many of the outboard version.

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.


Carl_Huff

Constant Q/Constant Bandwidth
« Reply #10 on: June 30, 2005, 11:56:47 AM »
From a practicle stand point the Barry Porter 'NetEQ' design that he gave away a couple years ago had the same features as the ones described here.  Has anybody here looked at that design?  I built two channels at the time and found it to be exceptional.

The distribution was complete with Gerbers, parts list and reasonable instruction.  I'd build it again.
Best Regards,
Carl Huff

chrissugar

Constant Q/Constant Bandwidth
« Reply #11 on: June 30, 2005, 01:23:08 PM »
I didn't see that schematic for years but if I remember correctly it used some tricks (many opamps) to avoid the use of revlog pots. Please correct me Carl if I'm wrong, is this the schematic with only 10K linear pots for all functions?

chrissugar
Christian Mike Sugar
        CMS-LAB

SSLtech

Constant Q/Constant Bandwidth
« Reply #12 on: June 30, 2005, 06:42:10 PM »
Dual linears can be used if it's a State-variable design, that's how Neve did it in the 51/55/V EQ which they called the "Formant" Equaliser. SVFs can also be built with rev logs. I think that Jakob's Calrec uses linears inteh same way that the Neve does.

Wien Bridges are a little different: that's not to say that you can't use alinear and still get the same scaling, but I've never seen it done. -If it can be done I'd love to learn how...  :thumb:

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.

Carl_Huff

Constant Q/Constant Bandwidth
« Reply #13 on: July 01, 2005, 02:58:17 AM »
Chris,

I don't have a schematic in front of me and I cannot easily get to it.  I've recently moved cross country and most of my stuff is jammed into a storage container.  Argh ....

It's been a few years since I built the circuit but I remember it being a straight forward SVF design.  Barry's way of combining the hi/lo shelving filters was interesting too.  I don't remember it having any more opamps (5532s) than other parametric designs.  What I do remember is that the pots had to be custom ordered (center tapped) and cost me a small fortune!
Best Regards,
Carl Huff

Rogy

Constant Q/Constant Bandwidth
« Reply #14 on: July 04, 2005, 06:03:55 PM »

chrissugar

Constant Q/Constant Bandwidth
« Reply #15 on: July 04, 2005, 07:40:54 PM »
Yes, it looks like my memory is still ok.
If you read the aditional document he used only 10K linear pots for all the functions.
http://users.pandora.be/Rogy/Barry%20Porter%20NetEQ/readme.pdf

chrissugar
Christian Mike Sugar
        CMS-LAB

SSLtech

Constant Q/Constant Bandwidth
« Reply #16 on: July 05, 2005, 09:28:18 AM »
Yup, there it is.

State-variable biquad, using a similar single-gang 'Q' solution to the one which Graham Langley favoured.

Center-tapped cut/boost pots.

U2a sees about an 800Ω load, since it's looking at 6x5kΩ in parallel (R13 and R2 both go to vitrual grounds, and the clockwise halves of the four cut/boost controls all show 5kΩ each to their grounded center-taps. It might make U2 work a little hard at high levels, but it's fine for lower level stuff. Personally I'd have thought of generating one or two more inverted signal feeds to share the load of driving the total number of cut/boost pots just to make sure that the op-amps don't get asked to do hard labour when the signal gets big...

I have a couple of thoughts about this design; it's essentially partly what I had in mind when I suggested the "Ne-Mek" EQ a while ago. Not that I thought it was particularly original on my part, but that this design really IS very DIY parts-sourcing friendly. The "Neve" part of the Nemek was the grounded-end linear frequency pots part, as used in the Neve V-series. The "Amek" part of Nemek was from Graham Langley's approach to controlling a single shunt leg of a feedback loop *AND* have the same resistor be the shunt leg of the input signal attenuation. That way, as the gain of the filter rises with 'Q', the input signal attenuates in a precisely-matched manner (down to the accuracy of the series leg resistors, which is bound to be a LOT better than inter-gang pot tracking!)

This particular design uses a much quieter way of avoiding the noise from unused bands. The Neve and Amek designs both permanently sum the outputs of the various filters in with the signal. The problem with this is that any bands that are being neither boosted or cut still contribute regenerated noise to the signal. As an additional curse, the approach of grounding one end of a linear frequency pot tends to be noisier than using reverse log pots, particularly at the high-frequency end of the range. With both Neve and Amek designs both, you can hear a faint filtered-noise sweep as you move the frequency controls of the neither-cut-nor-boosted bands, particularly in the high mid area. -This may not be too bad on a single channel, but on a console when you have several dozen EQs switched in, many of them left in the exact same "normalled" position for unused bands, the noise can become an issue, believe it or not. I've certainly heard things get quieter with all the EQs switched in & out with no boost or cut applied before now. This design does things rather better; when a band is at the center position, its output is attenuated to ground and the filter contributes nothing to the signal.

-However, unless I'm missing something, with all pots in full boost, looks like the signal might also see 4 x 2k2s in parallel, which would under an extreme circumstance bring the total load impedance to U2 well below 400Ω, unless I'm misreading it? (mind you, if you really do have all 4 bands at full boost, you're probably long-since abandoned all worship for sonic fidelity!  :wink: )

Keith
"A waist is a terrible thing to mind"
Quote from: PRR
Ah, but that was 1999; we don't party like that any more.


 

Related Topics

  Subject / Started by Replies Last post
7 Replies
3373 Views
Last post December 10, 2008, 07:27:52 PM
by PRR
5 Replies
2049 Views
Last post March 04, 2009, 04:02:23 PM
by DAN_000
0 Replies
4798 Views
Last post August 02, 2010, 07:59:21 PM
by owel
64 Replies
9338 Views
Last post May 06, 2018, 08:40:55 PM
by PRR