Constant Amplitude Phase Shift EQ design

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bigugly

Well-known member
Joined
Jun 27, 2004
Messages
220
Location
Burbank, CA.
Here is a schem I drew based on the Steve Dove article.

This is the PCB layout.


The C2/R7/RV1A and C4/R8/RV1B values can be changed to get different frequencies and sweep ranges.

Does anyone see any problems? I'm going to build and test this. If everything is good I'll offer boards if anyone is interested.

James

btw, the pot footprints are for Omeg parts
 
Do you really want RV1 to be a triple? I think not. The left-hand pot tied across the inputs of U1D would control boost/cut by itself, and the two pot sections connected to R7 and R8 will tune frequency.
 
Sorry, I didn't mention that RV1 is a dual concentric with the first two sections (the outer shaft) controlling freq and the last section (the inner shaft) controlling gain. My roommate recently received an order from Omeg of a triple deck dual concentric pot for panning and level on a project of his and we were both impressed with the quality so I figured I'd get some for this EQ.
 
[quote author="bigugly"]Sorry, I didn't mention that RV1 is a dual concentric with the first two sections (the outer shaft) controlling freq and the last section (the inner shaft) controlling gain. My roommate recently received an order from Omeg of a triple deck dual concentric pot for panning and level on a project of his and we were both impressed with the quality so I figured I'd get some for this EQ.[/quote]

Ah cool---that occurred to me after I made the comment.
 
By the way, in the attempt to realize some of the supposed benefits of using CAPS (constant amplitude phase shift) networks deployed as filter elements, as touted by Dove in his massive HB for Sound Engineers 1st Ed. article on consoles, but not discussed to the point of showing a CAPS equivalent to a state-variable filter that actually works, I played around for a while. I finally got an allpass boost-cut arrangement that differs from the shunt synthetic L-C approach above. Whether it has any advantages or not compared to the state-variable general biquad remains to be seen.

Forssell argues convincingly for the single sum-difference stage with multiple shunt L-C-R, rather than a long chain of stages. You do have the advantage that the noise in the synthetic LCs nulls out at the midpoint of the boost-cut potentiometers. You do however have the contribution of thermal noise from each pot.

Actually, feedback integrators don't scare me that much---I think some of Steve's objections are a bit exaggerated, particularly the peak current requirement one. In that case the current needed is only the magnitude of the input current to the integrator plus whatever the external load current is---it's not like the amplifier is driving a big C to ground and trying to produce a step function. And the benefit of feedback integrators is small common-mode swing, which you lose with the CAPS configuration.
 
one benefit of that topology is that all the filter sections need not be CAPS networks. the high band can be just a cap from the wiper of the boost/cut pot to ground! you could make a gyrator based LF shelf, or even use an inductor. so my suggestion is to make a few different filter boards including the one you already layed out.

mike p
 
tested the circuit last night and it works fine. I didn't give it any serious time on the Neutrik A2 I just wanted to prove the circuit first. I need to get the proper pots now. (That was the stumbling block last night, all my pots were linear and not the correct values)

I'm working on a gyrator based low shelf section with switchable bell and variable frequency. I've been trying to find something like this for the high band but have been unable to find a satisfactory solution.

I am going to start working this design into a finished PCB. It will have 5 bands, Hi - Shelf/Bell, Hi Mid - Bell, Mid - Bell, Low Mid - Bell, Low - Shelf/Bell. All bands will have variable freq. and Q(on the mid bands).

I'll keep this post updated with my progress.

James
 
Hey guys - I am looking for the design equations for the Dove C.A.P.S eq circuit.

I have the Forssell pdf and the chapter in the Handbook for Sound Engineers, which I think is the same as the original article (?).

Seeing as though bigugly and peter purpose have designed and used the C.A.P.S circuit - I was wondering if anyone had the equations for bandwidth and frequency range etc?

Cheers Tom
 
I don't have time to work it out at the moment, but I got as far as to revisit some of the errors in the HSE** article. I should quickly add that I love the overall piece and have great respect for Steve. Moreover he may have been the victim of some hasty typesetting and editorial screwups (the overall book has plenty, like that bogus formula for attentuators that came up in here a while ago).

On page 756, Figure 22-61, "Inductive Reactance Synthesis", we are actually shown a step-by-step approach to the series L-C resonator needed for typical EQ boost-cut topologies. The various lettered subsections of the figure are discussed in the text. There are some problems there as well (for example paragraph 6 of 22.10.11, where in the penultimate sentence the word "not" should be removed).

After the discussion about the CAPS network, parts of Figure 22-61 are pretty much unsupported by the text. And that's where some of the more serious errors come into play. 22-61F shows inductor synthesis with the follower-buffered CAPS network: "Bootstrap/feedback inductive synthesis using CAPS network". But the next circuit 22-61G is incorrect as representing an inductor with no series resistance. In fact the inductance has an equivalent series R of about 1/2 the bootstrap R. And no formula is given for what the equivalent L value is.*

Then after the two bits about the CAPS behavior at low and high frequencies, we jump to a full-on two-CAPS resonator in 22-61J, followed by an equivalent circuit in 22-61K that should be captioned "Equivalent circuit of J". But once again 22-61K lacks a needed series R.

I haven't worked out the equations for the Q (which has to be defined carefully), but I believe the simple answers for the rest of the resonator behavior are that f nought is 1/(2 Pi RC) where R and C are at the noninverting CAPS network opamps' inputs, and the equivalent series R of the L-C-R series network so synthesized is 1/2 of the bootstrap/feedback R. So when driven from a given source R, you can determine the depth of dip at resonance based on a simple voltage divider analysis.

BTW I assume that the inverting inputs of the CAPS circuits have equal value Rs as input and feedback.


*It appears to be L = (1/2)*Rboot * RC, in henries, where R and C are the CAPS phase-shift parts and again the inverting input Rs are equal. Note that for the two-CAPS resonator, the resonant frequency is not affected by the bootstrap/feedback R value.


EDIT: ** the first edition
 
Without doing the real maths but making some educated guesses and doing simulations, I find that the Dove two-CAPS resonator winds up producing an equivalent series R-L-C with R as advertized above (at 1/2 the Rbootstrap-feedback value), but Lequiv of half the value of the single-CAPS synthetic L circuit, namely (1/4)*Rboot*RC, and Cequiv of 4*RC/Rboot.

EDIT: Also now realizing that the restriction in the structure as shown in Figure 22-61J to unity gain is very limiting on Q. I see that later on when he actually uses the circuit in an example, he puts variable gain around the whole circuit.
 
The more I play with this the less I like it. I'm sure it works, but I am not seeing significant advantages over GIC designs.

The inductor synthesized in the single CAPS circuit is of decent quality, but the two-CAPS resonator needs a lot of help, at which point one wonders why one is still flogging the horse IMO.

I thought perhaps the resonator might turn out to have lowered internal voltage swings for moderate-high Qs compared to state-variable designs, but if there is an advantage it's slight. And SV has low common-mode swings (not so with GICs). With SV you do have to look carefully at opamp unity-gain stability and accumulation of parasitic phase shifts, but these are not as bad as the article tends to suggest.

It's a shame in a way that the OTAs (Operational Transconductance Amplifiers) never worked very well. Perhaps the time approaches that some enterprising outfit like THAT should consider a new product or two. Or maybe some existing products are already appropriate.
 
One thing that occurs about this stuff: I wonder if anyone has done a hybrid design with digitally-controlled high-res and low-glitch step attentuators or other variable-resistance techniques (like PWM), while preserving the basic analog signal path?

Forsell argues for the CAPS topology for the relative ease of adjustment, compared to his simple voltage follower synthetic inductor and series cap resonator. Although that topology is not well-suited to variation with just resistor changes, some could be.
 
[quote author="bcarso"]

It's a shame in a way that the OTAs (Operational Transconductance Amplifiers) never worked very well. Perhaps the time approaches that some enterprising outfit like THAT should consider a new product or two. Or maybe some existing products are already appropriate.[/quote]

The tradeoff with OTAs was always noise vs distortion. The CA3280, one of the better parts was something like 8 nV/rt Hz so a modern low noise input revisit seems like it could deliver a noticable improvement but if it takes more silicon area than a VCA just give me some cheap quad VCAs for lower price.

The THAT folks selling OTAs might be eating VCA business.

JR
 
If the current mirrors had provision for external ballasting Rs the noise of parts like the 5517/13700 could be a lot lower. And those lateral PNPs are always a drag.

In other, non-audio, areas of filter activity there's been a lot of development of gm blocks, but primarily to realize faster structures, not so much for higher precision.

EDIT: OTAs (and not talking necessarily current-programmable variable gm ones) make it a lot easier to realize floating circuit elements too.
 
Hey Brad and John.

Sorry for not coming back to the thread after you posted such insightful replies. I have been busy with work.

I really appreciate your help and have just had a read through the thread and things are starting to make sense.

I will draw up a schematic based on my application asap.
I grabbed a copy of the Dove articles - which are heaps better than the SED book.

So will mull them over further and come back. It's the overall bootstrap resistor value and Q math that have me puzzled right now. Also the 220k resistor at the input of the first C.A.P.S opamp.

If you were designing a four band swinging inputs EQ and wanted to have one band (LMF) fully sweepable between 100Hz and 3kHz, with a switched Q of either 0.8 or 3 where all other bands are broad inductor driven resonators - what would you choose?

For the C.A.P.S - and the circuit at the top of the thread I get RC combinations of the following for a freq sweep of 102Hz to 3.2kHz:

  • - Bootstrap R: 4.99K
    - Feedback R: 10k
    - C = 150nF
    - Freq. Pot: 10k
    - Series limiting R: 330 ohms
Cheers Tom
 
[EDIT]
Think I ballsed up the Q resistors, hadn't noticed they were log-revlog pots...will post updated version asap....if anyone is interested LOL
/EDIT]


Had a play in the simulator but its behaving funny.

These are the circuit values I ended up with:
[EDIT]
Circuit removed - Q values were wrong
[/EDIT]


Getting the right frequency sweep, did without the slugging resistor (33k) on the Q pot as I won't use a pot, but a switch instead for wide/narrow. Using the simulator I ended up with 8k45 for wide Q of 0.8 and 453R in parallel (would be switched in) for a value of 429.9R and a narrow-ish Q of 3.

Would love to know the math for that. Something feedback related obviously. The changes in positive feedback caused by the Q resistors have quite a large affect on the overall cut/boost range (resonance I guess).

The Dove article claims a variation of +/-1dB over the range of the slugged Q control... so with switched resistors there should be little variation right?

Here is a graph I generated from the LMF values in the Dove article (25 to 500 Hz):
2r5z5n6.gif

You'll notice a massive loss in overall cut/boost when the bandwidth is set wide. Is this correct for this topology? I thought the CAPS provided variable bandwidth independant of gain.

Buzz Audio use the CAPS topology in their MPE1.1 EQ... these are the graphs from it at two bandwidths, centred @ 1k.
mid1.jpg

mid2.jpg


Only a 1dB change in the gain... what's up with my simulator? LOL Maybe the Buzz isn't swinging inputs? How else can the CAPS be used?

These are the sweeps from my LMF values (100 to 3 kHz):
2zdzj87.gif


How much of an issue is the rise in H.F towards 1Meg, I ran some transient analysis on 10kHz square waves, wondered about stability, played with the 100pF cap around the swinging input amp, didn't result in much.

Also this thing gets screwy with more bands in.
I tried a 4-band with one CAPS section and three inductive ones.
The slopes looked all skewed, bandwidths all out of line - which I guess is due to interaction of the nearest bands.

I set centres at 50Hz, 500Hz, 3k2 and 10k... Dove mentions using centre-tapped pots for minimising this but there is a noise penalty.

For a four band EQ do you think it would be best to run 2 opamps and have a 2x2band cascaded stage to minimise overlapping bands?

Put LF and HMF on one and LMF and HF on the other?

Cheers Tom
 
[quote author="TomWaterman"]Also this thing gets screwy with more bands in.
I tried a 4-band with one CAPS section and three inductive ones.
The slopes looked all skewed, bandwidths all out of line - which I guess is due to interaction of the nearest bands.

I set centres at 50Hz, 500Hz, 3k2 and 10k... Dove mentions using centre-tapped pots for minimising this but there is a noise penalty.

For a four band EQ do you think it would be best to run 2 opamps and have a 2x2band cascaded stage to minimise overlapping bands?

Put LF and HMF on one and LMF and HF on the other?

Cheers Tom[/quote]

from what I remember of the Dove article he separated the bands in his example circuit. High shelf, mid, and lo shelf. then lo mid and high mid bands. each group running on their own op amp. check out the circuit diagram near the end of his discussion on eqs.
 
Yeah thanks bigugly - that's what I had read.

I think it may be the only way to get really solid curves...

Still purple audio aren't doing it and they have a 4-band EQ out called the Odd.

http://purpleaudio.com/Product/Odd.html
ODDv4Side.jpg


Looks like 4-bands with single input and output amp...TX I/O to me.

Cheers Tom
 

Latest posts

Back
Top