Help understanding the parametric EQ from Soundcraft Ghost

[mr coffee]

Hi all,
I have been studying the schematic for the High-Middle-Frequency  parametric EQ circuit from the Soundcraft Ghost, and there are a few things I just don't get in terms of design choices. Hopefully someone with more circuit design chops can explain a few things. In any case, it's an interesting variant from the textbook State Variable Filter to check out.

My questions are

1) Why the decision to use different values of capacitors in the two integrators in the SVF section?  Most textbook circuits seem to go for the same value of capacitor in each...

2) The Q control is not a textbook thing I have seen. What are it's merits?

3) What are the advantages\disadvantages of this circuit?

Any other thoughts about this design and it's merits (or lack thereof) ?

Attached is a redrawn schematic of one section of the Ghost Parametric EQ. Sorry I don't know how to make it show up directly in the post.

Thanks in advance

mr  coffee

 

[JohnRoberts]

Hi all,
I have been studying the schematic for the High-Middle-Frequency  parametric EQ circuit from the Soundcraft Ghost, and there are a few things I just don't get in terms of design choices. Hopefully someone with more circuit design chops can explain a few things. In any case, it's an interesting variant from the textbook State Variable Filter to check out.

My questions are

1) Why the decision to use different values of capacitors in the two integrators in the SVF section?  Most textbook circuits seem to go for the same value of capacitor in each...

Using different value caps for the two integrator section poles will spread them further apart shifting the BP bandwidth wider. I looked at doing that decades ago but never found it very useful. I did think about using different integrator pole tuning with digitally controlled resistors/pots because you could then vary the  bandwidth with even less noise gain. 

2) The Q control is not a textbook thing I have seen. What are it's merits?

By spanning the Q control across both inputs of the SVF HP op amp section allows you to command a wider range of Q while at the same time delivering lower noise gain, vs the fixed resistor to ground in one leg and variable to ground  in the other.  This way the one resistor to ground gets larger (for lower noise gain) while the other one gets smaller. 

I've done this before and it works well.

3) What are the advantages\disadvantages of this circuit?

parametric EQ are useful.

Any other thoughts about this design and it's merits (or lack thereof) ?

I wouldn't vary the frequency that way...(pots wired as rheostat in series). Pots are typically 20% bulk resistance accuracy, so LF end may vary from channel to channel, pot to pot.

I prefer to  use the pots as potentiometer (voltage divider to ground) so you mainly care about the ratio accuracy not bulk resistance tolerance. Then allowing the pot wiper to go all the way to ground at LF end and adding a resistor in parallel to define the LF pole which will be far more accurate (5% with cheap resistors). This way the pot tolerance/accuracy is completely out of the circuit at both frequency extremes.

Attached is a redrawn schematic of one section of the Ghost Parametric EQ. Sorry I don't know how to make it show up directly in the post.

Thanks in advance

mr  coffee

With SVF parametric there are many ways to skin the cat, and I have seen far worse implementations.

JR

 

[mr coffee]

John,
Thanks very much for the helpful explanations. You are always a goldmine of audio electronics knowledge. Apologies if some of my questions are obvious to those with a more rigorous background in filter design.

Using different value caps for the two integrator section poles will spread them further apart shifting the BP bandwidth wider. I looked at doing that decades ago but never found it very useful. I did think about using different integrator pole tuning with digitally controlled resistors/pots because you could then vary the  bandwidth with even less noise gain. 

So in this circuit, the different capacitor values in the integrators mean that at lower Q settings the bandwidth  is broader than it would be in the equal capacitor topology, but it will get narrow and sharply peaky like a regular SVF at higher Q settings, correct?

Is the shape of the bandpass curve at low Q settings different from that of an equal value SVF at a low Q setting?  I have seen cascaded lowpass and highpass filter sections used to make broad passband filters with ( IIRC ) flatter passbands and steeper slopes, and your comment made me wonder  if the different capacitors in the integrators might somehow behave in an analogous fashion?

Thank you

mr coffee

 

[JohnRoberts]

John,
Thanks very much for the helpful explanations. You are always a goldmine of audio electronics knowledge. Apologies if some of my questions are obvious to those with a more rigorous background in filter design.

So in this circuit, the different capacitor values in the integrators mean that at lower Q settings the bandwidth  is broader than it would be in the equal capacitor topology, but it will get narrow and sharply peaky like a regular SVF at higher Q settings, correct?

I never built one this way, but I suspect the whole Q range is just shifted broader.  I wouldn't try to go very narrow starting so broad but have no actual hands on experience doing this.

Is the shape of the bandpass curve at low Q settings different from that of an equal value SVF at a low Q setting?

AFAIK the shape is defined by the standard poles, just shifted broader (as if Q was adjusted broader).

I have seen cascaded lowpass and highpass filter sections used to make broad passband filters with ( IIRC ) flatter passbands and steeper slopes, and your comment made me wonder  if the different capacitors in the integrators might somehow behave in an analogous fashion?

Thank you

mr coffee

No... same standard one pole response...  You are describing cascaded multi-pole filters where under-damped and over-damped stages  are combined mostly for steeper skirts.  I killed a lot of brain cells designing these for BBD delay lines back in the '70s.  I even wrote my own crude filter design software running on my H-11 computer to calculate filter response with actual component values. A huge time saver. The printout wasn't pretty, I used the tab function on my dot matrix printer to make crude response plots.

An interesting cascaded filter response was a "staggered pole" crossover  our resident golden ear at Peavey designed for a GP plug in Xover module. He used a one pole slope at the crossover point for lowest possible phase shift in transition region, then stacked on steeper rolloffs away from the crossover point. A little hard to describe but a good sounding crossover.

I have used more than 2 pole SVFs. Back in the '80s I made an adjustable PA crossover that was switchable between either 3 pole or 4 pole (Linkwitz-Riley) alignment. Nowadays it is just crazy to think of doing that analog. 

JR

 

[mr coffee]

Hi John and all,

I keep studying this design, and I got to wondering if this design choice might be driven not only by the designer's desire to broaden the Q, but by the desire to reduce the difference in output levels at the BP and LP outputs to reduce the risk of overload clipping in the SVF?

Isn't the LP output level higher than the BP in the textbook SVF? Seems like I remember reading that somewhere sometime.

Would using different capacitor values in the two integrators change that?

@John,
Thanks for the suggestion about the merits of using the potentiometers in voltage divider mode instead of variable resistor. Makes sense. I know I have seen that in other designs, but I never really knew why.

mr coffee 

 

[JohnRoberts]

Hi John and all,

I keep studying this design, and I got to wondering if this design choice might be driven not only by the designer's desire to broaden the Q, but by the desire to reduce the difference in output levels at the BP and LP outputs to reduce the risk of overload clipping in the SVF?

I didn't discuss this to prevent SVF TMI overload, but the voltage gain for the HP,BP, and LP output are completely defined by the negative feedback resistor values. The caps only affect pole frequency tuning .

The HP gain is the resistor from HP output to it's - input. BP gain is resistor from BP output to HP stage + input (since it is inverted to the BP), and finally resistor from LP to HP - input. Of course it's a little more complicated than that with resistors from HP stage to ground, and input resistor, etc.

I mentioned I have seen far worse and one design I really didn't like varied the Q by boosting the feedback R in the bandpass stage (higher BP gain causes narrower bandwidth). Then padding the BP output down with a second pot section to scrub off the excess gain at narrow bandwidth.  This BP section at narrow bandwidth could clip 10 dB or more before the direct path. Clipping the BP is pretty audible, but another subtle SVF flaw is clipping the HP or LP stage. Even though not in the direct path, clipping there can increase the overall path distortion.

Isn't the LP output level higher than the BP in the textbook SVF? Seems like I remember reading that somewhere sometime.

I am not aware of a single textbook SVF parametric design. I found the SVF design equations in a chip company app note back in the '70s, and as I said there are many ways to skin that cat (I've done it several different ways myself for different applications). Back when I made the 4 pole SVF IIRC it was not stable with all stages running unity gain, the 2 pole SVF is stable that way (I am going from memory since I haven't messed with these this century or longer.)

One pretty "different" one I did was a parametric EQ kit for hifi-consumer market (1979).  I put part of the variable Q pot in series with the BP feedback resistor so that I got higher BP gain when I commanded narrower bandwidth. This was tweaked to provide a first order correction for the apparent loudness change when boosting wideband signals and you adjust the Q . I provided something like 20dB of boost for the narrowest bandwidth but only 6 dB or so boost at widest bandwidth setting.  The intent was to make it more intuitive for EQing prerecorded material (and lower noise gain than using only shunt Q controls.) Other parametric EQs I designed for studio or professional applications did not allow for any interaction between the parametric parameters. 

Would using different capacitor values in the two integrators change that?

@John,
Thanks for the suggestion about the merits of using the potentiometers in voltage divider mode instead of variable resistor. Makes sense. I know I have seen that in other designs, but I never really knew why.

mr coffee

Yup pots are much better at making ratios than for actual resistance tolerance.

JR

 

[L´Andratté]

Back when I made the 4 pole SVF IIRC it was not stable with all stages running unity gain,...

Hey John, out of curiosity, did you build a real 4-pole SVF (mixer+4 integrators with different feedback coefficients) or was it the two cascaded stages with global fb approach?

 

[JohnRoberts]

Hey John, out of curiosity, did you build a real 4-pole SVF (mixer+4 integrators with different feedback coefficients) or was it the two cascaded stages with global fb approach?

That was over 4 decades ago, but I think I used overall global feedback involving all the integrator stages in series, and even had a switch to select between 3 stage butterworth, and 4 stage L/R alignment (that switch tweaked a few components to get the Q/damping  right for both alignments as well as a polarity inversion between 3 stages and 4 stages). 

Don't expect too many details. I do recall that the 4 ganged (alps) frequency pot was a PIA to get good tracking over the full range (even using it the best way and paying up for a premium alps tracking spec). I had my technician measure and tweak the pots at 12 0'clock with a shunt resistor from wiper to top or bottom so they were dead on at full up, full down, and 50% rotation.

Like I said this was little crazy to do analog, but back then that was the only option practical. 

JR

 

[mr coffee]

Hi John,

Don't worry about TMI - I am trying to fully comprehend the trade-offs involved. Thanks so much for sharing. The explanation about the feedback resistors controlling the gain was actually quite helpful.

I mentioned I have seen far worse ...

Is the UREI 545 (attached) more what you would recommend (minus the old-timey op amps)?

Or can you point me to a parametric design you think represents better design choices?

Thanks very much.
mr coffee

 

[JohnRoberts]

Hi John,

Don't worry about TMI - I am trying to fully comprehend the trade-offs involved. Thanks so much for sharing. The explanation about the feedback resistors controlling the gain was actually quite helpful.

Is the UREI 545 (attached) more what you would recommend (minus the old-timey op amps)?

Better but still not as good as can be done.

Instead imagine shorting across the end limit resistors (R26 and R27), so at the LF end the frequency pot wiper is grounded so  completely out of the circuit. Then we need to add two resistors, one from the top of each pot section to the integrator op amp - input to provide a LF path. For a 10:1 frequency range use parallel resistor 51k  (10x the 5.1k).

The way they are using the pots the voltage divider formed by the 10k pot and 1k end limit resistors can vary with the pot's 20% bulk resistance tolerance, making the LF end not very accurate. 

Or can you point me to a parametric design you think represents better design choices?

Thanks very much.
mr coffee

I like all my designs....  8)  Only the old kit with the wierd Q-boost/cut interaction schematic is published (Phoenix Systems P-94). 

That Urie looks respectable enough. Use quality film caps for the integrators and newer op amps. If DC coupled the improved frequency pot approach can amplify DC offset errors, so modern op amps with better DC performance (and or caps in series with the 5.1k frequency resistors will help. The DC coupled 51k resistors will stabilize the DC feedback loop ).

JR

 

[mr coffee]

Hi John,
I googled the Phoenix P-94 and all I found in the first 3 pages was stuff about Phoenix brake bleeders, an old Analog Bucket Brigade Delay with a Sallen and Key 4-pole LPF in it that you designed for Popular Electronics way back when (so cool - do you know some guitar players are still building Panasonic BBD delays and don't like the modern DSP versions?),, and other miscellaneous stuff, but no Phoenix P-94 SVF design.

Don't want to bug you for it, but if it isn't too much trouble, could you give me a pointer to your P-94 SVF design? Maybe I could pick up some (more) useful ideas. I want to build one to go in a guitar to

Thanks again for sharing  your  wisdom about SVF circuitry! The TMI was just right ;D

mr coffee

 

[abbey road d enfer]

I keep studying this design, and I got to wondering if this design choice might be driven not only by the designer's desire to broaden the Q, but by the desire to reduce the difference in output levels at the BP and LP outputs to reduce the risk of overload clipping in the SVF?

Simulation shows that, for identical Boost/cut and BW (note my reluctance to use "Q"), the levels at the output of the summer and the 2nd integrator are significantly lower, which was to be predicted, but the output level at the 1st integrator is exactly the same, as could also be predicted.
Anyway, in boost mode the output level of the boost/cut stage is always higher than any of the others, so would be the first to clip in all circumstances. Caution must be exerted when notching out, though.
Maybe a thorough analysis of the noise performance may show some difference. You must remember that one of the factors that makes SVF topology criticized is that, even with boost/cut in neutral position, sweeping the frequency pot makes a swishing noise; that is considered by some as an undesirable addition to signal.

Isn't the LP output level higher than the BP in the textbook SVF?

In the usual implementation with two strictly identical integrators, the peak levels are identical in the summer and the two integrators; they are not centered at the same frequency, though.

Would using different capacitor values in the two integrators change that?

In the Ghost configuration, the levels of the summer and the 2nd integrators are about 6dB lower than at the 1st integrator.

Thanks for the suggestion about the merits of using the potentiometers in voltage divider mode instead of variable resistor.

It must be noted that there is a small penalty in terms of noise performance, notably when the frequency is set at its minimum. With a rheostat arrangement, the noise spectrum BW of the integrators is equal to the tuning frequency; with a potentiometer arrangement, the noise BW is always at the maximum.
For the very same reasons JR explained, I always used the potentiometer arrangement in my commercial designs, but I also spent countless hours working on the noise optimization of the integrators.

 

[JohnRoberts]

Hi John,
I googled the Phoenix P-94 and all I found in the first 3 pages was stuff about Phoenix brake bleeders, an old Analog Bucket Brigade Delay with a Sallen and Key 4-pole LPF in it that you designed for Popular Electronics way back when (so cool - do you know some guitar players are still building Panasonic BBD delays and don't like the modern DSP versions?),, and other miscellaneous stuff, but no Phoenix P-94 SVF design.

Yup, my flanger kit from the 1976 article was built by one of the guitar players in the original "Heart" group, and used on their first album.

Don't want to bug you for it, but if it isn't too much trouble, could you give me a pointer to your P-94 SVF design? Maybe I could pick up some (more) useful ideas. I want to build one to go in a guitar to

I'll try to scratch up a scan this weekend, but I repeat this design was targeted for Hifi final mix use, so there is an intentional interaction between Q and boost/cut (wider bandwidth get less boost/cut, narrower more).

Thanks again for sharing  your  wisdom about SVF circuitry! The TMI was just right ;D

mr coffee

Somewhere I have a set of the design equations (I think), a lot of my customers were college students who needed to modify my kit design for their EE course credit, so I helped them out with equations.  If i can find them I scan them in too.

JR

 

[PRR]

> I'll try to scratch up a scan this weekend,

If it is Popular Electronics, or several other rags, just find the issue date. They can be found here:

http://www.americanradiohistory.com/

 

[JohnRoberts]

> I'll try to scratch up a scan this weekend,

If it is Popular Electronics, or several other rags, just find the issue date. They can be found here:

http://www.americanradiohistory.com/

http://www.americanradiohistory.com/Archive-Poptronics/70s/1979/Poptronics-1979-09.pdf

page 47

JR

 

[mr coffee]

THANKS!!!!!!!!!!!!

mr coffee