Would like some opinions on that parametric EQ schematic

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

ricothetroll

Well-known member
Joined
Jul 5, 2005
Messages
325
Location
Bruxelles
Hi !
I found that schematic in a book about studio equipment. I would like some of your opinions concerning the design and the comparative quality of that circuit compared to other similar projects that can be found here (Calrec, etc.).

EQ.jpg


E (entrée) = input
S (sortie) = output
M (masse) = gnd
P1 : Gain
P2 (st) : Freq
P3 : Q

There's a text in the book explaining the design, feel free to ask questions i'll answer asap. Some other nice stuff to build in it, PM for the title (book in french).
 
This is a serial type parametric EQ based on non inverting universal filter cells.
There are several stages similar between them, each stage is made with one cell and one local summing amplifier that adds to the audio signal in stage's input with the audio signal from the cell.    

the typical cell of this eq is good but the problem is, like seem from this image, that all stages (cells and local summing ampplifiers ) are putted in series, so the audio signal go into too stages.

similar schematics are used in the 1960-1970 years

The eq can be improved for having all cells in parallel config.







 
hi ppa !

thanx a lot for your answer !

do you know a link to an example of such a "parallely working" EQ ? what kind of differences does that make audiowise compared to a "serie working" EQ ?

best regards.
 
> seem from this image, that all stages (cells and local summing ampplifiers) are putted in series, so the audio signal go into too stages.

No, only the stuff inside the dotted line is repeated. IC1a IC1b can serve any reasonable number of frequency bands, paralleled at (1) (2) and (3). If you already have a low-impedance source, IC1a can be omitted.
 
The neat thing about this implementation of the filter section is the use of linear rather than reverse log pots for the frequency setting. This works because the pot is used as voltage divider rather than a reostat (variable series element). A linear pot is easier to obtain and has better section to section tracking.

Note that the linear law of the 10k pot is slightly modified by the 2k2 resistor. The only downside is that such an arrangement is tricky for a very wide sweep range AND nice frequency scale. As shown here, the sweep range is roughly 1:10. Early Focusr*te products used this "trick" as well.

 
ppa said:
the typical cell of this eq is good but the problem is, like seem from this image, that all stages (cells and local summing ampplifiers ) are putted in series, so the audio signal go into too stages.
The eq can be improved for having all cells in parallel config.
This is very debatable.
Whatever the configuration (series or parallel), there will be he same number of biquads for a given number of bands. So let's take them out of the comparison.
In the series topology, there will be as many gain/boost opamps as there are bands. The total degradation of signal produced by these N stages will be: quadratic noise addition (4 bands => +6dB) and THD (anything between +6 and +12dB).
In the parallel topology, there is only one opamp, but its noise gain increases LINEARLY with the number of bands. So the degradation of signal increases by 12dB for a 4 band design.
In addition, in parallel topology, two adjacent bands interact in an "unpredictable" way. See attached curves. IMO, the series topology does what it should, i.e. the resulting curve is exactly the sum of the individual bands - very instinctive.
 

Attachments

  • param 2band series.jpg
    param 2band series.jpg
    182.9 KB · Views: 187
abbey road d enfer said:
ppa said:
the typical cell of this eq is good but the problem is, like seem from this image, that all stages (cells and local summing ampplifiers ) are putted in series, so the audio signal go into too stages.
The eq can be improved for having all cells in parallel config.
This is very debatable.
Whatever the configuration (series or parallel), there will be he same number of biquads for a given number of bands. So let's take them out of the comparison.
In the series topology, there will be as many gain/boost opamps as there are bands. The total degradation of signal produced by these N stages will be: quadratic noise addition (4 bands => +6dB) and THD (anything between +6 and +12dB).
In the parallel topology, there is only one opamp, but its noise gain increases LINEARLY with the number of bands. So the degradation of signal increases by 12dB for a 4 band design.
In addition, in parallel topology, two adjacent bands interact in an "unpredictable" way. See attached curves. IMO, the series topology does what it should, i.e. the resulting curve is exactly the sum of the individual bands - very instinctive.

I know well these problems , but I'm referring that several stages (like is in a series topology eq) in series sound not very transparent, even if THD and noise of is very low.
Moreover, isn't true that series eqs do the exact sum of the near band. In the parallel eqs the interactions are given by the different (and with polarity opposite) phase of the near band signals, but in the series eqs the problem of the near bands interactions are given by module of the signals. I agree with you that the sum signals is better (but not exact) with the series eqs but having a very long signal path like is in the series eqs is a problem bigger than the parallel eqs' band interactions. Infact, Massemburg and Maselec eqs don't use series topology but use parallel topology only, and these EQs are well knowed for their trasparency. I think Mr. Massemburg knows these problems, but he makes parallel eqs........and the first thing important is that sound engineerings use this eqs without big problems......
   




 
Hi !

Thanx for that interresting debate ! I stay quite silent because of my incompetence but I learn a lot !

So if i understood well we can do both parallel and serie EQ here, depending on if we connect the next stage using 1-2-3 or E-S-M, right ?

The book gives technical caracteristics :

transfert attenuation : 0dB (not so surprising :p)
total distortion : 0,009%
noise level at the output : -98 dBm (unweighted)
                                    -101 dBm (A)
                                    -88,5 dBm (CCIR)

I'll definitely try to build one when I'll have my 8 Green and 1176 finished ! If someone he also wants to try I can send the PCB layer by PM or the book title. That book s a guide for building you own console by E¤L¤E¤C¤T¤O¤R, really interresting stuff in it !

Best regards.
 
ricothetroll said:
So if i understood well we can do both parallel and serie EQ here, depending on if we connect the next stage using 1-2-3 or E-S-M, right ?
Yes you can.
The book gives technical caracteristics :noise level at the output : -98 dBm (unweighted) -101 dBm (A) -88,5 dBm (CCIR)
Voltage levels should be expressed in dBu (0.775v w/ no impedance specified - dBm's are related to a specific impedance -600 ohms for audio, 50 ohms for RF)
That book s a guide for building you own console by E¤L¤E¤C¤T¤O¤R, really interresting stuff in it !
The problem with Elektor (and all those similar DIY electronics magazines) is they never go to the deep end of the subject. Their designs work somewhat, but are never optimised. Major sources of noise in the biquads: R5, 6, 7, 8, 15, 16, 18, 19, 20 and P3 should be ca. 5x smaller. 5532's want to see less than 5 kohms at their inputs.
 
cuelist said:
The neat thing about this implementation of the filter section is the use of linear rather than reverse log pots for the frequency setting. This works because the pot is used as voltage divider rather than a reostat (variable series element). A linear pot is easier to obtain and has better section to section tracking.

Note that the linear law of the 10k pot is slightly modified by the 2k2 resistor. The only downside is that such an arrangement is tricky for a very wide sweep range AND nice frequency scale. As shown here, the sweep range is roughly 1:10. Early Focusr*te products used this "trick" as well.

Potentiometer are typically 20% bulk resistance tolerance so they are more accurate when used ratio-metrically with light loading. The tolerance issue with that schematic will be at the low frequency end of the adjustment range when the 20% pot resistance is one leg of the divider.  A similar but more accurate approach is to connect the bottom of the pot to ground so at the LF end the pot tolerance is completely out of the circuit, then parallel the resistor going into the integrator with a second larger resistor to the top of the pot. The LF extreme is set by just this larger resistor, the HF end is both resistors in parallel. Now you will have just the fixed resistor tolerance accuracy at both frequency extremes with the pot used just for ratios in between. The pot errors then will be factored by relative impedance and pot ratio accuracy.

JR
 
abbey road d enfer said:
The problem with Elektor (and all those similar DIY electronics magazines) is they never go to the deep end of the subject. Their designs work somewhat, but are never optimised.

I have been an Elektor reader since 1977. My first power amplifier was Edwin 40W with 2N3055 on the outputs  ;D. Round about the same time I built a three band EQ for a guitar amp which whistled like hell  when boosted the mid section. Obviously I had no design experience and could not understand why. So you are quite right that a little more explanation on the circuit operation would have helped. Overall it is a very good magazine but there are  surprises that the schematics are full of errors. But the horrors become in their translation into English language. I was once reading some valve stuff and the guy was using the convention for input impedance for internal resistance. If you are a beginner you'll have hell of a time with mixing thing up.
 

Latest posts

Back
Top