any suggestions for a 3band fully parametric equalizer

I want to make and rackable 2CH (stereo) eq.

desires:
3 bands
HP/LP filters 
perhaps "inspired" by a 1081
(though I'm basing that on the VEQ plugin being my favorite software EQ)

I'm new at this stuff, and can't seem to find a kit, so detailed instructions are a big big plus.
I have resolved to keep on trying until I start succeeding at projects like this.

thanks in advance!

If you want it fully parametric, there are basically three possible topologies: State-variable, Wien bridge and bridged-T. Each has its pros and cons. The main issues with parametric EQ's are noise optimization and procurement of control parts (pots). Wien bridge and bridged-T require dual revlog pots. SV can be implemented with dual LIN pots. Wien bridge and bridged-T are a tad easier to optimize for noise. SV requires very low noise opamps. On the net, you can find the schematics for Rane PE17, UREI545 and dbx905 (this one shows how a linear pot can be used)

How about three fifths of a Sontec clone? It offers fully parametric operation in each band with boost and cut, Q and frequency adjustments. This is a fairly simple eq and you can implement just one of the low band and high band filters, as in the original these were duplicated. Modify the low and high band switching slightly to give selectable peaking or shelving. Your ranges would be 8-800Hz, 80-8000Hz, and 120-25,600Hz. This eq runs all the frequency selective elements in parallel, bridging just one op amp so the audio path is very simple.

Here is a link to a site by Bauman which has the information on a unit he built. He uses very expensive switches and resistors to implement the reverse log frequency controls. He also provides for using discrete op amps in the frequency selective elements, which, in my opinion is overkill. I would suggest considering DOAs in the audio path, however. I would also suggest using dual reverse log 100K pots in place of the switches and resistors in the frequency controls. These parts are a little hard to find right now but that should improve in the future. There is information on P-Pro about Igor's and others versions of the five band Sontec but some of these versions use a less than optimal high band filter in the top band.

http://www.thediypill.phx.com.br/diy_sontec.htm

I am a fan of using the SVF approach. I don't see noise as a major issue if you use the right topology and decent opamps. Use a topology that doesn't cause the full passband to see the noise gain of the boost/cut. There are several different SVF approaches, some better than others. Boost/cut gain to just the EQ filter passband is not generally a noise problem (IMO).

JR

JohnRoberts said:

I am a fan of using the SVF approach. I don't see noise as a major issue if you use the right topology and decent opamps. Use a topology that doesn't cause the full passband to see the noise gain of the boost/cut.

Click to expand...

In order to achieve this, you have to connect the boost/cut pot after the biquad, and you need a pot with a center tap to ground or a three position switch giving boost/off/cut.

There are several different SVF approaches, some better than others. Boost/cut gain to just the EQ filter passband is not generally a noise problem (IMO).

JR

Click to expand...

My experience is that customers listen to the noise. Unless you implement the latter principle, noise from the SVF varies with the frequency and Q settings and it is considered as a bad thing; they reckon, not without reason, that in the flat position, there is still something in the signal path that they don't like.

I have designed several, and studied many more SVF based parametrics. There are many many ways to skin that cat.

I am not sure I follow your comment... The SVF is typically separate from the boost/cut, while I do recall one that varied the BP gain of the SVF to adjust the Q (quiet, but not great for operating headroom).

In general the quietest topology for the boost/cut section is subtractive. Cut is effected by adding an opposite polarity version of the subject bandpass, effectively subtracting it from the input. Boost is effected by subtracting that same bandpass signal from the feedback path. The noise gain of this approach is only slightly more than an inverting opamp and delivers as much boost/cut as you could desire. The noise coming from the SVF is BP filtered, and injected at only unity gain, independent of boost/cut.
-----
The next consideration is noise gain of the SVF.  While one could vary the Q very quietly by staggering the poles further apart or overlapping, that isn't practical using commonly available panel controls, so Q is generally adjusted by a shunt to ground at either the + or - input of the first opamp stage in SVF. I have found that instead of using a fixed resistor to ground at one port and variable at the other which must swamp out the contribution of the first, connecting the Q potentiometer's end terminals to the + and - inputs, with the wiper connected through a resistor to ground, gives a best case noise gain for a given Q (using that general approach). Using a relatively large value pot relative to the end limit resistor reduces noise gain further at modest Q settings.

There are a number of subtle ways to get in trouble with SVF so avoid high HP or LP section gains, because inadvertent clipping there could introduce distortion into the main signal path that may be hard to diagnose.

Starting with a quiet topology first and then applying quiet parts as needed is preferable to brute force use of uber low noise parts, as high noise gain has other consequences (phase shift, distortion, etc). 

JR

abbey road d enfer said:

If you want it fully parametric, there are basically three possible topologies: State-variable, Wien bridge and bridged-T. Each has its pros and cons.

Click to expand...

Or you can make one with Multiple Feedback Bandpass as frequency selective element. (Like one of three middle sections in this http://www.korganalogue.net/korgother/KLM62/Korg_KLM62_Resonators.jpg schematic). You need one inverter on input, being fed with boost/cut pot.

I have slight preference for MFB one for simplicity, and SVF one for flexibility. IMHO frequency selective blocks should be as transparent as possible (personaly, I dont care much about noise, but I do care aboout clipping of one of inverters in SVF topology. I would trade noise for headroom anytime). Also, my philosophy is that frequency pot should cover one decade, with aditional two or three position switch that would scale frequency range by switching different capacitor groups (lot of desk EQs I've seen allow frequency pots to go down to 600 ohm. I would stick with 20K to 2K range for most opamps, with capacitor switch for range, AND lower levels in filter).

cheerz
ypow

recnsci said:

abbey road d enfer said:

If you want it fully parametric, there are basically three possible topologies: State-variable, Wien bridge and bridged-T. Each has its pros and cons.

Click to expand...

Or you can make one with Multiple Feedback Bandpass as frequency selective element. (Like one of three middle sections in this http://www.korganalogue.net/korgother/KLM62/Korg_KLM62_Resonators.jpg schematic). You need one inverter on input, being fed with boost/cut pot.

Click to expand...

The MFB topoogy doesn't lend itself to easily tune the Q. You need to add a second feedback loop to make a variable Q. And the MFB is intrinsically noisier than SVF.

I have slight preference for MFB one for simplicity, and SVF one for flexibility. IMHO frequency selective blocks should be as transparent as possible (personaly, I dont care much about noise, but I do care aboout clipping of one of inverters in SVF topology. I would trade noise for headroom anytime).

Click to expand...

I may accept to trade 1dB of noise for at least one dB of headroom, within limits.

Also, my philosophy is that frequency pot should cover one decade, with aditional two or three position switch that would scale frequency range by switching different capacitor groups (lot of desk EQs I've seen allow frequency pots to go down to 600 ohm. I would stick with 20K to 2K range for most opamps, with capacitor switch for range, AND lower levels in filter).

Click to expand...

Agreed, I have a DG308 taking care of switching capacitors, and linear pots with law-bending loads.

recnsci said:

I have slight preference for MFB one for simplicity, and SVF one for flexibility. IMHO frequency selective blocks should be as transparent as possible (personaly, I dont care much about noise, but I do care aboout clipping of one of inverters in SVF topology. I would trade noise for headroom anytime). Also, my philosophy is that frequency pot should cover one decade, with aditional two or three position switch that would scale frequency range by switching different capacitor groups (lot of desk EQs I've seen allow frequency pots to go down to 600 ohm. I would stick with 20K to 2K range for most opamps, with capacitor switch for range, AND lower levels in filter).

cheerz
ypow

Click to expand...

In SVF you can predict when clipping will occur and manage it, but you need to look at all three outputs.

Using common pots to vary center frequency in (ground referenced) SVF integrators over a wide range can be sloppy due to 20% tolerance of bulk resistance in typical controls. Potentiometer ratio accuracy is much better than bulk resistance. I get the best results from using the frequency pot as a simple divider, lightly loaded, dividing all the way down to off (ground). With the wiper grounded, the LF limit is then defined by a parallel shunt resistor, so the pot is completely out of the circuit at both frequency extremes and accuracy there is limited by the tolerance of Rs and Cs used. Between those endpoints the pot tracking is IMO acceptable for a lot more than one decade of range, while decades make for simpler panel graphics with range switches.

JR   

abbey road d enfer said:

The MFB topoogy doesn't lend itself to easily tune the Q. You need to add a second feedback loop to make a variable Q. And the MFB is intrinsically noisier than SVF.

Click to expand...

Yeah, I've messed with both positive and negative feedback around input inverter/summing amp. I considered it for simple lowQ/highQ topology with similar behaviour as 1084 (if I remember that neve part number corectly).
I originaly used inverter->MFBPF for guitar pedal thingie with two band EQ, and built around Hex CMOS inverter (not exactly high fidelity stuff). As far as noise goes, well I'm interested in music like berlins' BasicChannel/ChainReaction stuff (for those who know this stuff). That means adding lots of noise intentionaly, and  regular boosts and cuts of 18dB. Thats why all preferences are strictly personal.

Agreed, I have a DG308 taking care of switching capacitors, and linear pots with law-bending loads.

Click to expand...

Thats other major benefit, you can fiddle with lin pot to behave nice over one decade

cheerz
urosh

JohnRoberts said:

In SVF you can predict when clipping will occur and manage it, but you need to look at all three outputs.

Using common pots to vary center frequency in (ground referenced) SVF integrators over a wide range can be sloppy due to 20% tolerance of bulk resistance in typical controls. Potentiometer ratio accuracy is much better than bulk resistance. I get the best results from using the frequency pot as a simple divider, lightly loaded, dividing all the way down to off (ground). With the wiper grounded, the LF limit is then defined by a parallel shunt resistor, so the pot is completely out of the circuit at both frequency extremes and accuracy there is limited by the tolerance of Rs and Cs used. Between those endpoints the pot tracking is IMO acceptable for a lot more than one decade of range, while decades make for simpler panel graphics with range switches.

JR   

Click to expand...

I consider that with (typical) small desk knobs, much more than decade of sweep feels fiddly. Agree with rest (I put series and paralell Rs to well define end points almost automatically).
Anyway, regarding original poster question, I dont know if somebody made a group project of it, but my first choice would be AMEK 2500 (or Barry Porter EQ, but that one needs IIRC center tapped pots, or whatever you call them).

cheerz
urosh

well after all this I'm left realizing that probably not smart enough to make one. I need a kit with good instructions, but I also want a good, recording quality EQ.
I saw this:
http://www.paia.com/proddetail.asp?prod=6760K

is there any better kits out there?

pixeltarian said:

well after all this I'm left realizing that probably not smart enough to make one. I need a kit with good instructions, but I also want a good, recording quality EQ.
I saw this:
http://www.paia.com/proddetail.asp?prod=6760K

is there any better kits out there?

Click to expand...

This is about closest to nobrainer kit stuff: http://www.gyraf.dk/gy_pd/calreq/calrec.htm
A friend (as well as half a million people on this board) has built it and it sounds good
There are some PCBs for sale right now: http://www.groupdiy.com/index.php?topic=31630.0
You will have to source all the parts your self (biggest hurdle is sourcing dual reverese log pots; OMEG makes them and I think I saw a company that sells them in small quantities)
There are lot (LOT) of threads on this board regarding making these EQs and troubleshooting, just search.

cheerz
ypow

http://www.johnhroberts.com/document.JPG

Here is a schematic and layout from a kit I designed and published back in the '70s (30 years ago). This design doesn't reflect later refinements I mentioned above, but is a good starting point.

One feature I designed into this EQ (for home Hifi use) that I didn't use in professional recording EQ, is an intentional interaction between the Q adjustment and amount of boost/cut.  This was IMO useful in home Hifi for eq'ing full range signals as it automatically provides more boost/cut for narrow Q settings, and less for wide Q settings, giving a first order normalization of loudness when EQing complex sources.

This interaction can be defeated by shorting the wiper of the Q pot to the BP feedback resistor. Some other tweaks will be required to rescale boost/cut and Q range. 

JR

JohnRoberts said:

http://www.johnhroberts.com/document.JPG

Here is a schematic and layout from a kit I designed and published back in the '70s (30 years ago). This design doesn't reflect later refinements I mentioned above, but is a good starting point.

Click to expand...

I built that one  a stereo version, my first serious 19" DIY & PCB-making experience. Not back in the 70ties (was a little boy back then) but found it in university-library somewhere in the 80ties.

Although I don't use it much anymore I don't have much reasons why I don't!

IIRC the full scanned article is around somewhere as well, I presume with the apporval of John.

This interaction can be defeated by shorting the wiper of the Q pot to the BP feedback resistor. Some other tweaks will be required to rescale boost/cut and Q range. 

Click to expand...

Hmm, might be a good add-on switch (assuming the tweaks are easily taken along on the switch as well). If you happen to have those additional tweaks documented somewhere then I'm all ears.

Bye/thanks,

  Peter

clintrubber said:

JohnRoberts said:

http://www.johnhroberts.com/document.JPG

Here is a schematic and layout from a kit I designed and published back in the '70s (30 years ago). This design doesn't reflect later refinements I mentioned above, but is a good starting point.

Click to expand...

I built that one  a stereo version, my first serious 19" DIY & PCB-making experience. Not back in the 70ties (was a little boy back then) but found it in university-library somewhere in the 80ties.

Although I don't use it much anymore I don't have much reasons why I don't!

IIRC the full scanned article is around somewhere as well, I presume with the apporval of John.

This interaction can be defeated by shorting the wiper of the Q pot to the BP feedback resistor. Some other tweaks will be required to rescale boost/cut and Q range. 

Click to expand...

Hmm, might be a good add-on switch (assuming the tweaks are easily taken along on the switch as well). If you happen to have those additional tweaks documented somewhere then I'm all ears.

Bye/thanks,

  Peter

Click to expand...

It might be more than a 2 pole switch to do it right...

As drawn the the BP gain of the SVP varies 3:1 (51k to 151k). The 10 dB gain change, affects both the max boost/cut and Q.

Defeating this will probably require changing values of at least two fixed resistors to scale Q appropriately for the new fixed BP gain, and perhaps add a resistor to the summing section to increase gain of signal coming from SVF to deliver full boost/cut.  Note these changes will increase the noise gain of this topology so will not be as clean as stock circuit.

IIRC I wrote up a set of design equations that I shipped with the kit, because so many college students bought these to satisfy a popular EE course requirement back then... The EE students were supposed to modify an existing design, so I suspect my design equation cheat sheet helped quite a few students pass that course.

If there is enough interest I'll dig through my 30 YO kit company papers to see if i can find a copy. Just not right now.

JR

 

What about a swinging input type with 3 CAPS resonators for para?

-T

TomWaterman said:

What about a swinging input type with 3 CAPS resonators for para?

-T

Click to expand...

If we're using the same terminology, a swinging-input EQ is based on an antiresonant circuit moving progressively between the + and - inputs of an opamp. A CAPS is a current in - voltage out bandpass filter. Turning it into an antiresonant impedance cost another opamp. What's more, it has a problem that most other filter topologies don't have: all parameters are interactive, i.e. Q and Amax change when you change frequency.

JohnRoberts said:

It might be more than a 2 pole switch to do it right...

Click to expand...

Thanks John for the response. OK, sounds like it becomes a complicated switch pretty quickly... a few toggle-style 4PDT's might still fit but anything bigger (like a multiple-deck rotary) will not, I pretty much used most of the front & rear panel.

If by chance you run into the equations, then that'd be interesting, but I understand, countless other things... please don't dig too deep.

Regards,

  Peter 

I'll dig through my papers this weekend. A passable version could probably done with 3 or 4 poles.

1 pole to defeat variable BP section gain
1 pole to add some more boost/cut
1 or two poles to shift Q to similar range

I'll dig up the note and you can run your own numbers.

JR