Choices for a simple transparent 5 to 10 band EQ (for vocals)

Started by jfrabat, May 19, 2020, 03:58:09 PM

Previous topic - Next topic

Mark Hammer

Quote from: jfrabat on May 22, 2020, 12:16:27 AM
Quote from: Mark Hammer on May 21, 2020, 01:30:42 PM
It's intended to show that the two 100k pots are each half of a dual ganged 100k pot.

Ah...  You learn something every day!  Good to know!
Often, schematics will do us the service of labeling dual-ganged pots in a manner like "VR1a / VR1b" to indicate that the two are halves of a single unit.  Not so in this case.  I guess they figured the dotted line would implicitly convey that information.

jfrabat

Quote from: rankot on May 22, 2020, 02:53:15 AM
If you're aiming to have a good quality and low noise EQ, all those capacitors in the signal path shall be polymer type, not ceramic! Those 10n (0.1uF) in power supply can be ceramic. It is also better to have regulated power supply with linear regulators.

My idea was to build this one:



I was planning to use film (brownies) capacitors (except for the electrolytic, of course; those will be Panasonic caps).  The only thing is that I was planning to change the 0.0039uF for 0.0033uF, as I do not have 0.0039uF available (not sure how much 0.0006uF will change the frequencies, but it is what I have).  I will replace 1N4001 diodes with 1N4004, but that should have no effect. Resistors will all be Metalic +-1% throughout.  Neutrik jacks and Alpha pots.
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

rankot

Quote from: jfrabat on May 22, 2020, 09:41:54 AM
I was planning to use film (brownies) capacitors (except for the electrolytic, of course; those will be Panasonic caps).  The only thing is that I was planning to change the 0.0039uF for 0.0033uF, as I do not have 0.0039uF available (not sure how much 0.0006uF will change the frequencies, but it is what I have).  I will replace 1N4001 diodes with 1N4004, but that should have no effect. Resistors will all be Metalic +-1% throughout.  Neutrik jacks and Alpha pots.
That's fine!
  • SUPPORTER
60 pedals and counting!

jfrabat

Here is what I am thinking:



EDIT: I noticed you cannot read anything above.  direct image link here.

It is basically the PAiA 9303 4 band equalizer with a power supply regulator circuit added.  If I went overboard, or if I am duplicating the protection, please let me know.  I also want to have 2 switches, one for power, and another to bypass the circuit (same wiring as a stompswitch, but with a toggle switch), each with an LED indicator (I am thinking blue would look nice!).

I also changed the .0039 caps for 0.0033 (because I have those!).  But I was wondering what effect will that have on the frequency.  Anyone with time on their hands care to teach me how I do that?  I tried reading online, but did not understand...
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

PRR

> 0.0039 caps for 0.0033

Frequency is 1/C.

1kHz becomes 1.18kHz.
  • SUPPORTER

Mark Hammer

Five semi-parametric bands is a little nuts.  I mean it is certainly ambitious, and aims high, but the question is whether it accomplishes anything you actually need.

However, if you feel 5 bands is necessary, I recommend including bypass toggles for each section.  The idea is that
a) Whatever hiss from sections you don't need to use can be avoided, and
b) You can zero in on one aspect of the frequency response at a time, identify what more may be needed, and then engage another section/band, and perhaps another; the toggle will let you easily compare with and without, without having to fiddle with the knobs.  Remember that the bands will overlap to some degree, so what you hear in the end is a joint product of what band A and B are both doing.


jfrabat

Quote from: Mark Hammer on May 22, 2020, 07:19:32 PM
Five semi-parametric bands is a little nuts.  I mean it is certainly ambitious, and aims high, but the question is whether it accomplishes anything you actually need.

However, if you feel 5 bands is necessary, I recommend including bypass toggles for each section.  The idea is that
a) Whatever hiss from sections you don't need to use can be avoided, and
b) You can zero in on one aspect of the frequency response at a time, identify what more may be needed, and then engage another section/band, and perhaps another; the toggle will let you easily compare with and without, without having to fiddle with the knobs.  Remember that the bands will overlap to some degree, so what you hear in the end is a joint product of what band A and B are both doing.

Sorry, you are right!  It is actually 4, but I messed up when I stitched all the print screens together (not sure how to save an image out of Eagle another way!)  But rest assure, it is only 4 bands...  The Hammond 1590Q will fit 4 bands nicely, but 5 would be too crowded!  Not to mention I have 4 gray and 4 tan knobs I plan to use; I do not have enough for 5 bands, unless I go to my aluminum knob stash, and I plan to keep those for guitar pedals! 

Sorry about that!

EDIT: I modified the above image and link to the correct one
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

jfrabat

Quote from: PRR on May 22, 2020, 06:18:04 PM
> 0.0039 caps for 0.0033

Frequency is 1/C.

1kHz becomes 1.18kHz.

OK, have some patience with me here...

So, we have 4 bands in the original schematic. 



Lets focus on the third one, which is the one I changed the cap value...

That one uses 2 X 0.0039 caps.  According to the schematic, that band controls signals from 450Hz to 8.5KHz.  The math does not add up...  Now here, the caps are kind of both in series and in parallel at the same time.  So how do I count C? 0.0078? Then 1/0.0078 = 128Hz, right?  If it is counted as 0.0039 then 1/0.0039 would be 256Hz.  In neither case is it the frequencies listed on the schematic (I imagine they are not exact, but these are not even close!).  So how do I calculate the frequencies I will be able to select in band 3?

I mean, I did use the 1/c to calculate a ratio, and when I use the 0.0033 caps, I get 500Hz-10KHz (well, 490Hz to 9.7KHz really, but who's being exact!), but I do not know how to get to that number with pure math...
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

PRR

> that band controls signals from 450Hz to 8.5KHz.

If it is variable 20:1, why even think about 1.18:1 cap change?

The filter is a Wien Bridge. The formula for F has root(1/(R*R*C*C)) and some pie.
https://en.wikipedia.org/wiki/Wien_bridge
https://en.wikipedia.org/wiki/Wien_bridge_oscillator
When R1=R2=R and C1=C2=C, the frequency of oscillation is given by:


The bridge works for unequal parts but 97% of designs use equal values for sanity. There is a variant with useful values of un-equal but only H-P had the brains to get anywhere on that.
  • SUPPORTER

Rob Strand

QuoteIf it is variable 20:1, why even think about 1.18:1 cap change?
22:1 to be precise.

10:1 range is a bit more repeatable to adjust.

For high frequency spans like that you will need to use Reverse-audio/Antilog taper pots.
(If you are willing to compromise with back to front frequency scales you can use
audio/log taper pots.)

If you intend to use linear taper pots then it would be wise to reduce the range somewhat.
Perhaps 4:1 or at a squeeze 5:1.
Send:     . .- .-. - .... / - --- / --. --- .-. -

duck_arse

your power supply makes no sense. what are you feeding in, and what are you expecting out?
You hold the small basket while I strain the gnat.

jfrabat

Quote from: PRR on May 23, 2020, 12:53:44 AM
If it is variable 20:1, why even think about 1.18:1 cap change?

So you mean don't worry about the change?  I did not get your meaning...

Quote from: PRR on May 23, 2020, 12:53:44 AM
The filter is a Wien Bridge. The formula for F has root(1/(R*R*C*C)) and some pie.
https://en.wikipedia.org/wiki/Wien_bridge
https://en.wikipedia.org/wiki/Wien_bridge_oscillator
When R1=R2=R and C1=C2=C, the frequency of oscillation is given by:


The bridge works for unequal parts but 97% of designs use equal values for sanity. There is a variant with useful values of un-equal but only H-P had the brains to get anywhere on that.

OK, you lost me again...  I tried understanding Wiki, but it just went over my head...

But after toying with Excel, I kind of figured it out (I do not get the EXACT values listed, but I am REALLY close those values!)

Band 1 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.000000047 X 104700) = 32.34Hz
Band 1 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.000000047 X 4700) = 720.48Hz

Band 2 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.00000001 X 104700) = 152.01Hz
Band 2 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.00000001 X 4700) = 3,386.28Hz

Band 3 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000039 X 104700) = 389.77Hz
Band 3 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000039 X 4700) = 8,672.76Hz

Band 4 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000022 X 104700) = 690.96Hz
Band 4 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000022 X 4700) = 15,392.16Hz

And then, according to the formula, my change would make the bands as follows:

Band 3 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 104700) = 460.64Hz
Band 3 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 4700) = 10,261.44Hz

If that is the case, I am 100% fine with this.

Quote from: Rob Strand on May 23, 2020, 01:20:53 AM
QuoteIf it is variable 20:1, why even think about 1.18:1 cap change?
22:1 to be precise.

In stock form, the range would be (with the above mentioned figures):
32Hz to 720Hz = 22.3:1
152Hz to 3,386Hz = 22.3:1
390Hz to 8,683Hz = 22.3:1
691Hz to 15,392Hz = 22.3:1

Mi change would make band 3 461Hz to 10,261Hz, so 22.3:1

Quote from: Rob Strand on May 23, 2020, 01:20:53 AM
10:1 range is a bit more repeatable to adjust.

If you intend to use linear taper pots then it would be wise to reduce the range somewhat.
Perhaps 4:1 or at a squeeze 5:1.

Well, i am kind of stuck because I only got 100K pots...  Sure, I could lower the range by increasing the resistor value in R, but I would lose my high end control if I do (for example, to go to 5:1, and keeping 100K pots, I would need to raise the resistor to 25K, which means the highest frequency would be 2.9KHz; if I go to 10:1, by using a 12K resistor, the highest frequency would be 6KHz.  I loose all the "airy" frequencies).  And the smallest cap I got (not counting ceramics) is the 0.0022, so changing cap is not the way either.

Quote from: Rob Strand on May 23, 2020, 01:20:53 AM
For high frequency spans like that you will need to use Reverse-audio/Antilog taper pots.
(If you are willing to compromise with back to front frequency scales you can use
audio/log taper pots.)

I have (in dual gang) Audio and Linear pots in 100K.  I intended to use linear; why do you recommend reverse logarithmic?

Quote from: duck_arse on May 23, 2020, 10:57:58 AM
your power supply makes no sense. what are you feeding in, and what are you expecting out?

I I was kind of following Rankot's advise... Remember, this is for vocals, not guitars!

Quote from: rankot on May 22, 2020, 02:53:15 AM
It is also better to have regulated power supply with linear regulators.

So I added this in front of the power supply to insure no electrical noises:

I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

PRR

QuoteBand 3 (low end): f1 = 1 / (2 X 3.14159265358979 X 0.0000000039 X 104700) = 389.77Hz
Band 1 (high end): f1 = 1 / (2 X 3.14159265358979 X 0.0000000039 X 4700) = 8,672.76Hz
my change would make the bands as follows:
Band 3 (low end): f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 104700) = 389.77Hz
Band 1 (high end): f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 4700) = 8,672.76Hz

It appears that changing 39 to 33 makes NO change of frequency.

And "Band 1" won't go away.

While I never trust Excel too far, this result is likely user/finger error.

You are asking for a very fancy "transparent" EQ, using only parts under your lock-down bench. Maybe it isn't that easy?

Throw together ONE band. Try it out. Does the opamp really make much difference? Do the pots turn the right way? Can you even hear 39 vs 33?
  • SUPPORTER

jfrabat

Quote from: PRR on May 23, 2020, 04:44:56 PM
It appears that changing 39 to 33 makes NO change of frequency.

And "Band 1" won't go away.

While I never trust Excel too far, this result is likely user/finger error.

Ah, yes, finger error.  The frequency with 33's is:

Band 3 (low end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 104700) = 460.64Hz
Band 3 (high end):
f1 = 1 / (2 X 3.14159265358979 X 0.0000000033 X 4700) = 10,261.44Hz

I corrected that and the BAND 1 error in the original post.

Quote from: PRR on May 23, 2020, 04:44:56 PM
You are asking for a very fancy "transparent" EQ, using only parts under your lock-down bench. Maybe it isn't that easy?

Throw together ONE band. Try it out. Does the opamp really make much difference? Do the pots turn the right way? Can you even hear 39 vs 33?

I am going to breadboard it as soon as I clear one breadboard; I have 2 projects on breadboards, so I need some space (one is almost done, so in a week or two I will start).  I will not make anything I do not breadboard first!
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

ElectricDruid

Quote from: PRR on May 23, 2020, 12:53:44 AM
If it is variable 20:1, why even think about 1.18:1 cap change?

PRR is right. If the band is variable over such a wide range, a small change in capacitor value really isn't important. You can tune it to get it where you need it anyway, so there's no problem. It only makes a tiny different to the very lowest and very highest frequencies you can get, but those aren't likely to be the ones you use most. There are other bands for that. So go for it, and don't worry too much.


Rob Strand

Quoteif I go to 10:1, by using a 12K resistor, the highest frequency would be 6KHz.  I loose all the "airy" frequencies).  And the smallest cap I got (not counting ceramics) is the 0.0022, so changing cap is not the way either.
When you reduce the range you need to place the frequencies so they are all in the region you want.  It's OK to use 100k pots.    The frequencies are then by changing the caps.   On a project of this scale it seems crazy holding back doing the right thing for the price of a few caps.     If you don't have the parts on hand you could use two caps in series to create a smaller value.

QuoteI have (in dual gang) Audio and Linear pots in 100K.  I intended to use linear; why do you recommend reverse logarithmic?
Professional equipment use reverse-log pots because it spreads the frequencies evenly over the range of the control.

For arguments sake, suppose you have a frequency control range 500Hz to 10kHz.   

From the point of view of having the frequency control being naturally centered the pot position 5 (on 0 to 10 pot scale) should be 2.2kHz.   That gives even frequency control 2.2 octaves up or down from 2.2kHz and would sound even to the ear. 

With a linear pot the center is about 1kHz.    The problem now is you have 0.86 octave span (500Hz to 1kHz) over the range of pot settings from 0 to 5 but a 3.5 octave span (1kHz to 10kHz) over pot setting 5 to 10.    In use that means the control gets very bunched up on the high frequency side.   As you get closer to 10kHz the bunching up gets worse.    It's not nice to use and it means every time you set-up the EQ on the high-frequency side it won't quite be the same as last time.

The reverse-log pot evens that out to be much more like the naturally centered case.
Send:     . .- .-. - .... / - --- / --. --- .-. -

jfrabat

Quote from: Rob Strand on May 23, 2020, 07:35:53 PM
[On a project of this scale it seems crazy holding back doing the right thing for the price of a few caps.     If you don't have the parts on hand you could use two caps in series to create a smaller value.

Good point...

Quote from: Rob Strand on May 23, 2020, 07:35:53 PM
Professional equipment use reverse-log pots because it spreads the frequencies evenly over the range of the control.

For arguments sake, suppose you have a frequency control range 500Hz to 10kHz.   

From the point of view of having the frequency control being naturally centered the pot position 5 (on 0 to 10 pot scale) should be 2.2kHz.   That gives even frequency control 2.2 octaves up or down from 2.2kHz and would sound even to the ear. 

With a linear pot the center is about 1kHz.    The problem now is you have 0.86 octave span (500Hz to 1kHz) over the range of pot settings from 0 to 5 but a 3.5 octave span (1kHz to 10kHz) over pot setting 5 to 10.    In use that means the control gets very bunched up on the high frequency side.   As you get closer to 10kHz the bunching up gets worse.    It's not nice to use and it means every time you set-up the EQ on the high-frequency side it won't quite be the same as last time.

The reverse-log pot evens that out to be much more like the naturally centered case.


Sold on the reverse audio tapered pots.

Where is a good source for dual gang reverse audio pots?  I looked in Mouser, but they do not have any.  Also, I was thinking about the frequency range...  I could go with a 10:1 ratio.  I would need to get caps for the base (33nF would be nice), but the rest I have.  So if I am going to order the pots, may as well order the caps.  I ran the numbers and this is what I could get with the caps I have:

Band    C (nF)   R1 (min)   R2 (max)   Lower Freq   Highest Freq   Band ratio
1         22.0nF   12,000Ω   112,000Ω     65 Hz              603 Hz        9.33:1
2         6.8nF   12,000Ω   112,000Ω     209 Hz           1,950 Hz       9.33:1
3         2.2nF   12,000Ω   112,000Ω     646 Hz            6,029 Hz      9.33:1
4           1.0nF   12,000Ω   112,000Ω   1,421 Hz      13,263 Hz        9.33:1

If I get 33nF, then it looks more like this:

Band     C (nF)   R1 (min)   R2 (max)   Lower Freq   Highest Freq   Band ratio
1          33.0nF   12,000Ω   112,000Ω    43 Hz             402 Hz      9.33:1
2         6.8nF   12,000Ω   112,000Ω    209 Hz          1,950 Hz       9.33:1
3         2.2nF   12,000Ω   112,000Ω    646 Hz         6,029 Hz        9.33:1
4         1.0nF   12,000Ω   112,000Ω    1,421 Hz      13,263 Hz      9.33:1

Although, the only reason to have to go down to 43Hz is to equalize instruments.  Vocals do not go below 80Hz, so I could make do just fine with 22nF (or I could use a 22nF and add a 10nF in parallel and call it good).
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).

Rob Strand

QuoteWhere is a good source for dual gang reverse audio pots? 
I think Small Bear does them.   In the past I bought them as spare parts from amp manufacturers, which can be a little expensive.     The last parametric I built was for bass and I used linear taper pots with reduced frequency ranges.

QuoteBand     C (nF)   R1 (min)   R2 (max)   Lower Freq   Highest Freq   Band ratio
1          33.0nF   12,000Ω   112,000Ω    43 Hz             402 Hz      9.33:1
2         6.8nF   12,000Ω   112,000Ω    209 Hz          1,950 Hz       9.33:1
3         2.2nF   12,000Ω   112,000Ω    646 Hz         6,029 Hz        9.33:1
4         1.0nF   12,000Ω   112,000Ω    1,421 Hz      13,263 Hz      9.33:1

All that looks OK to me.
Send:     . .- .-. - .... / - --- / --. --- .-. -

rankot

  • SUPPORTER
60 pedals and counting!

jfrabat

Quote from: rankot on May 24, 2020, 05:56:10 AM
C100k dual pots can be bought at Tayda for cheap.
https://www.taydaelectronics.com/100k-ohm-anti-log-dual-taper-potentiometer.html

Wanted the 90* ones, but no one has them, so I ordered those from Tayda.  Will start breadboarding soon (will have to wait for the pots, though!)

The other thing I was thinking was decreasing the R values (the fixed ones) in band 4 to get a little higher frequency there.  I was thinking 10K instead of 12K to go from 1,447Hz to 15,915Hz and have a bit more control over the "air" part of the spectrum...  What do you guys think? 1.4K~13.3K or 1.4K~15.9K?  Low end is almost the same (1,421Hz vs 1,447Hz), but I do gain a lot in the highs...
I build.  I fix.  I fix again.  And again.  And yet again.  (sometimes again once more).  Then I have something that works! (Most of the time!).