Vintage Blue Box 10-band EQ Build - Some Q's

[strungout]

Oi all!

So, I'm getting ready to breadboard this EQ, going by this schematic here: http://www.freestompboxes.org/viewtopic.php?f=19&t=3988#p139538 (will need to be logged in to see it...).

This is my layout:



Will fit a 1590BB nice and tight, lengthwise, and plenty of room for the hardware. If you happen to spot anything that looks like it might be a problem or looks weird (not asking to verify it against the schem ), please tell me. I plan on building a regulated power box with multiple outlets. For now I'll build a daughter board so I can run it off an 18vdc adapter.

After some research I have questions left:

-I'll be using 10k pots instead of the 15k. I know changing pot values will screw with impedance, but I'm not exactly sure how and how much difference the 5k would make?

-Should I bother using a trim instead of the gyrator resistor to ground to dial in the exact frequencies? I did some calculation and noticed they either come off short or above the noted frequencies (ex.: for 8kHz, I get 7.8kHz). Is that out of components/space restrictions or is there some logic behind those choices?

Any input appreciated!

[Rob Strand]

The "best" pot value isn't such a straight forward question.

Most commercial graphic equalizers which use Gyrators, like that circuit, use pots which have a special
"G"  taper(graphic-equalizer).   Some companies provided the same pot taper but with a different name/suffix.
The reason they use this taper is to stop the effect where nothing happens over 90% of the pot movement
the all of a sudden whoof, there is this enormous boost in the last 10%.  I call this control "cramping" or "bunching".

If you can source such pots off hand I would say the "best" pot value is about 20kohm.

Now if you cannot find such pots then the next best choice is a linear taper pot.
The "best pot" value is a balance between reducing cramping and band-interaction.
A lower value pot will reduce "cramping".  So a 10k might be better here.

Going further say down to 5k will further improve reduce "cramping" but what happens then is the bands start interacting more.
Normally would I play around a on a circuit simulator to see how bad it is.
Unfortunately I don't have that one entered into the simulator - well ... I can't find it ATM!.

(Another quirk of the old 10-band MXR is that the gyrator resistors to ground are quite low in value,
these also affect the response in an undesirable way.)

[R O Tiree]

That schematic at http://www.....org/viewtopic.php?f=19&t=3988#p139538 is wrong in several respects.

For starters, R29 and R30 should be before the Zener diodes, not downstream of them.

Secondly, the 4kHz, 2kHz and 1kHz circuits are identical. That is clearly wrong.

There's a better version (possibly... at least, it makes more sense in some ways) earlier in that thread over at the "other" forum in: www.&*%$£#@.org/viewtopic.php?f=19&t=3988

There are some values that George Gilbert admits he can't provide and he's gone with 50k lin sliders whereas the general consensus (including from our own Govment Lackey who posted in there as well) that they should be 15k and W-taper (same as G-taper according to sources I've seen?).

I'm messing around with this in sim and calculations to see if (a) the size of the pot matters much and (b) if it does, can we fake it with parallel resistors somehow (initial calcs look promising at getting around about 15k with a G/W-style taper... I'll get back to you shortly)

[Rob Strand]

For starters, R29 and R30 should be before the Zener diodes, not downstream of them.

Secondly, the 4kHz, 2kHz and 1kHz circuits are identical. That is clearly wrong.

Both items look OK to me.

(from fs.or post) It seems the original pots may well be 20k lin.

"p.s. slide pots seems to be 20K linear. "

[R O Tiree]

OK, it's looking like I messed up the axes on my graph in Excel... popping parallel resistors from pins 1 to 2 and pins 2-3 actually makes the "cramping" that Rob mentioned above even worse.

Trawling through Bourns, Alpha and other manufacturers' datasheets for slider pots, it would appear that the de facto standard is variations on a "B" taper... B1 is "perfectly" linear. B2 to B4 are varying degrees of the "G-taper" required. Quite where you'd find the damn things is anyone's guess. A manufacturer can afford to leverage Bourns, for example, into making a 15k B4 pot in bulk. A hobbyist can't and, because it's such a weird value and unusual taper, you're not very likely to be able to source them privately from suppliers whose only interest is in selling "popular" values/tapers.

Bottom line, I'd go with Rob's suggestion of 10k linear (B1) pots as the best compromise between cramping and band interaction for this particular circuit.

[R O Tiree]

@ Rob, the schem that Frank (strungout) linked to in his original post was one from a guy in Italy, calling himself "Fix_Metal" (posted 29 Mar 2011, 23:16).

http://www.******.org/download/file.php?id=12059&mode=view

That is the one that is clearly wrong for the reasons I said. I could insert the img code, but I'm not sure that people who aren't members over there would be able to see it and I'm also not sure that the mods here wouldn't be a little annoyed at me linking to that site and I'm pretty positive that the mods over there would be annoyed... Hence the thinly-disguised link to the image I'm on about in the post number that Frank linked to. You can see that C14, C16 and C18 are all identical (6.8nF), as are C15, C17 and C19 (150nF), along with R16, R18 and R20 (47k) and lastly R17, R19 and R21 (470R). Three gyrators with identical components cannot give three different peak/notch values?

The schematic at the start of that thread over on the "other" forum was from a chap called George Giblet (not Gilbert...oops), and it shows the resistors R29 and R30 ahead of the Zeners, where they should be, and sensible values for the 1, 2 and 4 kHz bands (posted 06 Jan 2009, 05:59). I've linked to George's here:



Govmt_Lacky's post about 15k pots is on page 2 of the thread and refers to the original MXR schems (albeit for the 6-band version). I suspect that MXR might have changed to 20k pots later in response to manufacturers' reluctance to make weird values, even if they had to make weird tapers (MBAs - gotta love 'em). Based on what you said earlier, I guess they found the slight increase in value did not compromise band interaction too badly, so they went with the flow, perhaps?

@ Frank - I think that, in an ideal world, with ideal components with 0% tolerance, then one might look to be a little more accurate with the gyrator values. As it is, about the best you're going to get with caps (at a reasonable price) is 5% or 10% tolerance. Back when this MXR circuit was built, I'd guess that 20% tolerance caps were the norm. I also think that it would be hard to tell whether a peak/notch was 7.8kHz, 8.0kHz or 8.2kHz purely by ear. Even 7.0kHz or 9.0kHz might be a tad tricky without an ABY test, flicking between the two... So, going with commonly available values and getting it there or thereabouts probably looked like a perfectly viable commercial decision.

[strungout]

Thanks for the replies.

The "best" pot value isn't such a straight forward question.

Well, I was more concerned about the effect the lower resistance would have on impedance (not sure about how to calculate it. Have to read up more.), not the taper. Though it IS useful information you provided.  I didn't know what exactly a g-taper was... I can live with a linear taper on rotary pot. What can I say, I'm gettho  

(...)but what happens then is the bands start interacting more.

Is this because of the difference in impedance?

(Another quirk of the old 10-band MXR is that the gyrator resistors to ground are quite low in value,
these also affect the response in an undesirable way.)

Can you expand on this?

(...)the 4kHz, 2kHz and 1kHz circuits are identical

Both items look OK to me.

Yes, I noticed that. Tho we don't seem to agree... If the gyrators have the same value, with the same pot value controling cut/boost, they affect the same band, no?
Fix_Metal noted that he measured the pots at 14k to 22k. I haven't calculated the freqs in regards to those values, I just check the freqs against the 1rst 10-band schematic posted by George Giblet and went with those values (you'll notice both schematic share the same values in the gyrator, except the 62.5 band with use a 120k res in one and 130k in the other, and 500 band that says 100n in one and 10n in the other. I suspect the latter is a typo. I'll recalculate them myself anyway before I commit this to perfboard. Just to make sure).

@ Frank - I think that, in an ideal world(...)

That's what I thought. I was wondering about band overlap, but then again, I might not be able to tell the difference by ear anyway. Good point.

[R O Tiree]

The pot values don't make any difference to the extreme gain/cut or the centre-freq of the peak/notch. They do, however, make a difference to the width of the peak/notch, so adjacent bands interact more with each other, or less... and they also affect the apparent "linearity" that our ears perceive. So, to paraphrase what Rob said above:

Small value lin pots = more linear-sounding response but more interaction.
Large value lin pots = less interaction but cramped response.

And you're right - 3 identical gyrators would all affect the same band. I'm not sure how that would work, though... I'm going to fire up my sim and find out.

[R O Tiree]

It appears that 2 identical gyrators gives around double the gain in that band. Three gives a little under 3x. Not sure why that would be so... maybe diminishing returns?

I've also been playing around with Jack Orman's Gyrator Calculator. The values specified in George Giblet's schem give centre-freqs within a gnat's doo-dahs of where we want them (plus or minus a very few percent) with Q values of about 2... around about 2/3 of an octave, then, which also seems reasonable. I ran my sim with pot values of 15k, 2k and 1k (I got bored with intermediate values... taking a long time to crunch the numbers). Here's the result:



I did the tests at 500Hz, as that's in the middle of the array, which struck me as a reasonable way to go, so there would be equal interactions either side of it from the other gyrators.

Note that the gain figures on the Y-axis are plotted on a log scale... with 15k lin pots, they're massively cramped, even on a log scale! No wonder they had to go with a weird taper... With 1k lin pots, however, the line is pretty much straight, which means that, since our ears are pretty much logarithmic in response, that's going to be quite good.

Despite someone at the other forum stating categorically (and I believed him) that pot size does not affect gain, it appears that it does... looking at commonly-available values, it seems to hit a minimum at 5k and increases either side of that. The increase with 2k and 1k pots is marginal (2.5% and 8% respectively) and about the same increase when you move to a weird value like 15k. Going insane and putting 2M pots in there gives almost a 100% increase... that's just getting silly, though.

I worked out that I could "normalise" the gain response by changing the value of the 2k4 at R6 in the Giblet schem so the output with 1k lin pots matched that with 15k pots. I then compared both schems with 3 adjacent pots maxed (500Hz and the 2 either side of it). There was only an 18% increase in gain with 1k pots compared with 15k ones - gain in that state was 5.165 with 15k and 6.125 with 1k. If my sim is telling the truth (and if I did my sums correctly), then I think I could live with that increased level of interaction if it meant that I could use readily available pots?

All of that said, it's only a simulation. There might be something that the sim is not quite modelling properly.

[strungout]

Whoa...

I'm not sure I understood all of it perfectly, but that seems like a lot of work on your part, so thanks!

What I got is that the 10k pot I'm using will look more like the 15k line, with the almost-vertical part of the curve being that "cramped" section that would make it harder to dial in the boost/cut (since the changes happening there are bunched up on a few degrees of rotation of the pot). So, I'd have a more even response in the rotation of my pot with a 1kB pot, if I find that the higher gain/band interaction doesn't make much difference to me.


I have part of the circuit on my breadboard -- input, boost/cut stage, output + 2 bands (500 and 2k) -- just to test it for running it on single supply (I'm not totally sure which points to connect to Vref). I'm getting some unexpected results when I cut. It seems to cut much more than the assigned band. Could this be just because I don't have all bands on it? Or maybe it's just the points I connected to Vref. Normally I should connect the RC filter of the feedback loop of the input stage, the resistors coming off the non-inverting input to Vref, right?

I'll check my components, connections again...

[slacker]

For single supply R1, R10 and all the resistors in that position for the other gyrators should be connected to vref. C3, C5 and R28 can be connected to either vref or ground it shouldn't matter which.

[R O Tiree]

I'd try it out with two 9V batteries "as stock" first to make sure it's working, then mod it. That said, the gain gets up to about 5 or 6 when you have 2 or 3 bands up high - that means even a 0.5V amplitude signal is going to clip like a bastid.

How about using a charge pump to give you dual supplies?

[strungout]

So, I spent some time on this today. I've built the rest of the circuit on my breadboard and am running it on 16.92v single supply. I have this unregulated multi-voltage adapter. The version that I've seen that works on the highest voltage is 15-0-15. The newest version of the schematic says it runs on 18vdc single supply (posted of same forum, different thread). Figured it's close enough. I've check and rechecked my connections and found a wrong value here and an unground IC there... I made a list of all my voltages at the ICs and everything looks fine, but I dunno.

That I can't really see much of a difference at 8kHz and at the 16kHz shelving is probable. I barely see one at 4kHz. Just enough that I know I'm not imagining it! But not seeing a difference at 31.2Hz and 2kHz leaves me wracking my brains... I've tried changing the pots for ones that were part of a gyrator I knew worked. I moved ICs around in case it was my breadboard getting funky... nothing. Maybe 12db is not enough gain for my ears, at least for those frequencies ( I did use my 'good' amp, Peavy Bandit 112, with the presence and highest gain setting maxed). Maybe it needs a lower Q value (lower = wider right?).

In the meanwhile, I'll try out some other values and try to see if there'S some other frency near the ones I mentioned that is gonna stand out more to me. That and build my regulated PSU... Could the fact that it isn't running on a +/- supply be that important? I don't have a charge pump handy but I have everything (I think) for my PSU. Including some monster caps (lol)...

Just in case:


(Note refer to breadboard layout).


Vcc 16.92v
Vref 8.46v
Ground 0.00v


IC1- OPA2134 (input/cut-boost)
1- 8.43
2- 8.42
3- 8.30
4- 0.0013

5- 8.41
6- 8.43
7- 8.44
8- 16.88


IC2- NE5532 (500Hz/Output)
1- 8.42
2- 8.38
3- 8.34
4- 0.0024

5- 8.39
6- 8.41
7- 8.39
8- 16.80


IC3- NE5532 (2kHz/4kHz)
1- 8.40
2- 8.40
3- 8.36
4- 0.0022

5- 8.41
6- 8.43
7- 8.42
8- 16.82


IC4- NE5532 (62.5Hz/125Hz)
1- 8.39
2- 8.38
3- 8.26
4- 0.0023

5- 8.31
6- 8.40
7- 8.39
8- 16.83


IC5- NE5532 (250Hz/1kHz)
1- 8.40
2- 8.37
3- 8.28
4- 0.0032

5- 8.24
6- 8.30
7- 8.29
8- 16.67


IC6- NE5532 (8kHz/31.2)
1- 8.27
2- 8.28
3- 8.25
4- 0.0025

5- 8.20
6- 8.28
7- 8.28
8- 16.61


Can barely tell that something's happening @ 31.2Hz (even with my 8 stringer). More @ 4kHz.  @ 8kHz and 16kHz, can't really say for sure.


Freqs (from component values) and Qs. I used the AMZ calculator.

32.3Hz         2.23
63.35Hz       2.43
127.9Hz       2.78
241.09Hz     2.07
490.32Hz     2.09
1028.71Hz   2.24
1981.32Hz   2.51
3874.54Hz   2.65
7781.31Hz   1.98
     

[strungout]

I've been experimenting with the circuit, raising the Q, the gain, made some frequencies parametric, etc. with varying success. A higher Q (ex. 3.41 @ 30Hz instead of 2.23) does make the selected frequency stand out more. A gain knob is pretty useful when notching. I used a 10k pot in the feedback loop of the cut/boost stage. At max gain and boos, things get clipped but that's a useful option to have. Notches get deeper. Parametric bands were fun but, in the end, I think I'll build this as is, plus the gain knob and a shelving switch for the 31.2 band. I wanted to check something with you guys:  basically, all I would need to do to have a low shelving is reverse the high shelf? Ie. res>cap>vref? Of course change the values...

EDIT: I ended up back on this page http://sound.westhost.com/project28.htm I'll do that, just add the switch to short the first cap from the wiper.

After I build the power supply, I'll build a parametric eq and play with both. I can figure out what I like and make a hybrid EQ...

[R O Tiree]

OK, a 6-string guitar tuned "normally" has it's low E string at about 82Hz. Top E string open is about 330Hz. Fret 24 on the top E string is about 1318Hz. Those are the "fundamental" frequencies but there are a lot of harmonics inherent in the guitar body, etc, which make a guitar sound like a guitar, not a trumpet or a flute, for example, but the strongest frequencies are the fundamentals.

This is why you're not hearing much effect much above 4kHz and not a lot in the 31.2Hz band either. If you relax the Q values, then there will be even more interaction between bands than there already is. Admittedly, an 8-string takes you down almost another octave to F# but 32Hz is almost half an octave below that F#, so that slider's running out of puff before you start...

If I was designing an Eq for guitar, I'd start at 41Hz, 82Hz, 165Hz, 330Hz, 660Hz, 1320Hz, 2640Hz, 5280Hz and 10kHz and leave it at that, with Q values at 2 as near as I could get them. Let's face it, most guitar amp speakers would seem to roll off at about 5kHz, so anything above 10kHz is wasted effort, isn't it?

Here's the maths:

Define "C1" as the bigger cap and "C2" as the smaller one.
Define "R1" as the bigger resistor (to GND in the schem) and "R2" as the smaller one (from opamp output to +ve input)

So, looking at the 31.2Hz gyrator in the Giblet schem:

C1=4.7µF
C2=100nF
R1=100k
R2=470R

The AMZ calculator takes the component values and solves for F and Q. I have re-arranged the formulae so we can tweak it...

Choose C1>>C2 (by about a factor of 50 to 100 in order to get higher R1 values as Rob Strand suggested above).

F is the centre freq of the band we're calculating.

Then:

R2 = 1 / (F * 2 * pi * Q * C1)

R1 = Q^2 * C1/C2 * R2

This will give you stunningly accurate values for R1 and R2. Bear in mind what we discussed above about capacitor tolerances, though (+/- 10% usually), so choose a close match for R1 and R2 from the E96 resistor values in this table, then plug those values into the AMZ calculator to see if the F and Q figures have drifted too far away from your design figures. Rinse and repeat for all the bands.

Hope this helps.

[strungout]

Hope this helps.

It does, thanks! I mostly suck at math because I have a shitty memory. I had no idea how to re-arrange the terms of the equation to use difference values. But I can follow directions!

You said "If I was designing an EQ for guitar (...)", it's dead on: this EQ is designed for guitar AND bass, at the very least. They had to compromise. A low B on a 5 string bass is around 31Hz. My 8 string is tuned GBEADgbe. Honestly, I haven't played with the 8th string much (had it for 3 months. I mostly chug on it and F# with .74 felt a bit too sloppy. Gauge, another thing I have to mess with...), so that could change. When I bought my RG321, I had first wanted to get a 7321. I kinda regretted it, so I figured this time I'd get the 8th!

The frequencies you provided might be better suited, not sure if I can test it right now tho, some of the values I ended up with aren't in my bins. I can put some in series and get them. I'll check this out today. That's a good thing about breadboards, have lots of room to series/parallel components.

I'm glad I decided I was too tired to build it the night before I read your post!

Here's the values I came up with:

Band     R1       R2    C1     C2      Hz     Q

41Hz     36000    820   2.2    0.22    42     2.1
 
82Hz     18000    430   2.2    0.22    82     2.05

165Hz    20000    470   1      0.1     164    2.06

330Hz    10000    240   1      0.1     325    2.04

660Hz    1000     120   0.82   0.47    662    1.95

1320Hz   11000    270   0.22   0.022   1328   2.02

2640Hz   36000    910   0.033  0.0033  2666   1.99

5280Hz   6000     150   0.1    0.01    5308   2

10kHz    15000    360   0.022  0.0022  9850   2.13

I stuck to 10% and 5% tolerances, cause I'm pretty sure my local electronics shop doesn't have smaller. But when I get some money I can order some through mouser, digikey, etc. The values do come out to be very precise. When you pick the right caps.

I've started another list, mostly through trial and error using the gyrator calculator, to see just how close I can get and the Q moves around more but it's using 10% tolerances. Not finished.
All that matters is Hz and Q, since were simulating Henrys, right?

[R O Tiree]

Try a wider split between C1 and C2. You need C1 at least 50 times bigger than C2. You've gone with 10x except for the 660Hz band, where it's less than 2x (which is why your R1 value is only 1k). Rob Strand mentioned at Reply#1 that lower values for R1 give some quirky effects.

[strungout]

Augh, that didn't get through completely. Back to the drawing board!

[strungout]

Ok, take two... I was so focused on the hard stuff, I forgot about the simple advice!

Band     R1     R2    C1     C2     Hz     Q   

41Hz     330k   430r  4.7u   22n    42     1.90

82Hz     180k   430r  2.2u   22n    82     2.05

165Hz    240k   470r  1u     8.2n   166    2.05

330Hz    120k   240r  1u     8.2n   328    2.02

660Hz    100k   150r  820n   4.7n   662    1.95

1320Hz   300k   180r  33n    820p   1317   2.04

2640Hz   150k   360r  82n    820p   2643   2.04

5280Hz   75k    180r  82n    820p   5285   2.04

10kHz    390k   240r  33n    82p    10000  2.01

It's late...

[PRR]

Maybe I'm over-simplified, but....

Can't you compute ONE filter, correct Q, any nominal frequency?

Then for other frequencies, leave the resistors alone, multiply/divide the caps to change frequency.

Band     R1     R2      C1     C2      Q  
330Hz    120k   240r    1u     8.2n    2.02 - Reference design
Octaves up:                        
660Hz    120k   240r    0.5u   4.1n    2.02
1320Hz   120k   240r    0.22u  2.05n   2.02
....
10KHz    120K   240r    0.03u  0.27n   2.02
Octaves down:
165Hz    120k   240r    2u     16.4n   2.02
82Hz     120k   240r    4u     32.8n   2.02
.....

Now step back and ask if these values are reasonable.

240 ohms is on the low side for common opamps.

8uFd (for 41hz) is kinda huge.

120K is not-large for FET-input opamps.

This suggests the impedance is generally rather low.

Re-try with 240K+480r or 480K+1K.

The 480K+1K leads to a 67pFd in the 10KHz band, which is rather high (you always have several pFd of sneak capactitance to unknown and maybe bad places). So that's rather high impedance.

Numbers like 220K and 470r may be a sweet zone.

Now you only have to buy two lots of good resistors.

And to some degree (assuming mostly octave spacings) you can series/parallel capacitors. Five all-the-same may be cheaper than three different odd values. With a bag of 1nFd you can make 2nFd 1nFd and 0.5nFd. (And they give nominal 2:1 ratios, not 2.2:1 or 1:0.47=2.12:1 of preferred values.)