GE-7 op-amp replacement

Started by kashmir, September 05, 2004, 07:11:40 PM

Previous topic - Next topic

kashmir

:?:

I have the newer model. It´s equipped with three JRC022D chips and two NJM/JRC2068 8-pin SIPs. I thought I could substitute the JRC022D chips for something better. Any ideas? TL072 maybe?

petemoore

Noise was the main complaint I had with GE7, which otherwise is a very nice and useful pedal to have.
 Would be interesting to see if one could get the noise floor down on this pedal. I would rather have this one as a Boss, those sliders make voice shaping a breeze, and would be very hard to do DIY.
 I'm thinking a good portion of the noise in graphics is caused by the # of passive components.
Convention creates following, following creates convention.

cd

There's a factory low noise mod for this pedal: there should be a 270pf cap across pins 6 and 7 of the JRC4558 chip.

Dirk_Hendrik

Apart from using low noise opamps a good way to get the noise down is by replacing all resistors with their metal film equivalent. While doing this it is worth recalculating the values of the 5% tolerance resistors in the gyratorcircuits to their exact value in the 1% tolerance range. By doing this the EQ will start to sound more musical as the octave spacing between the bands becomes more exact.
More stuff, less fear, less  hassle and less censoring? How 'bout it??. To discuss what YOU want to discuss instead of what others decide for you. It's possible...

But not at diystompboxes.com...... regrettably

cd

Quote from: Dirk_HendrikWhile doing this it is worth recalculating the values of the 5% tolerance resistors in the gyratorcircuits to their exact value in the 1% tolerance range. By doing this the EQ will start to sound more musical as the octave spacing between the bands becomes more exact.

Has anyone out there tried this mod?  Just got a GE-7 to futz around with, and for S&Gs I measured all the parts in the gyrators exactly (caps, resistors) and was surprised to see the frequency bands were way, way off!

100hz actually 118
200hz actually 243
400hz actually 509
800hz actually 983
1600hz actually 1986
3200hz actually 4222
6400hz actually ? didn't get the formula right for this one (no gyrator!  Assuming it's the usual 2piRC formula, that makes it 4131Hz according to the factory schem? anyway)

The formula I used was 1/2*PI*[SQRT(C1*C2*R1*R2)].  I recalculated the resistor values for 1% tolerance to get the frequency ranges back to where they're supposed to be, but will have to wait a while to get them in (mail order).  Then again I'm thinking my multimeter's cap measurement function is probably not all that exact - if it's off by a 10% range (I can't remember how exact it's supposed to be) that throws everything off as well.

mrsage

Quote from: kashmirTL072 maybe?
That'll help a lot.

TheBigMan

Yep, TL072 or NE5532s.  Mouser or Digikey has NE5532s in 8 pin SIP as well, so if the pinout matches you can replace all the opamps.  Might bump your current draw a bit with 5 NE5532s in there though.

cd

Forgot to mention I also replaced all the DIP opamps with NE5532s (replaced one 4558DD and three TL022s).  I didn't think it would make a whole lot of difference with the gyrators, since they're just simulated inductors, and I was right.  The noise is reduced slightly, but by design the pedal will be noisy, especially if you crank a single band all the way to +15dB.

Paul Perry (Frostwave)

[quote="cd100hz actually 118
200hz actually 243
400hz actually 509
800hz actually 983
1600hz actually 1986
3200hz actually 4222
6400hz actually ? didn't get the formula right for this one (no gyrator!  Assuming it's the usual 2piRC formula, that makes it 4131Hz according to the factory schem? anyway).[/quote]

Except for the last one, the frequencies are pretty close to exactly an octave apart. Maybe the input impedances of the op amps are giving some error?
When RC determined frequencies have to be spot on (eg in an analog vocoder) some say to measure the actual value of the capacitance & then have one or more resistors at 1%, and then calculate the value of the remaining resistor, and cobble the resulting oddball value from a parallel combo of standard 12% resistors.
There's another thing as well: does an oscillating filter actually oscillate right at the peak of the bandpass curve? Or does the 'peak' of the curve shift sideways somewhat as the resonance increases?

cd

Quote from: Paul Perry (Frostwave)Except for the last one, the frequencies are pretty close to exactly an octave apart. Maybe the input impedances of the op amps are giving some error?

I measured things again without the opamps in, and had to recalculate things for different cap values... then again, I'm pretty sure my meter is only +/-5% accurate:

92
189
392
723
1318
2483

Again mostly an octave apart, but 2483 ain't anywhere close to 3200!!  I'm going to do some white noise tests and see what happens.

Quote from: Paul Perry (Frostwave)
There's another thing as well: does an oscillating filter actually oscillate right at the peak of the bandpass curve? Or does the 'peak' of the curve shift sideways somewhat as the resonance increases?

That's what I was thinking.  How close do you have to be?  2483 is far from 3200, but if we're talking the difference between say 3150 and 3200, I doubt I could tell the difference.  I'm going to run some computer simulations until I get the exact parts in and see what happens.

cd

Alright, the results of the caps are in:



I fed a white noise source through my modded GE-7 (see above) to test whether the frequency bands were really messed up or not.  I used a couple of laptops and Adobe Audition's Frequency Analyzer.  You can draw your own conclusions, but I figure my cheap-ass meter is no good.  In the left hand side column is the white noise with each slider on the GE-7 down (frequency CUT).  The 100, 200, 400, and 800Hz bands are fuzzy, but you can definitely see a reduction in frequency there.  Now the 1600 and 3200 bands are interesting: the 1600 labelled band is actually very close to 1600 (approx. 1570-1640) however, the 3200 labelled band is more like 3400Hz - 3200hz falls way to the left of the center of the "dip".  The last one is the 6400hz cut, and as expected, it's just a simple high cut filter.  BTW the x-axis shows 0-11Khz in log form.

On the right is the sliders up for a frequency BOOST.  Again, the 100-800 ranges are fuzzy, but the 1600 and 3200Hz labelled bands are clear... and as expected, they line up perfectly with the "cut" ranges.  1600 is close (1550-1670) while the 3200 is way off (3400-3600).

It's not shown on the graphs, but the last slider (+/-15dB overall gain) is pretty close - the y-axis on the graphs each represent 12dB of volume.

On the whole, IMHO everything is close enough for rock n roll... I bet everyone here can tell the difference between a guitar note at 3200Hz and 3400Hz, but can you tell the difference between a boost/cut at that frequency?  If you can, congratulations, you have better ears than me!!