Need help with TH Custom Effects 3band parametric EQ

[FantomXR]

Hey people,

I've designed a PCB based on the schematic from TH Custom effects. Here it is:


I've checked thousand times if there is any difference between what I draw and what TH Custom Effects schematic says. Apart from the values for some caps and resistors I can not find anything!

So whats wrong with it? I have huuuuuge (white) noise on the output and the signal / noise ratio is almost 1:1. Very bad... also there is a very high frequency which goes up and down overlayed on the output signal.

For a test I removed everything but the high-band as well as the IC1 which is the input and output buffer. Same behavior. I removed the ICL7660, jumpered it and powered everything from 9V directly. Same behavior. What do I overlook?!

Here is also the board-layout:

There is a ground-plane on the back.

Any ideas?

Thanks,
Chris

[FantomXR]

Apart from that I have a very loud noise when switching the device on and off. Here is an audio-sample which also shows the signal / noise ratio:
https://www.dropbox.com/s/15734lfng1dvt4c/EQ_Test03.mp3?dl=0

PLEASE TURN DOWN YOUR SPEAKERS! It's very loud at the beginning....

[ElectricDruid]

Could we have a more readable copy of the schematic please?

I can't see the values on that one, or some of the tracks.

One thing I can't see is an electrolytic capacitor on the Vref midpoint supply/virtual ground.

Thanks,
Tom

[FantomXR]

Oh, sure! Please excuse me.

Here is a high-res picture of the layout where you can also read the net-names. I uploaded it to dropbox to prevent downscaling (which happens in my initial post):

https://www.dropbox.com/s/lg52t046h9akjs4/Rhodes_EQ_brd.png?dl=0

And here is the schematic:
https://www.dropbox.com/s/sd5ip05ao3gls57/Rhodes%20EQ.pdf?dl=0

I have added a electrolytic cap between VREF and GND. Still very noisy...

[antonis]

Still difficult to trace 'cause there are some discrepancies between Schematic & PCB..
(like Q pots shown as single resistors on schematic although they actually are wired as variable ones on PCB BUT they go to Gain pots middle lugs, via respective caps, where Freq pots should go..!!)

If it's just PCB items labeling typo, it's OK..

[FantomXR]

Both schematic and PCB are valid. PCB design is directly connected to the schematic in eagle. I can not make a change in one part without affecting the other.

Where do you see Q pots as single resistors? Both schematics and PCB have a trim-pot installed.

Please see here for the original schematic:
http://diy.thcustom.com/?wpdmdl=4743

[antonis]

Where do you see Q pots as single resistors? Both schematics and PCB have a trim-pot installed.
Please see here for the original schematic:
http://diy.thcustom.com/?wpdmdl=4743

Thanks but I'm aware of the "original" schematic..
(which has nothing to add on any conventional xyz band fully parametric EQ..)

If you, or any other guy, can see Q pots wired as variable resistors on your particular schematic, it's more than fine with me..

P.S.
If you also insist that Q pots MUST be connected (via a single cap) to Gain pots middle lugs (as it is clearly traced so on your PCB), it's again fine with me..

[FantomXR]

Please don't get me wrong. I'm happy to receive comments on that and I really appreciate your help.
But I think I still don't understand what you mean.

Isn't that a variable resistor (screenshot taken from the schematic):



I do not insist on something. If you have another idea or suggestion I'm happy to hear it. I just want to get it to work :-D

[niektb]

I see you only bypass your power supplies using large elco's, have you tried adding 100nF's across the power pins of every opamp? (and placed very close to those pins)

[FantomXR]

I skipped this simple rule for some reason :-( I will add those caps tomorrow and report back... although I can't imagine that this huge noise comes from missing bypass caps... but yeah, could be ;-)

[antonis]

Isn't that a variable resistor (screenshot taken from the schematic):

Of course it is ..!!

P.S.
My bad, on your schematic (the blurred one) I was only able to read a "Q" (which is the last letter of FREQ) and I wanted to point you both "false" naming and connection..

My apologies..

[ElectricDruid]

I haven't managed to spot anything wildly wrong or different between your schematic and the original. So I think the circuit is ok.

That said, years ago I tried building a EQ like this (same design, gyrators, fixed Q, 6 bands). I used a hand-drawn and homemade PCB. I never got the thing to work, and it also had noise problems. So I suspect that gyrator designs are a bit sensitive.

I've got lots of other comments about your PCB, but I'm afraid I don't think they'll fix your noise problem. Here goes:

One small thing is that C9 and C1, both large electrolytics, are in parallel. You don't need them both, so you can dump one of them.

There are a lot of tiny vias, which shouldn't be necessary on a reasonable simple circuit like this, and where you have plenty of space. Why so tiny? You've got space, so make them a bit larger. They'll be more reliable, and if they're not, you'll have space to poke a bit of wire through and solder it on both sides.
The Vref, for example, is created at the top of the board by R1/R2, then goes down a via, across the bottom for a tiny distance, then back up a via again to the rest of the Vref track. Why not use the two resistors to bridge over those tracks instead?

You've laid out the op-amps with the centrelines all vertical. If you rotated them 90 degrees so the centrelines were horizontal, you could run the power rails straight down the middle of the whole lot. That saves you all those little blue "droppers" (and more vias) from the 18V rail.

Finally, you've got several vias underneath other components. That makes it difficult to inspect them/check them or fix them if necessary.

Would a better board layout fix the noise problems? I don't know. Probably not, but then again, it might help.

[FantomXR]

Hi,

thanks for your comments.
I have absolutely no problems with via (it's only 14pc). The GND plane on the back is big enough to save the circuit from noise. Of course I could avoid them. But yeah... I don't think it's so super important in that case.
I make vias as small as the manufacturer allows. In general if I would need to make such an adjustment, I could also stick a wire at the source or the other side where the signal ends.

I don't want to use 0-Ohm-Resistors as bridges, because it needs time to place them... :-)

Since the PCB was made by JLCPCB and I don't need to make the PCB by myself, those design-tipps I of course appreciate. But since there is either much current going nor high voltage, nor fast signals, it's very uncritical as I discussed with another guy already. I've made tons of digital and analog PCBs already and I never had a case where PCB layout is such critical.

But again: Thanks for the tipps. It's always good to receive feedback from others and see how they would do it.

And back to topic: Do you have another idea for an EQ instead of gryators?

Honestly I don't need an adjustable frequency-band nor q-factor. If I set it one time, I don't need to change it anymore. So for me a set-and-forget-schematic would do it also.

Thanks again!

Best,
Chris

[ElectricDruid]

I make vias as small as the manufacturer allows.

Why push manufacturing tolerances if there's no need?

I don't want to use 0-Ohm-Resistors as bridges, because it needs time to place them... :-)

No, I meant the existing R1/R2 resistors. Those are your bridges.

I've made tons of digital and analog PCBs already and I never had a case where PCB layout is such critical.

Fair enough, neither have I. You might be looking at the first case. I doubt it too, though!

Do you have another idea for an EQ instead of gyrators?

Honestly I don't need an adjustable frequency-band nor q-factor. If I set it one time, I don't need to change it anymore.

Ok, how about this?:

http://sim.okawa-denshi.jp/en/OPtazyuBakeisan.htm

Use a MFB bandpass for each band. You can set the Q and Frequency you need and off you go (second form down the page). I recommend running a few iterations. Put the Q and Freq in first, and see what it gives you in terms of R and C values. Then tweak those values and put them back in to make sure you still get close to what you need. Rinse and repeat until you're happy.

HTH,
Tom

[FantomXR]

Why push manufacturing tolerances if there's no need?

Because it never was a problem. :-)

No, I meant the existing R1/R2 resistors. Those are your bridges.

Yes. You are right about it. This I will change for sure.

Fair enough, neither have I. You might be looking at the first case. I doubt it too, though!

Ha, maybe...

Ok, how about this?:

http://sim.okawa-denshi.jp/en/OPtazyuBakeisan.htm

Use a MFB bandpass for each band. You can set the Q and Frequency you need and off you go (second form down the page). I recommend running a few iterations. Put the Q and Freq in first, and see what it gives you in terms of R and C values. Then tweak those values and put them back in to make sure you still get close to what you need. Rinse and repeat until you're happy.

Thanks for the link. Looks good for cutting. But it can not boost as far as I can see.

[ElectricDruid]

Thanks for the link. Looks good for cutting. But it can not boost as far as I can see.

The Boost/cut is done by the amp at the top, not the gyrator or whatever else you're using for each band.

That said, I did get confused. The gyrators used in that EQ produce a notch response, not a bandpass (so the other way up). Plus a gyrator only has one "end" - it doesn't have an input and an output. So it's not immediately clear how you'd graft a bandpass filter in there instead. However, it is possible. I've pasted an example below. At this point, that's a completely different circuit though, so it's not exactly fixing the problem, just replacing it with something else!

[antonis]

So it's not immediately clear how you'd graft a bandpass filter in there instead.

It could be done, in case of relatively narrow band margins, by setting two neighbour frequencies (e.g. Low pot full CW & MId pot full CCW) with also Q pots set for the lowest value..
(well, not strictly an "ideal" bandpass but close enough..)

[FantomXR]

Hm...

I really don't know how to proceed on this.
It couldn't be that hard to design an EQ. I found that one:
https://www.elcircuit.com/2018/08/10-channel-equalizer-using-4558-ic.html

Maybe it's worth a try. The 4558 could be replaced by a 5532. I don't have 4558 in stock. But a bunch of NE5532. I could breadboard it an try it...

[FantomXR]

I just saw that it's almost the same circuit as the one from my initial posting :-D

There is one major difference though. Where my frequency-potentiometer connects on one side to the virtual ground, in this schematic it connects to "physical" ground....

[antonis]

Of course it does, due to bipolar power supply..
(no need for Vref..)

Any general purpose op-amp (either of BJT or FET inputs) in place of 4558 can do the job..

Note 1: I shouldn't utilize 1k input impedance if I wasn't sure for signal source low enough output one..
Note 2: I should either reverse LED Anode/Cathode or connect R29 to negative supply..