GEOFEX Simple Easy Graphic Equaliser PCB Layout

[Brymus]

No that didnt work 

[Brymus]

OK YAY It works it is verified !!!!!
It will not work with a TL072 so I am guessing it has something to do with the FET input.
A NE5532 works as does a RC4558,both w/o having to removing lug 1 or 3 and plug it back in.
The problem does persist with every TL072 I have (4 total)
SO the layout is verified as good with a dual op amp like the NE5532 or RC4558
I should note it has alot of gain,so a volume out could be useful but proper adjustment is a better choice.
And the layout is cramped,it was way harder to populate and solder than my Dr Boogie build.
But I would almost bet it wouldve worked fine with the 5k6 resistors in there...
I am gonna mess with it some more,but Rick's layout IS good and verified at this point.
Thanks Rick

[frequencycentral]

Phew! That's is such a relief Bryan, this has been like watching a thriller/detective movie! When you've finished messing post your results and I'll amend the layout accordingly. I'm thinking that your idea for a volume pot instead of R27 is a good one, as dinking with the frequencies could require a tweak to get unity.

And the layout is cramped,it was way harder to populate and solder than my Dr Boogie build.

All my layouts are that cramped.  

[Brymus]

Thanks again Rick.
I sent you a PM but yeah the volume out is a good idea,I think there is enough room already to have so people could build with or w/o a volume pot and throw the pull down resistor in there if they didnt want one.
I wonder if it needs a load resistor before the 1uf output cap ?
would a 1k resistor before C18 be a good idea?
Then have either a 10/100k pot or pulldown resistor?
I dont know enough about this stuff yet to say...

[slotbot]

No that didnt work  

ah sorry my bad.  


you might be able to get it to work with TL072 if you lower that 1meg.

i think that this resistor should be sized to the (-) input resitor||the feedback resitor to try to get rid of errors caused by offset currents. (assuming thats the issue).

so in this case i think it would be around 75k.... but maybe i will try it myself since you already did so much work 

[Brymus]

I tried a TL082 and it had the same issue that worked if you did the same unhook/re-connect trick.
As is I like the sound of a MC4558CN the best (low noise ,wide bandwidth) The TLC2272IP (1P?) sounded good too,another low noise dual opamp.
Followed by the NE5532 and RC4558
I'll post some pics later.
And Rick is working on a refined layout that looks really sweet !!!

Could anyone comment on what effect changing the 6k8s for 5k6s would have ?
I am also curious if there would be any advantage/disadvantage to using say 4401s or 3904s in place of the 5088s or if higher gain would be better ?
And finally why/what exactly is happening with the FET input opamps in this usage ?
Maybe PRR or RG ,anyone could explain this as I really want to understand this better.
Thanks Bryan

[slotbot]

what is the voltage at the + input side of the 1meg with one of the op amps that works?

[head_spaz]

Keep in mind that the NE5532 has a much lower output impedance... more drive capacity... which is REQUIRED because all of those pots are hooked in parallel.

[R.G.]

Sorry I got to this late. At least there still some mystery left.

Overall, the opamp should bias up through the 1M to Vbias. All of the active pins - both + and - inputs and output - should be within about 15mV of each other and the Vbias voltage. If that's not true, then something is wrong with the biasing or connections.

As a general rule for debugging, if you're looking at DC conditions, you can open any capacitor and not change what the circuit should be doing, because capacitors should be open circuits at DC.

We have a situation where the DC conditions on the opamp seem to be going south with a TL07x but not a bipolar opamp. So let's do some divide and conquer.

If it were me debugging this, I would temporarily remove (or lift one leg of) all the C1s and the input and output capacitors, and see if that changed anything. It should not. If it does, there is a problem with one of the capacitors or their soldering. If it doesn't, it clarifies that the problem has to do with the DC-connected resistors remaining. It's obvious that there is something different going on with the DC levels on the JFET input opamp.

That is downright strange, because the JFET opamp should be much less prone to DC loading issues no matter what levels of resistance are connected to it. In terms of DC, the 7x 100K pots have their wipers blocked by caps. You should be able to replace all seven of them with a 14K resistor and have the same DC levels. Slotbot and head_spaz have some good ideas about what to do/look at. If the preceding tests turn up nothing, I'd try lowering the value of the 1M bias resistor, then reading the TLO7x datasheets.

It may be a quirk of the TL07x devices. It is probably not the high impedance or low drive of the devices. High impedance is eliminated because the TLC2272 is a CMOS input device, even higher input impedance than JFETs. TL07x devices happily drive loads down to 2K, and the paralleled pots only get down to 14K at DC, much higher.

It *may* be a power supply decoupling problem or a layout related stability issue. I haven't checked the layout, but I would stick a 0.1uF from V+ to V- as close to the opamp pins as I could get them. Have you put a 'scope on it if you have one?

[PRR]

> Could I use 5k6 or 10k  in place of the 6k8 ? It looks like a simple voltage divider or would I have to adjust the 7 x 33k resistors too?

It isn't a voltage divider. I need more coffee to remember which-does-what, but a slightly different R will give a slightly different frequency, which isn't a real problem.

> shouldnt there be a bigger difference between the base and emiter voltages ?

0.6V between Base and Emitter. Maybe 0.5V, maybe 0.7V, depending how big the transistor die and how much current. Your 0.7V is not "wrong".

> theres hardly any sound
> {later} it has alot of gain

With pots centered, it should be UNITY gain.

> do you think input /output buffers would be advantages to the circuit ?

Assuming your source can drive a 150K load, there is little point in an input buffer. (Don't over-complicate.) The output "is" a buffer: the raw opamp output (with DC-block cap).

The Vref buffer or a BFC on passive Vref "may" be a wise idea. The scheme sure will work without, but there's a half-dozen hi-Q tuned circuits dumping different frequencies into "gyrator ground". It will work with all that interaction, but you won't get the exact F and Q you calculated (not that you ever do!).

That said.... EQs tend to hiss, and the high 150K impedance here is significant, whereas guitar-cord signals are small. When you have it working right, if hiss is an issue, you might want to complicate with an input booster, lower-Z EQ networks, and output attenuation.

But unless you have a solder-blot or we are all missing a layout error: It should work very well with BiFET opamps.

I do not like the 1.4V you were measuring on opamp pins.... that should not be.

R27 just bleeds stray charge from the output cap so you don't get a POP. It should not load-down the opamp or cap. Opamp can drive 2K without distress. 0.1u cap can drive 100K with 20Hz bass-cut. If the external load is also 100K this rises to 40Hz at -3dB or 80Hz at -1dB. OTOH if R27 is too large it take more time after turn-on for the POP to bleed off, and if C18 were a leaky electro the leakage would not bleed. So anything 100K-1Meg is probably happy.

C15 is a rail-cap. Assuming there is a battery or a big-cap in the wall wart, 10uFd here is fine. If 100u seems more common it is because 100u 10V is cheap-enuff and designers don't always use the smallest workable value. Here space-crunch suggests a smaller cap.

The 1Meg probably should be over at the other end of R26, where it meets C17. That was "wrong" in the original at 4QD-TEC. But the difference is so minor it is hardly worth changing.

[Brymus]

Thank You Paul and RG !!
I have to get some coffee and read this a couple more times but it looks like I have some work to do.
Thank you so much for answering all my questions,I really appreciate the time you guys put into helping me with this.
I'll be back...

[PRR]

> answering all my questions

I don't think "any" questions were answered.

It "should" work. And where RG crossed his fingers "may be a quirk of the TL07x", I'll go futher: the TLO7 should be fine. Possible differences: marginally higher offset against the quite-high DC gain, power rail bypassing, and marginal layout (it's a big hi-Z circuit with MHz parasitics).

> get some coffee

That's first. I also have to de-gunk a lawnmower and box some beams. Let it simmer.

[Brymus]

OK after a little board hacking it works...even with the TL072 
Unfortunately I didnt feel like spending the time to do each change and test so I did both at once,as IMO they were both (all 3) needed.
As I was dropping a half volt from my Vref to my bias voltage and it was suggested to try a lower value than 1m I made R25 120K.
Also PRR suggested moving it from its position to the C17/R26 junction and since I had to drill on the PCB and change the resistor I went ahead and did that.
I also added a 10n AC cap directly to pin 8 of the IC,RG suggested 100n but in another debug I had to go to 10n before it quit(CMOS IC) oscillating.
And it was suggested this build could be oscillating in the MHZ region so what the hay.
If someone could explain if theres any difference between using a 100n vs 10n here I might change it back if needed.
Another change I did was I had C13 and C14 swapped,so I swapped them back the way they should be while I had my iron hot,I doubt that had any impact on the issues at hand.But its now correct.

Here are some voltage readings,note I charged my battery pack back up so my V is now a little higher.
IC V in pin 8 = 10.62v pins 5,6,7 = 5.30v pin 4 = .2mv pin 3 = 5.16v pins 1,2 = 5.30
Transistors 1-7 collector = 10.61v base = 5.28v emiter = 4.72v

At this point I would just like to make sure everything is correct for anyone else building this,regardless of what layout they use.
I am going to use the new layout Rick did and draw another one to accomadate these latest changes.

So I would like to ask...Is there any advantage /disadvantage using transistors of lower / higher gain ?
As in 3904,4401,5089,ect in that would there be any advantage to making the layout big enough to use sockets ?
As is you will have to mount them high enough to clear the box caps and resistors,which is perfect if you use a heat sink on them.
There is some hiss but only with the higher bands on max,leaving 15k and 6k at less than max gain and the hiss is almost undiscernable.Still how would we remove the hiss design wise ?
Speaking of doing another layout...
Would anyone be wanting to socket their caps for changing the frequencies ?
Would anyone want more than 7 bands ?

I have a much greater appreciation at how much work goes into getting a design verified to the point its ready for cookie cutter/paint by number status.
And I had no part in the design or the first two layouts,just etching and testing, still alot of work.
here are some pics to show the compactness and new changes note the pick in the first pic >Before the latest changes you can see two caps in the wrong places.


After fixing it,you can see where I moved the bias resistor and added the filter cap >


A pic of the trace side it in pretty good shape considering how much soldering/desoldering I did the last two days >


A log view to show the mounting heights and doubled up resistors to get 6k8 LOL >

And a big thanks to everyone who helped with this >THANK YOU ,its almost done so just a little more feedback,thanks Bryan

[R.G.]

OK after a little board hacking it works...even with the TL072 

Hey, really good news. Congratulations.

As I was dropping a half volt from my Vref to my bias voltage and it was suggested to try a lower value than 1m I made R25 120K.

That's very odd. Half a volt over a 1M resistor is a current of half a microamp, or 500nA.The datasheet says these things use a max of 200 picoamps (0.2nA) at 25C, and only 7nA over the whole temp range. That means something there is stealing nearly 500nA of current. It can only be the 14K of input-to-input resistance and the offset voltage, or a flaw in the opamp or PCB. I would expect the bipolar opamps to be worse on this issue.

And again, you could easily enough drop this if you used a buffer in front of it. Is that a double or single opamp? Could a second half of a dual be used for an input buffer?

Also PRR suggested moving it from its position to the C17/R26 junction and since I had to drill on the PCB and change the resistor I went ahead and did that.

No opinion either way on that one.

I also added a 10n AC cap directly to pin 8 of the IC,RG suggested 100n but in another debug I had to go to 10n before it quit(CMOS IC) oscillating.
And it was suggested this build could be oscillating in the MHZ region so what the hay.
If someone could explain if theres any difference between using a 100n vs 10n here I might change it back if needed.

It's hard to say. For best results, the cap should be a low-ESR monolithic ceramic device. The parasitics on the cap mean that at some frequency in the low MHZ, the 100nF is likely to go resonant, then turn inductive. The 10nF should do that higher, at the cost of not being as low an impedance at lower frequencies. In critically decoupling memory arrays way back when, we used a combination of 10uF electros, 100nF ceramic and 10nF ceramics. Each chip got one of these, and we alternated values, while running the power traces parallel and on wide traces. For a single opamp package, you want the physical lead length from the V+ pin through the cap and into the V- pin to be minimized to eliminate as much trace inductance as possible. Fortunately and unfortunately, we often get away with ignoring high frequency power decoupling in effects. This makes us think we can always get away with it, which is not the case.

So I would like to ask...Is there any advantage /disadvantage using transistors of lower / higher gain ?
As in 3904,4401,5089,ect in that would there be any advantage to making the layout big enough to use sockets ?

I personally would not bother with sockets. I'd go back to Keen's second law: when in doubt, use a 2N5088. It has enough gain, voltage, and current capability to do almost any NPN job in an effect. That doesn't say the 3904 or 4401 won't work just fine. Most bipolar circuits are designed to require some minimum gain on the transistors. Above that, you get rapidly diminishing returns. Generally that number is about 100. So most signal NPNs work fine in any NPN position. The exceptions are almost all where the transistor has to withstand X current, or Y power dissipation.

how would we remove the hiss design wise ?

Go back and redesign for lower resistances. the series 150K is probably the big offender. If you dropped the impedance of every resistor by 10:1, it would probably be pretty good. But you'd then really need an input buffer to keep from overloading any guitar that by misfortune got hooked directly to this.

And a big thanks to everyone who helped with this >THANK YOU ,its almost done so just a little more feedback,thanks Bryan

You're welcome. I'm glad it's turning out well for you. 

[PRR]

While gunking the gas-tank, I wondered about the DC sensitivity.

Using R.G.'s "divide and conquer. ...remove ...all the C..s", the DC analysis reduces to this:



SPICE opamps have zero offset; I inserted battery V1 to emulate the real-world error.

The worst-spec TL072 is 10mV offset so I set V1 at 10mV.

The output error is 831mV! (Can be + or - depending on opamp error. They really do aim for zero, so the miss can fall either way.)

This can be understood as (1Meg+14K+150K)/14K = 83.14, the circuit gain referred to the 14K and the opamp inputs, times the 10mV assumed offset error.

But that is Not A Problem. We bias to "4.5V", but really we want to stay well away from the zero and +9V supplies. If "4.5V" is really 5.3V or 3.7V, the circuit still works (passes audio).

And I doubt many TL072s have the max 10mV error. I assume 5mV and seem to get <2mV on most chips. (We come a long way from having a mercury-cell on every opamp and trimming-out +/-100mV drift before every run.)

The 1.4V reading is sure like the 1Meg bias wasn't connected at ALL. The FET inputs will drift forever, or until the output jams, which IIRC is around 1.4V on the TL07s. The BJT inputs "might" leak their way into an "operating point", though I would have to see it to believe it.

The ideal gain for the transistors is simply infinity. Anything less gives lower F and Q. If you plan to make a lot of money, we can do the analysis. But you can't make a lot of money in audio, so use Keen's second law: when in doubt, use a 2N5088. Plagiarism/"research" reveals that many-many-many low-Q EQs use a transistor (not always a hi-gain one) and it works great. For hi-Q with precise action, consider opamps; but that's a lot of chips and at higher F it takes a great op-amp to beat a simple NPN.

> remove the hiss design wise ?

You can't remove hiss.

You can minimize hiss voltage with lower impedances. Make all R 10 times smaller, all C 10 times larger. That only takes hiss voltage down by factor 3.16.... well, not quite because now the TL07s hiss appears. Go another factor of 10 on impedance, now you need an amazingly low-hiss-V opamp with ability to drive less than 2K. NE5532 will do. And of course the input Z falls from ~~150K to 1.5K so you sure want to buffer it.

The other way to go is to increase the signal before going through a hissy stage. If we put a gain of 100 in front (and loss of 100 in back), the hiss appears to be 100 times less. Yay? Well, 500mV guitar times 100 is 50V signal. We are far past what a 9V batt can do.

"Serious" EQs tend to 10K impedances and 1V levels.

So indeed an input amp may be wise. It can show a nice high Z (1Meg?) to the source, and also use 22K+10K NFB network for gain of 3 (3.2) to boost-up lower guitar levels. Then the EQ has everything 10 times lower impedance. TL07 will still be fine. The output can conveniently be 22K+10K divider to get overall gain to unity. Or perhaps 220K+100K to present a 69K impedance to the next pedal (which may be happier with some source impedance, some are like that).

[Brymus]

Thanks RG for the reply and info on the 100n and 10n caps,very useful.
As well as the transistors and sockets ,I needed to know that as well.

And Paul, WOW...It amazes me how you are able to do the spice sims ,and the fact you actually take the time to do them for so many of us here in the forum,thank you for that!

And so that your effort isnt wasted let me ask,does the 14k (R3) represent the 7 100k pots in parallel ? or is that all the Rs from each gyrator and pots ?
I am almost certain the 1m was soldered properly ,I could be wrong,however it does work now with the TL072 even though the other opamps do still sound better.
Also I was curious if lowering all the Rs meant raising all the caps,(now I am sure) so now at a factor of 10 we are talking electros and slightly more board space,not really a problem.

Back to furthering the EQ,the hiss does bother me as I like the boosted upper bands, and to not get too much hiss you have to cut rather than boost so I think more work IS in order.
The input amp with a gain of 3.2 ,you are refering to is an opamp correct ?
There is enough room in a BB enclosure to lengthen the PCB for a couple more bands and another op amp.As the opamps can be orientated side ways and take up not much more room.
SO what to do with the other half of the second op amp ? Perhaps a post EQ boost ? Or an attenuating stage to remove hiss ? IDK,these are the two ideas I get from what you just posted.
And will polarity be an issue with the larger value caps needed to scale all the values ?
We are talking 10uf, 4.7uf, and 2.2uf electros to get the scale right and that would make the pots all 10K from 100k right ?
I will work on not only a layout to keep the original values,but one to add another opamp and the scaled values as well.
I should say I do like this EQ ,it could use another mid/upper band perhaps 4khz and 8khz ,adding 10 or 12 khz would then be 10 bands...
And make it pretty comprehensive.
So a few more questions ,and alot more work for me to do.
But really thanks so much for the help,Bryan

[Brymus]

Second version on my breadboard all scaled,less hiss.
Still no second op amp,definetly needed...
I didnt have enough 680Rs for the 14 x 6K8 so I used 454R ,yeah weird value, I got a bunch for really cheap along with some other odd values...
But anyway I didnt have enough 3K3 so I used 3k for the 7 x 33K.Pots are 10KL
I have 10uf, 4.7 uf, and 2.2uf electros on the 3 lower bands,not noticing any problems just put the - side towards ground.
The Vref was low with a 12K bias resistor so I used 1K, I left the voltage divider 4K7 x 2 for Vref.
R24 and R26 are 15K of course now
Not sure what the odd values are doing to the frequency the bands center on,math is not a strong point for me.
The formula is on the original 4QD schematic I linked to earlier if anyone would care to check what they are with the odd values.
The lows are still low,ect, to my ear (not the greatest test gear) the bands of simply moved up in frequency some.
I will add the second opamp tomorrow and mess with it more.
Right now the 6KHZ and 15KHZ bands dont seem to be doing anything ,and they are wired right I have triple checked them,
Maybe with a buffer/gain stage in front they will have more effect ,maybe they are centered too high in the bandwidth,the low and mid bands are definetly working right.
There is still hiss at max gains but at lower levels than the unscaled version,less than most pedals have at least.
I will report back more as it progresses.
Comments ? Advice ? would be appreciated thanks,Bryan

[John Lyons]

15k is almost to the extent of human hearing and may not even be reproduced by your amp.
If you hear a change in hiss with amp turned up (not playing, guitar off) then that would be
the extent of it I think. It's not a useful freq for guitar.
6K would be about the highest for guitar use I'd say...not sure whay that's not working.
Do the hiss test to check that one.

[Brymus]

Thanks John,I had thought that 15K was a little high for guitar use and suggested to Rick that we add 4Khz and 8Khz and change the 15Khz to 10Khz instead.Perhaps adding 4Khz and 8Khz and dropping the 15Khz would be better.
Here is the original scheamtic from www.4qd.co.uk if someone would PLEASE show me how to work the F and Q formulas ? I will then do them all myself.Are all the Rs supposed to be megaohms ? and what about the Cs what number would I use for say .047uf ?
If I understand right its R1 X R2 X C1 x C2 divided by 2 X Pi divided by 1 ,Is that correct ? SO would someone give an example
using the first band of the EQ as in the 60Hz band ? So I can see it longform and I should then be able to do it after that,thanks

[Brymus]

Come on surely someone knows how to translate the formulas so I can do them myself 
On a plus side with the buffer in there this EQ sounds great no noise at all,just need to set the F on the bands and work out a few more issues with the gain and impedences.
And it will be ready for others to build.