LM4562 is openly defying me

[ggedamed]

Good day gentlemen (and ladies),

One of these days I was housing an effect (a clone of DAM Ezekiel 25:17) for a friend.

As my building routine goes, first time I apply the supply voltage with as few active parts as possible - i.e. I take out all ICs from the sockets and measure the current. If the curent value looks OK, I measure voltages at the sockets pins (supply, reference, etc). If this is OK, I put back the actives, power the effect, measure the current, then listen to the effect. If everything is OK, I start boxing it.

So, in this case, everything went OK until the IC (LM4562) arrived in its socket. Then, the reference went to values around 900mV, along with the outputs. For fun I changed it with another LM4562, a TLC2262, then NE5532, OPA2132, RC25480. It worked with every IC - reference at around 4.5V, outputs following nicely and making expected noises. EXCEPT with LM4562.

Intrigued, I took out a breadboard and built the following, in an attempt to simplify the test circuit:



The result was the same - LM4562 brought the reference (VR) down to around 3.3V and the non-inverting inputs to less than 900mV. The supply voltage is not affected, the supply current is less than 15mA. According to the datasheet, the input bias current is less than 72nA, so the 45uA from the reference divider (R1 and R2) should be plenty. Besides, NE5532's input bias current is 10 times bigger and NE5532 works.

Out of desperation and ignorance, I made some more tests. If C2 is moved on R1, it works. If I temporarily shunt R4 + R5 and R7 (), it starts working. But the best part is this: if I violently and repeatedly strum the guitar, it starts crackling and then, it starts working.

What's happening? Glitches in the matrix?

[Digital Larry]

I see this in the LM4562 data sheet:

These devices have limited built-in ESD protection. The leads should be shorted together or the device placed in conductive foam
during storage or handling to prevent electrostatic damage to the MOS gates.

It's just a guess.  Have you observed the necessary precautions?

[ggedamed]

Thanks for your input. Actually, I didn't handle them as CMOS devices. Texas Instruments sent them in clear non-conductive plastic tube pieces, so I never gave a thought about this.

Meanwhile, I did found something that matches my adventures really well: LM4562, LME49710, LME49720 Start-up Behavior. Me thinks this is it - a LM4562 specific quirk and a designer who was not aware of it.

[Digital Larry]

Seems more likely than my wild-a#& guess.

[12Bass]

Oscillation?

[ggedamed]

Oscillation was my first guess, but the page linked in my previous post describes exactly what happened to me (and my breadboard).

[duck_arse]

well, seeing as you're trying things, how about a really low value of C2? and, put a resistor R in front the input and R the feedback, so its gain will still be one. there is an opamp somewhere, I can't remember #, datasheet says "no hard-wire as follower, put resistors!" words to that effect. couldn't hurt.

[anotherjim]

What happens to the bias voltage if the bias network is unloaded?
It actually seems like less bias current makes it worse - so that suggest that the leakage current in C2 is unbalancing the potential divider.

edit...
If the bias current is that low, then C2 can be a 100nF NPO ceramic.

[slacker]

This thread might be some help http://www.diystompboxes.com/smfforum/index.php?topic=108030.0.
I'm not sure there's any reason to use that opamp in that circuit though, I'd just use one of the ones that works and not worry about it, from the schematic I found it's just a buffer and TS style clipping stage so the choice of opamp won't make much difference.

[ggedamed]

well, seeing as you're trying things, how about a really low value of C2? and, put a resistor R in front the input and R the feedback, so its gain will still be one. there is an opamp somewhere, I can't remember #, datasheet says "no hard-wire as follower, put resistors!" words to that effect. couldn't hurt.

OK, I further simplified the schematic:



The power supply voltage is 8.95V, LM317 stabilized. I tried a 9.5V one, it is the same.
1. C2=10u and C2=1u:
   3 - 0.925V, second reading after I touch it with the DMM probe - 4.19V (no change until I touch it)
   1 - 0.875V, second reading after I touch it with the DMM probe - 4.30V (no change until I touch it)
   5 - 0.910V
   7 - 0.865V
2. C2 = ∞ (took it out):
   3 - 4.19V
   1 - 4.30V
   5 - 0.910V
   7 - 0.865V
So it seems that with 220k bias resistor and no reference filtering, it works.

Then I inserted resistors in the bias and the feedback (C2 remained 10u):



1. R3, R5, R8, R6 = 1k:
   3 - 0.931V, second reading after I touch it with the DMM probe - 4.19V (no change until I touch it)
   1 - 0.837V, second reading after I touch it with the DMM probe - 4.30V (no change until I touch it)
   5 - 0.916V
   7 - 0.865V
1. R3, R5, R8, R6 = 100k:
   3 - 0.929V
   1 - 0.822V
   5 - 0.919V
   7 - 0.821V

Then I went back to the first circuit, changed R4 si R7 to 100k and decreased the reference divider resisitors (R1 and R2). No change at 51k, 22k, 18k. It did work with R1 and R2 = 10k.

It looks to me that, because of the high startup bias current, LM4562 is not a good choice for guitar circuits. And the low bias input from the datasheet is only a theoretical value, unless you make the circuit to accomodate this quirk.

What happens to the bias voltage if the bias network is unloaded?
It actually seems like less bias current makes it worse - so that suggest that the leakage current in C2 is unbalancing the potential divider.

edit...
If the bias current is that low, then C2 can be a 100nF NPO ceramic.

With only the DMM as load, the voltage is 4.44V. You can see latest tests above, C2 is not the cause of my problem.

This thread might be some help http://www.diystompboxes.com/smfforum/index.php?topic=108030.0.
I'm not sure there's any reason to use that opamp in that circuit though, I'd just use one of the ones that works and not worry about it, from the schematic I found it's just a buffer and TS style clipping stage so the choice of opamp won't make much difference.

It works when both R1 and R2 are bypassed by capacitors AND the capacitor bypassing R1 is bigger, so the reference starts from the positive rail instead of the ground.
The only reason to want to use LM4562 I can think of is when you use the 4th position of the clipping switch, where is no diode in the NFL. In this position it is a somewhat noticeable difference between op amps.
On a side note, it should work in Sansamp GT-2, as the bias allows for more bias current.

[bool]

I don't think these opamps were made for high-impedance circuits.

The input bias compensation is to reduce errors (and current-noise) caused by input bias current and not to provide an option to artificially increase input impedance - and to work high impedance biasing networks.t

While these aren't the quietest parts around, I have mostly seen them in circuits with a couple-kOhm to a couple-10kOhm biasing/nfb networks.