Weird 1N5817G problem only in one situation. Help?

[midwayfair]

I was building a Flabulanche, which is a circuit I ... know pretty well

http://www.madbeanpedals.com/projects/Flabulanche/Flabulanche_2015.pdf

Schematic page 4.

I ordered a pile of this diode, 1N5817G from Mouser:
http://www.mouser.com/search/ProductDetail.aspx?r=863%2d1N5817G

Datasheet:
http://www.mouser.com/ds/2/308/1N5817-D-103200.pdf

I've been through the datasheet and I can't find anything that tells me what the "G" suffix means, but I wouldn't have expected it to cause any issues. I mean ... it's a diode, right? In any case, I didn't give it a second thought when I was ordering it.

It works perfectly fine for the charge pump part of the circuit. I get my normal voltage at the end (17.7).

The place where I'm having trouble is the D1 rectifier. (For clarity, this rectifies the output of Q3 and puts a negative voltage on the gate, dropping the gain.)

There's some positive DC on the cathode of the diode (about 3.3V, from the gate bias for Q4). Normally that's not an issue -- I grabbed a previous build and there's no DC on the anode. But that build had a different diode.

On this build, when I flipped the switch, I've suddenly got a bunch of positive DC on the anode! Quite a bit of it, too -- about 2.3V.

So I grabbed a 1N60P out of the stash, pulled the 1n5817 and replaced it ... and the DC is gone. There's the tiniest bit of leakage across it, but that's it.

I doubled checked and none of the 1N5817 are bad or measuring something they shouldn't. My multimeter's not reading anything from cathode->anode for Fv. There shouldn't be enough current or voltage to hit any reverse breakdown or anything like that.

Just to make sure, I tested one of the 5817 in the previous build, and they cause the same problem. I tested every other diode I have at home and no other diode causes the issue. It's just this baggie of 5817Gs. I don't have any other 5817s to test, but I have 50+ other types of diodes and I can't find any others that malfunction in this way.

I popped one of the 5817 in a quick test circuit (just a load and a DC input). They block DC flipped around backwards like they're supposed to and drop .2V again like they're supposed to. So they diodes don't appear to be bad.

What am I failing to consider here? Is there something unusual about the "G" designation of the 5817 that's causing this unique problem?

[Rob Strand]

I've been through the datasheet and I can't find anything that tells me what the "G" suffix means, but I wouldn't have expected it to cause any issues. I mean ... it's a diode, right? In any case, I didn't give it a second thought when I was ordering it.

The G suffix is just how they are packaged.   You can find a list on p6 of the datasheet.

There's nothing weird there.

To me it seems like the diode has high reverse leakage.  Can you measure the reverse leakage?  Maybe compare it against a diode you know is "good".

[midwayfair]

I've been through the datasheet and I can't find anything that tells me what the "G" suffix means, but I wouldn't have expected it to cause any issues. I mean ... it's a diode, right? In any case, I didn't give it a second thought when I was ordering it.

The G suffix is just how they are packaged.   You can find a list on p6 of the datasheet.

There's nothing weird there.

To me it seems like the diode has high reverse leakage.  Can you measure the reverse leakage?  Maybe compare it against a diode you know is "good".

This was a good idea. I measured on Ohms and got a mere 49K reverse. That seems wrong and absurdly leaky. Grabbed a couple other diodes and they're above what I can measure. I seem to remember 5817s measuring tens of MOhms reverse in the past. Does anyone have a 5817 they could measure for me?

[Rob Strand]

I just checked the datasheet.  See figure 9.  The leakage is very high 0.5mA(!!!) at 2V;  not defined below that.
Compare that to say a BAT46 which is 0.1 to 1 uA.

[Rob Strand]

If you build a little leakage test jig (Maybe the DMM ohms is good enough), you might be able to sort them and pluck out the lower leakage units. 

I'm pretty sure I have some of these.   Sorry I can't find the darn things.  I might have used them.
Unfortunately "cross DMM" ohms readings might not be a reliable way to compare diode.

[Rob Strand]

Maybe your not going crazy after all and something *is* different about those diodes.

Look at figure 2 from the IR 1N5817 datasheet, and also the tabulated leakage data. 
The plots show typical.   At 2V, 25C  these are only 1uA.
http://www.irf.com/product-info/datasheets/data/1n5817.pdf

The ON-semi devices are 70uA typ.

So different, like WTx.

(I have to apologize I read-off the 75C values from the On-semi graph; the temp labels aren't lined-up with the curves.)

[midwayfair]

Hem. I might order a few different ma'am ales of the diode and do some measurements but I think it's easier to just warn people to use a different diode like the BATs in the project. Thanks again for your help, Rob, it got me to bed before an absurd hour.

[PRR]

> Datasheet:

Agree. 0.3mA leakage at few-Volt is insignificant in a 1 AMP Power job but overwhelming for you.

This is the flip-side of the low forward voltage. Note 20mA forward at just 0.1V! Extrapolate 0.2V to 0.1V to 0.0V, it may be flowing 2mA at _zero_ Volts. (Yes, this is possible at non-zero temperature.)

Leakage (actually thermal current) may drop if you pack in dry ice. (Bar ice is no big improvement.)

Note also this is "typical". Leakage could be far higher and you have no reason to complain.

Thanks for the heads-up.

[midwayfair]

Leakage (actually thermal current) may drop if you pack in dry ice. (Bar ice is no big improvement.)

Someday I want to build a device that actually implements your suggestions for improvement this way

I will note that a single capacitor would have solved my entire issue -- simply putting a cap between the MOSFET gate and the tone control output would mean 0V of DC. I foolishly failed to account for the diode's leakage and just figured that there was enough resistance to insulate the gate against any extreme bias changes. There may be a hidden "feature not a bug" in cases where it works -- I know that a build I made that used a 1N5817 in that spot DID sound really good -- but this is what I get for a case of the simples.