WTF diodes?

[Unixon]

Hi,

While troubleshooting one of my projects I found that most likely all schottky diodes PMEG1020EJ.115 from NXP that I've got show about 80 Ohm resistance in both directions even in low-voltage measurements (0.285V applied), 30mV forward voltage drop. My first guess was electrical damage during operation or thermal damage during soldering, but it turned out that any new diode randomly picked from tape shows just the same behavior. Needless to mention, they don't work in circuit.

I wonder WTF are those? Are they plain fake or am I hugely miss or underestimate something including storage conditions?

[SeanB]

Bought from a reputable supplier or random seller on eFake? Likely they are a fake part.

[Unixon]

I got these from a good local supplier from which I buy many parts on regular basis and it is almost always OK. For the last few years there were few issues with supplying wrong parts (different package or one-letter different partname), but I can't remember anything being this plain wrong. The worst part of the story is that I still need different PMEG-series diodes for current and upcoming projects and most local suppliers get parts from the same next tier stock. Looks like I will have to buy small amount just to check if new parts are OK.

[uwezi]

I wonder WTF are those? Are they plain fake or am I hugely miss or underestimate something including storage conditions?

Schottky diodes should not be particularly sensitive to ESD and I also don't think that storage conditions can be a major problem.

But the specs of these diodes are a bit particular. The reverse current at reverse voltages as low as 5 V is stated as up to 2 mA. From Fig. 2 in the datasheet you can see that there is still a typical reverse current of 0.3 mA at a reverse voltage of 1 V.

Highest reverse voltage as 10 V. Depending on the test current of your diode tester, it might appear as if the diode actually was conducting in both directions!

What is your application? These Schottky diodes might have been the wrong choice from the start...

[cybermaus]

Maybe I am being incorrect, but you express the diodes in odd values: Resistance?

Diodes are usually expressed as forward volt drop versus forward current. Of course you can then calculate a resistance, but that is not the habit.

Also, if I take the specsheet, your "80 Ohm" at 30mV means 0.3mA
That is "below the chart" in the specsheet I found, so strictly speaking not in spec. If you extrapolate the chart, they are a bit lower, but to call fake on out of spec use is maybe a bit too easy. Or am I seeing this too simplistic and it is allowed to extrapolate to beyond the manufacturer chart?

Bottom line of the spec chart is 1mA, which seems to be 60mV drop, at 25 Celsius, but that seems to vary wildly with temperature, up to 160mV for -25 Celsius. Maybe you should turn up the heat a bit?

[Unixon]

uwezi

I need a low Vf diode for near-precise rectification to save as much working voltage range as possible at 5V full scale. I think I totally missed this reverse current issue while trying to minimize Vf and package size and keeping in mind limitation for forward current and reverse voltage. This little beast PMEG1020 is too leaky as compared to "working horse" MBR0520. I don't know if these diodes are really OK and just unfit, but they are too leaky anyway. I'll be looking for something else of the same package size or even smaller.

cybermaus

Yes, I know that speaking in terms of resistance in respect of a diode sounds weird, but many reverse biased diodes will give tens of MOhms to out-of-scale resistance while shorted or damaged diode will show something near 1k or below in resistance measurement. Measuring forward voltage drop is only a partial check.

[uwezi]

uwezi
I need a low Vf diode for near-precise rectification to save as much working voltage range as possible at 5V full scale. I think I totally missed this reverse current issue while trying to minimize Vf and package size and keeping in mind limitation for forward current and reverse voltage. This little beast PMEG1020 is too leaky as compared to "working horse" MBR0520. I don't know if these diodes are really OK and just unfit, but they are too leaky anyway. I'll be looking for something else of the same package size or even smaller.

I was a bit surprised by their leakyness as stated in the datasheet... The MBR0520 appears to much better in this respect.

[Zero999]

The capacitance is also quite high.

Is this just for a small signal voltage measuring, rather than power? If so, how about using a normal silicon diode and an op-amp to cancel out the forward voltage?

[T3sl4co1l]

Schottky diodes are pretty much fleshed out, there isn't much give-or-take in characteristics.  1N5819, MBR340 and so on are still pretty much current technology, even since they were introduced however many decades ago.

The difference between types pretty much comes down to a single parameter: a lower barrier has less forward voltage drop, but leaks more and breaks down at a lower voltage.  A higher barrier has higher breakdown and less leakage, but more forward drop.

The only relatively new things are superbarrier and TrenchMOS type schottky, which extend the reverse voltage and improve leakage current, at some expense to the forward voltage drop.

What are you trying to do?  If it's power rectification, use a synchronous rectifier (or just deal with it, 0.4V out of 5V isn't terrifically bad, even for battery operated equipment).  If it's for signals, use junction diodes and a precision rectifier circuit; if the additional forward drop is a problem (for instance, it won't go within Vf of the supply rails), choose a lower amplitude for the signal path so it's not clipping.

Tim

[Unixon]

Hero999

All available PCB area of a very tiny device is already 100% populated with parts in packages like 0402, SC70 and SOD323
I was already very lucky to fit my circuit there.

T3sl4co1l

No, it's just a low current signal rectification. Can't use any complex circuit as there's no room for it. A precise rectifier circuit with op-amp is more interesting as I already have one op-amp just before this diode, but I left this diode out of feedback loop so it won't mess up desired op-amp operation. But really, I don't need any complexity in there, this rectified signal is then goes to averaging RC-circuit, so I just want to save a bit more amplitude for small signals.

[T3sl4co1l]

Small signals!  Go for a BAT54 or BAT85 or something like that, every junk box should have them (right alongside -- but not to be confused with -- the 1N4148s)!

If you're going for simplicity over precision but still can't spare half a volt, there are tricks you can do with transistors to semi-null the offset.  A pair of emitter followers, cascaded, one NPN, the other PNP (or vice versa..).  If you use an RC network instead of a load resistor, you get peak detection for free, as long as you don't need more than Vebo (useful to about 5V) in reverse.  Which it sounds like would be fine?

Of course, two transistors and a couple resistors is already more pins and board area than another op-amp, so the "simplicity" gain is theoretical at best.

Tim

[Zero999]

Hod do you know a diode mess up the feeback loop? Perhaps there's a way round this. Post the schematic.

[Unixon]

OK, the problem is finally solved.

I've got few other diodes in small packages. PMEG4010BEA appeared just slightly better than PMEG1020 but it was still too leaky, NSR0130P2T5G and NSR0230M2T5G - were just fine as well as 3-pin NSR30CM3T5G. I decided to get along with NSR0230M2T5G.

[kingoftaurus]

Maybe im not understanding this post correctly.

On my multimeter there is a "Diode" option for testing diodes in particular.

Grabbing one randomly and testing it.. I get

.OL  (over load) in one direction and
.386 Volts in the other.

13kohm when tested with ohms
and .OL (overload) in the other direction.

I as a newbie.. am understanding that this is correct right?

[Mr. Coffee]

Maybe im not understanding this post correctly.

On my multimeter there is a "Diode" option for testing diodes in particular.

Grabbing one randomly and testing it.. I get

.OL  (over load) in one direction and
.386 Volts in the other.

13kohm when tested with ohms
and .OL (overload) in the other direction.

I as a newbie.. am understanding that this is correct right?

That's a typical reading with a good schottky. Different diodes and meters will vary the results somewhat though. The diodes because they vary in forward voltage and reverse leakage, and meters because they don't all apply the same current to read the forward voltage drop (your .386V) and the forward voltage drop changes with forward current.

[Mr. Coffee]

...Of course, two transistors and a couple resistors is already more pins and board area than another op-amp, so the "simplicity" gain is theoretical at best.

Tim

Maybe so, but at high frequency (the OP never stated, did he?) your method may well do the trick when the op amp precision rectifier can't keep up.

[Unixon]

Op-amps are LMV321, working at sound frequencies (10~10^4 Hz).

This thread is actually related to that one about op-amp circuit.

[Unixon]

Hod do you know a diode mess up the feeback loop? Perhaps there's a way round this. Post the schematic.

I tried to test the circuit in Qucs, but it refused even to converge to a solution when the diode was in FB loop.
This op-amp implements "Vout = 2*(Vin-Vcc/2)", then Vout goes via diode into low-pass filter (averaging circuit) after which the output signal is obtained.
I will post the schematic after checking and cleaning it a little bit.

[T3sl4co1l]

I'm guessing Qucs is a SPICE backend like all the others (at a glance, I can't seem to find what it's based on), in which case, the circuit doesn't "fail", you just need to refine your model (including simulation parameters) until it produces meaningful results.

Tim

[Unixon]

I'm guessing Qucs is a SPICE backend like all the others (at a glance, I can't seem to find what it's based on)

Yes, it is.

in which case, the circuit doesn't "fail", you just need to refine your model (including simulation parameters) until it produces meaningful results.

OK, I checked that. You're right, simulation parameters were not fine enough.

If I put the diode in FB loop, then I need to correctly compensate for the output signal shift that this diode produces.
I'm not sure what and how exactly should I adjust in op-amp circuitry to keep things linear and not clipped off.

[Unixon]

OK, here's the basic schematic. This device is a kind of envelope detector for an electret microphone with pre-amplification and extended output range.

[Mr. Coffee]

Up to 100kHz is still too high for many op amps using them as a precision rectifier ("super diode").
I would try to use the precision rectifier idea with a couple schottkys and an LT1222 op amp and set the voltage gain about 10.

Maybe something like this, (but maybe with a DC blocking capacitor and a 100k ohm input load resistor to ground after the capacitor). I think the pinout is correct, but I can't promise.

[T3sl4co1l]

A 500MHz amplifier, you must be kidding.

[Mr. Coffee]

A 500MHz amplifier, you must be kidding.

Not as much overkill as you might think at first glance. The effective slew rate takes a real beating in even a 2 diode precision rectifier circuit. Plus you want some wiggle room in gain bandwidth to go for even a lower Vf drop by boosting the op amp gain and attenuating the DC output (if desired) to get up and out of the square law quagmire for very low signals.
These are not expensive amplifiers, unless comparing it to a 741 or if you have to dig quarters out of ashtrays to buy some raman for dinner.