diodes 1N4001 - 1N4007: are they really all different?

[exe]

Hi!

Many times I've seen in datasheets that most parts of a product range share the same specs. For example: http://www.diodes.com/_files/datasheets/ds28002.pdf . Diodes 1N4001 -- 1N4004 share the same specs, as well as 1N4005 -- 1N4007. Does this mean there are only two types of diodes instead of eight? Otherwise I can't explain, say, how 1N4001 and 1N4004 have different breakdown voltages and yet the rest specs are all the same. Even on plots there are only two curves at most (I attached example). Or, may be, they didn't want to plot all eight curves and draw pessimistic specs?

If they are really different, which to buy? I just bought 1N4007 because they should be a good fit for all around

May be someone wants to conduct an experiment? They are pretty cheap .

[TimNJ]

I don't know for sure if they physically *try* to make 1N4001s and 1N4007s, but I would figure they make one type of diode and then bin them depending on what their reverse breakdown voltages are. However, I'm not sure what the manufacturing process is totally like. Either will work fine for low voltage rectification, but if you're trying to rectify mains, definitely stick to the 1N4007s.

[Rerouter]

On the original spec sheets the reverse breakdown increased as the number went up, with 4007 being the highest, now that manufacturing tolerances have been improved its probably cheaper for them to bin into 3 groups, 1n4004, 1n4007 or fail,

[exe]

bin them depending on what their reverse breakdown voltages are.

This is interesting. I though that reverse breakdown is a destructive operation. But it seems it's not if the current is low. So. they can do this as well.
Actually, my next question would be "are cheap semiconductors tested?". When I buy an expensive IC I expect it to work (at least anyhow). But for a cheapish diode... Do they really test each of them? I hope so.... Because right now I often manually test things like diodes just in case (under real current, not just by attaching a multimeter). Maybe I shouldn't...

bin into 3 groups, 1n4004, 1n4007 or fail,

Yeah, can be. Especially for fairchild diodes (http://www.farnell.com/datasheets/2014524.pdf) where all of them have same specs (I bought those).

Thank you, guys, now I know more

[Richard Crowley]

"are cheap semiconductors tested?".

Disclaimer: I work at a place where some of the most expensive semiconductors are developed and produced. 

However, even for the most basic, simple, "jelly-bean" diode, it is so trivial to test each and every individual die that it makes no sense to waste the effort and materials to package a known-bad die.  I would be astounded if they actually packaged parts before testing them.  It is much easier to test die while still on the wafer with automated equipment. 

We used to actually apply a tiny dot of ink to die to indicate bad ones that shouldn't be packaged. But automated wafer identification and tracking systems now have the good/bad die information contained in a networked database. So a wafer could have been fabricated but then tested in a place thousands of Km away, and packaged and sorted in yet other far-away locations.

To be sure, there is likely also post-packaging testing to confirm that the assembly process didn't kill the part. And it also seems likely here in the Real World that parametric testing and "binning" happens for economic reasons at the very least.

[prawncrackers]

i think the main difference is the diode capacitance. not critical if youre just using it for linear 60hz-ish rectification but for high frequency, high powered, high efficiency applications sometimes the high voltage diodes in a series have slightly higher capacitance, higher reverse current, slower reverse recovery and sometimes higher forward voltage drop.

i.e. using a UF5408 (3a 1000v ultrafast diode) for a 12v smps will be less efficient/heat up more than a UF5400 (3a 100v)...

theres a youtube video somewhere where a guy scope'd the performance of different diodes

[Cerebus]

I don't know for sure if they physically *try* to make 1N4001s and 1N4007s, but I would figure they make one type of diode and then bin them depending on what their reverse breakdown voltages are. However, I'm not sure what the manufacturing process is totally like. Either will work fine for low voltage rectification, but if you're trying to rectify mains, definitely stick to the 1N4007s.

The breakdown voltage depends on the active depth of the depletion region which in turn depends on the level of doping. On such a wide spread of breakdown voltages as the 1N4007 series they almost certainly have different diffusion levels for different wafers. So yes they almost certainly try to make a particular characteristic but they might not try to make the whole series - perhaps 3 points in it and rely on binning to turn that into 7 devices.

[Cerebus]

I though that reverse breakdown is a destructive operation. But it seems it's not if the current is low.

The mechanism is avalanche breakdown. That's the same as the primary mechanism in a zener diode (over about 5-6V reverse breakdown) which clearly can go on running in breakdown for ever. However, it's also the same mechanism as reverse Vbe breakdown and allowing that to happen even once can alter a transistor's characteristics for good. So whether reverse breakdown is destructive depends a lot on current density and device geometry and it's a bad idea to assume whether it's OK or not unless you know a lot more about the device in question.

[edavid]

That graph shows that, for that manufacturer, 1N4005-1N4007 are made with a PIN structure, while the lower voltage parts are PN.

The breakpoint between PN and PIN can vary for different makers.

Hams often use 1N4007s as cheap PIN switches.

[exe]

there is a youtube video somewhere where a guy scope'd the performance of different diodes

Could you please find the link? I couldn't find any. I tried "diode benchmark", "diode comparision", "diode performance", "diode oscilloscope test", etc. No luck

[TimFox]

Breakdown voltage (as a performance spec) is given for a specified value of reverse current:  an easy test to do with ATE.
Rated voltage is a maximum specification, where the manufacturer guarantees that the device will not fail for voltages below that value.
Most specifications are for one of these two types.

[Gyro]

I though that reverse breakdown is a destructive operation. But it seems it's not if the current is low.

The mechanism is avalanche breakdown. That's the same as the primary mechanism in a zener diode (over about 5-6V reverse breakdown) which clearly can go on running in breakdown for ever. However, it's also the same mechanism as reverse Vbe breakdown and allowing that to happen even once can alter a transistor's characteristics for good. So whether reverse breakdown is destructive depends a lot on current density and device geometry and it's a bad idea to assume whether it's OK or not unless you know a lot more about the device in question.

... and of course some rectifiers are actually rated for significant avalanche energy.

[prawncrackers]

Could you please find the link? I couldn't find any. I tried "diode benchmark", "diode comparision", "diode performance", "diode oscilloscope test", etc. No luck

sorry the title is a bit different hahaha 

[TimNJ]

I don't know for sure if they physically *try* to make 1N4001s and 1N4007s, but I would figure they make one type of diode and then bin them depending on what their reverse breakdown voltages are. However, I'm not sure what the manufacturing process is totally like. Either will work fine for low voltage rectification, but if you're trying to rectify mains, definitely stick to the 1N4007s.

The breakdown voltage depends on the active depth of the depletion region which in turn depends on the level of doping. On such a wide spread of breakdown voltages as the 1N4007 series they almost certainly have different diffusion levels for different wafers. So yes they almost certainly try to make a particular characteristic but they might not try to make the whole series - perhaps 3 points in it and rely on binning to turn that into 7 devices.

Good to know. My semiconductor physics is pretty weak (as you can tell).