Trouble finding non-schottky SMD power diode

[MustardMan]

Hi All,
I am having great difficulty finding a surface mount (preferably SMC package) power diode that is not schottky. I have a specialised application that requires the voltage drop present across a silicon PN junction (ie: 0.7 volts), but it seems that pretty much every manufacturer has gone schottky because of the lower Vf.

Does anyone have any pointers to non-shottky power diodes, preferably with a curve similar to the '2017 S10AL' diode shown below? I require 10 Amps. Even a leaded device would be fine!

Two images:
The 2017 datasheet for the S10AL (from digikey) shows a curve suggesting that the knee is close to 0.7 volts.
The 2019 datasheet from the manufacturers website shows that the knee is likely to dip under 0.6 volts (at 100mA it is already under 0.7 volts, wheras the "old" one was way above).

Cheer, MM.

[exe]

Put two schottky diodes in series? Also, what is the average current?

However, I advice giving us more details. May be the problem can be solved elsehow. I'm also a bit skeptical that the drop should be exactly 0.6-0.7V as it largely varies with the temperature. So, please tell us why a regular 10A

PS it seems the original device is still possible to buy: https://octopart.com/search?q=S10AL&currency=USD&specs=0 .

PPS current rating of 10A for such a small part is indeed very uncommon.

[TheUnnamedNewbie]

trusting something on the knee voltage of a single junction seems like a bad idea, since it is so temperature and current dependent. I think it would help to look at why you (think) you need it to work that way and then see if there isn't a better way around it.

[OM222O]

look up idea diode, basically fets as diodes.

Couple it with some sort of op amp and a normal fet to generate the 0.7V with the help of a low value resistor.

I wouldn't trust just a normal diode for this application.

[magic]

There are reasonable reasons to avoid Schottky's besides desiring the voltage drop. For example, their reverse leakage is terrible.

I require 10 Amps. Even a leaded device would be fine!

Ahem, that's 7W of power dissipation. No surprise it's hard to find SMD parts, but I'm sure there are many options in TO-220

Perhaps you could use some MOSFET's body diode? A diode-strapped power BJT? Such parts would be available in DPAK and similar packages, not sure if it's good enough for 7W.

[ejeffrey]

Look at bridge rectifiers.  Those are easily available with 10-100 amp current ratings with regular PN diodes.  Wire them up so all 4 diodes are in parallel.

Or use a power BJT wired as a diode connected transistor (collector shorted to base).  Or use a FET with an opamp driver if you have an auxillary supply for the opamp.

[Gyro]

Ahem, that's 7W of power dissipation. No surprise it's hard to find SMD parts, but I'm sure there are many options in TO-220

Yes, that's the reason you're not finding them - even with the most optimistic specmanship...

If there's a viable niche then somebody is likely to fill it, even if it's just for lower reverse leakage, but the dissipation (in a sensible cost package) is the killer.

[exe]

Ahem, that's 7W of power dissipation. No surprise it's hard to find SMD parts, but I'm sure there are many options in TO-220

Yeah, looking at specs, its thermal resistance junction-to-lead is 18C/W. So, dissipating ~7W would require an "ideal" heatsink in order to keep it below 150C. I conclude that the spec is unrealistic and average current should be much below 10A, or it will unsolder by itself . Therefore, I conclude, that any regular 3A diode in similar enclosure would do just fine (imho).

[magic]

Look at bridge rectifiers.
Wire them up so all 4 diodes are in parallel.

I don't think so

BTW, one problem with diode connected BJT is that reverse voltage is limited by the BE breakdown rating, which is typically about 5V and rarely up to maybe 15V.

[MustardMan]

Sorry for the delayed reply - I have now generated a schematic. Yes it has default diodes in the simulation, but they at least give an idea of what I'm trying to achieve.

The constraints:
- 240 volt 50 Hz supply
- Circuit current max = 10A (breaker/fuse)
- Access to only one load wire, no access to switch
- Able to detect a load down to 10mA (and not blow sky high at 10A)
- No external power (must run off it's own generated power - from whatever it sniffs off the wire)
- Space constrained (very, hence SMD preferred). The lower the part count the better

My idea (at this stage) is to use back-to-back diodes. The simulation has four diodes (in case I need more voltage on the output), but that would double power dissipation. As noted, 7W is already quite high, and for the diode I listed (S10AL) it is actually 11W at 10Amps (but shared between each diode, so "only" 5.5W each!). The 'sniffer' is followed by a full wave voltage doubler using the lowest Vf schottkys I can find.

With an input range of 1000:1 (10mA to 10A) I am consequently looking for a diode that has the most consistent Vf covering this range, and every schottky I've seen (so far) doesn't even come close to being consistent across that range. I can't use something op-amp (unless I can power it off the voltage it sniffs), and I had not considered mosfets - how would the part count go with a mosfet solution?

Cheers, MM.

[magic]

So you have a wire which is part of some mains circuit carrying up to 10A and you want to detect when it conducts at least 10mA?
And you are absolutely sure that no other power supply can be made available? So where is the output signal going?
Do you realize that you cannot connect the output to ground as drawn, or to any other fixed potential for that matter?

[MustardMan]

Yes, correct. Detect 10mA or above, up to 10A. No external power supply available.

The output could best thought of as an LED in an opto that indicates the circuit is active. In reality it is a little more complex than just an LED.

The only reason a ground is shown on my posted circuit is because Microcap requires it to perform a simulation: there is no (and will be no) actual ground reference.

I did not think of it till now, but you have made me think 'current transformer', and using its' output to drive "the LED". 1000:1 though, I would have to think about a core that could give me enough output at 10mA but not go up in smoke at 10A, or produce so much output at 10A that it would make clamping it down to something usable worse (dissipate more power) than the diode idea.

Cheers, MM.

EDIT: I don't like playing with mains, so I am leaning pretty heavy on the simulation at present. When it gets real I will be working with 24 volts (AC) to do real world tests... as long as my voltage is above the drop of the "diode sensor", the actual voltage should not be of much importance, only the current. Final stage, after *all* verification, mains connection.

[mikeselectricstuff]

I did not think of it till now, but you have made me think 'current transformer', and using its' output to drive "the LED". 1000:1 though, I would have to think about a core that could give me enough output at 10mA but not go up in smoke at 10A, or produce so much output at 10A that it would make clamping it down to something usable worse (dissipate more power) than the diode idea.

If you can get a CT to detect at 10mA without too much primary dissipation, then that may be a good choice - chances are the core will just saturate at higher currents, which shouldn't be a problem.
The challenge is to get enough voltage - current is less of an issue as you can use a micropower pulsing circuit if necessary to drive LED/Opto etc. if all you want is to detect on/off state, i.e. charge up a capacitor until you have enough energy to pulse the LED.

[MustardMan]

I am currently following two investigative paths...

First: Looking at datasheets for different TO220 diodes. None yet with a curve that looks as good as the S10AL, but I am rather dubious about that particular datasheet anyway so I have a few from digikey on their way.

Second: Dismantling a commercial RCD. Since they are sensitive to an imbalance of 30mA (or better) and can trip a solenoid at that sensitivity, and handle significant overload (more than 10A), I could run one deliberately unbalanced (one core wire) and see what comes out of it.

Both methods involve current detection, so I can easily get away with 24 volts and not have to play with mains. Whew!

I have not looked much at RCDs except out of interest. From the little I've seen, it is unclear if they output  a high voltage/low current to trigger the solenoid, but that is most likely. Hence I would also require a buck converter to step whatever down. I suspect the CT solution will be larger, but by how much remains to be seen.

Has anyone here done a proper investigation of the current transformer in an RCD? Everything I've seen so far on the 'net says things like "and this is the sense coil that is connected to this over my head circuit". Nothing about what waveforms/voltages/currents appear and when.

Cheers, MM.

[mikeselectricstuff]

Modern rcds use electronics, with a capacitive dropper low voltage supply, and don't power the trip device direct from the ct output

[Circlotron]

Plenty of TVS diodes come in SMC / DO-214AB size. Their forward characteristics may be suitable.

[MustardMan]

Modern rcds use electronics, with a capacitive dropper low voltage supply, and don't power the trip device direct from the ct output

Yeah - I bought one yesterday & pulled it apart this-morning. Very disappointing. A sense CT and some electronics. Despite the complication & added parts, I suppose that makes it cheaper because then all trip ratings simply use the same guts with a different trigger setting. Oh, for the good old days when such things were made with just a CT and a solenoid.
It might be still worth looking at, but I am now leaning back towards a silicon solution

I am not so constrained by height, so I can use TO220 style devices with a caveat of two. Any more and it just starts taking up too much space.

MOSFETs (back-to-back, using them only for their body diodes)... there are some reasonably priced ones out there, and the curves look nice and steep (eg: attached, 0.6v at 10mA at 25C). Extra circuitry to drive the gate for lower dissipation at higher currents will unfortunately not fly because of the additional space required.

SiC rectifiers have much higher Vf (eg: attached, 0.7v at 10mA at 25c) as do the body diodes in SiC FETs, and the curves are good until low amps. I could go higher current ratings to get the curves steeper for longer, but I thought the prices for the 'smaller' SiC devices were high - until I saw the prices for the 'larger' SiC devices!

Thanks to all the respondents for the ideas you have given me!

Cheers, MM.

[MustardMan]

 
Somewhat goose-like, he realises that a fairly simple way of determining if a diode is schottky or silicon is Trr, as there is no Trr with a schottky!

Very few manufacturers put data on their spec sheets that cover any sort of decent range (thanks to Diodes Incorporated for putting the time into doing a decent test on this device) and this curve shows the behaviour at low currents. I had initially wondered why there were extra bends at Tj=25 and Tj=-65... which is of course the diode knee. It is very likely that other diodes with a Trr (ie: PN junction diodes) perform the same.

I do find it interesting though that just before the 25C 'knee' the forward voltage is in the vicinity of 0.5 volts. I always thought a PN junction at 25C was 0.7 volts?

MM.

[Circlotron]

Why do semiconductor manufacturers draw diode graphs like that? Usually you want to know the unknown forward voltage drop at a known current, so the current (the independent variable) should be across the bottom and the voltage drop (the dependent variable) should be up the side! I can’t think of a case where you would apply a certain voltage to a diode and need to know what the resulting current is.

[Zoli]

Hi All,
I am having great difficulty finding a surface mount (preferably SMC package) power diode that is not schottky. I have a specialised application that requires the voltage drop present across a silicon PN junction (ie: 0.7 volts), but it seems that pretty much every manufacturer has gone schottky because of the lower Vf.
...

So, let's see what kind of options Digikey(just because is the most popular supplier 'round) has, when selecting diodes:
Surface mount: yes, is present
Voltage - Forward(V_fMax): surprisingly, this one is present, too
Current - Average rectified(I_o): this one get booooring...
So let's go, and see, what is available - but I think, you should do it for yourself.
Anyway, this post is not intended to "teach you a lesson"; the intentions are to give you some guidelines for the future - and thank you for reading it.

[MustardMan]

@Zoli
LOL. Yes, I did search digikey. In fact, that was where I got the first of those graphs I posted.

The error I made, and probably not stated very clearly, was I was under the impression that Vf for a silicon PN diode was 0.7 volts at 25C. After looking at dozens (maybe up around a hundred) of datasheets the only one I found that indicated a Vf of about 0.7 (at 10mA - a reasonable current) was that one from MCC. All the others I saw were about 0.6 volts. My (wrong) assumption was manufacturers had simply switched to schottky.

So yes, I was wrong, I had made some wrong assumptions, and I feel embarrassed about them.


But the question I now have is the one I stated above... the commonly accepted forward voltage drop of a silicon PN junction is 0.7 volts. So why are so many datasheets showing it lower? Is it?

@Circlotron
Agree. We were all (mostly) taught in school to have the driving variable on the x-axis and the driven variable on the y.
So, it may not be what is expected, but at least all manufacturers seem to be consistent.

Cheers all, MM.

[Dabbot]

But the question I now have is the one I stated above... the commonly accepted forward voltage drop of a silicon PN junction is 0.7 volts. So why are so many datasheets showing it lower? Is it?

It's simply a reasonable estimate. The actual value will be around this estimate, depending on device type, temperature and current.

[TheUnnamedNewbie]

Why do semiconductor manufacturers draw diode graphs like that? Usually you want to know the unknown forward voltage drop at a known current, so the current (the independent variable) should be across the bottom and the voltage drop (the dependent variable) should be up the side! I can’t think of a case where you would apply a certain voltage to a diode and need to know what the resulting current is.

I have always thought of them being more logical this way. After all, when I learn and think of semiconductor devices, I usually do so with I-V curves, where you look at current as a function of voltage. So I just 'know' diodes as having that specific curve, and 'want' them to be shown in that way (as it is also the way that it is shown on our SMUs and so on).