When physics takes a holiday

[Groovenut]

Hello!

I recently breadboarded this circuit with through hole parts. Transistors were 2N3906

It worked wonderfully for both over voltage and polarity protection. Disconnecting the load (a 470R resistor) in both conditions.

Now here's where it gets strange.

I then ordered a populated prototype smt board using MMBT3906

Upon testing the prototype smt pcb, I noticed that while the over voltage protection functioned properly, the reverse polarity protection did not. The circuit allowed about -1.5 volts through to the load

I then built a point to point version with TH parts, suspecting some "magic" in the BB, and it worked perfectly.

I then began swapping out parts on the prototype pcb to find the offender and it turned out to be the MMBT3906s

As a test I replaced them with TH 2N3906s and the prototype pcb functioned as the bread board and PTP had, perfectly.

I also tried the transistors mentioned in the article, FMMT718 with similar results to the MMBT3906.

I contacted OnSemi's Apps department and gave them the info on the circuit and behavior but they were a dead end.

Hopefully that's enough info. If not please feel free to ask.

Any ideas?

I also realize that inserting a series diode would give reverse polarity protection, but at the loss of .25 - .7 volts in addition to the voltage loss at Q1.

Since the TH circuit performed so well in both conditions, I had high hopes for the circuit as presented.

Interested in hearing the theories on this one. It's a head scratcher. Thanks for your time 

[PRR]

I respect your mystery but as a practical caveman:

> load (a 470R resistor) ... ... .. about 1.5 volts through to the load

1.5V at 3mA is <5mW. I can't think of any stage-worthy part that will burst from that little abuse.

[antonis]

Due to Q2 reverse breakdown voltage (V_EB), perhaps..??

[Groovenut]

I respect your mystery but as a practical caveman:

> load (a 470R resistor) ... ... .. about 1.5 volts through to the load

1.5V at 3mA is <5mW. I can't think of any stage-worthy part that will burst from that little abuse.

Thanks Paul. I appreciate your answer.

I figured any amount of reversed polarity on an opamp would cause degradation over time

I realize I labeled the 1.5 volts incorrectly in my initial statement. It should read -1.5 volts as it is during reverse polarity that this voltage gets through to the load.

[Groovenut]

Due to Q2 reverse breakdown voltage (V_EB), perhaps..??

Thanks Antonis,

Possibly. V_EB is listed at 5 volts on both 2N3906 and MMBT3906 spec sheets

At Q2, I am only measuring -0.75 volts at the base and -0.24 volts at the collector (emitter is grounded) in reverse polarity on the TH circuit.

It's an interesting puzzle as I wouldn't expect there to be much difference between TH and SMT of the same part.

[Rob Strand]

Interested in hearing the theories on this one. It's a head scratcher. Thanks for your time 

The protection circuit doesn't work well for reverse polarity protection.   It's OK for overvoltage.  Some years back I looked into all the flaws. There's two main points:  BE junctions will breakdown when reversed biased, say around 8V to 10V but it depends on the transistor.  So instead of the BE junction blocking it conducts.   Next is the BC junction of the main switch transistor.  When you reverse the supply that junction is now  "pointing' in a forward  direction; note the path is through the load so you need a load.

When you combine the two effects you find it's not a good idea using the BJT for reverse polarity protection.   The simple add-on solution is to use a series Schottky diode at the input.

The reason you might be seeing issues on the SMD devices is the BE junction breakdown voltage might be lower (agreeing with Antonis's comment).    What you see depends on your test voltages .  You need to cover a good range with points in between say +5V upto say 20V to 30V depending on what abuse you want to fend off.   However you need to do the same for the reverse voltages, -1V, -5V, -9V upto some desired abuse level.   If you stop at -9V the reverse protection might appear to work on say the 2N3906 and not the SMD.  If you push out to -15V to -30V you will probably see issues with both.  [Test with resistor loads before using a real circuit at -20V and -30V!]

IIRC, RG has a page or two on this.  Maybe one page on switches and one page on protection circuits.  I'm sure he covered the BJT version.  He may very well mention the problems with the BJT and reverse polarity protection.

I might have been thinking of this one, which might actually have a problem,
http://geofex.com/Article_Folders/cheapgoodprot.htm

[Groovenut]

Interested in hearing the theories on this one. It's a head scratcher. Thanks for your time 

The protection circuit doesn't work well for reverse polarity protection.   It's OK for overvoltage.  Some years back I looked into all the flaws. There's two main points:  BE junctions will breakdown when reversed biased, say around 8V to 10V but it depends on the transistor.  So instead of the BE junction blocking it conducts.   Next is the BC junction of the main switch transistor.  When you reverse the supply that junction is now  "pointing' in a forward  direction; note the path is through the load so you need a load.

When you combine the two effects you find it's not a good idea using the BJT for reverse polarity protection.   The simple add-on solution is to use a series Schottky diode at the input.

The reason you might be seeing issues on the SMD devices is the BE junction breakdown voltage might be lower (agreeing with Antonis's comment).    What you see depends on your test voltages .  You need to cover a good range with points in between say +5V upto say 20V to 30V depending on what abuse you want to fend off.   However you need to do the same for the reverse voltages, -1V, -5V, -9V upto some desired abuse level.   If you stop at -9V the reverse protection might appear to work on say the 2N3906 and not the SMD.  If you push out to -15V to -30V you will probably see issues with both.  [Test with resistor loads before using a real circuit at -20V and -30V!]

IIRC, RG has a page or two on this.  Maybe one page on switches and one page on protection circuits.  I'm sure he covered the BJT version.  He may very well mention the problems with the BJT and reverse polarity protection.

I might have been thinking of this one, which might actually have a problem,
http://geofex.com/Article_Folders/cheapgoodprot.htm

Hi Rob thanks for weighing in

I can confirm that when the voltages are pushed higher under RP there is voltage into the load as you surmised.

So apparently I just didn't push the TH circuit hard enough.

Thanks for all the details and explanation

[Rob Strand]

IIRC, the circuits that definitely work used MOSFETs for the switches and the other parts of the ckt can be either.   The body diodes of the MOSFETs always point toward each other to provide a natural blocking.     On my old computer I've got a whole heap of BJT and MOSFET circuits attempting to do both over voltage and polarity and off hand the BJT versions always had some issue.   

In short the series diode is simpler and you can use it with a MOSFET or a BJT.  The non-series diode version which uses the extra MOSFET version has less voltage drop.   If you look closely you can see the body diode of the first MOSFET acts like a series polarity protection diode and when the polarity is correct the MOSFET switch shorts out that diode and the diode drop is reduced.

There's quite a few examples out there (and maybe some old posts on the subject),
https://duckduckgo.com/?q=reverse+polarity+and+over+voltage&iax=images&ia=images&iai=https%3A%2F%2F1.bp.blogspot.com%2F-yvrNe0grMz0%2FU8K40NUUg3I%2FAAAAAAAADSA%2FYjH5BSjBbzc%2Fs1600%2Fvprotection.png

[anotherjim]

Any idea why the smd part might have a lower breakdown V?
I can see a reverse beta path via c-b junctions, but I can't see the effect the Zener has when it's acting as a plain diode - the bases might not go negative enough wrt the collectors but Vce is always a little lower than Vbe.
Just thinking... if an SMD BJT example happens to have a higher forward current gain, then it has higher reverse too?

[Groovenut]

Another interesting thing...

When simmed, the circuit behaves perfectly in both over voltage and polarity reversal, regardless of voltage input. Even at -30 volts input the leakage to the load is in the nV range.

Granted everything that happens in the sim isn't real world and the explanations put forth so far seem to bring physics home from holiday Time to leave this one alone and move on.

I have been using the single P channel MOS circuit as polarity protection and it's great. I wanted something more elegant than the crowbar shunt zener for over voltage, hoping for a non-destructive answer.

[antonis]

Any idea why the smd part might have a lower breakdown V?

I presume it involves molecular physics analysis..

[Groovenut]

IIRC, the circuits that definitely work used MOSFETs for the switches and the other parts of the ckt can be either.   The body diodes of the MOSFETs always point toward each other to provide a natural blocking.     On my old computer I've got a whole heap of BJT and MOSFET circuits attempting to do both over voltage and polarity and off hand the BJT versions always had some issue.   

In short the series diode is simpler and you can use it with a MOSFET or a BJT.  The non-series diode version which uses the extra MOSFET version has less voltage drop.   If you look closely you can see the body diode of the first MOSFET acts like a series polarity protection diode and when the polarity is correct the MOSFET switch shorts out that diode and the diode drop is reduced.

There's quite a few examples out there (and maybe some old posts on the subject),
https://duckduckgo.com/?q=reverse+polarity+and+over+voltage&iax=images&ia=images&iai=https%3A%2F%2F1.bp.blogspot.com%2F-yvrNe0grMz0%2FU8K40NUUg3I%2FAAAAAAAADSA%2FYjH5BSjBbzc%2Fs1600%2Fvprotection.png

Thanks for the info and link Rob. I'll have a look at those.

The combo over voltage polarity protection is primarily for voltage converter ICs ala ICL7660 so having an additional diode drop for RP protection isn't a big thing. Obviously from a design standpoint less is better, but 250mV isn't going to break the bank.

[Rob Strand]

When simmed, the circuit behaves perfectly in both over voltage and polarity reversal, regardless of voltage input. Even at -30 volts input the leakage to the load is in the nV range.

The breakdown voltage in the transistor model probably isn't correct.   Breakdown isn't modelled well in spice anyway but you should be able to show some representative behaviour.   You might need to tweak the model.  I know I did sim these things.

Any idea why the smd part might have a lower breakdown V?
I can see a reverse beta path via c-b junctions, but I can't see the effect the Zener has when it's acting as a plain diode - the bases might not go negative enough wrt the collectors but Vce is always a little lower than Vbe.

In reverse polarity, BC is forward and BE has reverse breakdown (like when using transistors as noise sources).

The lower breakdown comes from the size and the doping levels.  Given both the through-hole and SMD are blackboxes we don't really know what's going on under the hood.  You can only measure the outside the box effect, which is a lower BE breakdown voltage for the SMD.