Confirm TVS Diode Selection

[bostonman]

I am selecting a TVS Diode for two circuits that put out +12V DC and -12V DC.

If I'm reading and understanding the datasheet correctly (see attached), a good option is: 5.0SMDJ18CA

It doesn't conduct until 18V and begins conducting at 20V. The two parts I'm uncertain of pertain to the "maximum" conducting voltage of 29.2V and whether turn on spikes may cause the TVS to trip the series fuse I'm installing (hence why I'm going with the 18V version).

If an external voltage (maybe static electricity) spike beyond 29.2V, what does this mean for the TVS, it fails possibly open and doesn't cause the fuse to open?

[Jay_Diddy_B]

Hi,

You need to read the datasheet like this:

1) with a test current of 1mA the  breakdown voltage will be between 20 and 22V.
2) With a test current of 172A the maximum voltage will be 29.2V when using the 10/1000us pulse.



You need to know how much the circuit that you are protecting with the TVS can tolerate without damage.


The C in the part number is indicating a bidirectional part. In most cases yiou will want A not CA.



Regards,
Jay_Diddy_B

[bostonman]

The C in the part number is indicating a bidirectional part. In most cases yiou will want A not CA.

Oh really? I thought if the concern is static electricity such as touching a connector, the pulse could go positive or negative.

I have 5-0SMDJ-08202024 and thought they would be too large. Actually, I installed them thinking they were bi-directional and blew the fuse when I connected power while performing a loose bench test.

How do I know what kind of pulse to expect since the concern is static? I may have interpreted the datasheet incorrectly or expected a certain type of spec. I was thinking that at the constant 12V the TVS will not conduct and then something zaps it like static that's >12V and the TVS pulls it to ground thus blowing the (125mA) fuse .

[David Hess]

How do I know what kind of pulse to expect since the concern is static? I may have interpreted the datasheet incorrectly or expected a certain type of spec. I was thinking that at the constant 12V the TVS will not conduct and then something zaps it like static that's >12V and the TVS pulls it to ground thus blowing the (125mA) fuse .

A TVS is not a thyristor which latches on.  It is a zener or avalanche diode designed and specified for pulse applications.  It begins to conduct at its breakdown voltage and has a low impedance so current steeply increases as the voltage across it increases above its breakdown voltage.  It operates as a voltage clamp.

Typically they are used on input or output signals to protect against over-voltage conditions, including electrostatic discharge since they are very fast, however they can also be used to clamp supply voltages preventing damage in the event of power supply failure.  During a continuous overload, the TVS will likely fail and short out the regulated supply output, which might blow a fuse.  I would use the working voltage and not the breakdown voltage, taking into account the tolerance of the supply voltage.  For a 12 volt regulator, I would use the 5.0SMDJ12A which has a reverse standoff voltage (working voltage) of 12 volts.

When I refurbished my DM502 multimeters, I resocketed the dual regulator DIP, and added TVSes to the +5, +12, and -12 volt regulated supplies, as shown below.  I was worried that a regulator failure could cause damage to the ADC chipset which has been out of production for decades.

[bostonman]

A TVS is not a thyristor which latches on.

Yes, I know it doesn't latch. I meant more that at turn on, a spike may cause the TVS to pull current and blow the fuse. Although the TVS would be doing its job, turn on spikes are normal and not cause to blow a fuse.

I would use the 5.0SMDJ12A

Does this have enough headroom so it doesn't blow the fuse if a turn-on spike occurs? Also, how come Uni versus bi?

[Siwastaja]

Maybe you confusedly think that unidirectional TVS would only clamp in one direction? That's not the case - the unidirectional part always clamps to a very low voltage (around 1V, a diode drop basically) in the other direction. You choose a bidirectional part when your actual desired signal voltage goes in both directions - then the rated clamping voltages apply both ways, and it does not clamp "too early" in other direction.

[David Hess]

I would use the 5.0SMDJ12A

Does this have enough headroom so it doesn't blow the fuse if a turn-on spike occurs? Also, how come Uni versus bi?

A spike from what?  An ESD event is not going to blow any fuse, and if a regulated power supply is generating a spike like that, then the fuse should be blown because something is broken.

[rteodor]

How do I know what kind of pulse to expect since the concern is static? I may have interpreted the datasheet incorrectly or expected a certain type of spec. I was thinking that at the constant 12V the TVS will not conduct and then something zaps it like static that's >12V and the TVS pulls it to ground thus blowing the (125mA) fuse .

This would be a fun question for "General Technical Chat": What is the minimum fuse that you can blow just by rubbing your hair ?

[bostonman]

You choose a bidirectional part when your actual desired signal voltage goes in both directions - then the rated clamping voltages apply both ways, and it does not clamp "too early" in other direction.

Makes more sense.

A spike from what?  An ESD event is not going to blow any fuse

Maybe I worded things wrong, but it certainly managed to get a good laugh regardless. I meant a spike from the power supply turning on. The AC/DC produces +/-12V (along with 5V I think). My thought is that at power up the +/-12V a turn-on spike may occur that would exceed the turn on of the TVS.

The design I incorporated is a board (with the TVS) that takes +/-12V (edit: from the unit to power my board) in, goes through a fuse on each voltage, and then the TVS. If either voltage exceeds the TVS, then it would clamp the voltage, increase current, and cause the fuse to blow; certainly not an ESD occurrence.

[Jay_Diddy_B]

bostonman and the group,

Have look at this application note from ADI (Linear Technology):

https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf


This application note explains how you can generate a voltage transient from 'hot plugging' a power supply.

You can clamp the voltage with a TVS, a modest sized one, is all that is required. Or you can damp the transient by using an electrolytic capacitor that has some ESR.

There is no need to blow any fuses.


Regards,

Jay_Diddy_B

[David Hess]

Maybe I worded things wrong, but it certainly managed to get a good laugh regardless. I meant a spike from the power supply turning on. The AC/DC produces +/-12V (along with 5V I think). My thought is that at power up the +/-12V a turn-on spike may occur that would exceed the turn on of the TVS.

The design I incorporated is a board (with the TVS) that takes +/-12V (edit: from the unit to power my board) in, goes through a fuse on each voltage, and then the TVS. If either voltage exceeds the TVS, then it would clamp the voltage, increase current, and cause the fuse to blow; certainly not an ESD occurrence.

Good power supply designs do not overshoot and produce a spike, but I have encountered ones that did.

I repaired a Kenwood HF radio where the linear regulated power supply for the HF amplifier did that, and the design had a TVS across the regulated output to control it.  The failure was that the TVS eventually shorted, blowing one of the pair of parallel TO-3 output transistors as well as the emitter ballast resistors, or the reverse? Examination of the schematic showed a couple of design mistakes, including lack of parallel base-emitter resistors across the output and driver transistors to control leakage and charge, and poor control of the transconductance of the driver stage.  Adding the TVS was not a bad idea to protect the expensive HF power transistors, but it should not have been used to control power supply overshoot.

[bostonman]

I'll read that app note, thanks.

I don't have much experience with measuring startup voltages in power supplies to know if many spike at startup, but it makes sense they'd be robust enough to not overshoot.

Years ago I built a linear power supply for my C64 and incorporated a SCR (see attached) thinking I could keep if spikes occur, it would clamp and blow the fuse before any spikes that exceed the threshold damages the ICs in the C64. The power supply worked for about two power cycles before the fuse blew.

Long story short, I never figured out why the power supply spiked at turn on, but I eliminated the circuit which resulted in hacking the power supply. Since then, I've always had concerns of turn on spikes.

Note: the two 1k resistors in parallel is to discharge the capacitor when the C64 turns off.

[David Hess]

I don't have much experience with measuring startup voltages in power supplies to know if many spike at startup, but it makes sense they'd be robust enough to not overshoot.

If you are doing your own design then it should be tested under real world conditions.  Digital storage oscilloscopes are really good for capturing the startup characteristics.  Some designs, like the Kenwood one I repaired, will need added elements to either protect the load or to solve the startup overshoot problem.  On the Tektronix PS503A power supply, there is an added two transistor circuit which clamps the reference until the control circuits are fully powered to prevent overshoot on the outputs.

Long story short, I never figured out why the power supply spiked at turn on, but I eliminated the circuit which resulted in hacking the power supply. Since then, I've always had concerns of turn on spikes.

There are two common causes:

1. The control circuit may not be properly powered at first allowing the output stage to run away.

2. During startup, the output voltage is low for an extended time during which the control circuit may suffer from "windup", which is a common problem when an integrator is part of the frequency compensation.  Old designs may avoid this problem by using old and slow operational amplifiers which do not require extra external frequency compensation.

[Zero999]

You can use a TVS as well as a latching crowbar.

Say the supply voltage is 12V, 15V absolute maximum.

Set the crowbar to clamp at 16V to 17V and add a low-pass filter with a time constant of 100ms or so. Add a suitable fuse and low value resistor if the prospective short circuit current is too high.

The 18V TVS will now clamp, short, high voltage spikes, but if the device is subjected to continuous over-voltage then the crowbar will be triggered and the fuse will blow.

[bostonman]

Tonight I took startup (and power off) oscilloscope captures of both the +12V and -12V DC.

Everyone was correct, it didn't seem to have any overshoot and started fine (see attached pictures).

Is the 5.0SMDJ12A still a good choice since I'm seeing 13.7V max (per the voltages the oscilloscope measurement feature shows)?

I zoomed in on the scale and never realized the oscilloscope grid seems to not make sense. I'm counting 17 lines between the majors of 10V and 15V, so I'm using the measurement feature. Normally I just give a quick glance, so I never realized the wonky minor grid count.

Looking closer, I guess maybe the dots are just the grid and not actual minor grid lines that are used for measuring. Now that I looked again, I guess just count the majors which making it 1/4 steps.