TVS limitations-are they really that good?

Thread Starter

BB Kid

Joined Aug 22, 2017
10
Greetings all, this is my first post. I've used the website off and on for awhile, but today I signed up for an account and I hope to contribute as well as receive help.

I have been following the saga of the USB Killer (USB Kill.com) observing folly after folly in its evolution. It's use is controversial, but it's purpose of encouraging OEM's to include TVS in new designs is probably valid.

My question involves TVS diodes. They've been around for a long time, early CMOS chips had them built in, but they didn't offer much protection for many reasons-probably the most significant of which is the built in voltage limit associated with the substrate structure that is built into all CMOS devices.

So, today we have really small TVS packages that are not built into (onboard) the device needing protection.

There is much data on the web, but are TVS diodes all they are cracked up to be. They have to be installed in parallel as any resistance in series with a CMOS device severely reduces its throughput- due to the time constant from the chips input C and the series R.

The makers of TVS diodes have glowing reports of their usefulness, but….

Along comes the USB Killer-argh. It works fine to demonstrate the need for CMOS input and output over voltage protection. BUT-the model used to define the worst case static charge is far outdated and doesn’t address the power dissipation of the diode when a constant over current and voltage is applied.

TVS stands for 'transient' and the TVS diodes work well as long as multiple or continuous/malicious over voltage attacks are perpetrated.

So...my question, finally::>

How do these diodes dissipate power if the input voltage is high enough to make the diode clamp? In such an attack, don't these diodes melt from heat dissipation quickly??

With very low input capacitance associated with the TVS diode, the junction cannot be very robust. And, they're not mounted on heat sinks!!

How long can a 5V rated microscopic diode junction dissipate 100 watts if the clamping current is a constant 20A (for instance)?
 

nsaspook

Joined Aug 27, 2009
13,086
I don't know about the USB Killer but TVS diodes work just fine to effectively short power systems and to activated protection circuits in power supplies.
I just replace some TVS protection diodes from a large direct drive motor system. The spin controller fried while a large disk was spinning at 1200 rpm, This sent a large overvoltage to the driver DC buss for three other motor controllers. The TVS diodes clamped the DC supply causing a short that blew the DC buss class T fuses. This also vaporized the TVS diode but it saved the other amps from critical damage. Cleanup the burn-mark, replace the diode and boom, we are back making money.

http://www.mouser.com/ds/2/240/Littelfuse_TVS-Diode_1.5KE-41903.pdf
 

Thread Starter

BB Kid

Joined Aug 22, 2017
10
Thanks NSA.

What a coincidence, I think great minds think alike! Your reply came in as I was looking at Mouser, pricing TVS diodes! Dang, they are all in these microscopic packages, my eyes ain't what they used to be::>

I'm thinking I should buy some diodes and test them on actual circuits that use unprotected I/O.

Regards,

BB
 

dl324

Joined Mar 30, 2015
16,846
Greetings all, this is my first post.
Welcome to AAC!
My question involves TVS diodes. They've been around for a long time, early CMOS chips had them built in, but they didn't offer much protection for many reasons-probably the most significant of which is the built in voltage limit associated with the substrate structure that is built into all CMOS devices.
Technically, these aren't TVS diodes. Input clamp diodes are for low power ESD protection. TVS diodes are relatively high power diodes.
How do these diodes dissipate power if the input voltage is high enough to make the diode clamp? In such an attack, don't these diodes melt from heat dissipation quickly??
They are constructed using multiple matched zener diodes in parallel.
 

Thread Starter

BB Kid

Joined Aug 22, 2017
10
Hi DL/Dennis, thanks for the warm welcome, I'm glad to be here!

The early protection (circa 20 years ago) were indeed zeners. But, zener diodes are painfully slow to clamp and they don't clamp hard enough to protect the inputs of a cmos device. And, they have large input capacitance, so they do not clamp fast enough to protect the mosfets in the cmos devices.

In fact, zener diodes have so much input capacitance that they are routinely used a varicaps, which is a voltage controlled capacitor!

Modern TVS diodes have (typically) around 1 pF or less input capacitance and they switch to and from conduction in nanoseconds and conduct 100's of amps. The extremely low input capacitance is what enables the TVS to turn on so quickly. If a TVS diode were made up of parallel zeners, the input capacitance of the diodes in parallel would make them horribly slow. I think the TVS is not just parallel back to back zeners.

Even if TVS diodes are back to back parallel matched zeners! The types of packages they come in cannot dissipate the power applied with a sustained voltage that is in excess of the clamping/breakdown voltage (I think). The TVS diodes have small pn junctions, which is why and how they switch so quickly. And, (I think) they can't dissipate the power from a sustained attack.

I think a zener diode placed in parallel with a cmos logic gate or discrete mosfet would severely limit the switching speed and thus wouldn't be usable for many types of devices.

Unfortunately, the actual construction of the devices are highly proprietary so we (as users) don't really know how they are fabricated. I think LTSpice has zener and TVS models, it should be possible to quantify the extreme speed loss (and increased prop delay) that results if a zener is used in place of a TVS.

Back to the drawing board.

Regards,

BB
 

nsaspook

Joined Aug 27, 2009
13,086
There is nothing secret about the construction of TVS diodes. The exact recipes for a machine set are highly proprietary but the solid state physics are not.
Just look for patents for 'Low capacitance two-terminal barrier controlled TVS diodes'.
 

dl324

Joined Mar 30, 2015
16,846
The early protection (circa 20 years ago) were indeed zeners.
I have never seen a CMOS IC that used zener diodes for input protection.

This is what was typical 40 years ago:
upload_2017-8-23_11-45-14.png

Modern TVS diodes have (typically) around 1 pF or less input capacitance and they switch to and from conduction in nanoseconds and conduct 100's of amps. The extremely low input capacitance is what enables the TVS to turn on so quickly. If a TVS diode were made up of parallel zeners, the input capacitance of the diodes in parallel would make them horribly slow.
TVS diodes are designed to clamp transient over voltages. Some are designed to dissipate 10's of thousands of watts for brief periods.
I think the TVS is not just parallel back to back zeners.
They aren't parallel back to back; just parallel. Google them and see for yourself.
Even if TVS diodes are back to back parallel matched zeners! The types of packages they come in cannot dissipate the power applied with a sustained voltage that is in excess of the clamping/breakdown voltage (I think).
Look up the definition of transient and sustained. TVS diodes are not for sustained over voltage suppression.
The TVS diodes have small pn junctions, which is why and how they switch so quickly.
What makes you think the junctions are small and that switching speed is important? TVS diodes aren't for switching circuits.

upload_2017-8-23_11-58-16.png
upload_2017-8-23_11-57-56.png
 
Top