TVS and Avalanche Diodes

[m00dawg]

I started reading through epic thread again (which really should be sticky, I think) while specifically looking for circuit protection techniques. A crowbar circuit was discussed but I also ran into one design that put a diode between +V and GND after the regulator instead. After reading up on things (here, here, and here, see Jay2's comment to name a few), I think I have it figured out.

Mouser has an entire section devoted to TVS Diodes and it looks like they have fairly specific voltage targets. So I think this is the thing to use over just a 1N4004 or Zener diode? I have seen designs using just an 1N4004 and according to what I have found, using one may work, but could destroy the diode in the process as they are not built to sustain avalanche breakdown?

I also want to make sure I understand things right, during normal operation, the TVS diode stops current flow to ground but since current is now flowing through (like it would be in-line), there is no voltage drop on the supply. If the breakdown voltage is reached, current explodes through the diode in the opposite direction, causing a short to ground which then should blow a fuse. One thing I read is that avalanche diodes are also used in situations that need to generate noise. I assume that is only when the breakdown voltage is met or exceeded (and that they do not contribute adversely to noise otherwise)?

Anyone have any thoughts? I don't want to just stick a random diode somewhere without knowing the right kind to use and its intended purpose. The designs that use a 1N4004 in this way confuse me given what I have read about those diodes compared to TVS / avalanche diodes.

Attached is my current design with this in use. My plan is to make this a sort of template board that I can use and plug in various values for the resulting voltage and load. It's sort of the basis for a semi-modular PSU design I am working on for my band's live stage rack. Since my requirements can change as I add more components, trying to have something modular (at the expense of space) is worth it over trying to design one big PSU that has all the voltages I may ever need. Or so is the idea.

Thoughts or suggestions?

[Shuriken]

First of all i am not an expert. I hadn't heard of TVS before. But a TVS looks like to be protection from only a short spike.

I did do a bit of research on crowbar designs. You can make a simple crowbar with a zener and a scr. The downside of a zener is they only come in specific voltages (same as TVS as it seems) which makes it a bit inflexible. So there are are more intricate design which utilize an IC instead of a zener. Onsemi makes a couple of good ones.

But in the end it all comes down to how much it costs to replace what is behind the PSU. Is it worth the additional cost of the protection. The main reason i started to look into a crowbar is that 8 x SID is a lot of money. I don't want to blow it up with a 20 year old C64 psu.

Edited August 8, 2012 by Shuriken

[m00dawg]

I haven't seen the diode used as much for non-polar supplies but I have seen it used for the purpose I explained, including some documentation from a few college level lab courses so I think there is some purpose there for it. I tried busting a circuit simulator to see if I can determine what's going on. Without a way to emulate a fuse it was hard to fully validate it but I was able to break the simulator a few times with it so I think it works ;)

From what I have read, TVS tends to be an optional replacement for a classical crowbar design - the two are meant to do the same things it would appear. SCRs are similar (looking at the Wikipedia article you sent, they both involve avalanche breakdown). I think I was confusing these with thermistors so I'll have to look at those a bit more I think. Either design seems reasonable since they are both meant to control over-voltage (which should never happen unless there is a component failure?)

As far as cost of the protection, it's a pretty inexpensive addition so I think it's good to go ahead and put it in. For cases where failure is less of a concern, one can just back off on the tolerances to avoid accidental tripping. But for a couple bucks worth of parts, it's WELL worth it for the cost of those hard to replace chips like SIDs or the OPL3 chipset.

EDIT: Oh I forgot to ask - I was trying to look at some IC-based designs and wasn't able to find much when looking at Mouser. Do you have any suggestions for some reference designs?

Edited August 8, 2012 by m00dawg

[Shuriken]

I bought some MC3423 at reichelt. However that IC is EOL. I seem to remember someone found an up to date solution from ONsemi.It was technobreath as i suspected

Edited August 8, 2012 by Shuriken

[Hawkeye]

Me findz it very hard to believe, that there aren´t any modern linear vregs with built-in output protection?!?

Have got the zener triggered crowbar in my mb6582 psu, too, but as you say, the trigger voltage may be already a bit high... which logic voltage is considered fatal for SIDs?

Many greets,

Peter

[Shuriken]

Me findz it very hard to believe, that there aren´t any modern linear vregs with built-in output protection?!?

Have got the zener triggered crowbar in my mb6582 psu, too, but as you say, the trigger voltage may be already a bit high... which logic voltage is considered fatal for SIDs?

Many greets,

Peter

You used a 5.6V Zener for +5V circuit? That has a 5% precision. So max. Voltage passing through is 5.88V. Datasheet SID6582 says VDD=9V+5%VDC and VCC=5V+5%VDC.

Edited August 8, 2012 by Shuriken

[Hawkeye]

yupp, i think i can remember testing it and it being triggered just below 6v... so you think the SIDs would survive that for a short time? :nuke:

Edited August 8, 2012 by Hawkeye

[m00dawg]

yupp, i think i can remember testing it and it being triggered just below 6v... so you think the SIDs would survive that for a short time? :nuke:

Worthy of giving it some board space just for the chance is my thought :)

Wish I was able to find a good comparison between the Avalanche Diode as I have it versus a crowbar. The two seem the same but Wikipedia suggested that Avalanche Diodes can react a bit faster.

[Hawkeye]

Worthy of giving it some board space just for the chance is my thought :)

of course it is already operational :) only the best filtered, clean, safer power juice for the SIDs :)

The avalanches sound interesting, too :)

Edited August 8, 2012 by Hawkeye

[m00dawg]

of course it is already operational :) only the best filtered, clean, safer power juice for the SIDs :)

The avalanches sound interesting, too :)

Indeed! Worthy of testing anyway. Curious - how did you end up dumping 6V onto your circuit to test the crowbar?

[Hawkeye]

Indeed! Worthy of testing anyway. Curious - how did you end up dumping 6V onto your circuit to test the crowbar?

Tested the crowbar-circuit (the retro-donald-psu one) with a voltage-adjustable lab PSU... after successful test, connected the Vreg outputs to the location where i had attached the lab PSU for testing...

Many greets,

Peter

Edited August 8, 2012 by Hawkeye

[m00dawg]

Aha! Guess I need one of those as well as an oscope :P

[Hawkeye]

Hehehe, true :) (both were cheap, you can build a lab psu on your own for a few bucks, there are lots of schems out there).

A cheap second-hand electron tube osc is really something. Not that I use it a lot, but it can be handy for attracting cats :)

Edited August 8, 2012 by Hawkeye

[m00dawg]

Hehehe, true :) (both were cheap, you can build a lab psu on your own for a few bucks, there are lots of schems out there).

A cheap second-hand electron tube osc is really something. Not that I use it a lot, but it can be handy for attracting cats :)

Hah! Agreed! I guess for a 5V supply I can use a battery pack just as easily to hit 6V but I'll need something more variable for my 9V and 12V supplies.

SmashTV recommends this and it looks pretty much Win for the price, but I may have to settle on a used old-skewl one for a little while. Prices went up on eBay since the last time I looked though :/

[Hawkeye]

while the new one sure haz hot specs i am a real fan of those old tube based ones :) mine has only got 20mhz but that is ok... around 50 bucks or sometimes for free from those people upgrading to the new tft based units :)

[m00dawg]

while the new one sure haz hot specs i am a real fan of those old tube based ones :) mine has only got 20mhz but that is ok... around 50 bucks or sometimes for free from those people upgrading to the new tft based units :)

20MHz is enough to look at signals for audio applications though, no? I guess once you have to start looking at logic of fast switching components it might start to show its age? Mostly for me I need one for PSU stuffs, although I guess I could see it being useful to debug sick chips too.

[Hawkeye]

yes, if you want to debug modern microcontroller logic 20mhz is too low :) but as you said, totally sufficient for audio stuff and all kinds of old-school circuits :)

Edited August 8, 2012 by Hawkeye

[Shuriken]

Ok, i have been reading up a little bit on a TVS. Conclusion a TVS is not a replacement for a crowbar.

a TVS is meant as protection for spikes. So for instance an ESD. It can withstand a peak current for a very short time (max 1ms) after that period the amount of current it can withstand dramatically decreases as it heats up. And eventually will be destroyed. A TVS will only clamp at the rated voltage and never become a short like a crowbar.

A good example i found about the differences between the two: If you have a 15V supply and a 20A current. The TVS will basically have to dispensate 300W in a problem situation. Whereas a SCR will drop the current to 1A and has only 20W to dispensate. You will need a mighty heatsink to dispensate 300Wattz :wink:

In one of the datasheets i found that one of the applications of a TVS can be to protect a SCR from (false) triggering. So i guess it is not about one or the other. But they basically complement each other.

Edited August 9, 2012 by Shuriken

[m00dawg]

Ah that makes sense. Though 300W is a lot. I think for my design it would be closer to 50-100 max, although that's still quite high for such a small device, noted. Still, for what it does provide, it's pretty easy to just drop one in where it is.

Having both does make for an interesting situation though since now I have to make sure one triggers in certain cases over the other. *scratches head*

Well looks like there is more research to be done! Thanks for all the helpful info!

Edited August 9, 2012 by m00dawg

[Shuriken]

Ah that makes sense. Though 300W is a lot. I think for my design it would be closer to 50-100 max, although that's still quite high for such a small device, noted. Still, for what it does provide, it's pretty easy to just drop one in where it is.

Having both does make for an interesting situation though since now I have to make sure one triggers in certain cases over the other. *scratches head*

Well looks like there is more research to be done! Thanks for all the helpful info!

I agree the example is a bit over the top in comparison to our usual designs. But a 5V TVS looks like a pretty small device, with no heatsinking. So even 20W will be to much to handle. It will probably result in a big burn on the pcb :wink:

Tbh i think a TVS is only usefull in special enviroments. Like environments which have instable power. Or with mobile devices, which are not grounded. Which can be more vulnerable to ESD.

For a net powered MB6582 psu, i think a crowbar will be sufficient.

Edited August 9, 2012 by Shuriken

[m00dawg]

I agree the example is a bit over the top in comparison to our usual designs. But a 5V TVS looks like a pretty small device, with no heatsinking. So even 20W will be to much to handle. It will probably result in a big burn on the pcb :wink:

Probably so :) I'm actually looking at the ON MC34161 family of chips Technobreath mentioned in his crowbar circuit design gallery (linked up above) to see if I can use that to help with a crowbar. Pretty neat in that it can alert to over and under voltage, but I haven't quite figured out how to use them yet (and I don't know how to power it ;P regulated or otherwise).

[Shuriken]

Probably so :) I'm actually looking at the ON MC34161 family of chips Technobreath mentioned in his crowbar circuit design gallery (linked up above) to see if I can use that to help with a crowbar. Pretty neat in that it can alert to over and under voltage, but I haven't quite figured out how to use them yet (and I don't know how to power it ;P regulated or otherwise).

Not sure tbh. The MC3423 datasheet is quite clear. However reading the MC34161 datasheet i am not so sure it will or can do the same.

Edited August 9, 2012 by Shuriken

[m00dawg]

Not sure tbh. The MC3423 datasheet is quite clear. However reading the MC34161 datasheet i am not so sure it will or can do the same.

Looks like the application is mostly the same, although the newer chip has some nice features - and happens to already be in Eagle so that's a bonus :) Only concern I have is the accuracy of the voltage divider, especially with 12VDC. Can get high precision resistors, but still curious what the percent error might be. I suspect small enough to be useful. Using a voltage divider also means those only need to be changed to handle different voltage inputs, which is nice.

[m00dawg]

I think I finally figured out how to use the MC34161 along with the clamp. You can do some cool things like having under and over voltage thresholds but I have used it for over-voltage only. I would guess a discrete circuit that doesn't use a voltage divider is probably a slightly better design. I'm not sure how the IC compares to solutions that do use resistors, but at least I can swap the chip out if it fails.

I removed the avalanche diode since the IC does basically the same thing. I might put it back so it handles small voltage spikes while the IC handles larger ones, but I'm not sure.

Anyways here's the new schematic.