Help with my power supply schematic...

[Brian Roth]

The first "real world" crowbar I recall (back when the earth's crust was still cooling <g>) was used in the power supplies of the Ampex MM-1100/1200 multitracks. I just looked through my 1200 manual, and they indeed had the crowbar at the input to the regulator. It was a simple circuit...a zener diode feeding a series R/shunt C that in turn fed the gate of the SCR. A big power resistor (1 Ohm, 10 W) was in series with the SCR.

OTOH, the optional OVP modules available from Power-One, International power, etc for their linear open-frame modules do connect across the output terminals of the regulator.

Bri

 

[T-Dogg]

Sorry to keep dragging this thread up. Everytime I have this thing behind me, I wind up fumbling over a new problem. I want to impliment the suggested crowbar circuit on the 8v and 5v rails, and I'm wondering if the following sounds 'bout right:

--An SCR strapped across the 5v (or 8v) rail (before the regulator.) To trigger the gate, I was gonna take the regulated voltage feed through a Zener with breakdown of 6v (or 9v) to trigger the SCR gate... Are 6v and 9v too close to the average regulated voltage, that I might risk triggering an SCR under near-normal operating conditions? What's a good margin that sets it far enough outside the typical operating voltage, without allowing too much voltage to slide through?

--I'll need a big ol series resister between the +/- legs of the SCR to discharge the main filter caps. I'm guessing anything big and high-wattage will do here.

--The R/C network feeding the gate of the SCR is where I get really lost... What precisely does it do, and how does one choose values?

--Fusing the rails... I understand this happens before the SCR... So keeping in mind that my 5v and 8v rails are derived from 18v -- I'll really be fusing the 18v line. Is this good/bad/doesn't matter? I guess it's all about amperage, what would generally be a good amp for a fuse on a 5v or 8v line in a typical mic pre? I really have no idea what the current draw on these rails is, but probably small, right?

Big thanx guys!

 

[Svart]

Using a SCR is one possibility here but adds unneeded complexity IMHO. It would be simple enough to use some MOVs before the trafo, A TVS after the bridge but before the regs, current limiting resistors and/or resetable fuses after the regs.

My vote is for resetable fuses on the rails after the Vregs. This is by far the most used form of protection as they work faster than any SCR and will stay "open" as long as the short is there. this should save your regs while killing the rails by moving the protection to the output side of the PSU instead of allowing the short to possibly damage the Vregs and other parts.


and i'll give you a bit of info that I found out the hard way.. Please don't be tempted to use the cheapest Vregs you can find. No-name Vregs tend to die without provocation and will sometimes die immediately during powerup leading you to believe that there is a problem elsewhere.

Also some name brands have died on me without provocation too.. Beware of the letters S and T..

Motorolas tended to work the best when you could still get them.


:thumb:

EDIT:

Since I work a lot with SCRs in power switching, I'll give you some hints.

SCRs are not truly gate controlled although many lead you to believe this. They turn on but as long as a certain amount of current flows through them they stay in the conductive/ON state. This means they cannot be turned off via the gate. They can also be triggered to turn "ON" by exceeding a specified breakdown voltage regardless of gate current. They also exhibit the diode voltage drop so you'll need to account for the .5+v of drop.

SCRs are current driven but need a certain potential also. They MUST also see a potential and current on the anode in excess of the cathode before it can be turned "ON". this is specified in the datasheets also.

SCRs are slow in respect to other devices like MOSFETS, TVS and Picofuses.

there is much more but i think you'll get the picture.

:thumb:

 

[Brian Roth]

Svart.....ahhhhh...the issue was overvoltage "shut down" vs. "over current", unless I misread your post.

One HUGE mistake in many/most power supplies is ignoring the possibility of a regulator failure that, in turn, results in a HUGE raw DC rail pouring into the downstream circuitry.

I note that most of the 5VDC supplies (linear and switching) include OVP as a standard feature.

I'm in the process of working up some supplies for a Quad-Eight desk that used Unobtaium "hybrid" modules. The new supply will use International Power modules since I have the "official" start-up mods from them (Power-One won't return my phone calls), and I willl include the OVP add-ons:

http://gatewayto.com/cgi-bin/start.cgi/intl_power/products.html#ovp

Bri

 

[Svart]

ah ok I just thought you wanted general protection..

I would go for a TVS before the Vregs.. they also have single part crowbar devices rated in the amps *if* you must have it.

The addition of extra parts to "protect" can become overblown quickly.

First things first, your supply should be designed for worst-case scenarios. what is the absolute most current you will need. triple it. now build a supply based on that number. choosing a voltage to run at is secondary but shouldn't be more than 10v over the target voltage and no less than 3v over the target voltage when rectified. This ensures your Vregs are NOT seeing too much voltage and are not generating excess heat and are not going to see too little voltage to regulate. If you selected a large VA rating you not only will cut down on transformer-current related noise but you are guaranteeing that it will have sufficient current to support the voltage at any level of draw.

I say use TVS devices and MOV devices and you'll be more than ok without adding active circuit complexity.

as for Vreg failures that short HV downstream. yes that is possible, some also fail open too. However if you ensure that you aren't supplying too much voltage to the Vregs to keep them from heating up you are also going to ensure that you aren't supplying too much raw voltage downstream. output current limiting resistors and TVS on the output should be enough from there. reversed biased zeners a little over the intended rail voltage should help too if you are really worried about it.

so it should go like this: line in, fuse/breaker,line filter, switch(if needed), MOV(s), trafo, MOV(s)(if you really want them here), rectifier(s), filter caps, TVS(s), Vregs, filter caps, TVS(s), output picofuses.

between all that stuff and the design behind it you should have next to no failures unless you use cheap parts.

Research SIDACS too.. they might help too.

EDIT: a quick look at some TVS parts made by Di*des inc. state average response times of their TVSs at one picosecond. :shock:

 

[Brian Roth]

I have become very paranoid about overvoltage protection due to many bad experiences with equipment (including recording desks costing 5 or 6 figures when new) that had a failure in the regulator.

I fully agree with keeping the "raw" rail's voltage to a reasonable voltage relative to the desired output. One 32 x 24 desk I serviced had close to 30 VDC "raw" into the 16V regulators. Needless to say, when that failed, hundreds of ICs in the desk literally burnt to a crisp. That was indeed a VERY poor design which resulted in thousands of dollars in repair bills.

That was NOT an isolated incident, either. I have run into in many other times, although not to that extreme (more like "raw" 24 VDC appearing on a nominal 15 to 18 VDC rail to the equipment).

Nevertheless, allowing, say, 10 VDC above the target regulated voltage for the raw rail can still cause a bad calamity when the Absolute Maximum voltage rating of a chip is exceeded. Example. TL07x chips are as common as popcorn in LOTS of gear, and these have an 18VDC Absolute Max. rating on each rail. If these chips are normally powered at 15 to 17 VDC (a common usage), and the raw rail was, say, 24 VDC (well under the 10 V max discussed), you will be significantly above the Abs. Max. if a regulator fails.

This is even a larger issue with Olde TTL parts, which operated from a nominal 5VDC, with a 7 VDC Abs. Max. To ensure against loss of regulation due to AC Mains sags, and to stay above the dropout rating of the regulator, probably at least 10 VDC "raw" is necessary.

Hence why I am pushing OVP "crowbars". Perhaps a very LARGE zener (ie, high wattage) across the output could be an alternative. I guess I would need to see if I can calculate how many Watts it would dissipate with a shorted pass transistor. However, for some reason, all of the folks such as Power-One, International power, Condor, etc. use crowbars, either as "stock" in their 5 VDC linears, or as an option for their other models.

One other item not yet mentioned...proper thermal design. Once again, I've seen FAR too many pieces of equipment that had power supply components running hot as a firecracker (to the point where the green circuit board(s) turned brown).

Bri

 

[Svart]

:thumb:

I'm not trying to disprove or negate your design in any way!

For commercial devices that might see extremes an active device will work well if the budget allows. TVS technology has come a long way in the last few years and I am a huge supporter. there are low voltage high current TVS devices for 12, 15, 18, 20, 24 etc voltages up to 1500 watts.. more than enough to pulldown unruly voltages through a shorted closed regulator. These will also work in a thousandth of the time an SCR based feedback driven circuit will activate and do it's work.

 

[Brian Roth]

I will check those TVS devices out! The ones I recall were little things intended for clamping transient spikes vs. CLAMPING down an errant rail that may have been designed for 10 or 15 Amps, and have 1000's of microFarads of filter caps that have to be safely discharged.

Hmmmm...thinking out loud here....

IIRC, the TVS diodes are basically zeners? OK, let's assume a raw rail of 24 VDC and a desired regulated output of 15 VDC, and a TVS or Zener at the output of the regulator rated, say 18V. Assume now the pass transistor fails as a dead short, and the effective series resistance for the raw source (transformer winding R, series R of the rectifiers) is likely under 1 Ohm.

In my mind, the TVS/Zener's reverse resistance (ie, when conducting) will define how much current that part has to dissipate. And, it WILL have to dissipate that amount of power until/if the AC mains fuse pops.

Maybe there are some White Papers I need to read.......

Bri