Voltage Clamp with Alarm

[HarryDoPECC]

I restore vintage calculators, those of late 60s-early 70s can have several hundred early DTL and TTL ICs, many now unobtanium. Also generally have linear power supply regulators and at least one of my machines has dozens of failed ICs (some actually blown open) after a pass transistor failed short.

One way to protect from such events while maintaining originality is to put a volt clamp on vulnerable supplies.  The intent is to cause a hard short on overvoltage with beefy enough load so as to blow the original fuses, or sometimes add a new fuse if the existing fusing looks under-done.  For that matter blowing up the power supply is perhaps preferable to frying 200 ICs!  My design based on something from AoE, with the goal of a pretty well-defined and hard turn on.  This circuit (without R5, Q2, R8 and LED) has served me well enough, mindful that the load R, TIP122 and local environment will suffer until fuse blows or other help arrives.

My problem arose when I wanted to add an alarm for a particular situation.  Adding Q2 and associated parts gave a visual indication when the clamp was on. But if I substitute (R8 + LED) with a piezo buzzer, I get odd results.  (The buzzer sounds with 5-20VDC, drawing 20 odd mA, it tests open one way and high R the other, so I think its piezo with driver).

With the buzzer I get sound as the clamp turns on, and for a volt or so of overvolt, but as the input voltage rises even more the sound goes off, and will not return even if input taken to low level and then raised again.  I'm puzzled because the clamp action keeps the voltage on circuit pretty constant, and I'm not seeing how the increasing current thru R7/Q3 affects buzzer operation.  Is that buzzer adding another effect I'm unaware of?.

Hoping for some pointers to what I need to learn to get this working.  Thanks in advance.

[mikerj]

I think you have some more fundamental issues:
1) The whole circuit is powered from the fused side of the supply, if either fuse is blow it will become unpowered so there is no possibility of sounding an alarm or lighting an LED.
2) You have no latching action or even hysteresis in the circuit, this is quite likely to behave more like a shunt regulator and dissipate a lot of power in R7/Q3 if the voltage on the supply rises.

The circuit you need is called a "crowbar" and is typically implemented using a suitably beefy SCR to ensure that after triggering the crowbar remains active until the fuse is blown.  If you check the various TL431 datasheets they often contain an example crowbar circuit.

[HarryDoPECC]

Thanks for comments.
The typical use case is for the clamp to turn on hard until the fuse blows, then all is quiet.  Investigate later why power off with blown fuse. 
The alarm function is a special use case where the only fusing is on the primary side - if a regulated voltage drifts up a bit then the clamp could open a bit, increasing transformer load but without blowing the fuse. This is why I want the audible alarm that clamp is open(ing).
I use transistor and load resistor in preference to SCR because some of these old regulators can be slow to settle and a locking crowbar effect can be counterproductive, latching and blowing the fuse when the regulator would have settled. So I choise this scheme in preference to SCR.
I agree a bit unconventional, but I think suits my use case.

But I'm puzzled why it misbehaves if the visual warning is replaced with audible.

[mikerj]

If you are concerned about false triggering on small transient overvoltages then use heavier low pass filtering e.g. make C1 higher in value and add a suitable diode from the cap to the +ve rail to quickly discharge it when the unit is powered down to keep the transient filtering working in the event of a fast brownout condition.  However a latching circuit or large hysteresis is important, when Q3 turns on it applies a heavy load to the power supply so the voltage will drop, which may be the cause of your observations.  If there is enough phase delay in the system you might get an oscillator, or it might settle like a shunt regulator.  Crowbar protection has been around for a very long time and is a well proven technique, but not using a non-latching design with a linear element as the shunt.

Is it really necessary to have two fuses? If this is a single rail supply, which the circuit suggests, then a single fuse will provide all the protection you need and simplifies the implementation of an alarm. An LED or piezo sounder can simply be connected across the fuse and will sound when it is blown, the small amount of current required to drive it will be easily passed by the load.

[HarryDoPECC]

Thanks again, appreciate the interest

What I showed was the guts of my puzzle, but of course there is more...  This is a building block, sometimes using two or more if perhaps there is a + and - rail both vulnerable to pass transistor failure.  The schematic was also feeding a PCB layout, hence provision for fuse in either leg, high or low side according to application. But not both at same time.

I've been satisfied in the past with using these blocks on + or - rails that are independently regulated, sharing a GND, and no ill-effects if one supply is cut off. The clamp action will ride through a brief over-volt and then blow the fuse in the original circuit, or one that I provide.

Where it got messy was with a particular supply where +5 supplies 200+ TTL ICs, and -12 supplies a bank of MK4008 first generation static RAMs.  Both vulnerable to pass transistor shorts and wanting to protect these ICs.  But the original design regulates the +5 and then references the -12 regulation to the +5, so that -12 tracks +5. And if the +5 goes away then the -12 regulator loses its mind and cooks it output stage. Found that out the hard way. And +5 and -12 come from separate secondary windings. That is perhaps the reason why the original design has no fuses or other protection on the secondary side, only a fuse on the primary.

And that is why I wanted to add a (preferably audio) indication that the clamp(s) were open(ing). I'm not certain that shorting out any one of the secondaries is enough to blow the primary fuse in sufficient time to protect the transformer, and I don't really want to test that.  Also help with a deficiency of this approach, where the clamp may open partly, stressing the transformer and supply, but not enough to blow the primary fuse.  An SCR would avoid this and maybe I should use that method here.

But I'm still wondering why that piezo buzzer threw the whole thing off.  Was hoping someone would point me an an effect I was unaware of.