Golborne Vintage Radio

Many of us have systems that have several mains-powered boxes that have to be turned on & off together, but which may or may not have individual power switches. A case in point is my recent acquisition (written originally in 2011!) of a Quad 405-2 - fine if you are driving it off a switched outlet from a pre-amp, bit of a pain if you are not.

You can buy socket strips that have master-slave outlets where you connected the pre-amp/TV/whatever to the master outlet and the rest of your kit to the slave outlets, but I've found these unreliable and sometimes, if you have an SS pre-amp, not sensitive enough... the ones I've played with require a minimum of about 10W load to switch.

So, I did the following:

Based on an idea I'd seen before, I modelled the use of a zero-crossing solid-state relay (SSR) driven by the voltage drop across some a diode. There are two ways of getting sufficient voltage & current to drive the SSR - either use 3 diodes or use a small multiplier. The multiplier approach has the advantage of dissipating less heat under load at the cost of more of the larger components (capacitors) and the multiple-diode approach is the inverse of that.

[attachment=12336]

I've gone initially for the multiple-diodes route as I've built my controller into a Bulgin 10A/6 way IEC 320 C13 outlet and only had about 1" x 1.5" of space to play with. I've also fused the unit at 5A as the SSR (an S202S02 - cheap-ish and good) is rated at 8A - 5A should be plenty. The snubber values are calculated for an inductive load (transformers) as per the SSR data sheets.

Both models work well, with the RMS values through the SSR LED being about right, so I drew it up in Eagle and built it:

[attachment=12335]

The idea is really simple: When current is drawn by the master socket, a voltage drop of about 2.8V is developed across D1,D2 & D3 (3 x Vf). This is half wave rectified and used to turn on an SSR which in turn powers up the slave socket. D4 provides a return path for the other half of the cycle. D5 is a Schottky to limit forward voltage drop and thus limit losses (current though it is tiny). C1 & R1 are a snubber to provide some protection for the SSR if you are using an inductive load, e.g. an extractor

R3, C3 provide a voltage dropper for a "slave enabled" LED - D6 limits the reverse voltage over the LED to 0.6V (LEDs don't like reverse voltage) - R2 & D7 are the slave-enabled LED - all these components can be removed if you don't want the slave-enabled indicator. C1 & C3 should be X2 class...

The reason the snubber capacitor is a 10n X2/390R rather than something larger like 100n/100R is that Zc at 50Hz allows a good few mA through the snubber even when the switch is off - not good as it would be enough to light the "slave on" LED as well as leaking into the controlled kit. Using 10n provides good snubbing and low leakage.

When the master turns off, the slave turns off.

The whole thing is really simple and cheap and can be built into a power strip without much bother - I have a few of these around for various reasons; the one in the workshop has 3 masters and one slave - the slave is on the extractor, the masters are connected to various power tools. I also use them in the house to automatically turn off power amps when the preamp is turned off... You can split the strip in any ratio you like. The Sharp SSR are only about £3.50 at Farnell and are really neat devices - the whole thing probably costs less than £10 to build...

I really like this circuit

Looks really good

I've done similar things in the past, but "discretely" with an MOC3049 opto-coupler driving the gate of a triac directly. The MOC3049 takes care of the zero-crossing switching, and is dead cheap. At the time, I was stuck for space and couldn't find an SSR that was small enough, but the S202S02 would probably have done the job if I'd discovered it (was it around in 2004, or did I just fail to notice it?). The MOC3049 has gone obsolete since then, but there are many suitable equivalents...

I think I'd want to add a resistor feeding into the diode in the SSR. Yes, there might be a penalty in voltage drop terms (might need a 4th diode), but that poor LED is a slave to the Vf drops of D1, 2, 3 and 5. Can you predict exactly what the LED current would be in all circumstances? Not sure I could

Doing so would allow you to use a smaller capacitor, as its TC is currently set by the slope resistance of the LED (guess a few 10s of ohms), and because there is no "padding", the capacitor only has to discharge very slightly between charging pulses before it goes out (which is no-doubt why it ended up being so large in value).

Just 100 ohms or so might make the difference - the voltage dropped would swamp the variations in the Vf of the various diodes and the LED itself.

In the past, I've idly considered using a current transformer for this. Passively achieving the required sensitivity while still dealing with the highest current could be challenging, to say the least, but it might be a fun project

(26-06-2015, 01:23 PM)Mark Hennessy Wrote: [ -> ]...but the S202S02 would probably have done the job if I'd discovered it (was it around in 2004, or did I just fail to notice it?)...

It was around at the start of 2004 (just) - http://www.sharpsme.com/download/s202s02-epdf - see the rev date.

The S202S02 series seems to have been discontinued, though Farnell are still stocking them; The 16A version, S216S02, is double current (16A), the same size, in full production and is much the same cost as the S202S02at Farnell - http://uk.farnell.com/sharp/s216s02f/rel...dp/1618480

Not sure the resistor in the IR LED is necessary, but including it is no problem...

In the several years since I made these units, they've been completely reliable...

There should really be two fuses - one for the master @ 5A, (limit of the If of the BY500-600s) and one for the slave rated according to the scale of the SSR and the power strip you are using...

The BY550-600 diodes are not sold by some resellers - any reasonably good 5A or greater silicon rectifier with a Vf of around 1V should do...

Although the circuit has been reliable, I'll expand on my thinking in the hope of giving an insight into the way I think about these things. Hopefully it'll be of interest to someone, somewhere


Let's take the worst-case of 5A across the diodes. At that current, the BY550 could drop 0.9V at 5A. So that's 2.7V in total.

This is being reduced by D5 by 0.3V approximately*, so there is 2.4V available for the LED...

The forward voltage of the LED is 1.2 to 1.4V. At a forward current of 50mA, it might be as high as 1.5V. Trying to put more than 1.5V across it would clearly result in more current, but the graph stops at 50mA (which is the absolute maximum rating for the LED). In other words, if you try to "force" 2.4V across it, it wouldn't be terribly happy!

OK, the above is slightly simplified, but hopefully you can see my concerns

*In practice, the BAT81 has a very "soft" knee, which implies a high slope resistance (not untypical with a signal diode). In reality, it's protecting the LED, and if anything, this diode might be more vulnerable than the LED, given that it's only rated for 30mA....



Of course, the trouble with developing a circuit like this is the difficulty of measuring the current in the LED without changing it. The best I can think of - in lieu of a genuine Hall-effect current probe - would be a 1 ohm sense resistor and a differential 'scope probe (and an isolation transformer powering everything, of course!). A lot of DMMs would present too much "burden voltage" and reduce the current while making the reading. The EEVBlog "Microcurrent" adaptor would do a good job here, probably.

Having had a closer look at the datasheets for the diodes and SSR, I reckon that my earlier wild guess of 100 ohms should be more like half that value. Let's see how I got to that:

You need about 8mA in the LED, though more won't hurt. But let's assume 8mA for starters.

At low currents, the BY550 has a Vf of just under 0.8V (hence ~2.4V for 3 of them).

At 8mA, the BAT81 has a Vf of nearly 0.8V.

At 8mA, the LED has a Vf of about 1.2V.

So, 2.4V minus 0.8V minus 1.2V is 0.4V. Hence, we need a resistor of 50 ohms to get to 8mA. Call it 47 ohms.

How does that change at high currents?

With 2.7V across the BY550s, you'd have around 14 or 15mA in the LED if nothing else changed. In practice, the voltage lost across the BAT81 will increase slightly, but either way, ~14mA is perfectly safe for the LED and the BAT81.

As I said before, with no resistor, I can't put exact numbers on the current in the LED, but instinctively, I am sufficiently worried about it that I'd want to measure it before deciding that it's OK. Remember; any more than 30mA will endanger the BAT81, and any more than 50mA is doing the same for the LED. The 3 BV550s form a "stiff" voltage source. The BAT81 is providing some "softness". I haven't began to think about how the capacitor is influencing things (other than the fact that it will greatly increase peak currents seen by the BAT81). And because we're seeing half-wave current peaks, we might need to reduce the series R further to achieve an average of 8mA in the LED - but at least it's easy to measure the average DC voltage across the capacitor and calculate the required resistance. Surprisingly complex, these simple circuits!

Anyway, I hope this helps,

Mark

Hi Mark - thanks for the reply.

Its some years since I did this project - I modelled the design in LTSpiceIV and in practice the physical implementation behaved very closely to the model. I'll dig out the netlist if you want, but ISTR the numbers all lay within the SOA for the various active components.

I'll see if I can find it over the weekend, else I'll probe one of the live units - I do have an EEVBLOG µCurrent Gold units (very nice, aren't they ) - so I could use that...

Main issue is heat - With the Vf across D1-3 and a few amps going through it, it does get a bit warm due to simple P=IV heating...

Honestly, don't go to any trouble - I've wandered off into academic realms here

When it comes to simulators, they do have their uses, but in general, I'm in agreement with Bob Pease

Indeed, 5 times 2.7 is 13.5 watts! OK, that'll be about half that in practice, but even so... That's why the idea of a current transformer appeals. I've never tried to see if it can be done passively, however. If needed, some amplification could be powered by a capacitor dropper. I'm a little bit wary of capacitor droppers these days; although I've made many over the years, and they have been reliable with the right component choices, there are better ways.

Cheers,

Mark

I quite like LTspiceIV - If your models are good, it gives pretty good results - you do have to know what you are doing though - just hacking at SPICE is not a great plan, nor is blindly relying on the results.

That said, I'm using Ronald Dekker's uTracer combined with Derk Reefman's ExtractModel tool to generate LTspiceIV models that match real, measured, tubes. Both Ronald & Derk are serious serious people - they really understand this stuff. Derk is Phillips' Head of Magnets (the bit that does magnetic imaging etc.) and Ronald is a professor at iTU and at Phillips Research.

The nett result is that the SPICE models acurately reflect REAL, MEASURED rather than theoretical curves. Neat. The current common library is at : http://www.dos4ever.com/uTracer3/TubeLib.inc

No question that for larger loads and extended running times, a current-transformer (CT) approach would be far far better, especially if space/cooling is constrained. 100% agree on that.