Design check for a telephone ringer

A picture is worth a thousand words.

A bit under 200 in this case, but:

I omitted the ISP header because it's not important, and I'm lazy. If something is in the schematic backwards, I made a mistake in Eagle.. the circuit works fine.

I had it running for about an hour at 12V (Ringer muted, of course.) Heatsinks were a little warmer than ambient, but not painfully so.

R C

What state are the mosfets in, when the part is in reset? Add turnoff resistors on the gates.

Is your output high enough to turn those fets fully on?

Do you need a snubber across drain and source?

Did you really mean to leave the transformer and fets powered when the logic is off?

Why use a tiny-26 where a tiny-11 will do?

You need also, to stop the ringing current within a very short time of when the handset goes offhook, by sensing the loop current. I see no provision for that.

The ring voltage is added in series with the 48VDC as well, so there are considerations for the transformer at 100mA loop current, and what happens on the low voltage side of the transformer when you pulse dial, or go on/off hook.

Also interesting how the impedance of the transformer will affect the line, since it's in series with it.

There's a good book for this, "Understanding Telephone Electronics". It used to be a $5 book at Radio Shack, red cover. Now it's a hardback and much pricer.

In that schematic, the +12 V for the LED and transformer doesn't go through the power switch. Probably just a mistake in the schematic.

I agree that the inductive spike from the transformer might be a problem. I have observed spikes to over 100 V when switching 12 V to a standard automotive relay. If you have a scope you can tell in a second.

It's probably not a good idea to keep sending the ring signal after the phone has gone off-hook. This can be addressed by more hardware, or by user training.

What happens if the output (to phone) is shorted? Most likely you'll bog down your 12 V supply. You might want a fuse in the 12 V line against this possibility.

What kind of voltage does it end up generating? Spec is 86 V AC. Under this (down to maybe 40-50 V) isn't a problem but too much over might be bad.

Matt Roberds

Whatever state they were in before the gates were tristated. The only way to put the part into reset is turnon (reset for ~60ns, IIRC), or via the ISP header.

Good idea.. I'll add some weak pulldowns.

5V.. threshold on the IR510's is 2-4V. Fully on.. possibly not.

That's what I was wondering.

Both FETs are off during IDLE cycle. The center tap is 12V, but the +/- 12 legs are not conducting.

It was the only small form factor non-mega AVR I hand on the bench.. couldn't well use a '8515 or '8535.

Even with an 18V 2A power supply, as soon as the phone is picked up all you hear is a minor clicking in the handset. As I power it via a 12V 400 mA supply now, it's not an issue. There is no provision to provide talk voltage.

R C

Right.

25 V max open circuit with the heavy supply; ~20 V open circuit with standard supply. Less with a phone connected (22/17V) or shorted. Transistors are rated to 100V.

The users do turn it off after it is picked up; the ringing voltage does not appear to do anything to the phone except some minor clicking in the handset. I wouldn't hook a digital phone or modem up to it.

The transformer is not very beefy.. I'm not sure of the actual specs, but it's less than half an amp[3].

It draws a little more current, but not much. Max draw while ringing is 2A (roughly 50% of the time.. the other 50% is at 1 amp.) Voltage sags pretty far down during ringing.. all the way down to 7V with the weaker supply. If I can dig up one or two more big caps it may help smooth it a bit more.

Open circuit voltage is ~400 V PtP 12V beefy. 12V standard is 325 V. With a

0.6 REN load it's about 350/200V PtP. Note that the 70-90 V nominal AC is the RMS value. In the field, you also see actual voltages of about 50-150 V RMS. Rule of thumb conversion[1].. 0.707*(0.5x)=123/70.7 RMS. So if I go with the standard pack I'm fine; also the actual phone is 0.9 REN. If I wish to ring more than 1 phone, I'd probably need the beefy pack.

R C ([1]Yes, I know that's for sine waves.. but these aren't perfect square waves either. I suspect the transformer is attenuating the higher frequency components.) ([2] This is a random transformer I pulled from a broken cassette player; it might be better off being replaced by a known 12-0-12/120 transformer.) ([3] DC resistance of each leg of the secondary is about 1.8 ohms.)

Current tops out at 2A max drawn, shorted or not. A fuse wouldn't hurt.. might grab an inline automotive type around 2.5 or 3 amps.

There is no 48 VDC bias on the line. The only items connected are the telephone and the ringer unit. There is no provision for talking on this circuit.. it is strictly to ring the bell.

R C

How'd you measure this? Scope or meter? If scope, OK, but if meter, there might be higher voltages that the meter isn't catching...

has info on how to work it out, if you'd like.

1 A sounds like a lot for just a microprocessor sitting there, but then again I'm not that familiar with the Tiny26.

Look up the October 1989 "Hardware Hacker" column by Don Lancaster (available in PDF at his Web site, tinaja.com , the last time I looked) for a neat way to work out power supply filter cap sizes.

Hmm... 86 V RMS is about 86 x 1.414 or 121 V PtP for a sine wave or

86 V PtP for a square wave with 50% duty cycle. Going by your loaded value, you're delivering somewhere between 247 and 350 V RMS - this seems high to me.

I haven't stuck a meter on that many phone lines, but the ones I have have been 90 V RMS or lower. But these were all CO-served lines; PBX extensions might be higher.

Agreed, or at least test it per the link above to know what it can do.

Matt Roberds

Scope.

Describes how to measure the current ratings.. but not the ratio. However, there are other sites that do.. I'll look into it, but I don't have a variac handy.

2/1 A is during ring cycle only. Total current during idle is roughly 20 mA including the LED.. and I'm not using the sleep capabilities of the tiny26 at all. I suppose I could have rephrased that a bit.

The average draw while ringing is about 1.5 A, with a plateau at 2 and valley at 1. The ring cycle is 2 seconds on, 4 seconds off, which gives overall average of 500 mA.. which means I need to current limit the mosfets.

Thanks, I'll look into it. To be honest, I may be better off current-limiting the mosfets. The supplies are different designs.. the 'Beefy' one is 12V, regulated, and rated for 1A continuous. The standard one is 400 mA, unregulated.

Hmm.. doesn't look like an exact match as I'm feeding off an already fixed wall supply, but I'll read the whole thing. But all very interesting.. I'll have to read some more tonight.

It is a bit high.. esp. as I'm not testing it with the stage phone, which is a lot higher draw (mechanical, not piezo). The 200V PtP value is the important value, as it's the one with the weaker supply that I'm actually using with the circuit.

Also, this is not a true square wave output.. it looks very much like a square wave and a low-pass.

That's from my notes gathered from who knows where. It may be a bit dated.

It may be a 6-0-6 or 9-0-9. I may try running the circuit at 9V if I current-limit the mosfets to preserve the headroom on the 5V regulator. Hmm.. probably a single beefy resistor on the center tap, or two smaller ones between the mosfets and the transformer. Have to check the parts bin.

R C

Quick update: Stage tested it; works fine when the actors _hang up the phone_ properly when they're done... guess I have to add that off-hook detect after all, and perhaps ring a separate phone after all..

R C

You shouldn't _need_ a Variac, just a way to load the output. For the ratio, put a known AC voltage (from another transformer) into one winding and measure what you get at the other windings.

OK, that sounds a lot more reasonable.

For the off-hook detect, a phone with a REN of 1.0 is supposed to have a DC resistance of 50 megohms on hook. Off-hook, it drops to 200 ohms or less. My initial thought would be something like a relay with SPDT contacts on the "phone" side of the transfomer. One side of the phone always goes to the transformer, and the other side goes to the contact common. When you want to ring the phone, drive the relay such that the phone is connected to the transformer. When you're not actively making AC, drive the relay so that the phone is connected to a pullup to +5 V (you'll have to ground the other side of the high-voltage winding to DC ground) and run that into a port pin. When this value goes low, the phone is off-hook.

I am almost 100% certain that the phone company doesn't do it this way. I seem to remember reading that they have a small coil with two windings. One determines the frequency of an oscillator, and the other is connected to the DC loop to the phone. When the phone goes off-hook, a few tens of milliamps of current flows through one winding, which alters the characteristics of the coil enough that the oscillator frequency changes notably.

If the actors are really bad about hanging up the phone, then after a time-out, send that nasty loud staccato tone that the phone company uses when you've left a phone off-hook.

Matt Roberds