Need design for 120VDC 60mA constant current supply

I'm looking for a design for a 120VDC 60mA constant current supply. This is what's needed to drive old 60mA Teletype machines. The classic solution is to run a constant voltage supply through a

2K 10W resistor, but that approach dissipates 8W when idle, which requires a big resistor for a board-mount design. I'm currently using a big 10W thick-film resistor, and it works fine, but adds 1.5" to the board height and needs ventilation.

So I'm looking for a switch-mode constant current supply. This design from EDN looks promising:

But the components are only rated to 40VDC, and even the "HV" version only goes to 60V. Also, it's a design from 2002; there are probably better switching regulator components available now.

The On Technologies NCP3065 part (a constant current driver for LED strings) looks promising, but won't go to a high enough voltage.

Any suggestions?

John Nagle

e

National has a 100V buck, LM5009 I think.

could you change the drivers to do the current regulation?, something like an L6505 and two small high voltage transistors per two outputs?

-Lasse

Almost 100V.

Input max of 95V. Output max of 85V.

How about this:

Q1 is a small N-FET switching a 600V P-FET and charging the C via L and D. Fractional voltage sensing via R3 and R4. A small boot charge via R5 to startup the uC supply. One Sense input and One Enable output from the uC.

IR makes some HV Buck controllers for LED driving. heres one.

Available here.

Onsemi makes some to for operateing directly off the mains. You could probably use one of theres too. I forgot the part number but poke around there site, I rember reading an application about a low cost HV non-isolated buck controller they use for LED'S.

The key issue here is that the constant current and high voltage are required because the constant current feeds a big inductor, the 4H

55 ohm electromagnet of a Model 15 teletype. 3.3V is enough to get 60mA through the selector magnet in the steady state, but it takes a huge voltage to get a fast enough rise time to pull the electromagnet in fast enough. That's why these things are run off of such high voltages. The traditional loop voltage is 130VDC, and higher voltages were used when several machines were in series on the current loop. So, in fact, there's a huge voltage slew at the beginning of every ON bit.

Then, when theres a 1 to 0 transition, there's a huge inductive kickback (400V or more) as that big electromagnet dumps. That can be snubbed (1uF in series with 100 ohms works without cutting the fall time too much), but there's still a transient.

So plenty of headroom is needed on the component voltages.

There are so many switching regulator parts around that something should be available to do the job with a small parts count.

Here's the actual teletype involved, and the current circuit being used for driving it.

John Nagle

That looks encouraging. I'm not sure it can handle an inductive load that's being switched on and off; I need to look at the data sheets. This really needs a SPICE model. John Nagle

Hammy's was a really good suggestion. Along those lines, you could use pretty much any old switcher chip--just stack a HV MOSFET in cascode on top of the chip's output switch.

Cheers, James Arthur

e.

agle

ge

e

Can you do the switching on the ground side of the coil?

an avalance rated fet with the right voltage rating should be able to handle the clamping as well

-Lasse

Step down the mains to 90 vac. Fair radio sales has 90 vac transformers. Rectify and filter the 90 vac to 125 vdc

If you're driving a real electromagnetic TTY coil (an old clunker with no electronics inside) the current has to rise and fall fast to work properly. The EDN circuit will be way too slow. These things typically need a lot of voltage to bang the coil, and the switch-off has to be fast too... no clamp diodes allowed. A high-voltage h-bridge would work, but that's a lot of work.

You might save some power by doing this:

+120---------R1-------------------+ | +24?---------ak-------+-----------+ diode | | | | cap | | coil | | gnd | | | | fet to ground

with some sort of zener clamp maybe to protect the fet from flyback. R1 needn't be as big as the resistor you're using now... it charges the cap when the fet is off. The lv supply is whatever the coil needs steady-state.

John

Now that's a good point. The real problem here is that the initial voltage required to get the needed rise time is about 35 times the steady state voltage, because of that huge 4H coil inductance. What's needed is the energy to power that transient.

That energy will be required about every 20ms, given the baud rate (45.45 baud!). So we really don't need that much current from the

120VDC supply if there's a capacitor to accumulate energy between pulses. It might even be possible to get the necessary 120V with a step-up arrangement from a 5V or 12V supply, which would get rid of the need for a separate 120VDC supply. That would be an elegant solution.

But how to regulate all this?

John Nagle

I have no trouble at all sketching a 60mA switching current source with 120 volt compliance.

The difficulty is making a simple circuit that dumps the selector coil quickly.

A buck topology would energize the coil quickly, but decay through D1 will be very slow:

FIG 1. (SW1 controlled by current mode switcher.) ===== SW1 L1 (selector coil) / .-.-.-.

+120v >---/ ---o---' ' ' '---o------> feedback, to | | control ckt. --- .-. D1 ^ | | Rsense | | | === '-' GND | === GND

So here's an idea--adding MOSFET Q1 does make it dump quickly:

FIG 2. ===== SW1 L1 (selector coil) / .-.-.-.

+120v >---/ ---o---' ' ' '---. | | --- | |--' Q1 D1 ^ | |

Oh, better yet, add a diode from L1's 'cold' end to +120v, recycling the 'dump' back to +120v. That saves power, cuts heating.

Cheers, James Arthur

How about charging an inductor from +12 (or whatever) and dumping it into the tty coil to get the fast current rise? That's sort of a strange flyback power supply that only makes one shot now and then... namely 45.45 times a second.

I have one of those feelings that there's a very simple, very elegant circuit lurking out there in circuit space. If I wasn't deeply involved with a carnitas super burrito, I might scribble something.

And I thought an ASR-33 was slow!

John

On Mon, 09 Mar 2009 09:11:05 -0700, John Nagle wrote:

It's regulated by the fact that the finite energy stored in the capacitor. As the coil current increases, it drains the capacitor, reducing the voltage. Once the capacitor has been drained down to the 3.3V operating voltage, the continuous current is limited by the resistance of the coil and the resistor in the 120V lead.

Try the appended LTSpice model.

Version 4 SHEET 1 880 680 WIRE -160 -16 -272 -16 WIRE 48 -16 -80 -16 WIRE 208 -16 48 -16 WIRE 384 -16 208 -16 WIRE 448 -16 384 -16 WIRE 608 -16 448 -16 WIRE 384 32 384 -16 WIRE 448 48 448 -16 WIRE 608 80 608 -16 WIRE 208 144 208 -16 WIRE 48 176 48 -16 WIRE -272 224 -272 -16 WIRE 384 272 384 112 WIRE 448 272 448 112 WIRE 448 272 384 272 WIRE 496 272 448 272 WIRE 608 272 608 160 WIRE 608 272 560 272 WIRE 720 272 608 272 WIRE 608 304 608 272 WIRE 48 336 48 240 WIRE 560 384 432 384 WIRE 432 416 432 384 WIRE 432 528 432 496 WIRE -272 560 -272 304 WIRE 48 560 48 416 WIRE 48 560 -272 560 WIRE 208 560 208 208 WIRE 208 560 48 560 WIRE 608 560 608 400 WIRE 608 560 208 560 WIRE -272 608 -272 560 FLAG -272 608 0 FLAG 432 528 0 FLAG 432 384 signal FLAG 720 272 out IOPIN 720 272 Out FLAG 608 -16 supply SYMBOL voltage -272 208 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V1 SYMATTR Value 120 SYMBOL ind 592 64 R0 SYMATTR InstName L1 SYMATTR Value 4 SYMATTR SpiceLine Rser=55 SYMBOL cap 192 144 R0 SYMATTR InstName C1 SYMATTR Value 1µF SYMBOL res -176 0 R270 WINDOW 0 32 56 VTop 0 WINDOW 3 0 56 VBottom 0 SYMATTR InstName R1 SYMATTR Value 4.7K SYMBOL voltage 432 400 R0 WINDOW 3 24 104 Invisible 0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR Value PULSE(0V 20V 20ms 100us 100us 10ms 20ms) SYMATTR InstName V3 SYMBOL voltage 48 320 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V2 SYMATTR Value 4V SYMBOL diode 64 240 R180 WINDOW 0 24 72 Left 0 WINDOW 3 24 0 Left 0 SYMATTR InstName D2 SYMBOL cap 432 48 R0 SYMATTR InstName C3 SYMATTR Value 1µF SYMBOL res 400 128 R180 WINDOW 0 36 76 Left 0 WINDOW 3 36 40 Left 0 SYMATTR InstName R2 SYMATTR Value 10K SYMBOL nmos 560 304 R0 SYMATTR InstName M1 SYMATTR Value IRF1302 SYMBOL schottky 560 256 R90 WINDOW 0 0 32 VBottom 0 WINDOW 3 32 32 VTop 0 SYMATTR InstName D1 TEXT 264 -88 Left 0 !.tran 0 0.5s 0.3s

I though about precharging an inductor--you could pump 1 amp into a much smaller inductance and store an equivalent amount of energy, or you could use another 4H inductor and charge it to 60mA...

I don't immediately see how to control the flyback voltages though. Hmmm.

I also want to say a "series cap-thing switched to +120v", where the cap delivers a time-limited +120v pulse, then you reclaim the cap energy later somehow, but I don't have a clean topology for that offhand.

(An H-bridge works, but that's messy.)

James Arthur

R1 4.7K / 5W

+120v >---\\/\\/\\-----o-----------o-----o-------. | | | |_ D1 | [R2] [C2] _) L1 +4v >----|>|------o [10k] [1uF] _) | | | _) (4H, 55ohms) | '--o--' | --- C1 | D2 | --- 1uF '----|

That's not bad. The waveforms look good.

The 120V supply is supplying 25mA (3W) in the steady ON state, though. The 4V supply is only supplying 35mA (0.14W). So the 4V supply isn't really doing much work. The 120V supply should be the low-power one.

Ideally, the 120V supply should just charge up a capacitor used to provide the voltage kick needed to get the necessary rise time. Then a very low current

120V supply could be used, like a simple boost converter, and the whole thing could be run off a 5V supply. The usual drivers for older teletypes have a big 120VDC supply; I use a 200mA 120V open-frame supply, which is overkill.

John Nagle