[GreenKeys] New 60ma current loop driver board design

[John Nagle]

On 10/27/2016 06:57 PM, drlegendre . wrote:
> "This has roughly the same functionality as my previous board,
> but does not require an external 120VDC power supply.  It
> generates the selector magnet power from the USB port power."
>
> This comment threw me for a moment, until I read on, and came across
> this caveat:
>
> "  It works in LTspice, but may take some tuning
> in the real world.  More later."
>
> So you're counting on the limited duty cycle to bridge the gap between
> the 120 @ 60mA (7.5W) and the USB power of 5V @ 250mA (1.25W)? That
> seems like a trick, but I can see how it might work - but the loop
> length will have to be fairly short, eh?
>
> Is there any reason you're not using the full 2.0 spec, for 500mA? I'd
> think that would really help to close the power gap.

    There's no "power gap". There's plenty of power available.
You only need a lot of power for the first 3ms after a SPACE
to MARK transition.  After that, the selector magnet inductance
has been overcome, and it only takes 3-4V to push 60mA through
a 55 ohm magnet.

    The classic "120VDC through a 2000 ohm 10 watt resistor"
supply uses most of its energy to heat the resistor.
It draws 7.2 watts, of which about 97% goes to heating
the ballast resistor and 3% drives the selector magnet.
This design uses about 1 watt.

    The energy to initially pull in the selector magnet is
accumulated during the SPACE time in a 1uF capacitor.
That capacitor is discharged into the selector magnet on
a SPACE to MARK transition.  This is similar to how a
photoflash works.

    Here's the circuit modelled in LTSpice:

      http://www.aetherltd.com/public/misc/ttydriver22.asc

    You can download LTSpice for free from Linear Technologies.
It's a general SPICE simulator, but has models of most Linear
Technologies parts.  With this, you can explore the design.

    Here's a simulation with that model, with
the most output values graphed.	

      http://www.aetherltd.com/public/misc/ttydriver22.png

The green line is the current through the selector magnet.
The blue line is the voltage across the selector magnet.
The teal line is the data signal from the serial port.
After 20ms, there's a SPACE to MARK transition, and thereafter
it switches every 22ms, which is 45 baud.  Note the green
line. That's current through the selector magnet.  It
quickly ramps up to 60mA and stays there as the voltage
drops.  At turn-off, there's some ringing in voltage,
but the current drops fast, so the selector magnet will
release fast.  (Too much snubbing will slow selector
magnet release.)  This is roughly the current curve
you see with a classic power supply.

During the first 20ms, the data input is SPACE, and
the capacitor C1 is being charged up by the boost
supply.  U1 and its associated circuitry generate
a 100KHz 25% duty cycle square wave, which is used
to control the MOSFET on the low side of the transformer
L1L2.  When the MOSFET is ON, current flows in the transformer
and builds up a magnetic field.  It's on just long enough to
do that. When the MOSFET turns off, there's an inductive kick
as the circuit opens.  That alone produces about 40V at the
input to the transformer. The transformer steps that up to
over 120V.  On each cycle, a little more energy is transferred
into capacitor C1, which charges up to 120V in about 18ms.
A 120V Zener stops further charging.  That's where the
energy to pull in the selector magnet comes from.

This is a common boost supply design.  The only exotic
thing here is that the supply is designed to deliver a
roughly constant current of 60mA under load at a low
voltage once the capacitor has discharged.  So one
supply provides both the initial surge and the sustain
power.

			John Nagle