true/complement driver

I'd just use a 422 driver. You don't need the bi-directional nature of the

But you need to drive twisted pair. That sounds a lot like 100ohms (and a lot of power). If you have to go a long distance you can use a couple of drivers and add pre-emphasis.

How about half a 'HC74? D=1, CLK=/CLR= input (maybe a little delay so the /CLR goes high a few nanoseconds early).

Yeah, an honest 5 volts will be 50 mA. If I source terminate, half that.

The customer wanted me to drive the pairs at 24 volts! I refused, mumble mumble about dissipation and EMI and stuff.

RS422/485 drivers swing way less than r-r, so most honorable customer may not go for them.

The PWMs will (somehow) control about a megawatt of RF pumps for a CO2 laser. I think it's possible for me to blow the whole thing up if I do something wrong.

John

A 74AHC1G08+74ACH1G00 (2 * SC-70 packages) will give complimentary outputs within 0.6ns typically. Load the G00 with ~15pF to get the nominal error down to about zero.

I think that's way better than the typical skew on complimentary 485 driver outputs. But not ESD safe to directly drive cables.

Best regards, Spehro Pefhany

That isn't impossible at 100kHz, just a lot of power. EMI problems? Maybe.

Why not drive something reasonable, like an honest RS422, and convert to what he wants at the other end?

Unterminated they will. At distance and 100kHz, not likely. ;-)

Ohhh! Can I watch?

Go for it anyway...no EMI with differential twisted pair...

junk

That thing should be interlocked out to the max against every possible failure....

You can get a nice signal with

For sufficient bandwidth, use a transmission line transformer (the primary and secondary wires twisted together and the bunch coiled around a ferrite core, please observe polarity when connecting the secondaries).A 28 Vdc power supply should be sufficient, however, using other transmission ratios, the DC voltage could be reduced.

At such power levels and frequencies, there are galvanic isolation and grounding/bypassing issues.

Have you considered using fiber optics to carry the PWM signals ?

On a sunny day (Thu, 21 Apr 2011 16:46:45 -0700) it happened John Larkin wrote in :

I did just that in the eighties to drive I2C data and clock to a number of remote stations, ignoring the ACK, so one way, speed figures, except I used + and - 12 IIRC, or 24 ? of course with universal PNPs and NPNs. At 100kHz what is a few nano seconds. The Joerg (tm) way, just use some universal si (TUP and TUN).

Is there any way to get your hands on a 2X clock? If so, then a flip-flop would do it. :-)

Good Luck! Rich

go crazy and use ADSL drivers ?

-Lasse

Yup, the run could be 50 meters or so, between floors of a building, in an industrial environment. All sorts of people are nervous about this.

Weird, but I'm making a dozen or so fast timing signals on fiber, but they want the PWMs as differential electrical drives. Life would be great if there were no customers!

John

Sounds like a job for a multi-part gate driver. 24V as you mentioned later is a bit tougher. That may be possible if you abuse some stereo class-D audio chip.

Both would be pretty cheap and have the muscle to drive TP.

TI SN75114/SN75115 line driver/receiver is propably what you still find in Siemens NC machines interfacing with quadrature encoders. That combination worked fine for me in that environment at 800kHz (didn't test any higher). Might be a candidate for your job.

Regards Werner Dahn

Pere