I see your point. I hadn't thought about EMI problems that would be exacerbated by having the diode located far from the inductive load, so I now understand the need for mounting it close. In the case where there are long wires between the switch and the inductive load, a diode near the switch (built into the ULN series, but not into most other devices) is also desirable.
While it is built in to the UNL series it is also a junk silicon diode not meant for switcher frequencies, hence I still say that if you insist on using them that way add good schottkys instead of relying on the slow ULN silicon.
I would have to agree with Ubergeek63 on that one. If you test the motor with the diodes not connected you will see that the motor is running much smoother and with more torque.
In spite of all the litterature suggesting using diodes, I do not use them in a few drivers that I made using ULN2803 chips:
They have been running for hours and hours and hours without showing any problems, not even loosing steps. Of course, I rund them at 360 RPM wich is slow. I suppose fast diodes would be a must for more sophisticated drivers.
Both the ULN2003 and ULN2803 are rated for 5V operation. The clamp diode voltage is not given, it should be 6.8V/ 0.5W. Following ubergeek63's advice, using a FET will handle the current without alot of heat, and all FETs can share a common heatsink with insulating hardware for each FET. You could try IRF630 or IRF640.
Both the ULN2003 and ULN2803 are rated for 5V operation. The clamp diode voltage is not given, it should be 6.8V/ 0.5W. Following ubergeek63's advice, using a FET will handle the current without alot of heat, and all FETs can share a common heatsink with insulating hardware for each FET. You could try IRF630 or IRF640.