ULN2803A

[Lindsey]

Hi to all, Just joined the forum and would ask for some feedback please from anyone who may have noticed perhaps some similarities to these: I am using a ULN2803A controlled off the outputs of a MCP23017 I/O expander, this drives the ULN2803 input directly with a 3.3volt bias. I also have another circuit driving the ULN2007 chip from 5v 74HC logic chip. Both the ULN IC's have the same internal circuit but they are in two different boards - no connection to the other. Both the ULN devices sink coil current of a 24dc/80ma relay, 1 x relay per channel. On many occasions two adjacent channel relay coils are required to oscillate simultaneously. Over the period of a year and for several years now the 5v board freezes and now with the MCP23017 board only 12 months old it has demonstrated a similar occurrence, both very seldom mind you. Needless to say I have exhausted everything I can think of to identify the cause but seem to end up starring at this ULN relay driver chip. After sitting for many hours with a crow on the input of the 5v ULN2007 and adjusting to various time divisions I started to notice a 10 - 20 nanosecond blip appearing on the ULN input pin in excess of 40volts dc. Strewth!! Obviously this is coming from the relay coil's back EMF. - I do have the freewheeling diodes common pin of th eULN connected to relays 24vdc source.
I am beginning to form an opinion that the ULN 2803/2007 may have on certain situations of oscillating outputs of relays a latency of coil back emf suppression which is passing through the chip to the logic IC's before it and stalling them. I would like to know if any one has had a similar experience with this relay driver?
Thanks
Cheers
Lindsey

 

[Harald Kapp]

Not from experience with this particular chip. This chip is afaik not known for particular problems.
But: your observation of a nanosecond spike at the ULN's input sounds like a problem with the ground system. Looking at chapter 8.2 in the datasheet one way for the observed pulse to happen is via the diode connected between point E and B.
If your ground connection to the driver is not rock solid, current pulses on the output may create voltage pulses at the input via the inductance of the ground system. Remedies:
- bypass capacitor from Vcc to GND right next to the driver's pins
- solid grounding system, preferably using a GND plane

An additional series resistor from the MCP to the ULN and a zener diode from the ULN's input to GND to limit possible overvoltages may also help. Although a zener diode may be a bit on the slow side. It is worth a try.

 

[AnalogKid]

Agree with HK. Tack a 0.1 uF ceramic cap directly the ULN's GND and COM pins with the shortest possible leads, and see if the transient waveform is altered (or eliminated). If (hopefully) attenuated, tack on another one to see if things improve more. If yes, then you have a solution path to retrofit existing boards and update the design.

While a heavy ground connection to a driver chip is something most people see the need for, consider the COM pin. First, the suppression diodes are not very fast, so I'm not surprised that nanosecond effects are present. Second, the trace to the COM pin is an inductor, raising the impedance between the relay coil and the +24 source at nanosecond speeds. This pc board trace needs to be treated as a ground connection, with all of the low impedance tricks you can use.

ak

 

[Lindsey]

Thanks Gents for your replies.

All those points are helpful and appear warranted in the application of the ULN chip especially as an inductive driver.
Little bit more testing to get on with now, in the meantime if anything else comes to anyone's mind please feel free to let me know, always interested in experienced advice.

Cheers

Lindsey