I was called into to look at a board that has a DAC7846 -16 bit D/A) , which has a history of "failures" according to the previous "tech" (they always replace DAC to fix the "problem" ). They seem to replace these quite often... at $50 buck a chip, this gets expensive.

For some reason (i suspect noise) they piggy-back (manually solder) a 470uF Cap across Analog and Digital ground of the DAC.

Has anyone ever seen this done before... a cap between the two grounds?

The power supply inputs (+15v,-15v,5v) to the DAC only have 47nF caps each (mounted next to the DAC)... seems a little on the small side?

 

I've not personally seen that, but I can understand what it does.

First off, you can't think of ground as some perfect universal zero voltage point, or area underneath your circuit. Ground itself needs to be built out of something, and that something will have some resistance, and possible some transmission line effects too, so that you can see a real voltage difference between ground "here" and ground "over there."

In a system using both analog and digital parts there is usually an "analog ground" and a "digital ground" as the digital stuff will typically do nasty things to ground.

By it's very nature a DAC sits in between the analog and digital domains of a circuit, and it needs to be connected to both grounds. Now where these two grounds connect is something that keeps designers up for nights on end, and there is usually only a "not as bad as the other places" to connect them together, and not an ideal point.

So what happens is the first prototype boards seem to pass (my first boss/mentor told us "anyone can make one of anything work"), but given a normal production spread when you start cranking them out some will fail.

As the job progresses, things that "seem to work" get added in. Such as that cap.

I would bet dollars to donuts that tech was tossing out perfectly good $50 parts, perhaps at best he's sorting for parts that have a higher tolerance for the noise.

All a cap there will do is help equalize the noise between the grounds, and as long as it induced no other problems is probably OK. However, a piece of wire may be better!

470uF seems like too much: not that more capacitance is bad that high a value is going to drop out at higher frequencies: small value ceramic caps are good at high frequencies, large value electrolytic caps are best at low frequencies. Many times you will see both caps, a large value with a small value parallel to it: this gives the best combo of low and high frequency utility.

I agree 47nF seems a bit small also. I would try getting a 0.1 to add on top (assuming this is an SMD board).

 

Ernie,

Thanks for your response... you make some really good points. I'm little weak in Analog theory (haven't used those brain cells in a long time). This board is a very old design (at least 20 yrs. in production).

All thru-hole parts, no SMD. The board is about the size of a large pizza with the all the toppings:

CPU,Ram, Xilinx, Relays, Relay Drivers, ADC's, DAC's, Two Crystals, Voltage Regulators, Two digital chip oscillators (square wave),
Two analog component oscillators (saw tooth), Two heater driver circuits and 25% of the pc board is dedicated to an RF circuit....I probably left some stuff out. Oh yes... there's +750 and -750 volts on the board as well.

And this board gets installed on a VME bus.

I have tubes of at least 100 DAC's that were "rejected" in the past.

I'll look at exploring the cap values for starters... thanks for your help.