Bypassing problem

[Devil_H@ck]

I've started routing the input board of my preamp and I've run into a little problem. There are 8 SSM2404 chips on the board and they all need a +12v, -12v & dgnd connection. The +12v & -12v pins are 7.5 mm (300 mils, 0.3 in) appart and the dgnd pin is on the other side of the chip, straight across the +12v pin.

If I decide to bypass the usual way (+12v to dgnd, -12v to dgnd) I'll have to make the board bigger, because of all the extra traces needed. If I use SMD caps, I still have to make it bigger because the caps occupy space where otherwise traces would run. But then I noticed the spacing between the +12v & -12v pins. It's the same spacing as the MKP caps I always buy. I think they are 100nF, but it might be more, anyway, that's probably enough, since only 2 chips will be active at a time and they only need 5mA max (0.9 typical) (that's for +12v, -12v is even less). So, the question: how much worse is it to put a cap between +12v & -12v instead of doing it the usual way?

Keep in mind that the SSM2404 is basicly just a switch, that bypassing isn't even mentioned in the datasheet (not even in the basic configuration schematic) and that the max current draw for +12v & -12v combined will be 11.2 mA (that's for the entire circuit).

[officeboy]

I did this on a board I was just working on and it' caused a lot of oscillation. Pulling the capacitor solved it immediately.

But that was with TL072’s

[Devil_H@ck]

Yes, but the TL072 are opamps. I wouldn't even thing of doing this with opamps. The SSM2404 is a quad analog switch.

So, anyone who ever tried this on a chip like this?

[Devil_H@ck]

Bump?

[janneman]

I think the purpose of decoupling is to try to make the supply stable against whatever is the signal reference. That usually is ground, so I would recommend to decouple to ground if at all possible. How much bigger would the layout be? A few tenths of an inch extra tracing would be worth the extra decoupling I think. The switch resistance is depending on the supply voltage, so noise and ripple on the supply will modulate the on-resistance. That may be significant or not depending on the circuit, but I would decouple to ground with .1 uF.

Jan Didden

[Devil_H@ck]

Ok, a bigger layout it is then.

[cpemma]

Wouldn't it be relatively easy to add a bit of trace to mount SMD caps under the board?

Oops, sorry, missed "If I use SMD caps, I still have to make it bigger because the caps occupy space where otherwise traces would run. " Non-smd with insulated legs?

[Devil_H@ck]

I'll find a way to do it .

[richie00boy]

So, the question: how much worse is it to put a cap between +12v & -12v instead of doing it the usual way?

It's not that one way is worse than the other, they are totally different things. Bypassing to ground addresses common-mode issues, whereas bypassing across the supply addresses differential-mode issues.

[Prune]

So then, should both be used to deal with both issues in a general power supply? In my case, +/gnd feeds digital part's regulators through a CLCLC filter, and +/gnd/- feeds discrete analog part. How do I best wire PSU capacitors for the analog part, + to gnd and - to gnd, + to -, or both? I think there was in one of the tubecad.com articles a mention of when to use which, but I can't find it anymoe.