On a 74HC595 shift register, the power lines are on separate sides of the chip. I've added a capacitor from one side to the other by going over the chip (I'm soldering to protoboard) but this is not the most elegant way to do this as the cap is actually 'on top' of the IC. As the cap needs to be as close to the IC as possible, should the cap be on one side with a jumper cable to the other? Does this extra couple of inches make a difference if it's placed between the chip pin and the incoming power?
Everything that I have read, around here, about the .1 uf decoupling cap on the595 says that it should be as near to the + as possible. Whenever I use 595s I have jumpers from the -(8) to OE(13) and from +(16) to CLR(10), so there are a few places you could place the cap, and im not really sure how much it matters if you are soldering the cap directly to the chip.
This is my current 595/cap setup.
The first ones I soldered up were from pins 10 to 13, but then I figured out that 13 to 16 was a little closer, even if it was a little more difficult to place.
Graynomad:
Holy crap! Did you need some more current and the TPIC6B595 shop was closed?
Rob
Hey, it's a hobby, the rules can be bent to do what works with what you have. It's nice to sometimes not have a boss to say "why the hell did you do it that way for".
That was a technique we used in the 70s to double up on computer memory!
As to the question from the OP, yes the closer it gets the better it is. However you won't notice any difference immediately when going from just OK to good.
It is best to use surface mount capacitors, they have no leads at all.
The rule about decoupling close to the supply rail only holds if there is a ground-plane - otherwise it just joins supply and ground close to both pins. An inch or so shouldn't matter, but the shorter (and wider) the leads the more effective (lower stray inductance between chip and its capacitor). The reason PCB traces for power are wider than signal lines is only partly to carry more current, but also to reduce inductance.
A wide PCB trace running opposite a ground-plane = low inductance, a fine wire floating around in mid air away from ground signals = high inductance [typical breadboard!]
Actually it reminds me of how we used to add RAM to the Exidy Sorcerer computer. IIRC they had a bank of 8 4k-1 DRAMs, there was a spare RAS (or was it CAS) signal but no provision for another bank of memory.
So I piggy-backed 8 chips with the RAS (or CAS) pins bent out and ran a flying wire to the 8 pins.
Worked a treat and gave me 8k RAM, so much I didn't know what to do with it
Would it make any difference if I just had one big cap, instead of one cap on each chip? what would be a suitable replacement for 6 little caps? or would 1 cap be sufficient for all?
Its not for a lot of current, so much as its for 48 bits, and it is piggybacked, just like we used to do with ram chips.
The other side of the chips have the first 7 leads sticking straight out, and a bent up LED lead soldered to pin 15 aligned with the other output pins. I also have a 32 bit stack with sip headers soldered on, but I havnt found connectors for it yet.
Its very convenient that the - and + are on opposite corners, that helps for stability. its also real convenient that most of the outputs are on one side, and the inputs are on the other. I should be able to make them as a plug-in module.
The decoupling cap should be close to the supply pins if there is a ground-plane or not.
Imagine you have a 555 timer flashing an LED, and it is connected to the supply via one meter long thin wires. ( which will have some resistance and inductance )
Every time the LED switches on, there is a pulse of current drawn from the supply, and there will be a momentary voltage drop on the vcc and ground wires ( but not on say a third wire going to the trigger or reset pins )
If the +ve pulse on the ground pin was higher than 1/3 vcc ( OK this is an exageration ) then the chip would retrigger.
By having a cap on the supply pins, the surge is supplied by the cap, which is recharged slowly between pulses .
I wasnt sure about that, but i had assumed that if sinking, we didnt need the caps, but I wasnt sure.
So are these true statements?
If you are sinking only, then you dont need de-coupling caps.
1 cap will work for many chips, if it is large enough (higher voltage?)
decoupling caps dont have to be soldered directly onto the chips pins i there are large enough traces or wires, and not too far away (an inch?)
Will any type of cap work for decoupling? can I use electrolytic or tantalum? I assume ceramic caps are used because they are inexpensive.
Fungus, FWIW, I generally prefer SMT to THT. I dont have much trouble with either, but smt is faster, easier, and often ends up looking nicer.
fungus:
Are you using those chips as current sources or current sinks? If they're sinks you don't need so much decoupling.
No, same currents are changing and inducing EMFs everywhere. You might want to place the decoupling
capacitor in a different place (close to the ground pin on the chip), and the presence of a ground plane
but no supply plane will affect the symmetry of the situation, but Maxwells equations still apply!
So are these true statements?
If you are sinking only, then you dont need de-coupling caps.
No total rubbish.
1 cap will work for many chips, if it is large enough (higher voltage?)
No. The voltage rating has nothing to do with it. Large meaning more uF then also no.
decoupling caps dont have to be soldered directly onto the chips pins i there are large enough traces or wires, and not too far away (an inch?)
It depends on how effective you want them to work. The further they are away from the chip the less good they do, due to the inductance in the lead and the track.
Real engineers keep telling me I can never have too much decoupling.
