Connect with us

decoupling caps placement

Discussion in 'Electronic Basics' started by tempus fugit, Dec 21, 2008.

Scroll to continue with content
  1. tempus fugit

    tempus fugit Guest

    Hey all;

    I've got a circuit that uses 3 4049 inverters. On this IC, the V+ is on pin
    8 and the ground is on pin 1. I know that the decoupling caps need to be as
    close to the IC as possible, but how can I connect 1 end of the cap to V+
    and the other to ground when the pins are so far away? Is it sufficient to
    connect 1 end of the cap to V+ and the other to a nearby ground node, or
    should the cap be connected close to the actual ground pin of the IC? Also,
    do I use 1 cap for each IC? If so, (the ICs are fairly close together)
    wouldn't the IC "see" the caps as being the paralleled value of the 3 caps,
    thus reducing the available capacitance? I was going to use 0.1uF for the
    value of each decoupling cap. Would it also be wise to use a larger (1uF or
    higher) cap in parallel?

  2. ian field

    ian field Guest

    Sometimes I've seen capacitor strips that can be cut to length, but never
    did find a supplier.

    Some types of IC socket have an open centre that you can mount a small
    capacitor diagonally, normally 0.1uF non-electrolytics (very low ESR) are ok
    if you decouple every chip as long as you fit a few electrolytics
    distributed around the board.
  3. Jamie

    Jamie Guest

    In our day we call them by pass cap's and as far as i'm concerned,
    they still are.

    As on your question. I've found that the by pass cap really helps
    in removing load variations from a previous path from another component
    that may also be taxing the rail and causing unstable voltages. Placing
    by pass caps between the components as you go along normally cures it.
    Unless you're trying to work with high freq R.F. design, I don't
    think having a little lead way isn't going to hurt any.

    1 uf or more for load swing variations and ~ .01 non inductive type
    for R.F. issues.
    Many times, you'll see a combination of 2 types in a single location."
  4. Andrew Holme

    Andrew Holme Guest

    They're not like resistors. Parallel capacitance adds:

    C1 || C2 || C3 = C1 + C2 + C3
  5. neon


    Oct 21, 2006
    .1 is a good cap value anything bigger will not do you any good inductance comes into play. don't worry about puting those things into up to the pins the caps do have leads.
  6. There are many ways to do this depending on how your board and circuit
    are layed out.

    For instance, on say 2 layer boards it's common to have a "power
    strip" going underneath the IC that feeds ground and power to a whole
    row of chips end-on-end, so you can put the cap at the end of the chip
    in this case.

    When you go to multiple layer board you usually have a ground plane
    that provides a nice low impedance ground path for you. So in this
    case you would put the cap next to the positive power pin and then to
    the ground plane.

    Other circuit topologies may dictate something different again.
    Ideally it should be the ground pin of the IC. But in your case any
    nearby ground node will almost certainly do. A 4049 is not a fast
    device, so it's not likely to be at all critical.

    What you are after is (simplistically) the shortest electrical "loop"
    path between the positive power pin, through your decoupling cap, and
    back to the ground pin on the chip. The shorter the better.
    That's the general rule of thumb, yes.
    It's not just the capacitance, it's the inductance (and resistance) of
    the entire power/ground "loop" that matters. It's a complex thing.
    This is complex area and has to do with all sorts of factors.
    Generally, if the datasheet for your device does not specifically say
    so, then one cap will be sufficient. If it's critical, then the
    datasheet might recommend two or three caps of different values and

  7. Eeyore

    Eeyore Guest

    With a 4049 and its slow speed it really doesn't matter. You mean a CD4049 ?

    I saw one guy build a complete CMOS processor board using 4000 series and god
    knows what micro, probably running at ~ 1 MHz and it had about 3 decoupling
    caps spread around. Now that DID have problems but the fucking clown who sold
    it claimed that CMOS "didn't need decoupling" ! Jesus Wept. We liked the
    product and offered to relayout the board for free so he could supply a
    reliable model but he turned us down !

    The company folded of course. Both supplier and customer.

  8. Eeyore

    Eeyore Guest

    RS ( Radiospares ) and others do a socket with an integrated
    decoupling cap. Costs an arm and a leg though !

    But you don't need it for 4000 series. OTOH if it's a double sided board place
    on the reverse side two traces from the power pins to the middle of the chip and
    put an SMT cap there. But he is worrying about nothing.

  9. Eeyore

    Eeyore Guest

    You mean YOUR day and btw you don't use apostrophes for plurals.

    NO. They're called DEcoupling caps as opposed to coupling caps. Because they do
    DE opposite ! ;~)

  10. Eeyore

    Eeyore Guest

    Thank goodness this is the basics group ! I don't expect it'll be long
    before we see graduates claiming the same though.

  11. Eeyore

    Eeyore Guest

    Heck, I don't even always bother with one decoupling cap per chip with HC
    logic !

    Now consider a 40MHz '8051' with the power on pins 20 and 40 !

  12. Eeyore

    Eeyore Guest

    But it's 4000 series ! It's almost ANALOG !

    In fact some analog circuits do use 4000 series.

  13. tempus fugit

    tempus fugit Guest


    Sorry, but I wasn't thinking - forgot they add up opposite to resistors.

    Thanks for correcting me on that.
  14. IanM

    IanM Guest

    Some general comments on effective grounding and decoupling pitched at
    hobbyists (pros will know how much they can save by cutting back on
    both, but for the rest of us, the small additional cost is easily repaid
    by time saved debugging a glitchy circuit):

    By ground, I and most other dabblers in low voltage electronics mean
    chassis, common rail, 0V, battery negative, Logic negative supply etc,
    NOT a hard wired connection to a copper stake in the earth.
    Electricians are different - when they say ground, they mean ground!

    If building on veroboard or solderless breadboard, make sure you have
    reliable power and ground rails. Except in exceptional circumstances,
    DONT wire them point to point. (low level audio is an exceptional
    circumstance with special layout and decoupling requirements as are high
    gain RF amplifiers and high power circuits in general.)

    Most solderless breadboards have long rails of contact positions on each
    edge, USE THEM (but beware of boards with a split between two halves of
    the rail half way down it, they need a link inserted). Local decoupling
    can be put over the top of critical ICs. Have an electrolytic capacitor
    across the power rails where the power comes onto the board. Somewhere
    between 10 and 100uF axial type is a good choice for most circuits on

    If building digital or RF circuits on veroboard, you need a *solid*
    ground rail. 2 or 3 tracks tied together at intervals along the board
    is about right, or use the board crossways and run a heavy bare solid
    copper buss wire along the top tacked down at intervals to any tracks
    you want grounded by a little loop of bare wire over it, soldered
    through two adjacent holes and to the copper buss wire.

    Power rails can be done the same way but are usually less critical.
    On veroboard, if you are using thin kynar wirewrap wire or similar for
    your signal wires, DONT use it for power and ground.

    All decoupling caps should lead as directly as possible to the ground
    buss and to the + supply pin of the chip in question. 0.022uF to 0.1uF
    disk or resin dipped ceramic capacitors are generally suitable for
    individual chips. The chip should be located for the most direct ground
    connection possible. If there is more length of wire than the width of
    the board between the power supply and the board, put an electrolytic
    (typically around 100x the individual ceramic capacitors) accross power
    and ground where the supply wires go. Add an additional electrolytic
    decoupling cap for every 10 chips.

    The aim is to keep everything happily stable with respect to the ground
    bus with any bounce from one IC switching *NOT* getting into other ones.

    With some care with layout, bread boards are good for circuits operating
    at up to a couple of MHz and veroboard up to a few tens of Mhz.

    In the case of the OP's 4049 logic circuit, *ALL* 4000 series logic is
    slow and low power so needs minimal decoupling. If there is no other
    power consuming circuit on the board, a single 0.1uf ceramic located
    short as possible supply and ground wires radiating to the three chips
    will almost certainly be fine. If there are LEDs, a speaker or relay or
    other high current loads, add an electrolytic capacitor as well.
    However, I wouldn't build it that way unless I needed minimum size/weight.

    If one is etching double sided PCBs or building complicated processor
    boards etc., one is out of the 'dabbler' category, even if still an
    amateur, so had better have a PROFESSIONAL understanding of layout
    grounding and decoupling, othewise you will get PRO sised grief as
    alluded to by Graham (Eeyore) in his first reply.
  15. Jamie

    Jamie Guest

    Go fly a kite you little fairy..

    You don't know the difference between the two, it's obvious.
  16. tempus fugit

    tempus fugit Guest

    Thanks for your reply Ian.

    I am building on a non-etched board - one of those perfboards with little
    copper pads you solder to (I assume this is the same as veroboard). What are
    some of the special layout and decoupling requirements you mentioned in low
    level audio circuits? I don't have any actual audio on this particular
    board - it contains 3 voltage regulators (for 5, 9, and 12v) along with the
    4049's in question, 2 8 bit latches, and a ULN2003 relay driver. The relays
    are located on a separate board and switch audio signals from guitar fx
    pedals in and out of the audio signal path.
  17. whit3rd

    whit3rd Guest

    Yes, that IS humorous. CMOS has good logic margins, it can tolerate
    more noise than TTL, but it VERY MUCH needs decoupling.

    The decoupling capacitors are required for the transient current
    needs of the 4049; probably milliamps and sub-microsecond times,
    so an inch of wire is not a real blockage (unless you use very fine
    wire), either in terms of resistance or inductance. You need
    wired-in decoupling within a couple of inches, but not within

    It's not really 'slow speed' that matters, but the transient current
    requirement. Fast ECL doesn't have any power-current change when
    the state changes, and doesn't need much capacitance to decouple.
    I've seen a chip (a self-clocking LED counter/driver in CMOS) with
    nearly 100 mA spikes; it took LOTS more decoupling than usual
    for that particular chip. This kind of problem occurs in CMOS when
    a slow-slew-rate input signal is allowed, like in a mixed-signal

    Multilayer circuit boards are easier to decouple than random-wired
    prototypes, because the ground/power layers are low inductance
    conductors. The 'as close as possible' recommendation is
    only a guide to prevent oscillatory feedback.

    Catchphrase of the RF beginner:
    "My amplifiers oscillate and my oscillators don't"
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day