Connect with us

How inaccurate is a 555 or 7555 REALLY?

Discussion in 'Electronic Design' started by WildIrish, Dec 2, 2006.

Scroll to continue with content
  1. WildIrish

    WildIrish Guest

    Two questions. I'm a novice with 555's, some minor experience.
    I need to develop a very small 555 timer circuit that runs on a 1.5v AA
    battery.

    1. Which 555 is out there that runs that voltage and small package?
    (smd would
    be fine as long as It has solder pads i can solder to under microscope)

    2. I plan to calibrate it up to run at 420hz. I'm still unsure if 555
    is accurate enough,
    I have read some Maxim and National spec sheets and I'm unclear on
    accuracy.

    How inaccurate is it truely in percentage at 420hz? I really can't
    afford more than .1%
    inaccuracy, is there a way to reach these levels of accuracy using 555?
    special
    R and C to make that happen?

    The only reason I want to use 555 is it seems very tiny (space is an
    issue) and
    rather simple, but maybe there's another more accurate way? I'm open
    to suggestions if they use minimal # of components.

    Thanks a bunch for all your help in advance!
    -Brendan
     
  2. You could hand match R and C to potentially get that, but then it'll
    change with temp and other parameters. So practically you won't
    achiveve 0.1%

    If it's just a fixed frequency you want then use a crystal oscillator
    and divider. That can be made as small as or smaller than a 555
    solution.

    Alternatively use a tiny microcontroller in an SO-8 or maybe SOT-23
    package. Internal oscillators are only 1% accurate or so, but if you
    use an external crystal you'll get what you want. Use software timing
    or a PWM to give you the frequency you want.

    Dave :)
     
  3. TLC551, SOIC-8. Not really small.
    Expecting much better than ~1% over reasonable temperature changes,
    supply voltage changes etc. is probably optimistic. Not only the chip,
    but capacitors and other parts will change.
    You could step the 1.5V up to 3V using a switching regulator and then
    use a CMOS chip or a microcontroller with a crystal timebase, and
    avoid the calibration.

    Best regards,
    Spehro Pefhany
     
  4. mri_bob

    mri_bob Guest

    actually the good old 555 is stable to .005% per degree C so your first
    task to get .1% is to keep it in the range of +/- a few degrees C. then
    you have to turn your 1.5v into the minimum operating voltage of 4.5v,
    easy enough to do with a maxim charge pump chip which is tiny and smd and
    needs only a couple of small caps. then you have to have your resistor and
    capacitor be cermet and polystyrene so they will have a good enough
    tempco. if you do all this you could easily get .1% out of a 555. the chip
    is the least of your problems with an RC time constant, the problem is
    temperature coefficient. you can get extremely low tempcos by combining
    positive and negative tempco components to get a single one with almost
    zero tempco.

    i agree this is a job for a microprocessor. you can get a PIC in an SO-8
    package and to get .1% stability or 1/1000 you would need something like a
    watch with a stability of 1 second every 16 minutes. a pretty easy thing
    to do, as you can see. when you want precise stable timebase related
    functions a computer is the easiest and best way to do it.

    -bob
     
  5. Guest

    metal film
    polyphenylidene sulphide
    Fat chance. You'd have to stqabilise the voltage as well as the
    temperature - the divider inside the 555 is a series of three
    voltage-sensitive diffused silicon resistors.

    I spent a month or so back in 1974 on exactly this problem, and
    discarded the 555 very early on.
    For a individiual capacitor - the temperature coefficient of wound
    capacitors depends on the mechanical tesion in the dielectric, and the
    tolerance on te temperature coefficient is all over the place.

    Polyphenylidene sulphide sufrace mount capacitors are not wound (as far
    as I know) and do better, but they still don't hack it at the 0.1%
    level.
    10MHz crystals are sold with a +/-0.1% absolute tolerance. 32768Hz
    watch crystals aren't that good (if I remember my Farnell catalogue
    corectly.
     
  6. mri_bob

    mri_bob Guest

    of course crystals need to be trimmed to get them exactly on frequency, but
    they do stay there. When was the last time you heard of a digital watch
    that could not keep time to better than a minute and a half per day? .1%
    accuracy and stability is easy with a crystal system.

    as for capacitors, a cheap film capacitor is about 200 ppm per degree C so
    if you are keeping it within the 5 degrees you need to get .005% out of
    the 555 you will have .1% stability right there. you can do a lot better
    than a cheap film capacitor.

    as for the voltage, i said you needed a voltage converter to run the chip,
    and these are regulated. the 555 uses 2/3 of the supply voltage as the
    comparator switch point, so it has a lot of immunity to power supply
    voltage changes.

    you can see that there are several variables with the 555 that are close
    to .1%, so by the time you add them up it would be difficult, but the last
    poster obviously has never combined parts with different tempco directions
    to get one that is better than either. i have.

    the last poster made it sound like getting .1% accuracy out of a
    microprocessor based system was impossible. he needs to go look at the
    accuracy of a $2 digital watch again. And if you could not get an RC time
    constant to stay stable to .1% your 100 MHz FM radio would drift 100 kHz
    and you would be on another station every time the room changed
    temperature. let's get real, huh?

    -bob
     
  7. Phil Allison

    Phil Allison Guest


    ** By whom ?

    All the ones I see have 20, 30 or 50 ppm accuracy.

    Equates to 0.002%, 0.003% or 0.005 % tolerance.



    ** They have 20 ppm @ 25C accuracy too, I'm afraid.




    ...... Phil
     
  8. WildIrish

    WildIrish Guest

    ..002% would be good enough for me.... but i don't know where to find
    examples
    on making a tiny quartz crystal divide down to 420hz in a tiny package
    (I don't have more than about 3/4" box to fit things into.) or any
    other component
    oscillators that can reach 420hz with the least number of components.

    Any basic circuit diagrams online that someone can point me to would
    be greatly appreciated! Sine wave output preferred, but I think square
    will do
    if that ads complexity (I bet it would). Again please remember I'm a
    novice
    with minor experience :)

    Thanks so much,
    ~Brendan
     
  9. Eeyore

    Eeyore Guest

    You won't get a 420Hz crystal. Why 420 btw ?

    Your best bet is use a tiny microcontroller with a standard crystal and use that
    to generate a crystal derived 420Hz. That involves some ( simple ) programming
    though.

    Graham
     
  10. Phil Allison

    Phil Allison Guest

    ** Groper alert.


    ** Then it is time to realise that YOU are not up to the task.

    The size and performance criteria you have come up with make it a job for an
    expert.




    ........ Phil
     
  11. Phil Allison

    Phil Allison Guest

    "mri_bob"

    ** I got news for you pal - FM radios do NOT use RC oscillators.

    They DO use temperature compensated LC oscillators and even then the
    temperature drift is still so bad that in nearly all cases an "automatic
    frequency control" ( AFC ) circuit has to be used to keep the radio on
    station.

    Modern hi-fi FM receivers use frequency synthesisers and so have the
    accuracy and stability of the crystal used in that circuit.



    ........ Phil
     
  12. Guest

    Dream on - the 2/3 ratio depends on the diffused silicon resistors, and
    they drift around with voltage.
    May be you have done it with one resistor and one capacitor. Try doing
    while paying attention to worst case tolerances, which is what you have
    to do for even small volume production, which is what I was being asked
    for back in 1974.
    Rubbish. anything with a 10MHz crystal will do 0.1% without any effort.
    As Phil Allison has pointed out, getting 10MHz crystals stable to
    better than 0.1% over their temeprature range is just a matter of
    paying more money. Phil claims that better than +/-0.1% watch crystals
    are also available, and he's rarely wrong on that sort of detail -
    mymemory suggestsed that there was a tendency to rely on the human
    wrist to keep the temperature range at the crystal within a fairly
    narrow range.
    They are pretty awful. The slightly more expensive watch I use needed a
    session with a good frequency meter to pull the crystal onto 0.5
    seconds per day (+/-0.01%), but that was some twenty years ago.
    Ever heard of automatic fequency control? Getting 100hkHz accuracy out
    of an FM radio depends on dividing down the local oscillator to
    somethihg like 10MHz and using a phase-locked loop to keep the local
    oscillator locked to a decent crystal - once you have found the station
    you want, the system then locks the local oscillator to the station
    with some kind of automatic frequency control circuit, and the
    frequency control voltage does change with room temperature. Try to
    understand the technoligy you pontificate about before you claim to be
    getting real.
     
  13. Guest

    32768Hz divided by 78 gives you 420.103Hz, which is 0.0244%.

    That is a seven bit divider - which you can set up with a single
    74HCT40104

    http://www.ortodoxism.ro/datasheets/philips/74HCT40103D.pdf

    which is availalble in a 10mm by 6.2mm surface mount package - rather
    smaller than your 19mm square target space

    10MHz divided by 23810 gives you 419.992Hz which is rather better.

    If you poke around the standard crystals available off the shelf from a
    broadline supplier like
    Farnell you should be able to find something gets you closer to 420Hz -
    10MHz is pretty much the the worst possible place to start.

    Dividing by 23810 calls for a 15-bit divider - two 74HCT40103 parts in
    series, or a single programmable logic part which you would be able to
    find in a 44-pin square package - the first example I came up with
    occupied a 12mm square., which is about the same as a pair of surface
    mount 40103 parts. Phil Alison should be able to show me up by finding
    something smaller.
     
  14. Phil Allison

    Phil Allison Guest


    ** Phil said no such thing.

    If you must quote me out of context, use my words and not your own.



    ** Phil actually said that standard watch crystals are made to 0.002 %
    accuracy or 20ppm at 25C.



    ** Yep, 20ppm = 52 seconds a month drift.

    Only if the crystal is frequency trimmed and the watch worn on the wrist can
    time keeping get seriously accurate.



    ** Errr - 0.01% = 8.6 seconds per day .

    0.5 second per day = 5.8 ppm or 0.00058 % !!!!




    ** AFC is not necessary with synthesised FM tuners and IME simply not
    sed - standard crystals are plenty stable enough with around 1 ppm drift
    per degree C.

    Even radio scanners work fine with narrow band FM ( 25 kHz wide) at UHF
    frequencies without AFC.



    ........ Phil
     
  15. mri_bob

    mri_bob Guest

    i did not make up the .005% spec for the 555. It is on the first page of
    National's data sheet. the poster said he wanted .1% and i said the 555
    would not be the controlling factor.

    as for prototype/production issues, the guy said he wanted to make one, he
    didn't say 1000. your experience in a production environment is
    irrelevant.

    i said crystals were good enough. i don't know why you are arguing that
    point? you seem to be misstating my position on that one just so you can
    disagree with me.

    i said .1% is a minute and a half a DAY and a $2 watch is better than
    that, then you rant about needing to go to a lot of trouble to get .5
    seconds per day and say that is .01% Well, there are 86,400 seconds in a
    day so .01% is 8.6 seconds. You are off by over an order of magnitude and
    you are berating me saying i don't know what i'm talking about?

    and you do not need automatic frequency control in a radio to keep it from
    drifting 100 kHz. nor do you need a PLL. a cheap simple fm ham receiver
    with no AFC will not drift that much by an order of magnitude. to quote
    you: "Try to understand the technology you pontificate about before you
    claim to be getting real"

    bob
     
  16. Phil Allison

    Phil Allison Guest

    "mri_bob"

    ** The Natsemi data book gives the following for the LM555C:

    Temperature drift: 0.015 % *per degree C* for the astable mode.

    Drift with Supply: 0.3% per volt.


    The 555 will certainly be a *major source* of frequency drift.

    You MUST allow a for at least a 25 degree C change in temp - ie from 10 C
    to 35C.

    So, even with *perfect* Rs and C sand a rock steady supply, temp drift
    will be 0.375 %.



    ** That is absolute BOLLOCKS !

    A well made, variable 100MHz LC oscillator will drift by around 5 kHz per
    degree C.

    However, once the gear has warmed up and if the room temp is stable, there
    is no temp change going on !!!

    If there is NO temp change, then there is of course NO frequency drift
    !!!

    This the DUMB error you are making.



    ........ Phil
     
  17. Robert Baer

    Robert Baer Guest

    Ignoring the insufficent supply voltage, the basic inaccuracy of the
    555 design rests on the offset voltage of the comparitors combined with
    the mismatch of the internal resistor divider ratios.
    Crudely speaking, 0.1% cannot be achieved with one 555.
    If you want a relatively accurate frequency, look to crystal or
    tuning fork oscillators using discrete SMT parts.
     
  18. mri_bob

    mri_bob Guest

    please let's go back to my original answer and let me explain what i was
    trying to say. when the guy said he was a novice and wanted to build
    something with a 555 that was stable to .1% the FIRST thing I said was
    that he would have to keep it temperature stable to a few degrees to have
    any chance, and that if he did that it would not be the 555 that would
    prevent it from being done. i was trying to help him to understand the
    variables involved in the analog implementation by analyzing the best that
    might be done with common analog parts, then I said that there were
    several sources of .1% error that would add together and so it would be
    very difficult to do what he wanted. I then told him that a crystal based
    microprocessor circuit would be a better choice.

    bill then attacked my response, ignoring my first statement about needing
    to control the temperature tightly. he said watch crystals were not good
    enough to get .1% accuracy. I pointed out that a cheap watch was easily
    accurate to .1% because that was a minute and a half a day. he proceeded
    to come up with an argument against watches being that accurate by grossly
    miscalculating the seconds in a day.

    my last statement that a 100 MHz FM radio would not drift 100 kHz without
    AFC is true from my experience with older radios with no AFC after they
    have warmed up. I should not have made the statement about "get real" and
    i apologize. I was reacting to what i felt was an unjustified attack
    against my attempt to explain to a novice the various sources of error
    when trying to get an analog circuit stable to .1% I did not mean to
    insult anyone.

    -bob
     
  19. Phil Allison

    Phil Allison Guest

    "mri_bob"

    ** True - the actual temp range figure is +/- 3 degrees C for +/- 0.1%
    frequency variation.

    So totally impractical in the real world.


    ** He does that kinda stuff a lot.



    ** Nice to see you read my explanation.

    No temp change = no frequency drift.



    ** No offence taken here.

    Frequency stability of oscillators is a non trivial topic.



    ...... Phil
     
  20. Arlet

    Arlet Guest

    If you can afford to use a 3V lithium cell instead, there are a few
    microcontrollers in SOIC-8 package that have a 32.768 kHz xtal
    oscillator option. Divide by 78 will get you 420.10 Hz. While you can't
    avoid the 30 usec jittter, you can improve long term accuracy by
    dividing by 79 once in a while (use a phase accumulator to add up phase
    error until you get a whole cycle)
     
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

-