Comparing caps for same-same nanofarads

[petemoore]

  Where to get 3n3 capacitros @ <2% tolerances?
  and 33n's also
  Is this an SMT Device?http://www.avx.com/docs/Catalogs/ccog.pdf
  Another whacky post.
  http://www.quadtech.com/products/family.asp?id=2
  OUCH, cap meters...I'll have to check my piggy bank...I would guess someone would want to rent theirs...lol.
  It would seem larger uf cap values take longer to discharge from x voltage than smaller values.
  Is there a way I could get my stopwatch out and measure the voltage drop time constants of 3n3 capacitors to get better matched ones...how long it takes for voltage to drop a certain amount using a resistor to slow the discharge?...not so much to hit precisely 3n3 every time, but to find caps near that value, but closely clustered.
  I did find some 33nf 1% in the 650V monster size range, but I'd have to pay like pal @ 6.53 bux a pop x6. Then I'd still have the 3n3's to source.
  Perhaps someone with a capacitence meter sells small capacitors, sorted from 'quality bin' to be within 1% or 2% comparitive values...all near 3n3?
  Or...some other way of sourcing suitable capacitors for an active crossover which recommends the use of 1% [or 2%] tolerance?
  I called parts express and mouser...but found 3n3's only in ceramic @ +/-10% tolerance.
  http://nov55.com/cap/fig1.htm
  Looks simple enough...lol...
  But here's the rest of it !
  http://nov55.com/cap/cap1.htm#1.

[petemoore]

http://www.quadtech.com/products/family.asp?id=2
  These look pretty good, I didn't see a regular % tolerance, but a bunch of temperature/value stuff on them, and those figures look 'tight?
  Maybe see which ones win the races to x voltage charge from 0v, with some type of comparator which lights an LED when a winner is determined?...then the same type of test to see which ones discharge the fastest?..
  or some way to get a batch of suitable caps for my hipass xover?
  Phase shift and distortion in a crossover are beginning to look alot less undesirable.

[puretube]

mmmh... can`t remember exactly how i did it,
but somewhere I seem to remember in the back of my head about finding several
different quadruples of identical caps (to 0.1%) from a grabbag once,
by hooking up a voltage divider with the CUT (capundertest) and a resistor,
and feeding that with a constant AC voltage/frequency (mains),
and measuring the voltage drop across either of these 2 components...

for a rough orientation I used a table of approximately "to be expected" voltages
by inserting a "known" tightly tolerated polystyrol cap as a "reference" in the testclips,
and then going thru the hole bunch.

the exact value wasn`t that important;
but identical 4 was (or where it 8 identical ones for vocoder filters?);

of course the cheap multimeter wasn`t better than 2 or 1%, but I did find very
closely tolerated values that way, that indeed went a couple of digits right of the decimal point...

[petemoore]

  the exact value wasn`t that important;
but identical 4 was
  From what I can gather, that's exactly where I am, exact match is best, close is good, whatever I threw in there seems to work so I can't define any differences yet.
  It would seem there is a way to determine cap value ratios without buying a many-thousand-dollar meter, your way sounds pretty good, a whole lot better than swapping the 4+ positions with various values using caps marked for example 332.
  I don't know whether studying your post would be enough to cause me to further investigate, in the form of a build though.
  Or...someone with the fascility to do so, sorting nice sounding 10% ers into groups of 4 or 6 with tight ratios, near the desired value, similar to matching Jfets for phasers or Ge's for FF's..that and some field testing to see what size ballparks the applications have [ie...perhaps 2% caps in my application get 99% of the desirable results of 1% caps according to reports].
  With these four positions socketted, and copious HF response sound, I'm in a good position to do some testing...[of course I haven't been able to even source 1 nicely matched set of 4 caps near 3n3 so that's out for this week].
  To slightly quell my curiousity...
  While shopping for fancy caps, I  found four cleargloss crossfiberglass yellow looking capacitors marked 332 !
  They had some electrolytics in the neighborhood marked 10v..and others marked 6.3v..but, I figured they are probaby 16v or 25v because of their size and availability at the time of the toy keyboard's mfr.
  Of course I tried them out.
  This PA is now very hot !
  Yes, I am impressed with myself, but humbled by the help-work I'd like to thank everyone for.
  Besides that though, the performance is outstanding!
    Tube amp 15w drives 2x12''s full range except the tone knob is turned down, the greenbacks have a 5khz high mark for freq response.
  LM3886 is driven by the active 3.1Khz Xover, using a stereo mixing board to preamp the hi and low freqs.
  OT and adjectives aside, big chuckles on the thick full low-mids-mids response [no thunderous bass, yet], and hot HF output with this system, dynamic response on the mic is to write home about!  I might be done working on it for a while...nope, you know I'm going to be doing more work on the active just to see, as well as other amp/speaker testing !!! For now it is high performance ! [some 'predicted' hiss in the HF is to be expected/about average for the HF 'heat' level], the high end has plenty of 'reserve heat', that's what I wanted..more high end than I could blow HF drivers with, 100w worth seems to be more than adequate. Sweetest singing PA I've heard since...I've heard some really nice PA's..I'm getting 'cheep thrills' !

[Nasse]

Wheatstone bridge perhaps

[petemoore]

  http://en.wikipedia.org/wiki/Wheatstone_bridge
  Yes, that'd be this one I believe.
  And perhaps pushing HF AC into the shown circuit, as I'm reading I see capacitence and AC mentioned..along with 'extreme accuracy'..this is worth the read regardless...when in doubt, compare if possible.
  and Maxwell's bridge
  http://en.wikipedia.org/wiki/Maxwell_bridge oops that one's a stub'.
  Here's an interesting one

http://www.efunda.com/designstandards/sensors/methods/wheatstone_bridge.cfm
  [/list]
  But I'm not really catching on to anything about capacitor measurement, and I would copy the equations and try working some of them, but I doubt I'm the one now to crack the cap testing delimma.
  Is there a way to set up a scope to set up a crossover with matching caps?
  I don't have one, we used to, who knows,  but I'd have plenty of other things I would want to try with one.

[R.G.]

555 timer with precision resistors to determine frequency makes the capacitor the thing which determines. In fact, non-precision resistors let you match them, if not guess the tolerance.

Capacitive dividers?
Two caps in series divide AC voltage just like two resistors do. If the voltage is half, the caps are matched.

Current ramp?
I = constant and small. I=Cdv/dt, so C=I*dt/dv, or C = current times the time it takes to change by X volts. Again, if all you're doing is matching, I does not have to be precision. But your voltmeter DOES have to need much less current than the source provides, so a FET input opamp or comparator is a good "voltage now" trip.

Frequency resonance with an inductor?
F = (1/2*pi)*SQRT(L*C), again, L can be anything if you can measure frequency.

RC divider as Ton said. Good for accurately measuring the Cap if you have 1% resistors.

Charge sharing? Charge C1 to 5V. Short C2. Open C2, connect C1 to it. Q (stored charge) = C (capacitance) times voltage, so
C1 has charge in it of Q=5*C1. When you connect C2, the same charge is there because C2 was at zero. So the new voltage must be V = Q/(C1+C2). If C1 and C2 are equal, the voltage is half. Can be made to measure with precision. If you have a 1% cap of any value, charge the precision cap to X volts. Then connect to the cap under test. The known cap has charge Q = Ck*V. When you connect it to the unknown cap Cu, the voltage changes to V = Ck*V/(Cu+Ck).