Why do DMMs have a capacitor on their AC measurements?

Is a capacitor really necessary? And wouldn't it make measurements frequency-dependent?

-
 The difference between AC and DC is blurry and frequency-dependent. 100Hz is definitely AC, whereas 0Hz is definitely DC. Is 0.000001 HZ AC, or drifting DC? If you're designing an instrument, you have to make the call. Pick an impedance and capacitor value such that there is a flat frequency response down to such and such a Hz where you have -3 dB roll-off. – Kaz May 17 '12 at 4:15

A series capacitor acts as an high-pass filter, also known as AC-coupling capacitor. The reason is that for DC current, a capacitor is like an open circuit, whereas for higher frequencies (depending on C and the other resistances) it becomes more like a short circuit, if it's big enough.

The measurements will be indeed frequency dependent, but if the frequency is high enough the signal will fall in the right part of the curve, which is approximately flat.

So it's an useful tool, but to be used carefully. The same happens with scopes, where leaving the AC coupling on when measuring signals will distort them, like this square wave:

-

A series capacitor is not necessary to measure AC.
A capacitor to ground may be used to filter the derived DC.

Some meters do not utilise series capacitors.

The meter in the previous question MAY use a capacitor somewhere in the circuit but it is at least partially DC coupled as it reads 2VAC for every 1VDC when DC is applied. I have seen this behaviour frequently in the past with analog meters. The meter in the previous question may be an analog meter.

Use of a series capacitor allows DC components to be easily eliminated. This may or may not be desirable depending on the circuit used.

A series capacitor will cause some low frequency roll off but as the circuit can easily be very high impedance overall a modest value of capacitance will allow almost 'flat'measurement at mains frequencies. "AC" is usuallty 50 or 60 Hz in the large majority of cases.

If one will allow of the use of the dread Gargoyle in an answer then this link will be most informative regarding typical practice

Below is the circuit of an entry level "yellow' multimeter as found in zillions all over. Circuit diagram quality is only about 0.27 Olins but sufficeth for this task. It appears to use a simple half wave rectifier with no series capacitor. C_unreadable at left centre to the right od D_unreadable aoppears to be smoothing of sorts. [Far better circuit both in circuitry and in Olins is here and many more via gargoyle image link above.

Note: Use of Gargoyle image link here seems a very very good and appropriate use of the facility in an answer. If any disagree please advise and my second will call. [ :-) ].

-
I like the use of Olins to evaluate the quality of schematics! I think the components are C1 and D1 though. – clabacchio May 3 '12 at 14:46

If the AC signals you want to measure are pure AC, without DC component, it would be better not to have the series capacitor. You would only get a wrong reading when measuring DC.
If your AC signal is superposed to DC the capacitor will block the latter and only let the AC pass. The circuit which converts the AC to it's average will give a wrong reading if there's a DC component present.
And yes, the result is frequency dependent. A DMM will be tuned to measuring mains frequencies, i.e. 50Hz or 60Hz.

-

In fact, and now that you ask, a "true-RMS" volt/ampmeter should not have a capacitor in series, because the true RMS of a DC input is that DC value itself, and a series capacitor would just make that reading impossible.

Anyway, to answer your question, yes, a capacitor makes the measurement frequency dependent.

-
 False: if you want to measure AC, like the knob says, you can't leave the DC component, because the purpose of the AC function is exactly to filter that and have a more accurate measurement of the AC component – clabacchio♦ May 3 '12 at 14:33 @clabacchio - False false. RMS is the net heating effect of the signal components. It makes less than no sense to remove an arbitrary part of the signal. For RMS measurement purposes "DC" is an arbitrary part. DC level is the mean level such that voltage-time areas are matched above and below the reference line. That reference line may be meaningless and/or unknown without the meter "arbitrarily setting it. – Russell McMahon May 3 '12 at 14:45 @RussellMcMahon But if (as the OP says) you want to make an AC measurement, you simply don't care about the DC component, whether it contributes to the RMS or not. You simply want the AC, and get rid of DC; it's what the AC stands for. – clabacchio♦ May 3 '12 at 14:49 @clabacchio For a Fluke 114/115/117, Auto-V LoZ, AC mv and AC amps are dc-coupled, AC voltage is ac-coupled, and all ac functions and Auto-V LoZ are true-rms. What I'm saying is that all "true-RMS" choices should be dc-coupled, because otherwise the "true" there is just a lie. – Telaclavo May 3 '12 at 14:49 The point is that nobody mentioned true-RMS, but simply AC measurement, which usually (look at scopes, for instance) means that the input will be AC-coupled. – clabacchio♦ May 3 '12 at 14:51