# Is my Precise Capacitor Ratio Repeatable?

I have designed a 1Mohm, 0Hz-100MHz fixed attenuator circuit (15pf/300pf). I am using 1%, C0G capacitors. There is also a TS5A2053 analog switch connected to the output of the attenuator. The switch will add 18.5pf of capacitance to ground (typical). I will layout the PCB to minimize stray capacitance and I will tune the values to that layout.

My question is, when this ckt is produced in volume how much variation in the capacitance ratio should I expect to see from unit to unit? And what are the causes of this variation?

Obviously the capacitors (+/-1%) themselves will yield +/-2%. The resistors are negligible. C0G should not change much with age? Would soldering change the value? How much variation should I expect from the analog switch (only "typical" is specified)?

Edit: Sorry, I was vague about the schematic; but to be specific, it is a simple voltage divider - two resistors and two capacitors.

Edit: I have found several sources which say that the capacitance of a mosfet does NOT vary with temperature. So, the capacitance of my analog switch will not likely vary with temp - I think only the switch's initial capacitance will vary due to process differences (thickness of the diffusion channel?).

• Can you supply a circuit diagram. You talk only about capacitors but then what does the 1Mohm refer to? Why is the effect of whatever resistors you have negligible? You can't get down to 0 Hz with only capacitors. Are you trying to build something similar to an oscilloscope X10 probe? – Barry Aug 11 '15 at 0:38
• The resistors are negligible because they have less than 0.1pf of capacitance. I have confidence that the resistance will be as expected. Thus the question is only about the capacitance. And, yes, the circuit is for the input of and instrument like application. I am trying to avoid mechanical relays and manual adjustments. – PaulB Aug 11 '15 at 2:21
• You're thinking a bit backwards. Your specs should be in terms of the response you're looking for, and then you need to think about the components you need to achieve it. – Scott Seidman Apr 25 '16 at 20:53
• How very condescending. Your comment is in no way helpful. I asked the exact question that I meant to ask. I have not chosen components arbitrarily. – PaulB Apr 30 '16 at 0:51

The 1% capacitors have a ratio accurate to +1.92/-1.89%.

Possible variation in the input capacitance of the switch is hard to estimate without more knowledge of the processes, a WAG might be 10-20% of the nominal value to 2-3 $\sigma$. At 20% it would be adding +/-1.2% to the error.

Maybe consider using a lower capacitance switch such as a USB switch- some of them are in the 7pF nominal range. You may also see some variation of the switch capacitance with voltage which could cause some distortion of the waveform.

Of course whatever you have on the other side of the switch is going to add to the capacitance etc.

Good C0G capacitors should have negligible voltage coefficient and should be very stable with temperature (< 30ppm/K) and time. Do check the detailed information (not just the data sheet) from your proposed supplier.

• Thanks. I looked at USB switches; but I haven't found any in leaded packages that are also low cost. BGAs and QFNs would greatly complicate assembly and at low volumes would increase my cost. – PaulB Aug 11 '15 at 2:32

Without a schematic it is difficult to asses the problems your circuit could have.

Anyway, what I see as a potential problem, as far as both the repeatability and the precision is concerned, is the analog switch.

I doubt that the 18.5pF figure is specified to the same precision as your other parts (1%). Moreover, that capacitance may also drift sensibly with temperature (it's a semiconductor device) in an unknown way. That's why in many high-end instruments precision attenuators for low-level signals are often made with micro-relays.

• Thanks. Yes I am trying to avoid mechanical switching and manual adjustments. – PaulB Aug 11 '15 at 2:46