Hello, I'm reading up about bypass capacitors and I'm having trouble understanding why large caps are used for low frequencies and small caps are used for high frequencies. I've covered capacitive reactance and I get the inverse relationship between Xc and f/C, but the explanations tend to tack the large cap-frequency bit at the end. Apologies if this has been covered or I have overlooked something. Thanks for the opportunity to post! JC

 

It is about the parasitic properties (mostly inductance) of real capacitor types; electrolytic vs tantulum vs ceramic vs film, etc. For example, spiral-wound, aluminum foil electrolytic capacitors make great power supply filter capacitors at 120Hz, but they suck at bypassing RF or even SMPS switching frequencies of a few tens to hundreds of kHz.

 

You might find more detail helpful.

Analog Devices tutorial (pdf)
Decoupling Techniques

A thread on All About Circuits:
Decoupling or Bypass Capacitors, Why?

 

You might find more detail helpful.

Analog Devices tutorial (pdf)
Decoupling Techniques

A thread on All About Circuits:
Decoupling or Bypass Capacitors, Why?

Thanks very much!

 

If you have fairly high switching rates of signals that are modulated you sometime need up to three bypass capacitors for proper operation (send and receive the cleanest signal). I'm building a light-link tester that normally runs at 375000 BPS on 1 MHz transceivers but to test the links I have 50mhz transceivers so I can characterize the loss slope to see if the receiver photo-diode, transmitter emitters or cables of the OEM devices are going bad in remote loop-back mode. With a PIC 45K80 usart at 64 MHz fosc I can can step the serial BPS up to several MHz (5+) while looking for framing errors from bad serial return data and use the CTMU module to look for the actual loop time in ns. (send pulse start to receive pulse stop timing)

To reduce ringing, improve signal generation and detection I have electrolytic, tantalum and metal film caps at the light-links with tantalum and ceramic caps on the uC for bypasses. I have a ring at about 22 MHz on the receive signal but that's pretty good for a Vector board with no ground plane and 30 gauge wire point to point.

PIC RS232 slew rate on the top (slow even with SLRCON = 0), link receiver of the bottom. (fast)

This is where old high bandwidth analog scopes shine but the fancy ones have their place too.