Home Up Updates Current Products Prior Products  no longer available Documents Book Software Updates Softrock Lite 6.2 Adventures in Electronics and Radio Elecraft K2 and K3 Transceivers
 
Bypassing Effectiveness
I recently ran across an analysis by G4OEP of the
effectiveness of multiple bypass capacitors. His work, available at
http://g4oep.atspace.com/caps/caps.htm, looked at the effectiveness of
parallel capacitors, measured by their transient response.
I thought it would be useful to repeat G4OEP's work in the
frequency domain, by examining swept frequency responses of various bypassing
configurations over the range 300 KHz to 3 GHz.
The technique I used is to measure the loss between
two BNC connectors with various capacitor combinations shunting the signal
to ground using a home made test fixture. The shunting loss is measured by
an HP 8752B vector network analyzer. 0 dB loss corresponds to a BNC FF
adapter.

The test fixture. The height of the BNC center conductors
above the ground plane is approximately 1/4" (6mm).


Fixture Reference Loss
The test fixture is on the crude side and itself shows significant loss above a
few hundred MHz, as can be seen by the following sweep. The horizontal axis is
log and the major divisions are 1, 10, 100 and 1,000 MHz. At 2 GHz, the loss in
this simple test fixture is about 3 dB. If we take the 1 dB point as the cutoff
for reasonably acceptable results, we can trust the data up to about 1 GHz.

82uF Electrolytic  Long Leads. 
The first bypass we try is an 82uF electrolytic capacitor,
with long leads (3/8", or about 9mm).


If we set 30 dB loss as the threshold of acceptable
bypassing, the single electrolytic capacitor is effective up to about 7 MHz,
with 25 dB loss achievable through 10 MHz.


82 uF Electrolytic  Shorter Leads

Our first attempt at improving the effectiveness is to reduce
the lead length to the minimum possible, or about 1/8" (3mm) for this fixture.


Shortening the leads helps, with the 30 dB point now 20
MHz. At low frequencies where the lead inductance is negligible, both test
configurations show about 42 dB shunt loss.


82 uF Electrolytic paralleled with 0.1uF.
A common strategy to improve bypassing effectiveness is to
parallel large value capacitors with smaller value capacitors. One potential
problem—the reason G4OEP conducted his transient tests—is a concern that the
stray inductance will resonate with the smaller capacitor and thus reduce the
bypassing to ineffectual levels near the resonant frequency. G4OEP tested for
resonance by looking for ringing when the bypass filter arrangment was subject
to a fast rise pulse; our examination will be more direct by looking for
resonant effects in the frequency domain.

0.1uF shunting the electrolytic. The total lead length is
under 0.25" (6mm).


We see definite signs of resonance in this configuration,
with a notch at 8 MHz.
Still, paralleling the 0.1uF capacitor provides
significantly more effective bypassing than the electrolytic standing alone.
At 100 MHz, the single electrolytic provides about 17 dB shunt loss, whilst
the parallel combination increases the shunt loss to about 26 dB at 100 MHz.
There's only a small increase in worst case bypass effectiveness with this
combination.


82 uF Electrolytic paralleled with 0.1uF and 1000pF
If paralleling a 0.1uF helps higher frequency response,
perhaps paralleling a 1000 pF will offer even more improvement.

Three paralleled capacitor bypass configuration.


We see the 1000pF has added a second resonance, around 50
MHz.
Comparing the three capacitor arrangement with the two
capacitor configuration, we see at around 50 MHz, the two capacitor
arrangement is slightly better. However, at 100 MHz, three parallel capacitors
offer better performance.
The specific resonant frequencies and overall
performance are, of course, highly dependent upon lead length and circuit
layout, so take these measurements as only indicative.


4.7uF Tantalum Capacitor.Tantalum capacitors
offer less selfinductance than conventional electrolytic capacitors and should,
therefore maintain bypassing effectiveness to higher frequencies, all else being
equal.

4.7 uF tantalum capacitor installed with minimum possible
lead length.


Since the capacitance value is only 4.7uF, we don't expect
nearly as much shunt attenuation for low frequencies compared with the 82uF
electrolytic.
We see less than 30 dB attenuation throughout the
frequency range, but note that 20 dB attenuation is possible up through 60
MHz. This is a bit better than the 82uF electrolytic, and the tantalum
is a much smaller package.


Surface Mount Capacitors  Series 0.22uF.If
lead inductance is part of our problem, suppose we try a leadless, surface mount
capacitor. In this case, in order to physically fit, I've made a series stack of
two 0.22uF 1206 surface mount capacitors, for a net of 0.11uF.

Not very pretty, but it works.


As expected, we don't see much attenuation at low
frequencies, since the total capacitance is only 0.11uF.
There's a clear resonance at 9.5 MHz, and our arbitrary 30
dB cutoff point is about 43 MHz, with 20 dB attenuation available through 150
MHz or so.


Series Ferrite BeadAn alternative to
shunting capacitance is series inductance. In this case, we'll try a multiturn
ferrite bead, Steward 28C02360EW10, 3turn device, with a series impedance of
998 ohms at 100 MHz.
http://www.steward.com/web_info/CADPrints/Sales/28C02360EW10F.pdf.
Ferrite beads are intentionally made of low Q material so that they offer both
series reactance and resistance to high frequency signals and to minimize
resonance effects.

The data sheet shows that at 100 MHz (the bead's
selfresonant frequency), the bead appears as almost a pure resistance, with
zero reactance. This is possible because the ferrite material used has high
loss. You would not chose this material for a resonant circuit, but it's ideal
for supressing EMI or bypassing undesired frequencies.
Above its 100 MHz SRF, the bead looks capacitive, probably
due to turntoturn parasitic capacitance. Still, the data sheet demonstrates
useful series impedance up to 1 GHz.


3turn ferrite bead in series.


Not much effectiveness until we reach 50 MHz or so. The
bead remains useful up through 1 GHz, however.


Bead with 2x 0.1uF bypass capacitorsIn
addition to the series ferrite bead, it's possible to bypass the leads with
capacitors as well, thus combining shunt and series actions. We start with 0.1uF
bypasses on both sides of the bead.

Ferrite bead with 0.1uF bypasses on both sides.


Now we are starting to get someplace—30 dB or more
attenuation over the range 1 MHz through 400 MHz, with a best case attenuation
exceeding 70 dB.


Bead with 2x 0.1uF and 2x 1000pF Bypass Capacitors
We can extend the parallel capacitor strategy by adding
1000pF capacitors to both sides of the ferrite bead. This should improve the
high frequency loss.

Pifilter with paralleled shunt capacitors and series
ferrite bead.


Adding 1000pF capacitors improves the shunt loss to at
least 30
dB up through nearly 1 GHz.


ConclusionThere's no single approach to
bypassing, as we must start with an understanding of the frequency range over
which we wish the bypassing to be effective. And, we must know the impedance of
the circuit being bypassed. As the impedance drops, the bypassing loss of a
given shunt capacitance decreases, but the effectiveness of series inductance
increases.
The network analyzer
plots presented here are taken with a 50 ohm impedance. If the circuit being
bypassed is a low impedance power feed, then different component values will be
necessary. Likewise, a high impedance control circuit may call for different
component values. For example, component values for bypassing RS232 serial data
lines require knowledge of the data speed as well as the RF frequencies to be
protected against.
The data also supports the long accepted practice of
paralleling different value capacitors to extend the effective bypass frequency
range. Although in some circumstances resonant circuits can be created, the
overall effectiveness still is improved over a single value part.
The longheld practice of keeping lead lengths as short as
possible is certainly valid and should be followed, as well as using components
suitable for the frequency band to be bypassed.
However, resonances and bypassing effectiveness are
critically dependent upon circuit layout and associated strays. Accordingly,
take these measurements as representing only concepts and suggestions, not hard
and fast rules.




