Discussion:
Anticapacitor?
Karel Kulhavy
2005-12-07 11:04:07 UTC
Permalink
Is it possible to make a passive network from coils and capacitors that
would approximate negated reactance of 6pF capacitor in the frequency
range from 1MHz to 10MHz?

It has to be passive device because I want to use it to compensate out
capacitance of photodiode in low-noise broadband preamplifier.

CL<
Dan McMahill
2005-12-07 12:36:03 UTC
Permalink
Post by Karel Kulhavy
Is it possible to make a passive network from coils and capacitors that
would approximate negated reactance of 6pF capacitor in the frequency
range from 1MHz to 10MHz?
It has to be passive device because I want to use it to compensate out
capacitance of photodiode in low-noise broadband preamplifier.
Short answer is probably not. Depends on the rest of the circuit. If
you have a low resistance driving the 6pF so it's only a minor pain,
then yes. If the 6pF is driven by a high resistance, then no.

Longer answer is Bode showed how to compute such things in his book
"Network Analysis and Feedback Amplifier Design". I doubt the problem
he solved is 100% identical to yours, but it is similar. Basically what
he proved is that for a circuit with a resistive source driving a
capacitive load, no matter what LC network you stick in between, you
can't exceed a certain value for the integral of gain over frequency.
If you like math and circuit theory, Bode's original proof of this is a
pretty straightforward read. Fano then expanded and generalized Bode's
result, but Fano's paper is a harder read.

Bottom line, you can always do such things (with ideal components)
perfectly at one frequency and the wider the frequency range, the worse
you will do no matter how many elements you throw at the problem.

-Dan
Bob Paddock
2005-12-09 00:41:40 UTC
Permalink
Post by Karel Kulhavy
Is it possible to make a passive network from coils and capacitors that
would approximate negated reactance of 6pF capacitor in the frequency
range from 1MHz to 10MHz?
It has to be passive device because I want to use it to compensate out
capacitance of photodiode in low-noise broadband preamplifier.
You can use a Common-Base transistor to isolate the photodiode capacitance.
National Semi. shows a circuit doing this based on their LMH6642 op-amp.


http://www.national.com/dt/LMH_Qual.pdf see page #22 for the Common-Base isolate,
LMH6642 data sheet shows the circuit as well.

Karel, you might want to had this link to your web site, gives good overview of photodiode
issues:

http://www.national.com/onlineseminar/2004/photodiode/PhotodiodeAmplifers.pdf
http://www.national.com/onlineseminar/2004/photodiode/photodiode_transcript.pdf

The 'Anitcapacitor' reminded me of the 'Nonlinear Capacitor' , THAT Corp. App. Note 103:

http://www.thatcorp.com/datashts/600034-1.pdf

I always thought it was cool solution in search of a problem...
Bert Douglas
2005-12-09 02:44:29 UTC
Permalink
This is for Karel.

Part of your problem is that you are using baseband signalling over the
optical link. That is, if ethernet signal is high, LED is on, if ethernet
signal is low, then LED is off.

Since this signal was not designed for an optical link, it is highly
unlikely that it is optimum for the task.

An optical link is likely to have significant changes in signal strength
over relatively short periods of time. On ethernet this doesn't happen.

So you might want to consider re-encoding the ethernet bits to a different
signal pattern on the LED.

If you apply direct spreading you can get some code gain. And since the
interesting cases are always at the margin where there is a lot of noise,
you can add some FEC (forward error correction).

In really high noise cases, you can lower the effective data rate and use a
longer spreading sequence. It is better to get a little data through than
to get no data through. This document may be helpful:

http://public.ccsds.org/publications/archive/131x0b1.pdf
joeft
2005-12-20 19:37:05 UTC
Permalink
Two details I'm pretty sure I've seen documented, but can't pin down now
that I need them:
1) Is there an explicit "no connect" symbol that I can/should place in
the schematic to prevent gnetlist from thinking I've forgotten a connection?
2) How do I place a pin label in a symbol that will indicate an active
low or inverted signal (i.e. how do I put a "bar" across the signal
name?) I know I can put a dot at the pin to indicate an inversion, but
I would like the pin name to match the manufacturer's documentation.

Thanks,

Joe
sdb-jPU6KpCpLHWsTnJN9+ (Stuart Brorson)
2005-12-20 19:55:14 UTC
Permalink
Post by joeft
Two details I'm pretty sure I've seen documented, but can't pin down now
1) Is there an explicit "no connect" symbol that I can/should place in
the schematic to prevent gnetlist from thinking I've forgotten a connection?
misc -> nc-left, nc-right, nc-top, nc-bottom.

Caution: I have seen situations where this symbol creates a net
called "no_connect" (or "NC", or something like that). If you use
more than one, all the no connects pins get connected together --
which is exactly what you don't want! Make sure you inspect your
netlist if you use this symbol!!!!
Post by joeft
2) How do I place a pin label in a symbol that will indicate an active
low or inverted signal (i.e. how do I put a "bar" across the signal
name?) I know I can put a dot at the pin to indicate an inversion, but
I would like the pin name to match the manufacturer's documentation.
Lots of schematic capture packages lack this feature. Gschem lacks
it, AFAIK. (I could be wrong, however) The industry standard way to
handle this is to label the pin OUT_N or OUT_L or OUT#. In
particular, the OUT# nomenclature seems to be very widespread in PCI
bus systems, while the OUT_N usage is more generic.

Stuart
John Supplee
2014-08-30 12:49:42 UTC
Permalink
Post by joeft
Two details I'm pretty sure I've seen documented, but can't pin down now
1) Is there an explicit "no connect" symbol that I can/should place in
the schematic to prevent gnetlist from thinking I've forgotten a connection?
2) How do I place a pin label in a symbol that will indicate an active
low or inverted signal (i.e. how do I put a "bar" across the signal
name?) I know I can put a dot at the pin to indicate an inversion, but
I would like the pin name to match the manufacturer's documentation.
Thanks,
Joe
This is a really late reply, but thought I would put it in here anyway for
those who might still find it in a search.

To put a bar over a word you need to enclose it with escaped underscores
like so:

\_ON\_

This will produce ON with a bar over it.

Loading...