```   IN  ----+--\/\/\/\/\--+----\/\/\/\/\---+
|             |                |
|             |     |\         |
|        C    +-----|-\        |
|                   |  >-------+---- OUT
+--------||---+-----|+/
|     |/
\
/
\  R3
/
\
|
-----
---
-

This circuit will shift phase from 180 degrees at DC to 0 degrees
at very high frequencies.  Frequency response is flat, gain is unity.
At the frequency where X(c) = R3, the phase shift is +90 degrees.  The
phase shift is not uniform, but this can be managed as follows.

Wideband 90 degree phase shift networks use several of these cascaded
one after another, in two separate chains, with their 90 degree
frequencies spaced according to some formula which I do not know.  The
spacing allows the nonuniformity of the high frequency side of one
network to compensate for the nonuniformity of the low frequency side
of the next network in the chain.  With several networks in each chain,
this results in a constantly changing phase shift over a wide frequency
range, with constant gain.  It is the DIFFERERNCE between the outputs of
the two chains which has a relative phase shift of 90 degrees.

The 90 degree frequencies of the two chains are staggered.  Sweeping
from DC up, the chain with the lowest "90 degree" frequency has
progressively more shift until it reaches a little more than 90 degrees,
whereupon the other chain is beginning to shift phase, and from then on
until the highest "90 degree" network frequency, the phase difference
between the two chains tracks at 90 degrees relative.  Above that
frequency, the phase difference between the chains gradually falls off.
I seem to recall that there were three or four networks in each chain for
a 100 to 10,000 Hz, 90 degree PSN.

There are a few other ways to accomplish a 90 degree wideband phase
shift.  An older circuit by R. Dome covered 150 to 5000 Hz, if I
recall, with several (5?) of the following cascaded networks in each of
two chains, using vacuum tubes.  Again, the 90 degree phase shift exists
between the outputs of the two chains.

B+
^
|
\
/
\ R1
/
\
/
|
+---------+
|         |
-----      --- C
IN ---------- - -      ---
-----       |---------  OUT
'   |       \
|       /
|       \ R3
|       /
|       \
|       |
+-------+
|
\
/
\
/ R2
\
/
\
|
-----
---
-

And, there is a passive, multiple cross-linked network of numerous
resistors and capacitors, which can produce a decent differential 90
degree shift over a desired range - but I'm not going to try to create
an ASCII rendition of that!  It's in recent issues of the ARRL handbook
(ARRL, Newington, CT), although it turns out that this design is not as
forgiving of component tolerances as the handbooks said.  Evidently,
it wants matched caps and matched resistors, but their exact values are
not ultra-critical.

Finally, I've read about a wideband 90 degree PSN which uses a
multiple-tapped delay line and several shift networks.  Unfortunately,
I have never examined the design.

Bob Bruhns, WA3WDR, bbruhns@li.net

```

Date: 8 Jul 1996 03:08:09 GMT

Original Subject: circuit sugestion need