The best audio phase shift network

(1/1)

Paul:
Hi all,
All the best for the end of season festivities.
Which is the best form of audio phase shift, for best sideband rejection and overall performance- The  RC Polyphase or the IC based all pass filter ?

Dave Haupt:
I vote the IC all-pass.

Here's why:
I found several examples of the polyphase network on the web and entered the values in LTSpice and ran them.  Results:

Phase and amplitude balance are pretty darned delicate...if you use 5% caps and 1% resistors (very easy to find), you cannot guarantee more than about 25dB undesired sideband rejection.  That's quite poor.  You can, of course, use a capacitance meter to match the capacitors and get close to 40dB.

Output Z is relatively high, so you need to follow it, in many designs, with buffer stages.  This gives you four more stages of variability in gain and phase to deal with.

The IC all-pass has relatively simple design equations and better yet, a free designer.  Jim Tonne, W4ENE wrote this marvelous free windows app called "QuadNet", available here:

http://tonnesoftware.com/

You tell it how many IC stages you want to cascade, the bandwidths over which you want it to operate and it spits out the right schematic for best possible performance.  It then allows you to define component tolerances and runs a Monte Carlo analysis.

Pull the values into LTSpice and run it and you can tweak values and do your own variations.

After back-and-forth between Spice and QuadNet, I settled on a network that gives me 35dB guaranteed minimum opposite sideband rejection over 300-3000Hz, using 5% caps and 1% resistors (two cent parts in 0603 size) and four op amp stages.  My modulator requires all four phases, so I took the 0 and 90 degree outputs and fed them through two more op amps configured as inverters...to make sure that did not add another batch of uncertainty, I used higher-cost 0.1% resistors for those stages.  This is a fully manufacturable network with no hand-trimming.

As an experiment, I also designed a ten op-amp network using QuadNet, to cover 100Hz to 15kHz, simmed it and then built it.  As it was a one-off experiment, I "made" the capacitors, starting with one slightly undersized, then paralleling a much smaller one, to land within 1% of the desired value.  I built two of these, one in a modulator and one in a demod.  The audio quality that resulted was better than one would ever think SSB was capable of.  Music was really very good, if you managed to get the TX and RX oscillators all phase synchronous...which is possible if you're using PTS synthesizers synched to the same clock.

So, if your goals are "easy to make work right" I'd say the IC op amp design.

Now, there is a movement within ham radio towards "Minimalist art", and they frown on using ICs.  In that case, the polyphase network is easier to implement...and that movement within ham radio is happier with hand-matched parts than manufacturability, as the focus is on home-made artwork.  BTW, I'm not being critical - I admire that movement and its primary advocates.


Another anecdotal piece - there have been, over the years, a few tube-type phasing rigs, that used surprisingly simple phasing networks.  Also a currently-published six meter SSB transceiver design on a JA ham's website with a surprisingly simple network.

I have plugged any "surprisingly simple" network I could find into SPICE and discovered they share a lot of similiarities.  First, they've got good suppression at only a very narrow set of audio frequencies.  Over my usual range of 300-3000Hz, few of them beat 10dB opposite sideband rejection.  The few that do-depend seriously on odd value parts that are within 0.5% of the design values.

If you want to read up on a bunch of phasing method rigs that have proven to work well and repeatedly, find any of the last 10 years' writings of Rick Campbell KK7B.  A very practical EE instructor at Portland State, Dr. Campbell really knows how to distill a topic into its most meaningful necessities.  He's also one of the co-authors of "Experimental Methods in RF Design" and plenty of his phasing method designs are found in that book, a recommended read.

Sorry to prattle on so long, but it's a topic I've been living with and in for a little while....

73,

Dave W8NF

Clark McDonald:
These days, I have been experimenting with DSP...

Navigation

[0] Message Index