SSB

Hello,

Just want to build a SSB modulator with MC1496 chip.
The phasing method (second way) is desired.
Well, the phasing method needs a Hilbert transformer to change the
audio signal by 90 degrees. I saw the below circuit in an electronics
book, the author mentioned that the circuit produce a 90 degree phase
shift with an 1% error from 30Hz to 16KHz .
Most of the filters I have seen for this job are using more
complicated circuits
But with much lower band width!
Do you think the below circuit is able to generate the 90 degrees of
phase shift over that band width?

http://img251.imageshack.us/my.php?image=untitledwc7.jpg


<a href="http://img251.imageshack.us/my.php?image=untitledwc7.jpg"
target="_blank"><img src="http://img251.imageshack.us/img251/9660/
untitledwc7.th.jpg" border="0" alt="Free Image Hosting at
www.ImageShack.us" /></a>

Thanks

 

Well, it certainly wouldn't take very long to slap one together and
look and see!

From my limited knowledge of actual design, I'd say that it does
look like phase shifters, but I have no idea how to analyze the circuit
completely.

Maybe you could use one of those simulator programs; then you wouldn't
have to buy any parts or a protoboard.

What's the application? Homework? ;-)

Good Luck!
Rich

 

I recommend you download a spice package, like LT spice. I've entered
the +45 degree phase shift circuit on my simulator and the amplitude
response is not close to flat (15 dB) and the phase response is very
nonlinear. There are some series resistors in the schematic. I'd say
something was lost from the original, but IMHO, this design won't
work.

Frank

 

A doable way to get a 90 degree signal is to use
a PLL. A PLL adjusts the VCO such that the resulting
signal is 90 degrees to the reference. 30Hz to 16kHz
is a bit much for a single VCO stage. It'd be doable
with a difference VCO though.

Rene

 

Methinks the OP was talking of the All-Pass approach to phase shifting
the AUDIO for use in an SSB modulator.

PLL's do 90° phase shift only with analog phase detectors... which
have terrible useful frequency range, before harmonic locking and
other nuisances.

The best way with a PLL is to run the VCO at 4X, use a div4 in the
feedback loop, with edge-matching phase detector, then peel off the
90° with an extra div2 from within the div4 chain.

...Jim Thompson

 

The op needs a linear network that shifts a wide range of signals by
90 degrees, simultaneously, while preserving the amplitudes of each
spectral component.

The schematic he posted is trickier than the standard active allpass,
so I can't say if it would work to his specs. Williams&Taylor's filter
book has tables for the more conventional allpass:

8 opamps give 1.4 degree error over a 573:1 frequency span

10 opamps 0.4 degrees, ditto.

where each opamp has one cap and three resistors.

So his 4 opamp thing may be dicey.

John

 

That won't work.

He needs the phase shift network for audio, which implies a complex
waveform of varying frequency and amplitude. Even if a PLL can track
that, it will strip off any amplitude variation.

Michael

 

I'm sorry I don't have it handy, but there is an old paper that has a
family of low pass filters designed such that the phase difference
between the filters is 90 degrees. This is not as good as having a
real Hilbert transformer, but this is real life. ;-) The paper is
stone age. I got it out of Terman, which is where Jim ought to be
handing out rather than Sunnyvale pubs! [Stanford is nice enough to
let any fool use Terman library. You can't check anything out, but you
can certainly copy papers to your hearts content. You can also use
their on-line database searches from dinky Macs in the library. The
results from the search can be mailed off campus.]

Not quite as optimal as that stone age design, but it seems to me you
could take two bessel filters and shift the corner frequencies to get
90 degrees phase shift. [I haven't tried this, but it seems
reasonable.] There are many off the shelf Bessel designs, even
switched cap. The drawback I see is the amplitude will be off
slightly. The stone age design had decent amplitude matching between
channels. The networks were sloppy lowpasses.

In a datacom circuit, an octave worth of 90 degrees phase shift was
fine. In your case, you are trying to do over 9 octaves. I don't see
how you expect this to be simple. Also, you will probably have trouble
with any circuit that isn't a ladder filter since ladders have the
least sensitivity to component tolerances.

 

This circuit as shown has an error. The branch consisting of a 1n cap
and a 47k and 3.3k in series is shown going from the - input to ground.
This branch should in fact go from the + input to ground.

Then the entire circuit gives a phase shift difference of 90 degrees, +-
2 degrees from 30 Hz to 16.6 kHz. The +45 branch on top is nearly
all-pass, with an amplitude deviation of +.009 dB, -.069 dB. The -45
branch on the bottom has an amplitude deviation of +.014 dB, -.066 dB.

The usual op amp phase shifter has only 1 capacitor per stage. This two
capacitor stage has 2 pole-zero pairs per stage, at the cost of greater
sensitivity, but I think it will be ok in this relatively low performance
circuit.

 

No he is talking audio for an SSB transmitter, so amplitude
can change significally....
The standard diagram I have somewhere from the RSGB handbook for this...
There are a zillion of those audio 90 degrees phase shift ciruits on the web.
Here for example, 100% passive:
http://web.telia.com/~u85920178/tx/p-ssb.htm#af

 

A 2 degree difference gives a 35 mV of a sine wave against 1 V , side
band attenuation.
About 28x, so 20log(28) is about 29 dB or so, not very good for
sideband
suppression.
I'd want to see 50 dB or better

 

SSB modulators do not phase shift the baseband. And the word is
"bandwidth" and not "band width."

 

Yes I am thinking that it is not working with that wide bandwidth.
I have seen the below 90 degrees phase shift circuit which is working
from 20Hz to 20KHz with 1% of error.
it uses 12 op amps to do the job while the first circuit is using just
4 with the same circuitry

http://img187.imageshack.us/my.php?image=image22zr8.gif

 

Just goes to show that the circuit the OP found wasn't designed for voice
communication over a radio link!

30 Hz to 16.6 kHz is a little excessive for that application, I think.

 

The OP should also have a look at:

http://webpages.charter.net/wa1sov/technical/allpass/allpass.html

 

You'll note that it's not the *same* circuitry; it's only similar.

The first schematic you posted uses 2 capacitors per op amp, while this
one only uses 1 per op amp. With 2 capacitors per op amp, this circuit
could be reduced to 6 op amps instead of 12.

This circuit obviously came from:

http://webpages.charter.net/wa1sov/technical/allpass/allpass.html

The author of that page explains how to design these circuits. His example
circuits are designed for a 20Hz to 20 kHz bandwidth. I would think that
for radio communications something like 300-3000 Hz would be appropriate.

Following the procedure from the Bedrosian paper, decide how much unwanted
sideband suppression you want, and that will determine how many op amps you
will need.

 

Actually some do.

'Width of the band used' makes no difference, and in fact the Americans have a
tendency to split words, unlike the Dutch and Germans who seem to want to write
very long words.

And quote what and who you are replying to.

\~~~~~~~~~~~~~~~~~~~~/
\ water for flames /
\ to follow /
\______________/

(I was lying, it no water but petrol ;-) )


______
[ exit ] click here to close window of opportunity.
------

 

If that's the case and his highest frequency component is only 16KHZ,
then I would say a simple digital scheme sampling at 4x16K=64KHz with
simple sorting and inverting every other sample to produce perfectly
quadrature data streams to D/A may even be within range of an 8-pin
PIC-or two, dunno. Almost anything beats that ancient mess of analog
filters.

 

Yes, indeed, well for SSB there is basically 2 methods, the phasing
method (as discussed here), and teh filter method (more frequently
used commercially i think), in the last one a crystal or
some other steep filter simply filters out the unwanted sideband.
Very old hands here may remember XF9A and XF9B crystal filters.....

You are right, it seems digital is the way to go, is PIC fast enough?
FPGA is!
Look what this guy did:
http://www.microtelecom.it/ssbdex/ssbdex-e.htm
I would like him to publish his HDL code

 

Actually I just viewed that earlier. Note that he does not phase shift
the baseband, aka signal labeled "input," in the diagram. A nice
digitalization of the standard Weaver SSB modulator.