Post subject: DPSK Demod - Is there a better way?
postPosted: Fri Aug 27, 2004 11:39 am
Joined: Fri Aug 27, 2004 11:25 am
Can anyone offer me a technique for DPSK
demodulation alternative to the classic phase comparison of the signal
with a delayed version of itself?
The DPSK demodulation is to
be done at IF (~70MHz), and the bit period is 500ns. I have successfully
implemented a solution using a SAW delay (Band Pass device), but several
problems are associated with this, such as the need to overcome the
excessively large insertion loss, the cost and the fact that it is a
highly specialized device.
Ideally a delay of 500ns @ 70MHz is required,
and this is not an easy item to find.
My initial thought was
to use a shorter delay (as with practical devices, Bandwidth is inversely
proportional to delay) meaning that a passive delay line of say 100ns
could be used. The result of this would be the generation of 100ns pulses,
which could then easily be stretched back to the required 500ns using
digital circuitry. On closer inspection of delay lines available, it
seems that no bandwidths of greater than 3.5/Td (=35MHz for 100ns) exist.
In any case, the devices in question are actually meant for digital
applications and not what I envisage.
Thinking along the same
lines, I could halve the delay effectively doubling the bandwidth, but
this will produce narrower pulses, which due to finite rise/fall times
may not trigger the detection circuitry.
Am I missing something
here? The idea of DPSK is to avoid carrier recovery circuitry, but this
could be the only reliable route to go.
Post subject: Delay
Unread postPosted: Tue Aug 31, 2004 10:55 am
Don't know anything about DPSK demodulation, but 20+ years ago there
was a method of FM demodulation that employed a delay line - basically
a coaxial line. This was patented (probably expired) and used on some
Microdyne telemetry receivers.
If you could use a coax delay
line, you certaintly would not have the large insertion loss of a SAW
postPosted: Fri Sep 03, 2004 8:17 am
Thank you for the reply.
If a coaxial delay is obtainable as an off-the-shelf item, it may be
an option. Assuming it is made up of a length of some coax line, the
length of line required for the delay in question is excessive, which
I expect would make the device bulky. I omitted from my original enquiry
that space is also a major issue. (16 – 32m of coax for a delay of 80
– 150ns – Assuming Teflon - dielectric e=2.03. Using a PVC cable - e~8,
this length can be halved, but this is still too much)
I will explore the option.
Unread postPosted: Mon Sep 27, 2004 12:19 pm
an important point: You don't need DC to 70 MHz bandwidth - just sufficient
bandwidth centered at 70 MHz. This suggests an all-pass LC circuit with
the requisite delay at 70 MHz +/- X only. The program S/FILSYN from
ALK Engineering can do this kind of design.
Unread postPosted: Thu
Oct 07, 2004 10:34 am
A belated thanks for the response.
Your idea is an interesting one - My initial research had been to see
if a lumped LC solution was possible and when I found that theoretically
a 30th order circuit only gave a GD of 43 ns (Butterworth assumed where
in practice a Bessel would be better for its linear phase relation),
this approach was immediately discarded. My assumptin had been that
the GD of a LP with cutoff @ 70MHz would be similar to that of a BP
at this frequency with the same no. of poles (1/2 the order of the LP
- ie the same no of reactive components)
I have subsequently
found that this is not the case and larger GD is available from BP circuits.
It was then found that a 15th order BP @ 70MHz (30 poles as per the
LP cct mentioned earlier) gave a theoretical GD in the order of 300ns.
Better, but not enough. The other requirement not previously mentioned
is +/-10ns delay accuracy and stability
So in conclusion, the
BP circuit, although offering greater GD as compared to an equivalent
LP, is still inadequate. To manufacture a 7th order BP repeatably and
reliably is already a tough ask, so tens of orders is definitely unrealistic.
Fri Oct 08, 2004 8:32 pm
Well, let's try again. :)
devices can have significant delays - have you checked with people like
RF Monolithics or Sawtek?