BPSK modulation of fast pulses

[oopere]

I have set up a SRD-based pulse generator that outputs pulses between
0..5-10 V with leading edges of ~100ps to a 50R load. Pulse repetition
frequency is 20 MHz. Input pulses are 5ns wide with 2ns edges. The
output pulse is being radiated by an UWB antenna (cone monopole, horn, ...).

Now, I am trying to find a way to BPSK modulate this pulses (simple
inversion), preserving the fast leading edges. Up to now, I have
considered two alternatives:

1. Use a mini-circuits diode-quad + transformer based mixer, where the
LO signal switches the diodes.

2. Build a (discrete) pin-diode quad, feeding the adequate bias through
resistors.

I would appreciate your thoughts on these or other alternatives!

Since mixers are usually not specified in the time domain, I would
expect that the mixer approach would lead to severe time-domain
degradation. But perhaps it is ok if the mixer works up to 10 GHz? Also,
power specification is somewhat unclear with this kind of signals (?).

Otoh, switching narrowband (i.e. around a carrier) RF signals with
pin-diodes is well-known. How about the time-domain degradation that
should be expected? The edges will certainly be smoothed, but how much?
Any particular experience with specific diodes?

Pere

 

[Joerg]

I have set up a SRD-based pulse generator that outputs pulses between
0..5-10 V with leading edges of ~100ps to a 50R load. Pulse repetition
frequency is 20 MHz. Input pulses are 5ns wide with 2ns edges. The
output pulse is being radiated by an UWB antenna (cone monopole, horn,
...).

Now, I am trying to find a way to BPSK modulate this pulses (simple
inversion), preserving the fast leading edges. Up to now, I have
considered two alternatives:

1. Use a mini-circuits diode-quad + transformer based mixer, where the
LO signal switches the diodes.

2. Build a (discrete) pin-diode quad, feeding the adequate bias through
resistors.

I would appreciate your thoughts on these or other alternatives!

Since mixers are usually not specified in the time domain, I would
expect that the mixer approach would lead to severe time-domain
degradation. But perhaps it is ok if the mixer works up to 10 GHz? Also,
power specification is somewhat unclear with this kind of signals (?).

Otoh, switching narrowband (i.e. around a carrier) RF signals with
pin-diodes is well-known. How about the time-domain degradation that
should be expected? The edges will certainly be smoothed, but how much?
Any particular experience with specific diodes?

Why would edges be smoothed? PIN diodes when properly used act as almost
ideal RF switches. There are also integrated RF muxes for 2.45GHz apps,
5GHz, and such, from the major brands. However, so far I have preferred
the multi-source nature of PIN diodes. I don't like to lock parts of a
design into one particular manufacturer.

 

[Joerg]

Yike! At 10 volts? The diodes would probably break down.

How about these?

http://www.nxp.com/acrobat_download/datasheets/BAP1321-02_N_2.pdf

[...]

 

[Joerg]

I have set up a SRD-based pulse generator that outputs pulses between
0..5-10 V with leading edges of ~100ps to a 50R load. Pulse repetition
frequency is 20 MHz. Input pulses are 5ns wide with 2ns edges. The
output pulse is being radiated by an UWB antenna (cone monopole, horn, ...).

Now, I am trying to find a way to BPSK modulate this pulses (simple
inversion), preserving the fast leading edges. Up to now, I have
considered two alternatives:

1. Use a mini-circuits diode-quad + transformer based mixer, where the
LO signal switches the diodes.
Yike! At 10 volts? The diodes would probably break down.

How about these?

http://www.nxp.com/acrobat_download/datasheets/BAP1321-02_N_2.pdf

[...]

Nice part, but the carrier lifetime is too long for a 20 Mhz data
rate.

Yeah, true, even for these power levels it's a bit highish. The BAP63
would be around 300nsec, the BAP65 is under 200nsec, same price range:

http://www.nxp.com/acrobat_download/datasheets/BAP65-02_N_4.pdf

 

[[email protected]]

If you really need no degradation at all of the edge speed, maybe build two
of these things, and put the antenna of one of them upside down.  You can
then put the positive pulses through the first transmitter and the negative
pulses through the other... Actually you could put both antennas inside a
waveguide or similar, protruding through opposite faces, that way the
pulses could be made to appear to radiate from the same location regardless
of polarity.

Chris- Hide quoted text -

- Show quoted text -

this entire discussion about BPSK of impulses has me scratching my
head...

it seems for the case of impulses, it is not realy PHASE modulation
but rather POLARITY modulation...

and even with conventional carrier based signals using doubled
balanced mixes as modulators, that is also really polarity modulation
and not really 180 deg phase modulation..

Purists are always correcting people when they say 240 Volts 2 phase
AC power when they really mean 2 polarity.... Strictly speaking 180
deg phase shift is NOT the same thing as polarity inversion but it is
close enough in many cases.

So for the special case of BPSK created with a DB mixer,,....I'm
starting to think that the P should actually stand for POLARITY and
not PHASE. For QPSK etc, the P is obviously phase.

And I guess if you made a BPSK modulator with a delay or complex IQ
multiplier instead of a DB mixer, you could not tell the difference
between the two signals at the receiver anyway...

Its probably 6 of one and 1/2 a dozen of the other but interesting to
think about over a beer.

Mark

 

[oopere]

If you really need no degradation at all of the edge speed, maybe build two
of these things, and put the antenna of one of them upside down. You can
then put the positive pulses through the first transmitter and the negative
pulses through the other... Actually you could put both antennas inside a
waveguide or similar, protruding through opposite faces, that way the
pulses could be made to appear to radiate from the same location regardless
of polarity.

Chris

The idea of using two different antennas may be worth a try for testing
purposes. It should work nicely in the broadside direction of the
resulting "array". However, I did not get your idea on how I could make
the antennas seem to radiate from the same direction.

Pere

 

[oopere]

this entire discussion about BPSK of impulses has me scratching my
head...

it seems for the case of impulses, it is not realy PHASE modulation
but rather POLARITY modulation...

and even with conventional carrier based signals using doubled
balanced mixes as modulators, that is also really polarity modulation
and not really 180 deg phase modulation..

Purists are always correcting people when they say 240 Volts 2 phase
AC power when they really mean 2 polarity.... Strictly speaking 180
deg phase shift is NOT the same thing as polarity inversion but it is
close enough in many cases.

So for the special case of BPSK created with a DB mixer,,....I'm
starting to think that the P should actually stand for POLARITY and
not PHASE. For QPSK etc, the P is obviously phase.

And I guess if you made a BPSK modulator with a delay or complex IQ
multiplier instead of a DB mixer, you could not tell the difference
between the two signals at the receiver anyway...

Its probably 6 of one and 1/2 a dozen of the other but interesting to
think about over a beer.

Mark

You are right, BPSK is just a synonym form "sign inversion keying",
sometimes also called PRK (phase reversal keying)!

Pere

 

[oopere]

Why would edges be smoothed? PIN diodes when properly used act as almost
ideal RF switches. There are also integrated RF muxes for 2.45GHz apps,
5GHz, and such, from the major brands. However, so far I have preferred
the multi-source nature of PIN diodes. I don't like to lock parts of a
design into one particular manufacturer.

Thanks for your input! For the pulses not to be degraded I need (almost)
constant gain _and_ (almost) linear phase up to 10GHz. From what I have
read so far, designs based on switches with PIN diodes are usually not
characterized taking phase into account. Multiplexers behaving well up
to 10 GHz are also not so common...

Pere

 

[oopere]

Thanks for your input! For the pulses not to be degraded I need (almost)
constant gain _and_ (almost) linear phase up to 10GHz. From what I have
read so far, designs based on switches with PIN diodes are usually not
characterized taking phase into account. Multiplexers behaving well up
to 10 GHz are also not so common...

Pere

Also, multiplexers may not like the pulse amplitude!

Pere

 

[oopere]

Yike! At 10 volts? The diodes would probably break down.


You could possibly build a coaxial inverter (I posted one recently)
and switch the junctions with some gaasfets or pins.

There's probably some balun-like thing that would invert or not invert
depending on whether a pin is grounded. The grounding device could be
a beefy phemt maybe. Hmmm, that's almost equivalent to the previous
suggestion.

Phemts do seem to behave roughly like jfets, as far as analog switch
biasing goes.

Interesting, but not a simple problem.





coax in
=======================--------x---------+
|
---------------------------+---x---------+---->out
|
======================= | +---->
| | |
| | |
x x gnd
| |
| |
| |
gnd gnd


where the x's are switches of some sort and the coax is longer than
the pulse width. Add a ferrite maybe.


John

Thanks for the info, John. I also have the feeling that some king of
balun arrangement would be able to provide the inversion. But, to be
honest, I have not been able to understand them in a rigorous way, i.e.
to find a model that fully explains the role and differences of all
those "gnd" symbols.

Pere

 

[oopere]

John Larkin wrote:

I have set up a SRD-based pulse generator that outputs pulses
between 0..5-10 V with leading edges of ~100ps to a 50R load. Pulse
repetition frequency is 20 MHz. Input pulses are 5ns wide with 2ns
edges. The output pulse is being radiated by an UWB antenna (cone
monopole, horn, ...).

Now, I am trying to find a way to BPSK modulate this pulses (simple
inversion), preserving the fast leading edges. Up to now, I have
considered two alternatives:

1. Use a mini-circuits diode-quad + transformer based mixer, where
the LO signal switches the diodes.
Yike! At 10 volts? The diodes would probably break down.

How about these?

http://www.nxp.com/acrobat_download/datasheets/BAP1321-02_N_2.pdf

[...]

Nice part, but the carrier lifetime is too long for a 20 Mhz data
rate.

Yeah, true, even for these power levels it's a bit highish. The BAP63
would be around 300nsec, the BAP65 is under 200nsec, same price range:

http://www.nxp.com/acrobat_download/datasheets/BAP65-02_N_4.pdf

Thanks, Joerg. But all these have very limited isolation even at 3 GHz.
Perhaps I am overlooking something but I think they would not work!

Pere

 

[Joerg]

On Fri, 13 Feb 2009 11:11:22 -0800, Joerg

John Larkin wrote:

I have set up a SRD-based pulse generator that outputs pulses
between 0..5-10 V with leading edges of ~100ps to a 50R load.
Pulse repetition frequency is 20 MHz. Input pulses are 5ns wide
with 2ns edges. The output pulse is being radiated by an UWB
antenna (cone monopole, horn, ...).

Now, I am trying to find a way to BPSK modulate this pulses
(simple inversion), preserving the fast leading edges. Up to now,
I have considered two alternatives:

1. Use a mini-circuits diode-quad + transformer based mixer, where
the LO signal switches the diodes.
Yike! At 10 volts? The diodes would probably break down.

How about these?

http://www.nxp.com/acrobat_download/datasheets/BAP1321-02_N_2.pdf

[...]

Nice part, but the carrier lifetime is too long for a 20 Mhz data
rate.

Yeah, true, even for these power levels it's a bit highish. The BAP63
would be around 300nsec, the BAP65 is under 200nsec, same price range:

http://www.nxp.com/acrobat_download/datasheets/BAP65-02_N_4.pdf

Thanks, Joerg. But all these have very limited isolation even at 3 GHz.
Perhaps I am overlooking something but I think they would not work!

Well, you can't get around the capacitances. Only politicians can try to
make those "go away" with some fancy creative budgeting

You'll need T-structures and such to yield good isolation. OTOH
PIN-diodes are very cheap, so that would not be a big deal.

 

[Joerg]

Probably best to throw it onto SPICE and take a look, faster than
crunching numbers on the old HP calculator. I've done lots of diode
switches where phase was very critical but it was baseband stuff, under
50MHz.

Also, multiplexers may not like the pulse amplitude!

Yep, 10V is a stretch. You could visit the web sites of the better RF
vendors such as Sirenza and Hittite but my feeling is that you'll end up
with diodes here. Or maybe a 2nd SRD generator.

 

[JosephKK]

Also, multiplexers may not like the pulse amplitude!

Pere

I do not know of anything that will properly preserve those fast
edges. I suspect using two SRD pulse formers connected in opposite
polarity through a broadband combiner (oops may be tractable). Hmmm
how about feeding the SRD pulse former into the center tap of a
wideband transformer, with a pair of oppositely driven gaasfets
connecting one end or the other to ground, then the secondary has
polarity inversions. Well maybe, those fast edges make things tough.


gnd---------fet_sw------------
)
) -------
) (
) (
--------[pulse gen]---[SRD]----------------- (
) (
) (
) -------
)
gnd--------fet_sw--------------