rx frontend & filter
Jan Wagner
i'd like to build a frequency controlled filter into an rf preamp.
It's for a direct conversion receiver 0-50MHz, and the LO frequency should
control the input rf bandpass filter center frequency. The filter
pass band bandwidth is fixed.
Are switched capacitor filters any good for this? Or a Tayloe mixer style
setup working as a bandpass filter? Maybe someone has already tried one of
these? (google showed some postings on the topic but not much else than
speculations)
Or are there some "really simple" frequency-controlled RF bandpass filter
alternatives which I just missed?
Thanks!
- jfw
--
*************************************************
high voltage at http://www.hut.fi/~jwagner/tesla
Jan OH2GHR
Richard Hosking
good filter at carrier frequency. You can mix the signal back up to the
carrier or an IF frequency with a second mixer in reverse. I am not sure how
it goes at harmonics - presumably it will have a spurious response as the LO
drive is 4 digital signals. I am designing just such a receiver at present
Rod Green (VK6KRG) has tried this concept on 7 MHz with a Tayloe mixer
frontend and an IF at 455 KHz. You get SSB rejection of about 40 dB without
any special balancing of the 4 signal paths. The beauty of this setup is
that there are no phase shift networks in the audio path. He claims MDS for
20dB SINAD was -110dBm and mixer 1 dB compression was about 0 dBm. (FST3125
mixer with no RF amp) It has good adjacent signal performance due to the
inherent attenuation of off channel signals - in short the ideal reciever
topology!
Has anyone else got experience with this setup?
Richard
Jan Wagner <jwa...@lyta.hut.fi> wrote in message
news:Pine.OSF.4.50.030317...@lyta.hut.fi...
Jan Wagner
On Mon, 17 Mar 2003, Richard Hosking wrote:
> We have done some work on the Tayloe mixer - it appears to act as quite a
> good filter at carrier frequency.
Thanks!
It's a good filter on the audio freq side, but I'd need the quadrature
outputs. And input RF amplification, to use my existing dual AD831 I/Q
mixer board (dead-weak reception with no preamp).
My "idea" is:
RF ___
/ \
------/ \-------------------
^
| filter f_c == LO I or Q
| bandwidth = 200 kHz, maybe
|
+-> rf preamp --> 2 x AD831 mixer --> I & Q AF
^
|
LO I and Q
About the Tayloe mixer:
> You can mix the signal back up to the
> carrier or an IF frequency with a second mixer in reverse. I am not sure how
> it goes at harmonics - presumably it will have a spurious response as the LO
> drive is 4 digital signals.
Probably lot's of digital clock feedthrough - switch gate capacitances -
into the output. But these seem to be below -30dB
(http://www.hut.fi/~jwagner/dc-rx/N-path-SC_bandpass.pdf)
Maybe it works. OTOH, odd, two mixers for down and then up conversion, and
_then_ a RF preamp, and after that, quadrature mixer.
Highly unconventional... ;-) Well, if I can't invent anything ingenious
of my own, that'll have to do. Though I've some doubts about the
performance of such a setup.
But the Tayloe mixer (or N-path filter) gave an idea - modify it into a
band reject filter, so all out-of receiving band signals are terminated
on the output side of the mixer. And amplify the remaining in-band signal
from the input side of the mixer. I'll give that one a try.
> I am designing just such a receiver at present
> Rod Green (VK6KRG) has tried this concept on 7 MHz with a Tayloe mixer
> frontend and an IF at 455 KHz. You get SSB rejection of about 40 dB without
> any special balancing of the 4 signal paths. The beauty of this setup is
> that there are no phase shift networks in the audio path. He claims MDS for
> 20dB SINAD was -110dBm and mixer 1 dB compression was about 0 dBm. (FST3125
> mixer with no RF amp) It has good adjacent signal performance due to the
> inherent attenuation of off channel signals - in short the ideal reciever
> topology!
Sounds very good. 8-)
It didn't use the 455kHz IF for a second upconversion, though? ;)
Ok - any thoughts on how to use the AD9854 for this? Are four ultrafast
analog comparators sufficient? (square up the I, Q, inverted I and Q?)
If yes - have you used fast comps in your setup? NE521, other?
many thanks!!
- Jan
Richard Hosking
In answer to some your points
Jan Wagner <jwa...@kosh.hut.fi> wrote in message
news:Pine.OSF.4.50.030318...@kosh.hut.fi...
Yes there is clock feedthrough
In the first mixer this is irrelevant is the output is audio.
The second is at 455 KHz output (4 paths in - 1 out) There will be clock
feedthrough here but the signal is product detected in the third stage back
down to audio using the same clock and the feed through ends up as DC.
The signal is translated from signal to audio to IF and you can filter it
there. There is an SSB image down about 40 dB on the prototype The image is
the inverse of the wanted signal ie - USB -> LSB and vice versa
In practice the receiver seems sensirtive and birdies dont seem to be a
problem.
> Maybe it works. OTOH, odd, two mixers for down and then up conversion, and
> _then_ a RF preamp, and after that, quadrature mixer.
> Highly unconventional... ;-) Well, if I can't invent anything ingenious
> of my own, that'll have to do. Though I've some doubts about the
> performance of such a setup.
>
>
> But the Tayloe mixer (or N-path filter) gave an idea - modify it into a
> band reject filter, so all out-of receiving band signals are terminated
> on the output side of the mixer. And amplify the remaining in-band signal
> from the input side of the mixer. I'll give that one a try.
>
This should work except that the feedthrough will give you a carrier at LO
freq. The impedance of the mixer seems to fall off rapidly either side of
the LO freq. If you used the same LO for a direct conversion scheme I guess
this wouldnt matter
>
> > I am designing just such a receiver at present
> > Rod Green (VK6KRG) has tried this concept on 7 MHz with a Tayloe mixer
> > frontend and an IF at 455 KHz. You get SSB rejection of about 40 dB
without
> > any special balancing of the 4 signal paths. The beauty of this setup is
> > that there are no phase shift networks in the audio path. He claims MDS
for
> > 20dB SINAD was -110dBm and mixer 1 dB compression was about 0 dBm.
(FST3125
> > mixer with no RF amp) It has good adjacent signal performance due to the
> > inherent attenuation of off channel signals - in short the ideal
reciever
> > topology!
>
> Sounds very good. 8-)
>
I had the theory *slightly* wrong here. The circuit effectively translates
the wanted signal spectrum to IF via audio.
You then product detect the audio after appropriate filtering etc. The
beauty is you can use an IF filter without IF image issues.
> It didn't use the 455kHz IF for a second upconversion, though? ;)
>
> Ok - any thoughts on how to use the AD9854 for this? Are four ultrafast
> analog comparators sufficient? (square up the I, Q, inverted I and Q?)
> If yes - have you used fast comps in your setup? NE521, other?
>
The AD9854 does not seem to have any immediate application unless you want
to use a single output at 4X LO and divide it down in a Johnson Ring counter
to get 4 phases
Richard
Jan Wagner
and thanks for the clarifications! I'll be off to reading & building.
Just a quick last comment:
> > Ok - any thoughts on how to use the AD9854 for this? Are four ultrafast
> > analog comparators sufficient? (square up the I, Q, inverted I and Q?)
> > If yes - have you used fast comps in your setup? NE521, other?
> >
> The AD9854 does not seem to have any immediate application unless you want
> to use a single output at 4X LO and divide it down in a Johnson Ring counter
> to get 4 phases
Actually, if you draw the DDS I & Q output signals on paper, then you'll
notice it counts "10 11 01 00 ..." - a 2-bit Gray code counter.
So there's no need for 4 x LO or 2 x XOR pulse doublers, and no extra
counter chip required either. Yes, the multiplexer outputs won't be
conveniently ordered in a linear 0-90-180-240deg fashion, but just
swap two wires on the output and it's all sorted out. :)
cheers,
Richard Hosking
50% - you will have to square up the two outputs and sum them in a quad NAND
gate . But sure - good idea.
Richard
Jan Wagner <jwa...@kosh.hut.fi> wrote in message
news:Pine.OSF.4.50.030320...@kosh.hut.fi...
Jan Wagner
On Thu, 20 Mar 2003, Richard Hosking wrote:
>> Actually, if you draw the DDS I & Q output signals on paper, then
>> you'll notice it counts "10 11 01 00 ..." - a 2-bit Gray code counter.
>
> 50% - you will have to square up the two outputs and sum them in a quad NAND
> gate . But sure - good idea.
It's well possible I'm missing a major point here, but...
I: 1 1 0 0 (sin)
Q: 1 0 0 1 (cos)
This full 11, 10, 00, 01 count sequence is gone through in a _single_ LO
cycle => multiplexer lines are changed at 4 x DDS LO. Like in the Johnson
counter setup clocked at 4 x target freq.
No repetitions inside the "11,10,00,01" sequence => each of the four mux
outputs will have a duty cycle of 25%, and their frequency is equal to
the LO/DDS frequency. That's what is required for the Tayloe detector. I'd
think. Probably. ;o)
The AD9854 has one fast comparator on board, so squaring up the other of
the I or Q channels requires just one external fast comparator like a
900MHz SPT9689. And then the square wave outputs go directly to the
FST3253 mux.
Flames/comments? ;-)
wastebin
Wagner <jwa...@kosh.hut.fi> writes
>This full 11, 10, 00, 01 count sequence is gone through in a _single_
>LO cycle => multiplexer lines are changed at 4 x DDS LO. Like in the
>Johnson counter setup clocked at 4 x target freq.
>
>No repetitions inside the "11,10,00,01" sequence => each of the four
>mux outputs will have a duty cycle of 25%, and their frequency is equal
>to the LO/DDS frequency. That's what is required for the Tayloe
>detector. I'd think. Probably. ;o)
>
>The AD9854 has one fast comparator on board, so squaring up the other
>of the I or Q channels requires just one external fast comparator like
>a 900MHz SPT9689. And then the square wave outputs go directly to the
>FST3253 mux.
>
>Flames/comments? ;-)
Well I think you're right Jan. It's the demux internal to the FST3253
that gives the 25% duty cycle.
As far as I can see the Tayloe mixer is simply a neat implementation of
two SPDT mixers switched in quadrature.
Can anyone see any reason why the Tayloe mixer would be better than two
simple SPDT mixers? It seems to me that as the filter capacitors see the
signal for twice as long in the classic mixer and would give similar, if
not better, results.
I've built both and I can't hear any difference.
--
Martin G0PFT
Dan Tayloe
each, plus an additional signal splitter loss in front of that, for
perhaps 7 to 12 db of loss overall. The Tayloe mixer does it all in one
shot and has only 1 db of total conversion loss, plus off signal
rejection capabilities to boot.
- Dan, N7VE
Dan Tayloe
talking about here. However, the Tayloe detector seems to be more
sensitive to LO noise than a balanced diode mixer, leading to poor
performance above 14 MHz.
Using a good clean LO, noise is not a problem. VXOs, Phased locked VCOs
or VFOs are a much better LO for this detector.
- Dan, N7VE
Richard Hosking
data selector. If you use the 3253 then the chip itself will do the
selection - I guess this seems the logical choice.
The "Gray Code" issue seems quite important - it can make 20-30 dB
difference in noise according to our experiments.
Richard
Jan Wagner <jwa...@kosh.hut.fi> wrote in message
news:Pine.OSF.4.50.030320...@kosh.hut.fi...
>