Tayloe detector as narrowband tunable filter
Richard Hosking
I am working on mixers with Rod VK6KRG using the Tayloe detector which is a
"commutating" mixer using multiplexers.
A proposed circuit has one input and 4 outputs from a Fairchild FST3253 1 to
4 analog multiplexer. The input signal is multiplexed to the 4 outputs by a
1:4 duty cycle clock which switches each in turn. Each ouput is filtered by
a large capacitor. In the Tayloe system these 4 signals are fed to an audio
phase shift network for SSB detection.
In this system the 4 outputs (which are at phase 0, 90, 180 and 270 degrees
at 0 Hz) are fed to a second multiplexer set up in the opposite direction to
the first. The clock for this is at the same frequency as the first and the
original signal is restored. Of course, the clock for the second set of
mixers could be on a another frequency - in this case the result is a
frequency conversion with 0 Hz as an intermediate frequency. Both Rod and I
have breadboarded this circuit and some interesting results emerge.
(1) Frequencies more than a few KHz from the clock carrier are quite
significantly attenuated, with maximal attenuation being 20-30 dB.
Presumably the input impedance of the Tayloe detector is maximal around the
clock frequency and falls off rapidly on either side.
The filter at audio appears to determine the RF response
(2) This means that the detector exhibits good strong signal handling as off
channel signals are attenuated
(3) The original signal to clock relationship is preserved - tuning the
input siganl acrss the clock at the input produces the same at the output.
(4) Of course there are some problems - there is carrier feed through and an
opposite sideband image.These could presumably be limited by mixer balance.
In our prototypes noise was a problem which limited sensitivity
The circuit can be used as a narrowband tunable filter over a wide range.
Some questions:
(1) Has this been done before?
(2) Has anyone modelled or measured the input impedance of the Tayloe
detector?
(3) What would be suitable circuits/solutions to overcome the problems of
images, carrier feedthrough and noise?
Thanks
Richard
JLB
Filter.
Maxim, among others, sells the things in 8 pin dip packages. Look in
the Digikey catalog.
Do a web search on Switched Capacitor Analog Filter. You're sure to
find a bunch of info on it.
Jim
N8EE
"Richard Hosking" <zapri...@iinet.net.au(remove the zap)> wrote in message news:<3c750a48$0$13...@echo-01.iinet.net.au>...
Jim Pennell
Well, in the typical application, it is working against the antenna input
impedance of 50 ohms as part of the effective R/C lowpass effect. The
actual impedance of the switching device is something like 2 or 3 ohms so
the antenna reactance is a major part of the lowpassing.
Which means that the lowpass frequency rolloff which occurs in the
capacitors to ground after the switching device will vary depending on the
antenna impedance. It can be a major change. For example, if one goes
from a resonant antenna to a random length wire.
If using an RF preamp then depending on the design of the preamp, it may
provide a fixed impedance into the detector but even so the lowpass is a
function of the impedance of the RF amplifier and not the switching device.
Some of the articles I've read mention 200 ohms, but that is the result
of the 50 ohm resonant antenna being connected to a given capacitor 1/4 of
the time, hence 200 ohms effective impedance.
So if you have an off-resonance antenna, say for a multiband Rx of some
sort, the antenna may be, let's say 200 ohms, then that is multipled by that
1/4 multiplexing on interval and the tayloe detector is now forming a
lowpass with effectively 800 ohms resistance working against the shunt
capacitor to ground.
For that sort of general use, you have to make the capacitor small enough
so that the lowpass rolloff won't bother the desired audio band even under
the worst case antenna you are willing to allow.
Which means that when you happen to tune the detector to some frequency
where the antenna is resonant, the lowpass won't roll off very much.....
This means the detector works best in a narrow band, resonant antenna
application, where you can set the integrating capacitors to cause lowpass
rolloff just above the desired audio bandwidth.
Jim Pennell
N6BIU
Michael Black
> Sounds to me like you just invented the Switched Capacitor Analog
> Filter.
>
> Maxim, among others, sells the things in 8 pin dip packages. Look in
> the Digikey catalog.
>
> Do a web search on Switched Capacitor Analog Filter. You're sure to
> find a bunch of info on it.
>
> Jim
> N8EE
>
Mixer" here (and I found a schematic) was a certain resemblance to
the sort of circuitry used in Coherent CW. I probably noticed it
because looking at CCW schematics, there was something nagging me
at the back of my mind. There was something there that I thought could
be extended....
I've always been curious about how Dan Tayloe came to the circuit.
What prompted him to give it a try? I've not seen anything about
this (though perhaps I missed it). Because seeing commutating filters
(the first issue of QST that I ever saw, April 1971, had such a filter)
and seeing those CCW schematics made me think, though I never tried
anything and there was only supposition (and vague enough that I can
no longer remember exactly what).
In CCW, you have two data paths. In each, there is a mixer (or was
it a sample and hold?) being clocked at the input frequency, but
the clock to each channel is in quadrature. So you get a DC signal.
You then have two more mixers (or again was it a sample and hold?),
clocked at audio frequencies. Again, the clocking is in quadrature.
It's been a while since I've looked at descriptions (and sadly, none
of my books have CCW in more than passing), but the first conversion sure
looked like what you'd see in a phasing receiver. The second conversion,
back to audio, seemed to do the job of the audio phase network, The
result was singal signal, but a narrow bandwidth.
The best article in the ham magazines about commutating filters was in Ham
Radio magazine, circa 1978, I think by Hank Olson. Unless I'm garbling
it, the cover features Wes Stewart's split-band voice processor. I
had always meant to dig up some of these articles and post a bibliography,
because of the similarity between commutating filters, and CCW, and the
Tayloe Mixer.
Michael VE2BVW
Richard Hosking
The filter effect works at RF
For example, on a breadboard test with a clock of 3.5 MHz, the ouput peaked
with an input frequency of 3.5 MHz (ie at the clock frequency) and with some
setups (eg having an input resistor to increase the R in the RC network) a
very sharp peak could be obtained (6 dB down at inputs of 3.5 MHz +/- 20 KHz
or so)
Moreover the filter can be tuned to any frequency simply by tuning the first
mixer clock. Further filtering can be performed by making the second
multiplexer clock at some IF (say 455 KHz) and the signal is then
upconverted from zero to 455 KHz without image problems. (apart from
"inband" images due to quadrature/SSB inbalance)
Richard
"JLB" <jim.bo...@lintek.aeroflex.com> wrote in message
news:8d5ec8a2.02022...@posting.google.com...
Richard Hosking
the circuit
We have used a preamp with a predictable input impedance - eg a grounded
gate FET circuit whose input impedance is proprtional to Gm*drain current.
I am not sure what this looks like as a source to the mixer however.
Presumably the best circuit would be a low impedance source for the mixer
with the filter R in the RC set by an input resistor to the mixer.
Richard
"Jim Pennell" <NoS...@killspam.com> wrote in message
news:a54bd1$m3l$1...@slb7.atl.mindspring.net...
Clifton T. Sharp Jr.
> I've always been curious about how Dan Tayloe came to the circuit.
To save Dan the trouble of retyping it:
Message-ID: <3B2C28F6...@home.com>
From: Dan Tayloe <dta...@home.com>
Newsgroups: rec.radio.amateur.homebrew
Subject: Re: Tayloe Mixer
Date: Sun, 17 Jun 2001 03:50:15 GMT
"Clifton T. Sharp Jr." wrote:
>
> Dan Tayloe wrote:
> > I saw Phil Rice's article for the first time about a year ago. It is
> > not the same, but it is similar. However, as you have pointed out, his
> > circuit was not useful for either weak signal nor high dynamic range
> > receiver work, where as mine has been optimized for both.
>
> The question I've most wanted to ask you since I first saw this nifty
> circuit: what got you started in this direction? What were you looking
> for or experimenting with that eventually led you to do this design?
Actually what got me started in this direction was the availability of
the TI 74CBT3253 high speed analog mux and a 3000 mile (Phoenix to New
York) cross country drive. I had lots of time to do nothing but think
and ponder. The 74CBT3253 is a zero delay analog mux (as opposed to a
high delay digital gate) used to connect microprocessor address and data
lines to different banks of memory. It was never intended for analog
use, but it works quite well for that task.
I knew about audio switched capacitor filters using an input resistor
and four capacitors which were switched at a 4x rate. The input side to
the mux ends up exhibiting sharp bandpass behavior, based upon the R/C
time constant of the single R and multiple Cs used.
The idea of an active RF bandpass filter on a receiver front end is what
got me started. Since we now had a *really* fast dual 4:1 mux part,
perhaps I could do some active RF bandpass filtering using the switched
capacitor filter in front of a receiver mixer rather than a simple fixed
tune bandpass L/C.
The really neat thing about this is that if I combined the VFO and IF
BFO signals together at a 4x rate, this narrow active RF bandpass filter
would track the receiver tuning. This should enhance receiver
performance on congested bands such as 20m (contests) or on 40m with big
adjacent SWBC signals.
I pondered this and the fact that if I was using a really narrow rf
active filter, the signal I really wanted was at an offset of 600 to 800
Hz, and if I tried to do 4x at that offset frequency, I would likely be
injecting a really strong signal on top of what I wanted to listen to.
I pondered this aspect quite a bit.
In order to produce the necessary 4x clocking for the rf switched
capacitor filter, it would be necessary to combine the BFO and VFO
together and then produce a composite 4x signal (maybe using a PLL or a
bunch of exactly synchronized DDSs... a bit complex but manageable) to
drive the filter.
In order to simply test the idea, I decided upon a scheme to use a
direct conversion receiver as the test bed. Then it would be simple
since the 4x signal used for the RF switched capacitor filter could be
divided by four and used for the direct conversion mixer VFO.
Much simpler! A direct conversion receiver would not be optimum, but it
would allow me to test the idea before trying the more complex 4x
clocking trick on a superhet receiver.
It was then that I started thinking (I was still driving) about what the
signals looked like on the switched capacitor filter at zero beat to an
incoming RF signal, which would be four equidistant points on a
sinewave, stationary. Then I thought about what they looked like when
not exactly on frequency: each of the four outputs would drift slowly
tracing a sinewave at the difference frequency, each formed of RF
samples taken 1/4 cycle apart, each output then being 90 degrees apart
from each other.
I suddenly realized that the switched capacitor topology could be used
as a detector (no separate direct conversion mixer required), that it
naturally produced all four phases of output, had bandpass
characteristics, and that the input RF signal was being "integrated" on
to each of the four capacitors through the input filter R, so that only
difference frequency appeared and not the sum. Lastly, I realized that
the new detector seemed to capture the peak RF voltage, and thus have a
low detection loss (under 0.9 db conversion loss).
Since I have read about the R2 phasing receiver, I began to work on
receiver designs based upon that approach (two branch I and Q audio
combined using 90 degree phasing shifting to given single sided
reception). I have been working on incrementally making improvements to
the basic design for several years now. I am on my 4th generation
design, and there are still ideas to try out and improvements to made.
I am trying out new modifications this weekend.
I never returned to pursue the RF active front end receiver filtering
aspect. There is only so much time in the world to experiment! It might
not work the first time, but problems are simply engineering
opportunities.
- Dan Tayloe, N7VE; Phoenix, Az; Az ScQRPions
--
Britney Spears' Guide to Semiconductor Physics
<http://britneyspears.ac/lasers.htm>
Richard Hosking
"Michael Black" <blac...@cam.org> wrote in message
news:6447bcd3.02022...@posting.google.com...
> jim.bo...@lintek.aeroflex.com (JLB) wrote in message
news:<8d5ec8a2.02022...@posting.google.com>...
> >
> Mixer" here (and I found a schematic) was a certain resemblance to
> the sort of circuitry used in Coherent CW. I probably noticed it
> because looking at CCW schematics, there was something nagging me
> at the back of my mind. There was something there that I thought could
> be extended....
>
> I've always been curious about how Dan Tayloe came to the circuit.
> What prompted him to give it a try? I've not seen anything about
> this (though perhaps I missed it). Because seeing commutating filters
> (the first issue of QST that I ever saw, April 1971, had such a filter)
> and seeing those CCW schematics made me think, though I never tried
> anything and there was only supposition (and vague enough that I can
> no longer remember exactly what).
>
> In CCW, you have two data paths. In each, there is a mixer (or was
> it a sample and hold?) being clocked at the input frequency, but
> the clock to each channel is in quadrature.
At this point the circuits are similar
So you get a DC signal.
> You then have two more mixers (or again was it a sample and hold?),
> clocked at audio frequencies. Again, the clocking is in quadrature.
This is equivalent to the second multiplexer except that in our circuit the
second clock converts from DC back up the signal frequency or to some IF. In
the CCW case the IF is at audio (but not 0Hz). Presumably this has the
effect of shifting the CW carrier by an amount equivalent to the second
clock
>
> It's been a while since I've looked at descriptions (and sadly, none
> of my books have CCW in more than passing), but the first conversion sure
> looked like what you'd see in a phasing receiver. The second conversion,
> back to audio, seemed to do the job of the audio phase network, The
> result was singal signal, but a narrow bandwidth.
>
> The best article in the ham magazines about commutating filters was in Ham
> Radio magazine, circa 1978, I think by Hank Olson.
So it has all been done before...
Unless I'm garbling
> it, the cover features Wes Stewart's split-band voice processor. I
> had always meant to dig up some of these articles and post a bibliography,
> because of the similarity between commutating filters, and CCW, and the
> Tayloe Mixer.
>
Michael VE2BV
Richard VK6BRO
JLB
> I think you misunderstood the setup which is difficult to explain in text.
> The filter effect works at RF
> For example, on a breadboard test with a clock of 3.5 MHz, the ouput peaked
> with an input frequency of 3.5 MHz (ie at the clock frequency) and with some
> setups (eg having an input resistor to increase the R in the RC network) a
> very sharp peak could be obtained (6 dB down at inputs of 3.5 MHz +/- 20 KHz
> or so)
> Moreover the filter can be tuned to any frequency simply by tuning the first
> mixer clock. Further filtering can be performed by making the second
> multiplexer clock at some IF (say 455 KHz) and the signal is then
> upconverted from zero to 455 KHz without image problems. (apart from
> "inband" images due to quadrature/SSB inbalance)
>
> Richard
>
I'm not an expert on such circuits but I beleive that a switched
capacitor filter will respond at the clock frequency, and at harmonics
of the clock frequency. If the input signal is less than half the
clock frequency you can use it as a low pass filter, with a rather
sharp cut-off, if I can recall correctly.
What is usually done is to put an analog filter before the thing,
allowing you to select which 'mode' you want it to work in.
Jim
N8EE
Richard Hosking
I guess the Tayloe circuit could be looked at as a quadrature switched
capacitor filter
So the source of some of our noise may be Audio getting through the Tayloe
detector
Perhaps we should put an analogue high pass filter in front of the first set
of mixers
Richard
"JLB" <jim.bo...@lintek.aeroflex.com> wrote in message
news:8d5ec8a2.02022...@posting.google.com...
Carl R. Stevenson
effectively the same as, the Weaver modulator and
a LOT of other prior art which builds on the Weaver
modulator ... basically, mix (or sample, which is the
same thing, effectively) down to baseband, lowpass
filter, then mix back up to some other frequency ...
if the phases are arranged properly, it's "single signal"
(ssb) and the "LO" for the 2nd mix (back up from
baseband) can be arranged to make the recovered
signal fall at audio, thus making it a demodulator.
Carl - wa6vse
"Richard Hosking" <zapri...@iinet.net.au(remove the zap)> wrote in
message news:3c75c2b7$0$38...@echo-01.iinet.net.au...