Tayloe Detector / Polyphase SSB questions
Hans Summers
I'm interested in views on the following questions.
Considering a receiver consisting of input tuned circuit, Tayloe detector,
polyphase audio phasing network, and audio filterting/amplification:
1. Is an RF amp necessary/desirable ahead of the Tayloe detector? Or am I
right in thinking that you want to minimise IMD products by doing all
amplification after the detector?
2. Sometimes people use 2x clocking, sometimes 4x clocking. Presumably if
you use 2x it is important to get an exact mark/space ration of 1, which
could be difficult, particularly in a multiband receiver. Is the 4x clocked
version therefore always superior to any 2x clock design?
3. I have seen a few circuits using the Tayloe detector followed by a
phasing network. In some cases, people subtract the Tayloe 0 and 180
outputs, and the 90 and 270 outputs ahead of the phasing network. E.g.
"QRP2001 Receiver" in Sprat #101. Why? Why not simply feed each of the 4
Tayloe outputs directly into the polyphase network as in G3OGW's 4-path
Polyphase D-C receiver (Tech Topics, RadCom May2001). The 4-path method is
said to give big improvements over 2-path by cancelling the unwanted
sideband inside the polyphase network, this reducing errors / component
tolerance sensitivity.
4. Is amplification needed between the Tayloe detector and the phasing
network? I understand the input to the phasing network should be low
impedance, and the phasing network likes to drive a high impedance load at
its other end. So can one just put 4 unity-gain low noise op-amps before and
again after the phasing network? In that case all the gain in the receiver
would be in the subsequent AF stages.
5. On the subject of the phasing network itself, all phasing networks I've
seen use a fixed value of resistor in every stage, with different
capacitance values in each stage. Since close-tolerance resistors are a lot
more easily obtainable than close-tolerance capacitors, doesn't it make more
sense to use identical value capacitors throughout the network? Then you can
buy a large number of the same capacitors, and select close matching sets of
4 for each stage.
73's
Hans Summers, G0UPL G-QRP #7802
Dan Tayloe
accomplish. Do you want best possible IP3 performance? Do you want
best possible sensitivity? Do you want best possible opposite sideband
rejection? Do you want minimum power consumption? Are you looking at a
simple, medium performance radio?
Hans Summers wrote:
>
> I'm interested in views on the following questions.
>
> Considering a receiver consisting of input tuned circuit, Tayloe detector,
> polyphase audio phasing network, and audio filterting/amplification:
I personally have not used a polyphase audio because my understanding is
that they are rather lossy. 40 db? If it is indeed that high, there
are engineer consequences to be paid depending on your design goal. In
addition, it uses series resistors. Resistors are noise sources that
can further degrade the sensitivity of the receiver.
>
> 1. Is an RF amp necessary/desirable ahead of the Tayloe detector? Or am I
> right in thinking that you want to minimise IMD products by doing all
> amplification after the detector?
RF preamplification can improve the receiver sensitivity. I have heard
that 40m never has less than -116 dbm of noise. Since 50 ohm noise is
about -148 dbm (500 Hz), that means your receiver could be -120 dbm
sensitive (28 db noise figure) and still hear all there is to hear on
the band. I know that on 20m, one of my -143 dbm receivers (3 db S+N/N
MDS) could hear weak but readable cw signals that a -139 dbm receiver
could not hear at all. So on some bands, particularly 20m and higher,
additional sensitivity is useful.
Think of IM not as a function of just the detector, but rather as the
signal saturation of the rf preamp, followed by the detector, followed
by the first audio pre-amplifier. The detector can take 4 volts pk-pk
if a 5v device is used. As such the first audio pre-amplifier will
likely run out of voltage headroom (assuming it has reasonable gain)
before the detector will. Thus the audio preamp will be the stage that
sets the IP3, not the detector.
You can make a very good estimate of the IM point of the receiver by
simply taking maximum possible pk-pk voltage output of the audio
preamplifier and factoring out the gain of the rf preamp (if any), the
audio preamp gain, and the 1 db loss of the detector, and calculating
what input signal level will saturate the audio preamp output. As a
first order estimate, subtract about 23 db from this and you will have a
fair estimate of the IP3 of the front end.
More RF gain can give better sensitivity, as long as 4v pk-pk is not
exceeded at the detector input, but any gain either in the RF preamp or
in the audio preamp will reduce IP3. Without a rf preamp or a rf step
up transformer, I figure that -143 dbm sensitivity (5 db NF) is
achievable with almost +40 dbm IP3. Playing with the numbers, I think
-146.5 dbm (1.5 db NF) is achievable with about +20 dbm IP3.
I have a transceiver that emphasizes low current consumption above all
else, and at 11 ma at 12v achieves -137 dbm MDS (350 Hz, 3db S+N/N) and
+20 dbm IP3. It is all a big bunch of tradeoffs.
>
> 2. Sometimes people use 2x clocking, sometimes 4x clocking. Presumably if
> you use 2x it is important to get an exact mark/space ration of 1, which
> could be difficult, particularly in a multiband receiver. Is the 4x clocked
> version therefore always superior to any 2x clock design?
I use 4x because it is simple. I don't think there is really any
difference in using a 4x frequency source or a 2x frequency source. The
stability of both should be about the same. The neat thing
is that once you have a frequency source for 4x of 28 MHz, divide by 2
and you have 14 MHz, divide by 2 again and get 7 MHz, etc. However,
each "divide by two" also halves the available frequency range.
I have also used dual 1x sources, i.e., a pair of DDSs (AD9851s)
initialized to be 90 degrees apart to get the two bits to drive the
detector.
> 3. I have seen a few circuits using the Tayloe detector followed by a
> phasing network. In some cases, people subtract the Tayloe 0 and 180
> outputs, and the 90 and 270 outputs ahead of the phasing network. E.g.
> "QRP2001 Receiver" in Sprat #101. Why? Why not simply feed each of the 4
> Tayloe outputs directly into the polyphase network as in G3OGW's 4-path
> Polyphase D-C receiver (Tech Topics, RadCom May2001). The 4-path method is
> said to give big improvements over 2-path by cancelling the unwanted
> sideband inside the polyphase network, this reducing errors / component
> tolerance sensitivity.
The R2 type phasing strip has unity gain which is attractive when the
goal is high sensitivity or high IP3. I get a minimum of 45 db of
opposite side band rejection, which is pretty good. I understand that
the polyphase can do 60 db of rejection, but you have to design around
the associated high loss. It is difficult to use a polyphase network
and still have a receiver with both good sensitivity and IP3.
> 4. Is amplification needed between the Tayloe detector and the phasing
> network? I understand the input to the phasing network should be low
> impedance, and the phasing network likes to drive a high impedance load at
> its other end. So can one just put 4 unity-gain low noise op-amps before and
> again after the phasing network? In that case all the gain in the receiver
> would be in the subsequent AF stages.
You will need an audio pre-amp with gain if you want to keep the
sensitivity intact. For a really noisy band like 80 or 160m, you might
not care as long as you are not trying to use a high loss receive
antenna such as a beverage. for the same sensitivity, the polyphase
network needs more gain (much more?) to overcome the polyphase loss,
which will negatively impact the IP3, but on the other hand, the
opposite sideband rejection might be 15 db better.
- Dan Tayloe, N7VE; Phoenix, Az; Az ScQRPions
Hans Summers
Thanks for the detailed answer to my questions.
"Dan Tayloe" <dta...@home.com> wrote in message
news:3B676B0D...@home.com...
> I guess the answer to your question depends on what you are trying to
> accomplish. Do you want best possible IP3 performance? Do you want
> best possible sensitivity? Do you want best possible opposite sideband
> rejection? Do you want minimum power consumption? Are you looking at a
> simple, medium performance radio?
My priorities are
1. Simplicity and ease of setting up. This is why I am considering DC rather
than a superhet. This is also why I like the Tayloe detector.
2. Good sideband rejection. For this reason I am interested in 4-path
polyphase as opposed to 2-path audio phasing. I am not worried about the
cost of precision resistors and capacitors for the phasing network, or the
number of stages (6 or even 8).
3. Good IP3
4. Sensitivity (last because I am thinking only of 14MHz max)
I don't care at all about power consumption!
> >
> > I'm interested in views on the following questions.
> >
> > Considering a receiver consisting of input tuned circuit, Tayloe
detector,
> > polyphase audio phasing network, and audio filterting/amplification:
>
> I personally have not used a polyphase audio because my understanding is
> that they are rather lossy. 40 db? If it is indeed that high, there
> are engineer consequences to be paid depending on your design goal. In
> addition, it uses series resistors. Resistors are noise sources that
> can further degrade the sensitivity of the receiver.
That is interesting. Does anyone know of any references where the losses in
a polyphase network are mentioned? I was not aware that they were considered
lossy. There is a diagram on page 15 of June 1995's QEX in an article titled
"Polyphase Netowrk Calculation using a Vector Analysis Method", which shows
an attenuation of <0.5db in the optimised network, and <6db in a constant
resistance network. Since I am as yet unable to understand the maths in this
article I may be mistaken.
I too worry about the noise introduced by the resistors but I have no feel
for how significant this would be at HF.
I was thinking of a simple Huff & Puff stabilised VFO as per G3DXZ's design,
see the Tech Topics column of RadCom, Sep 2000 (reproduced on my site, see
http://www.hanssummers.com/radio/huffpuff/contents.htm). At 56MHz I would be
able to divide to have a 4x clock for bands 14MHz, 7MHz, 3.5MHz and 1.8MHz.
A lot seems to hinge on the losses of the polyphase network. If I have
understood the QEX June 1995 article correctly, these losses can be made
very small. I was interested in the 4-path rather than 2-path for the extra
sideband suppression, I guess as long as the losses are small this is still
valid. Since I am interested in 1.8 - 14MHz only, from what you have said I
think I could get away with having no RF preamp, at least at first.
Thanks again for the discussion
Hans G0UPL
Phil Rice
> A lot seems to hinge on the losses of the polyphase network. If I have
> understood the QEX June 1995 article correctly, these losses can be made
> very small.
Many years ago I simulated a 6 stage polyphase network using PSpice. The
losses in "mid-band" were around 15 to 18dB.
( See http://ironbark.bendigo.latrobe.edu.au/~rice/ssb/ssb.html
The polyphase network is figure 4. )
73 de Phil VK3BHR
Richard Hosking
I did some work on the polyphase network a few years ago (see Elelctronics
and Wireless world March 1994 and Jan 1996)
I found the the 8 pole audio network had about 20 dB loss if it was followed
by voltage followers.(high impedance) More importantly perhaps, the out of
band losses were much less than in band meaning that the follwoing stages
could be overloaded by out of band signals. These factors may limit the
receiver dynamic range if you did not carefully design for them. You could
reduce losses by summing all ouputs. Sideband suppression was very good at
up to 60 dB and you could get an wide effective audio range of 200Hz to10
KHz
I used a single capacitance value and varied the resistors for a RF version
of the network which was OK to about 30 MHz. At this frequency strays and
circuit tolerances become a problem with only about 20 dB of sideband
suppression. Performance was quite good to 10 MHz however. I think I used 10
nF caps with resistors in the range 33-150 ohms, all surface mount to reduce
size.
I would be interested to hear of your results
Incidentally I have a VFO stabilizer board based on Eamon Skelton's design
at my website
see http://www.iinet.net.au/~richardh/VK6BRO.htm
Richard
"Hans Summers" <Hans.S...@Tudor.Com> wrote in message
news:9k99gg$37uba$1...@ID-61331.news.dfncis.de...
Hans Summers
"Phil Rice" <ri...@ironbark.bendigo.latrobe.edu.au> wrote in message
news:rice-ya02408000R...@news.latrobe.edu.au...
Thanks for the link. I'm going to try simulate a polyphase network myself,
trying the geometric progression of resistor values as shown in the QEX
article, to see if it really does reduce the losses. I'll let you know what
I find.
Hans, G0UPL
Hans Summers
Hi
"Richard Hosking" <zapri...@iinet.net.au(remove the zap)> wrote in
message news:3b6b8f95$0$32...@echo-01.iinet.net.au...
> Hans
> I did some work on the polyphase network a few years ago (see Elelctronics
> and Wireless world March 1994 and Jan 1996)
> I found the the 8 pole audio network had about 20 dB loss if it was
followed
> by voltage followers.(high impedance) More importantly perhaps, the out of
> band losses were much less than in band meaning that the follwoing stages
> could be overloaded by out of band signals. These factors may limit the
> receiver dynamic range if you did not carefully design for them. You could
> reduce losses by summing all ouputs. Sideband suppression was very good at
> up to 60 dB and you could get an wide effective audio range of 200Hz to10
> KHz
>
> I used a single capacitance value and varied the resistors for a RF
version
> of the network which was OK to about 30 MHz. At this frequency strays and
> circuit tolerances become a problem with only about 20 dB of sideband
> suppression. Performance was quite good to 10 MHz however. I think I used
10
> nF caps with resistors in the range 33-150 ohms, all surface mount to
reduce
> size.
I was interested in the single capacitance value, because resistors are
available more accurately than capacitors, so I thought of purchasing a
whole load of same-value capacitors and matching them in 4's. But then I
read the QEX article which recommends that to minimise losses in the
network, one should use a geometric progression of resistance values from
stage to stage. This implies varying capacitances too.
> I would be interested to hear of your results
> Incidentally I have a VFO stabilizer board based on Eamon Skelton's design
> at my website
>
> see http://www.iinet.net.au/~richardh/VK6BRO.htm
I've seen your stabiliser before. Nice work! I linked to it from my
Stabiliser page: http://www.hanssummers.com/radio/huffpuff/contents.htm
>
> Richard
>
Thanks for the discussion, I'll definitely let you know of any results.
Hans G0UPL