EMRFD Message Archive 6496

Message Date From Subject
6496 2011-07-22 06:37:02 Tamás Fábián Tayloe detector with diodes
Dear Group,

few months ago I successfully built an SDR transceiver (It's one of
YU1LM's). The hart of this receiver is of course a Tayloe detector
implemented with some sort of analogue switching IC and divide-by-two
ICs.

The transceiver works from DC all the way to the high HF bands. The
big dream of mine is being able to work CW / SSB up to the 2 meters
band.

As I figured out, the only limitation is the divide-by-two and the
switching ICs. I don't really understand why "almost everybody" (based
on the info that appears on various HAM radio web pages) trying to use
canned (slow) switches instead of brewing one from fast shottky
diodes. On the contrary, while very fast divide-by-two circuits are
also hard to find, they probably even harder to homebrew.

So the problem is to implement a fast divide-by-two circuit, or a 90
degree phase shifter that accurate and has a large frequency range.

My questions are:

- As I mentioned, I think it should not be particularly hard to
homebrew a fast switch. Am I right?
- Is there anybody who ever tried to (succeeded in?) building an SDR
with a Tayloe (or whatever) style mixer for VHF / UHF? (Singe I/Q
conversion, no downconversion before Tayloe stage) Was there exotic
hardware involved?
- Is it hard to build (or buy) a really fast (few GHz) divide by two circuit?
- Is it hard to build a circuit that compares the phase of two signals
and generates an error signal that can be used to adjust a phase shift
network to achieve 90 degrees of phase shift in a broad range of
frequencies?

Thank you in advance,

HA5FTL
6497 2011-07-22 06:53:37 Dan Mills Re: Tayloe detector with diodes
On Fri, 2011-07-22 at 15:36 +0200, Tamás Fábián wrote:

> As I figured out, the only limitation is the divide-by-two and the
> switching ICs. I don't really understand why "almost everybody" (based
> on the info that appears on various HAM radio web pages) trying to use
> canned (slow) switches instead of brewing one from fast shottky
> diodes.

Some of those canned switches manage low single digit ns switching times
(granted I am thinking about the single DPST stuff in SC70 packages).

> - Is it hard to build (or buy) a really fast (few GHz) divide by two circuit?

There are some very fast ECL parts out there but the logic levels are
WEIRD. Don't forget also that as we are talking square waves here, you
need board layout that can handle first 8 or 10 harmonics to get
reasonable signal quality.

> - Is it hard to build a circuit that compares the phase of two signals
> and generates an error signal that can be used to adjust a phase shift
> network to achieve 90 degrees of phase shift in a broad range of
> frequencies?

The phase comparator is called an xor gate (or a balanced mixer will do
as well), RF phase shift networks are tricky if you want broadband and
are usually best avoided.

The easiest way to skin this cat is to use a pair of DDS chips fed from
a common clock and just program the phase control registers differently,
no fast dividers needed.

Regards, Dan.
6498 2011-07-22 07:01:15 Lasse Re: Tayloe detector with diodes
Dan Mills skrev 2011-07-22 15:52:
>
> There are some very fast ECL parts out there but the logic levels are
> WEIRD. Don't forget also that as we are talking square waves here, you
> need board layout that can handle first 8 or 10 harmonics to get
> reasonable signal quality.
>
This might do in lower VHF:
http://www.potatosemi.com/potatosemiweb/74Logic.html
6499 2011-07-22 07:45:45 Tamás Fábián Re: Tayloe detector with diodes
Hi Dan,

> - Is it hard to build a circuit that compares the phase of two signals
> > and generates an error signal that can be used to adjust a phase shift
> > network to achieve 90 degrees of phase shift in a broad range of
> > frequencies?
>
> The phase comparator is called an xor gate (or a balanced mixer will do
> as well), RF phase shift networks are tricky if you want broadband and
> are usually best avoided.
>
> The easiest way to skin this cat is to use a pair of DDS chips fed from
> a common clock and just program the phase control registers differently,
> no fast dividers needed.
>

Wow. Simple, but effective. :) I'm wondering if two Si570 -s could be
configured as such.

Thanks!


[Non-text portions of this message have been removed]
6501 2011-07-22 10:00:59 Dan Mills Re: Tayloe detector with diodes
On Fri, 2011-07-22 at 16:45 +0200, Tamás Fábián wrote:

> > The easiest way to skin this cat is to use a pair of DDS chips fed from
> > a common clock and just program the phase control registers differently,
> > no fast dividers needed.
> >
>
> Wow. Simple, but effective. :) I'm wondering if two Si570 -s could be
> configured as such.

You could probably lock a Si571 to a Si570 in phase quadrature fairly
easily, an xor gate between the two outputs will if lowpassed produce a
tuning voltage for the 571 that will tend to Vcc/2 at 90 degrees.
The details are left as an exercise for the student.

The nice thing about a DDS pair as opposed to the Si570 is that you can
supply an external clock that is common to both parts so with a little
interconnection to sync the two chips internal states to make the
IO_UPDATE happen synchronously (The chips have pins to do this), you can
trivially program any phase offset you like.

I have to admit to having been contemplating a direct conversion
spectrum analyser done in exactly this way, get a baseband I/Q pair,
digitise with a 24 bit audio AD (Actually gets maybe 20 bits!), then
into a PC.... A linear 20 bit input is good for 120db total power range
and by doing fft and averaging before taking the log performance should
be respectable.

Most of the RBW and VBW stuff could be taken care of at baseband (and
quite a lot of it in the PC in software).

Direct conversion makes the tracking generator output really easy, and
with a little careful modulation of the TG output could even make a
decent vector network tool.

Anyone for a 1 eurocard spec. an with tracking generator (DC to say
250Mhz or so), not my current project, but maybe the one after next?

Regards, Dan.
6502 2011-07-22 10:02:00 Tim Re: Tayloe detector with diodes
If you want a 90 degree phase shift across a broad bandwidth, a quadrature divider (aka "2 stage johns
6504 2011-07-22 10:59:49 Tamás Fábián Re: Tayloe detector with diodes
Dan,

thank you again for the info.

As searching for a suitable DDS I found that there are certain types (e.g.
AD9854) that have quadrature outputs. Great. :)

On 22 July 2011 19:00, Dan Mills <dmills@exponent.myzen.co.uk> wrote:

> **
>
>
> On Fri, 2011-07-22 at 16:45 +0200, Tam�s F�bi�n wrote:
>
> > > The easiest way to skin this cat is to use a pair of DDS chips fed from
> > > a common clock and just program the phase control registers
> differently,
> > > no fast dividers needed.
> > >
> >
> > Wow. Simple, but effective. :) I'm wondering if two Si570 -s could be
> > configured as such.
>
> You could probably lock a Si571 to a Si570 in phase quadrature fairly
> easily, an xor gate between the two outputs will if lowpassed produce a
> tuning voltage for the 571 that will tend to Vcc/2 at 90 degrees.
> The details are left as an exercise for the student.
>
> The nice thing about a DDS pair as opposed to the Si570 is that you can
> supply an external clock that is common to both parts so with a little
> interconnection to sync the two chips internal states to make the
> IO_UPDATE happen synchronously (The chips have pins to do this), you can
> trivially program any phase offset you like.
>
> I have to admit to having been contemplating a direct conversion
> spectrum analyser done in exactly this way, get a baseband I/Q pair,
> digitise with a 24 bit audio AD (Actually gets maybe 20 bits!), then
> into a PC.... A linear 20 bit input is good for 120db total power range
> and by doing fft and averaging before taking the log performance should
> be respectable.
>
> Most of the RBW and VBW stuff could be taken care of at baseband (and
> quite a lot of it in the PC in software).
>
> Direct conversion makes the tracking generator output really easy, and
> with a little careful modulation of the TG output could even make a
> decent vector network tool.
>
> Anyone for a 1 eurocard spec. an with tracking generator (DC to say
> 250Mhz or so), not my current project, but maybe the one after next?
>
> Regards, Dan.
>
>
>


[Non-text portions of this message have been removed]
6505 2011-07-22 11:34:59 AD7ZU Re: Tayloe detector with diodes
> - Is it hard to build a circuit that compares the phase of two signals
> and generates an error signal that can be used to adjust a phase shift
> network to achieve 90 degrees of phase shift in a broad range of
> frequencies?

AD8302 will provide relatively accurate phase and amplitude measurement between 2 inputs.
There are several VNA designs that have used this part.
 
The phase measurement appears to be most accurate at the 90 degree point.
8302 has a built in reference which could be used to calibrate the inputs
here is the info:  http://www.analog.com/static/imported-files/data_sheets/AD8302.pdf
 
I have not used this part but ran across the info a few months ago.
 
 
Randy
AD7ZU
6506 2011-07-22 11:53:55 dave Re: Tayloe detector with diodes
I can do 70 cm with my UHFSDR. But, it doesn't use a Tayloe mixer. It uses
DBM's. For the divider, I use MC10EP52 LVPECL dividers. They are rated to 4
GHz and cost about $4 each.

Dave - WB6DHW
<http://wb6dhw.com>

6509 2011-07-22 19:25:33 Ashhar Farhan Re: Tayloe detector with diodes
There are several approaches you could take. The easy one is to use a
transverter to get to 2M from your HF rig. That's the classic way to
go higher in spectrum.

From 2m onwards, coax can provide a very decent way to introduce phase
lag. I used it as the only option available to me (I was doing it
blind, without any instrumentation). Around 19 inch of coax should
work (trim until you no longer hear the beat).

There are several options on the analog side. Diode mixers with hybrid
couplers are described by rick in emrfd in precisely such an
application. That's my personal preference : to go analog as far as I
can.

There's yet another way to produce 90 degrees phase across a wide
range. Simply lock a vco to your frequency source. The PLL, by it's
nature provides 90 degree phase difference. I say 'simply' because it
will probably work even with injection locking or first order loops.

- farhan VU2ESE

On 7/22/11, Tamás Fábián <giganetom@gmail.com> wrote:
> Dear Group,
>
> few months ago I successfully built an SDR transceiver (It's one of
> YU1LM's). The hart of this receiver is of course a Tayloe detector
> implemented with some sort of analogue switching IC and divide-by-two
> ICs.
>
> The transceiver works from DC all the way to the high HF bands. The
> big dream of mine is being able to work CW / SSB up to the 2 meters
> band.
>
> As I figured out, the only limitation is the divide-by-two and the
> switching ICs. I don't really understand why "almost everybody" (based
> on the info that appears on various HAM radio web pages) trying to use
> canned (slow) switches instead of brewing one from fast shottky
> diodes. On the contrary, while very fast divide-by-two circuits are
> also hard to find, they probably even harder to homebrew.
>
> So the problem is to implement a fast divide-by-two circuit, or a 90
> degree phase shifter that accurate and has a large frequency range.
>
> My questions are:
>
> - As I mentioned, I think it should not be particularly hard to
> homebrew a fast switch. Am I right?
> - Is there anybody who ever tried to (succeeded in?) building an SDR
> with a Tayloe (or whatever) style mixer for VHF / UHF? (Singe I/Q
> conversion, no downconversion before Tayloe stage) Was there exotic
> hardware involved?
> - Is it hard to build (or buy) a really fast (few GHz) divide by two
> circuit?
> - Is it hard to build a circuit that compares the phase of two signals
> and generates an error signal that can be used to adjust a phase shift
> network to achieve 90 degrees of phase shift in a broad range of
> frequencies?
>
> Thank you in advance,
>
> HA5FTL
>

--
Sent from my mobile device
6511 2011-07-23 03:27:25 Tamás Fábián Re: Tayloe detector with diodes
>
> There's yet another way to produce 90 degrees phase across a wide
> range. Simply lock a vco to your frequency source. The PLL, by it's
> nature provides 90 degree phase difference. I say 'simply' because it
> will probably work even with injection locking or first order loops.
>
Farhan, I _almost_ thought about this, but instead of a complete PLL a phase
comparator and a tunable phase shift network probably would do the trick.


[Non-text portions of this message have been removed]
6512 2011-07-23 03:31:36 Tamás Fábián Re: Tayloe detector with diodes
Dave,

that's fine, I'm not insisting on tayloe mixers that much.

The UHFSDR looks great, basicly the same thing I'd like to achieve in
function. ;)

On 22 July 2011 20:53, dave <dave@wb6dhw.com> wrote:

> **
>
>
> I can do 70 cm with my UHFSDR. But, it doesn't use a Tayloe mixer. It uses
> DBM's. For the divider, I use MC10EP52 LVPECL dividers. They are rated to 4
>
> GHz and cost about $4 each.
>
> Dave - WB6DHW
>


[Non-text portions of this message have been removed]