EMRFD Message Archive 4599

Message Date From Subject
4599 2010-04-28 10:13:57 g0kla Quadrature Splitter
Hi all,

I'm on the air with my 40m version of the R2/T2. Seems to work well, with both ssb and cw contacts.

I'm thinking about a 20m version and that got me thinking of a 20/15/10m radio. I could build a core r2/t2 radio like I have for 40m and build a set of converters as front ends. That seems to be the common approach.

Instead, because I like the clean nature of a single conversion design, I've been thinking how to generate the necessary quadrature local oscillator signals for each band. The r2/t2 LO quadrature splitters are band specific. I would need one for each band and then switching for all the IQ signals. Sounds like a headache.

I've seen SDR designs, like here http://www.flex-radio.com/Data/Doc/qex1.pdf, that use 7474 flip flops to create the LO quadrature clock. This drives a Tayloe Detector, rather than a pair of diode mixers.

This looks broadband and ideal.

Are there significant drawbacks (beyond square wave harmonics and promising Dan its for hobby purposes only) before I start to prototype this?

thanks,
Chris
G0KLA
and now also AC2CZ
4601 2010-04-28 11:07:00 Dave - WB6DHW Re: Quadrature Splitter
Chris:
In my UHFSDR board, I use a pair of LVPECL flip flips(rated at up to 4
GHz) to generate quadrature signals at 1/2 the input frequency. These
signals signals then directly drive 2 DBM's. A DBM will generate its
own odd harmonic responses anyway. 74AC74's work up to about 200 MHz. I
use 74LVC74's(rated up to 350 MHz) on my 995x board.

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

g0kla wrote:
> Hi all,
>
> I'm on the air with my 40m version of the R2/T2. Seems to work well, with both ssb and cw contacts.
>
> I'm thinking about a 20m version and that got me thinking of a 20/15/10m radio. I could build a core r2/t2 radio like I have for 40m and build a set of converters as front ends. That seems to be the common approach.
>
> Instead, because I like the clean nature of a single conversion design, I've been thinking how to generate the necessary quadrature local oscillator signals for each band. The r2/t2 LO quadrature splitters are band specific. I would need one for each band and then switching for all the IQ signals. Sounds like a headache.
>
> I've seen SDR designs, like here http://www.flex-radio.com/Data/Doc/qex1.pdf, that use 7474 flip flops to create the LO quadrature clock. This drives a Tayloe Detector, rather than a pair of diode mixers.
>
> This looks broadband and ideal.
>
> Are there significant drawbacks (beyond square wave harmonics and promising Dan its for hobby purposes only) before I start to prototype this?
>
> thanks,
> Chris
> G0KLA
> and now also AC2CZ
>
>
>
>
>
>
>
4603 2010-04-28 12:35:40 Chris Trask Re: Quadrature Splitter
>
> I'm on the air with my 40m version of the R2/T2. Seems to work well, with
both
> ssb and cw contacts.
>
> I'm thinking about a 20m version and that got me thinking of a 20/15/10m
radio.
> I could build a core r2/t2 radio like I have for 40m and build a set of
converters
> as front ends. That seems to be the common approach.
>
> Instead, because I like the clean nature of a single conversion design,
I've been
> thinking how to generate the necessary quadrature local oscillator signals
for
> each band. The r2/t2 LO quadrature splitters are band specific. I would
need one
> for each band and then switching for all the IQ signals. Sounds like a
headache.
>
> I've seen SDR designs, like here
http://www.flex-radio.com/Data/Doc/qex1.pdf, that
> use 7474 flip flops to create the LO quadrature clock. This drives a
Tayloe
> Detector, rather than a pair of diode mixers.
>
> This looks broadband and ideal.
>
> Are there significant drawbacks (beyond square wave harmonics and
promising Dan
> its for hobby purposes only) before I start to prototype this?
>

The only drawback is that the input signal to the 7474 quadrature
generator is four times that of the output signal. I see that someone has
already posted a reply and has suggested some high-speed flip-flops for
this, so that may not be a problem.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
4604 2010-04-28 18:16:07 svss_editor Re: Quadrature Splitter
Hard to beat the digital divider method...

But, I think there was some interesting discussi
4605 2010-04-28 18:39:55 Tim Re: Quadrature Splitter
There's a zillion choices, but my choice was to build a combinati
4613 2010-04-30 12:14:47 g0kla Re: Quadrature Splitter
Yes, I've definitely considered a splitter per band. I'm not a band hopper right now because I only have 40m :) but I would like to be able to switch bands easily, or at least that is what I am trying to design.

Unfortunately I don't get to play with digital stuff at work, so I'll play with some of these ideas and see what I come up with.

Chris
G0KLA
AC2CZ

4615 2010-04-30 20:34:25 svss_editor Re: Quadrature Splitter
I had a chance to take a look. It was in EMRFD. There is some very interesting discussion in there about RF or LO phase shifting. Several approaches are discussed. The wide band phase shift network that isn't so wideband in amplitude followed by limiting certainly sounds like it would make for interesting experimentation.

The specific circuit I was trying to recall was described as 6.8 to 11MHz. (Not single band, but not an octave, either.) The text suggested it as only suitable if used with front panel amplitude and phase trim controls.

73
Drew
kb9fko

4616 2010-05-01 10:09:30 Tayloe Dan-P26412 Re: Quadrature Splitter
I think this would be interesting to try. I used a 90 phase shifter
(C/L/C 1/4 wave lumped - Xc = XL) with a trimmer in the NC2030, and it
was optimized for exactly one frequency on the band. Having something
that is "broadband" might at least maintain a more consistent phase
shift within a given band.

- Dan, N&VE

________________________________

4617 2010-05-01 12:33:07 KK7B Re: Quadrature Splitter
Hi Drew, Dan and all,

Here is a rambling commentary on LO and RF phase shift networks.

I've tried many different approaches to LO or occasionally RF signal path phase shifting, and there are many more that haven't yet made it onto my bench. The ones I described in EMRFD nearly a decade ago are approaches that I had used more than once in successful rigs. There are several considerations beyond 90 degrees of phase shift and equal amplitude, so a good approach for one application may be a poor choice somewhere else. Here are some of my current thoughts.

As Dan pointed out, the quarter wave line--either a transmission line or one of the lumped element equivalents--has 90 degrees of phase shift at only one frequency, so opposite sideband suppression degrades rapidly on either side. That may be OK for a rig designed for a favorite part of a band. But quarter wave lines have another property that isn't often mentioned in the context of phase shift networks: they are impedance inverters. A splitter driving a zero length line and a quarter wave line works fine in a pure 50 ohm environment, but if the terminations are not 50 ohms, the signal amplitudes will be different. A worst case scenario is a quarter wavelength 50 ohm line connected to directly to an input of a CMOS IC. The near-open circuit at the IC input transforms to a near-short circuit at the other end of the quarter wave line. Quarter wavelength 50 ohm lines need 50 ohm termination resistors at the IC....but that requires a lot of LO power to get enough voltage swing. I have used quarter wavelength lines at microwaves and sometimes into diode ring mixer LO ports....but not often and not recently. Their main feature is that they are easy to understand in block diagram form, so I use them frequently on the white board when introducing phasing systems.

The various digital phase shift network schemes suffer from similar termination issues, but work very well as long as interconnections are short enough that we can ignore transmission line reflections. They would seem to be ideal for highly integrated systems on chip. As soon as interconnecting lines are more than a few cm long at HF, one needs to be concerned with terminations for transmission lines carrying signals where relative phase is important. It is easier if the signals are sine waves, because then the termination only needs to be good in the band of interest. With square-wave signals, we need to use flat terminations over at least a decade above the fundamental to keep the phases of all those odd harmonics lined up. A few degrees of phase error at the 5th harmonic can upset the zero-crossings, and it doesn't take much termination impedance or line length variation to get a few degrees at the 5th harmonic. In high performance phasing systems with transmission lines between the LO system and mixers, I use a lot of low-pass circuitry and 3 dB resistive pads to keep the impedance near 50 ohms and the harmonic levels down.

I generally build single-band radios, often optimized for a single mode, and sometimes even optimized for a single operating event: field day; portable VHF contest; or straight key night. That approach allows me to customize everything for the application, without the compromises inherent in multi-band, multi-mode, do-everything radios. Imagine adding a passenger seat to an Indy 500 race car...that's what you do when you add a band to an optimized single band radio. For a single band radio, the single twisted-wire quadrature hybrid originally described by Reed Fisher decades ago in QST has many advantages. When used in the LO path, It offers good performance over the SSB and CW portions of any band above 75m without adjustments. On 80/75m, I'd optimize it for half of the band, or perhaps follow it with a pair of diode limiters low-pass filters, and buffer amplifiers. That would provide near-zero amplitude and phase errors over an octave.

I designed the wide-band quadrature hybrid described in EMRFD for 6.8 to 11 MHz for a binaural short-wave receiver. That's a great radio, with calibrated bandspread on two ham bands. If I remember right, the maximum phase error at mid-range is about 3 degrees, with smaller errors in the 40m and 30m ham bands. Amplitude errors are less important with networks connected to the LO ports of mixers. Performance of that broadband network is good enough for a binaural receiver or a general purpose phasing receiver, but for serious opposite sideband suppression, front-panel phase and amplitude trimmers would be useful. They are easy to add: see the microR2 design in the 2008 and later ARRL Handbooks for an amplitude and phase trim circuit using trimpots that goes in the audio signal path after the LNAs. If I remember correctly, I first encountered that phase and amplitude trim circuit in some work by Byron Blanchard N1EKV.

The old RC-CR LO phase shift network I used in the original R2 receiver in January 1993 QST is not a bad choice for driving a pair of CMOS inputs. I'd like to see some more work on that in the context of some of the excellent mixer work being done with various CMOS switches. I believe the RC-CR network may have been used in some of the circuits using the CD4066 as a mixer published in Sprat in the 1970s. I suggest putting it right at the CMOS inputs with zero transmission line length. Polyphase networks may also be used above audio frequencies...

Using the quadrature phase shift network in the LO path in the block diagram has the obvious advantage that amplitude unbalance is suppressed in the mixers. But the obvious solution isn't always best. There are a number of good arguments for putting the quadrature hybrid in the RF path, particularly in radios optimized for a narrow frequency range.

I am encouraged by all the experiments and interesting approaches. Keep up the good work and keep posting these things. This is a community of ideas, and the best ideas are ones that start interesting discussions. Those are often interesting concepts that didn't quite work as expected.

Best Regards,

Rick KK7B

4618 2010-05-01 14:52:46 ajparent1 Re: Quadrature Splitter
Hi,

What is clear is simple methods are ok for a limited range of frequency spreads.

There are elaborate methods using PLLs that lock at 90degrees
and other hardware that is elaborate to construct.

However, There is nothing to say one could not make a switched
LO, quadrature system for every band using multiple twisted wire quadratures.

I've elected to run the phasing system as a fixed IF at one
frequency and optimize there and use converters (transverters
for TRX) to get to other bands. This has it's liabilites
with dynamic range if care is not taken but also some simplicities.


Allison