EMRFD Message Archive 6469

Message Date From Subject
6469 2011-07-20 08:56:45 Dan Mills Cartesian loop exciter?
Hi all, I just found this group.

My name is Dan Mills, QTH is the UK and I am working on a cartesian loop
transmitter project.

Just wondering if anyone hereabouts had attempted such a thing for the
HF bands?

I have found a solution to the loop stability headache caused by the PA
group delay (at least for solid state amplifiers), and am currently in
the middle of designing a prototype exciter to test the concept.

The technology potentially allows for both some very interesting
efficiency tricks and for significantly improved 2 tone IMD.

Regards, Dan (2E0CHE formerly M6ATV).
6470 2011-07-20 10:05:18 KK7B Re: Cartesian loop exciter?
Hi Dan,

Welcome. That sounds like an great project, and I'm sure there are folks on this group with interest in that area. I'm familiar with the technology from my work, but for those
6472 2011-07-20 11:24:39 Dan Mills Re: Cartesian loop exciter?
On Wed, 2011-07-20 at 17:00 +0000, KK7B wrote:
> Hi Dan,
>
> Welcome. That sounds like an great project, and I'm sure there are
> folks on this group with interest in that area. I'm familiar with the
> technology from my work, but for those on the site who may be
> unfamiliar with cartesian feedback, do you have an on-line reference
> to suggest? I think maybe Gary Breed published something in High
> Frequency Electronics...

I don't have a copy of HF electronics to hand, and there is remarkably
little on line that does not concentrate on cmos microwave amps for
phones or tetra, and much of that needs IEEE library access (I just love
studying at the open university sometimes, their electronic journal
access rocks).

The following is a short rather handwavy explanation of the technique (I
will forgo the math!).

Cartesian feedback is an envelope feedback linearisation technique
suitable for radio transmitters where the baseband is readily available
as an I/Q pair. It differs from the polar loop in that the baseband
phase and amplitude information are not separate but are encoded in the
levels and relative phases of the I/Q pair.

The RF bits are basically a quadrature modulator taking an I/Q pair (and
ultimately via various filters and conversions and such feeding a PA),
and a power sampler at the output of the PA feeding a quadrature
demodulator producing a demodulated I/Q pair, an error amp then closes
the loop and the IMD within the loop bandwidth drops massively.

All very neat, but there is a gotcha, the PA has group delay potentially
equal to tens of radians of phase shift at the top end of HF, and the
product detectors mean that this phase shift is reflected in the
returning I/Q pair from the demodulator. Unless this is compensated
somehow the loop will only be stable for those frequencies where the PA
has little RF phase shift.

Note that this is an RF issue, in terms of the highest modulating
frequency the PA group delay amounts to nothing very much
(fortunately).

My design gets around the worst of the group delay problem by running
both the modulator and demodulator clocks from a pair of DDS chips which
are fed by a common master clock. The DDS chips have a convenient way
to adjust the phase of their outputs so the relative phase of the
forward and reverse path local oscs can be adjusted in software to
minimise the error term before the loop is closed, a phase detector and
small micro take care of the rest.

The alternative method is to do all the I/Q processing in DSP with
hilbert transforms and such to allow it to be fixed at baseband, but
that is not trivial because you tend to find that the group delay
through the AD and DA means that you now have a baseband loop stability
problem!

On my current board all the RF happens at a fixed 9Mhz (Because it
seemed a reasonable first IF in a transmitter and is slow enough that
switching mixers are reasonable), but the up and down conversion is not
exactly rocket science.

There is a wideband HF design in the literature that reports better then
-60dBc two tone IMD at reasonable power levels (more then 20 years ago),
amateur transmitters have NOT IMHO kept up with the improvements in the
state of the art in receivers.

Ultimately I want to add some games involving variable PA bias and
variable supply voltage to the finals to the mix to try to take
advantage of the reduced requirements for PA linearity.

Negative feedback is cool.

Regards, Dan.
6473 2011-07-20 11:50:43 Chris Trask Re: Cartesian loop exciter?
>
> Hi all, I just found this group.
>
> My name is Dan Mills, QTH is the UK and I am working on a cartesian loop
> transmitter project.
>
> Just wondering if anyone hereabouts had attempted such a thing for the
> HF bands?
>
> I have found a solution to the loop stability headache caused by the PA
> group delay (at least for solid state amplifiers), and am currently in
> the middle of designing a prototype exciter to test the concept.
>

Tucker described the useage of Cartesian feedback for linearizing
mixers. And I have seen it described in some Soviet papers.

The group delay can be very serious if the amplifier is narrow-band.
And you have to limit the useage to one stage at a time.

Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
6474 2011-07-20 12:24:18 Dan Mills Re: Cartesian loop exciter?
On Wed, 2011-07-20 at 11:46 -0700, Chris Trask wrote:

> Tucker described the useage of Cartesian feedback for linearizing
> mixers. And I have seen it described in some Soviet papers.

Very good mixers for the HF band we have these days, it tends to be gain
stages and oscillators (PNDR) that cause fits.

> The group delay can be very serious if the amplifier is narrow-band.
> And you have to limit the useage to one stage at a time.

For typical wideband solid state 'linear' amplifiers we see in amateur
service, and for the typical modulation bandwidth we tend to use, the
group delay is not that much of an issue. At least my old redifon mel
module from a rally (two push pull stages, BLW50/BLW96) does not on a
few experiments show excessive group delay for an envelope feedback
application.

I would not want to try it with a single ended tube amp with a high Q
tank circuit however.

RF as opposed to envelope feedback would also be a whole other matter,
THAT would not work well around more then two stages (and even one can
be an exciting prospect sometimes).

It would be a major problem if the modulation bandwidth was more on the
lines of 'video' bandwidth, but for something which is effectively DC to
< 10Khz modulation, so say a 50Khz loop bandwidth to get most of the
benefit it does not look insurmountable to me.

Regards, Dan.
6475 2011-07-20 18:34:08 Bill Carver Re: Cartesian loop exciter?
"There is a wideband HF design in the literature that reports better then
-60dBc two tone IMD at reasonable power levels (more then 20 years ago),
amateur transmitters have NOT IMHO kept up with the improvements in the
state of the art in receivers."



And THAT comment on transmitter quality has to be the biggest understatement in ham radio. Look in any 20 year old Motorola RF data book and you see lots of bipolar amplifiers with third order IMD 30 dB BELOW ONE TONE. And today, you see rigs reviewed in QST with IMD 10 dB worse, masked by 6 dB in their use of dB BELOW PEP in the specification. And generally without a hint in the pages of text about how bad the rig under test is.

W7AAZ
6477 2011-07-20 19:30:47 Dan Mills Re: Cartesian loop exciter?
On Wed, 2011-07-20 at 19:33 -0600, Bill Carver wrote:
> . And today, you see rigs reviewed in QST with IMD 10 dB worse, masked
> by 6 dB in their use of dB BELOW PEP in the specification. And
> generally without a hint in the pages of text about how bad the rig
> under test is.

Yep, and how we have some of the guilty parties making a big thing out
of running stupidly high bias AB so they can get class A "Up to 75W" and
other such nonsense.

Transmitters however only have to be heard by the OTHER GUY, so nobody
notices apart from the general splatter, noise sidebands and such which
are always some other guys fault.

Also, ever notice how nobody seems to do transmitter sequencing
correctly, Guys you switch the relays, WAIT for the relays to stop
bouncing, switch the bias on, WAIT for the bias to stabilise AND ONLY
THEN start the keying envelope, if you start bringing the envelope up
before the PA is fully online you get a leading edge that sounds like a
fecking spark transmitter (Further note, leave the PA online until the
envelope has fully decayed). Morse is NOT supposed to be audible 10Khz
away from the carrier (Especially not when I paid good money for that
narrow KVG filter).
Yes I know it makes full break in harder, deal with it.

Hell to start with just not running the transistors all the way into
saturation, (RF speech clipping does not mean what some evidently think
it does) and making sure the PA sees a reasonably BROADBAND termination
would help, diplexers really are not rocket science.

I start to think that even just doing all the basics right would put a
rig so far ahead of the curve that actually closing the loop becomes
pointless.

Still if I am going to go to the trouble of building PA3AKEs awesome
front end for the RX, I may as well try for a transmitter equally far
ahead of the pack.

I may be a lowly 2E0 (Roughly General class for the Americans), but at
least my morse is only 250Hz wide!

It is instructive to plug your TX into a 60+db pad and tune across it
with a good receiver while holding the paddles over.

Regards, Dan (Grumpy young man).
6479 2011-07-21 04:51:16 ha5rxz Re: Cartesian loop exciter?
"Guys you switch the relays, WAIT for the relays to stop
bouncing, switch the bias on, WAIT for the bias to stabilise AND ONLY
THEN start the keying envelope, if you start bringing the envelope up
before the PA is fully online you get a leading edge that sounds like a
fecking spark transmitter (Further note, leave the PA online until the
envelope has fully decayed). Morse is NOT supposed to be audible 10Khz
away from the carrier "

This can be done, but in return don't expect full break-in above thirty five words a minute. You also need to bear in mind settling time for the PLL or DDS if you are operating split, this could be greater than the switching time of a reed relay.

HA5RXZ
6480 2011-07-21 09:14:00 Dan Mills Re: Cartesian loop exciter?
On Thu, 2011-07-21 at 11:51 +0000, ha5rxz wrote:

>
> This can be done, but in return don't expect full break-in above
> thirty five words a minute.

If you are running full break in you may well be using a diode TR switch
anyway, but even with relays if you use a high compliance voltage
current mode drive to the coils (or just a big cap across a dropper
resistor, you can with careful relay selection , get the pull in time
down to a very few ms. A zenner in series with the catch diode will
massively accelerate the collapse of the flux and the relay opening.

At T_dot being approximately 1200/WPM, that puts a dot at 100 WPM as
being 12ms.

If pull in takes say 2ms and bringing the bias up takes say 500us then a
12ms dot equates to the receiver being disconnected for 16.5ms, and
available for 7.5ms between dots (assuming that the AGC can cope).

Realistically at that rate you are breaking in between letters or words
not between dots, but it can be done. With a diode switch the situation
is better again obviously, and for us mere mortals, QSB is really not an
issue.

Nope, still no excuse for 10K wide CW.

> You also need to bear in mind settling time for the PLL or DDS if you
> are operating split, this could be greater than the switching time of
> a reed relay.

This is where DDS are a huge win, they reprogram VERY quickly and start
producing the new frequency a couple of sysclock cycles after io_update.
You usually even have a pin that can select a different set of
programming parameters in hardware. Further there is nothing stopping
you overlapping the frequency change with the other delays.

One nice trick is in the way you generate sidetone, do it immediately,
don't wait for the RF, as soon as the key goes down, start making
sidetone, the processor can keep record how much extra key down it needs
to add on the end to match what the user is actually doing.

Regards, Dan.
6489 2011-07-21 20:10:32 Bill Carver Re: Cartesian loop exciter?
I agree Dan, if everyone had a rig that was state of the art rather than the product of a bean-counter/marketing driven company, the noise floor of the bands would drop. If the operators had their heads on straight at least. It is not that difficult to make a SSB rig with third order IMD down 45 dB from one tone.

But independent of that, making a TX that had IMD and splatter 60 dB down is a worthwhile endeavour.