Comments on European VHF discussion reflector
about receiver problems and LA6LCA's RX concept
Subjects:
Re:
VHFDX: LA6LCA remote controlled 2m transceiver
Re: VHFDX: Receiver Linearity - and more
1) G8GSQ Steve writes:
On Monday 09 June 2003 11:30,
J M Noeding wrote:
> It has been some
requests for LA6LCA remote controlled high level 2m
> transceiver. Now I've got some details written down, hope to
correct it
> later
>
> /la8ak/d12.htm
I don't mean this to appear critical, but is there any point in
having such a
high IP3? Doesn't phase noise (from even the best rigs) get you
long before
the intermods?
Steve
2) Leif SM5BSZ writes:
Hi Steve,
> I don't mean this to appear critical,
but is there any point in
> having such a
> high IP3? Doesn't phase noise (from even the best rigs) get
you
> long before the intermods?
If you are careful, you can get noise
sidebands at -145dBc/Hz from
conventional amateur equipment at 100kHz separation. This is the
performance of the LO of TM255E for example. (The tx is not of
this class due to some design error)
In case you are interested in EME (like I am) you want the lowest
possible noise figure which means that you have a mast mounted
preamplifier that places the noise floor from the antenna at
a power level somewhere around -150dBm to allow for a 4dB noise
figure at the receiver input. The antenna noise should be 20dB
above the noise floor of the receiver. This means that a signal
that degrades your system noise floor by 3dB has to be at -5dBm
at the receiver input.
If you have two signals, both of them at -11 dBm, reciprocal
mixing
would not be a problem if one is 100kHz away and the other
200kHz.
To have their third order intermodulation at the noise floor in
1kHz bandwidth, it has to correspond to -120dBm at the receiver
input. The IP3 you need is then +43.5dBm !!!!!!!!!
It IS justified to make IP3 as high as you can possibly make it:)
73
Leif / SM5BSZ
3) GW4DGU writes:
I agree with Steve. Very high linearities, although nice to have,
aren't
really necessary in most situations. Most of the time, at
VHF/UHF, an
overall receiver input intercept of 0dBm is more than adequate.
We solved
that problem (and more) many years ago. It's noticable, though,
that modern
low-noise devices - even some of those touted as high intercept -
aren't
significantly more linear than the old stagers.
I looked at the topology Jan-Martin advocates when I was
designing the muTek
GFBA series LNAs over 20 years ago. It wasn't super-impressive.
The major
problem is that feedback isn't applied over the whole amplifier,
so in fet
amplifiers, channel non-linearities aren't included in the loop
(it's even
worse with BJTs). The muTek amplifiers had (have?)
non-dissipative feedback
applied over the whole amplifier, and gave significantly sub-dB
noise
figures (about 0.5 - 0.6dB, if the switching losses were
disregarded) with
an input intercept of better than +10dBm. I now know how to do
very
significantly better, (I've designed sub-dB LNAs with input
intercepts well
beyond +30dBm professionally) but it would be wasted on the
amateur bands,
because transmitted signals simply ain't good enough.
The real problems nowadays are in bad LO design, and transmitter
noise and
intermod. outputs. Except in some very broad, and basic ways
these
parameters are rarely, if ever, mentioned in amateur reviews.
From what
little is published, it's clear that modern amateur transmitters
fall down
in several areas.
Frequency stability is now a serious consideration. Anyone who
has operated
off the moon using JT44 will have noticed that some stations give
significantly fewer errors than others at a given signal
strength. Why?
Well, not only do some people drift (JT44 needs stabilities of
<5Hz/30s,
preferably significantly better than that) but it also needs a
transmitter
spectrum with the majority of the energy contained within a
couple of Hz.
This level of performance isn't too difficult to achieve, but
many modern
transceivers - particularly those of the all-singing, all
dancing, variety -
don't hack it. My IC706 (which I don't use on VHF!) gaily wanders
around a
70Hz band at 144MHz in a typical 5 minute period, making it
unusable even
for CW EME, although it seems that there are a few well known 2m
EME CW
operators - and I'm not naming callsigns - who seem to think that
a game of
chase the signal is somehow character forming! Some of the more
expensive
multiband/multimode transceivers seem nearly as bad, even when
fitted with
TCXOs. See the latest Radcom. What's fine for casual SSB/CW may
not be good
enough for serious VHF/UHF operation, nowadays.
The width of the transmitted carrier is also significant. A well
designed
DDS + phase-locked translation loop synthesiser can give very
good results,
but it's quite a difficult trick to pull-off well, particularly
at UHF.
Trying to do it with a VCO tuning tens of MHz at VHF or UHF, will
inevitably
result in poor phase noise outside the loop bandwidth, and the
reduced loop
gain will result in more spurious FM within the loop bandwidth,
resulting in
a degraded close-in spectrum.
Acceptable figures for transmitter linearity, in amateur terms,
are still
based on 1950's requirements! While we can just about get away
with
intermods of -20dB wrt pep (which seems not to be untypical of
100W
solid-state PAs on 12V) when using unprocessed ssb, even a few dB
of speech
processing starts to make an increased demand on transmitter
linearity.
Modulation schemes such as COFDM, BPSK/QPSK, or QAM start to
place some very
severe demands on the transmitter line. Currently people
experimenting/using
with these schemes on the bands are simply turning down the wick
in an
attempt to stop the signals spreading, but with suitable
transmitter design
that really shouldn't be necessary. Some form of transmitter
linearisation
is almost mandatory in modern professional transmitters employing
non-constant amplitude modulation. Even modest handhelds for
TETRA have to
have an intermod performance of around -60dBpep. Linearising
amateur
transmitters could result in much cleaner sounding bands -
particularly when
signals are strong. The same would be true if we spent more time,
trying to
reduce transmitter noise outputs, but to do that we need a
standard. The
current RSGB VHF Contests Committee attempt, dating from the
1970s, is
meaningless!
I didn't mean to write this rant when I started! It's also moved
some way
from the original topic. However, it's clear that the bell and
whistles on
modern transceivers haven't really led to improved perfomance. I
certainly
wouldn't spend my money on a modern 'all-band' transceiver if I
wanted to
use it for serious operation, at vhf/uhf. They seem
strictly for Mr & Mrs
Mouse.
Some of the HF-only transceivers look a bit better, but I'm not
sure if I'd
want to spend £2 - 3k or more on a transceiver which wouldn't be
used to its
full potential. The
Elecraft box looks very interesting, but could be
totally screwed-up by a poor - or even typical - transverter. Applying
transmitter linearisation to a transverter-based transceiver
could be fun
for a masochist.
SM5BSZ's Linrad
software and hardware may well be pointing the way to a
better future. It would be hard to better the concept for it's
intended
purpose , but it's very application-specific - at the moment.
My inclination would be to design very good single conversion
transceivers
for the individual bands, probably operating under PC control.
and to use IF
sampling DSP following a phase-linear crystal filter. They would
only need
to tune in relatively wide frequency steps (like the DSP-10) with
interpolation taking place in the DSP. For a number of reasons I
wouldn't
use a zero, or near-zero IF topology.
Enough! I shouldn't have started this! If anyone wants to discuss
anything
further, I'm happy to do so.
73
Chris
GW4DGU
4) SM5BSZ:
Hi Chris and All,
> The real problems nowadays are in bad
LO design, and transmitter noise and
> intermod. outputs. Except in some very broad, and basic ways
these
> parameters are rarely, if ever, mentioned in amateur
reviews. From what
> little is published, it's clear that modern amateur
transmitters fall down
> in several areas.
The effort required to do something about
this unfortunate situation
does not seem utopic to me:)
Last weekand I made measurements on several transceivers at the
Scandinavian VHF meeting. It was not unexpected to find that most
units had what I would call obvious design errors causing poor
spectral purity. At the moment I spend most of my time trying to
do
something about it, the ultimate goal is to have the measurement
standards changed in amateur reviews. The reason I came into this
problem is that when the standard tests are performed on DSP
systems,
the resulting performance data may become utterly misleading.
The measurement is reproducible and standardized - but it is
not relevant. The measured property of the test object is not
the thing that the measurement is supposed to quantify.
Example: BDR as measured by ARRL is not blocking dynamic
range as it is defined in text. The weakest signal a receiver
can receive without significant degradation due to a strong
signal simultaneously present. ARRL just measures blocking
which is something else (and not relevant).
> My IC706 (which I don't use on VHF!)
gaily wanders around a
> 70Hz band at 144MHz in a typical 5 minute period, making it
unusable even
> for CW EME, although it seems that there are a few well
known 2m EME CW
> operators - and I'm not naming callsigns - who seem to think
that
> a game of chase the signal is somehow character forming!
If you drift by less than 50Hz per minute,
your signal is good enough
for 144MHz EME CW. Surely JT44 will not work, but there are
several
EME stations who have much worse stability than this.
Stability is fine (of course) but not really a problem with
modern rigs.
> The width of the transmitted carrier
is also significant. A well designed
> DDS + phase-locked translation loop synthesiser can give
very
> good results,
> but it's quite a difficult trick to pull-off well,
particularly at UHF.
> Trying to do it with a VCO tuning tens of MHz at VHF or UHF,
will
> inevitably
> result in poor phase noise outside the loop bandwidth, and
the
> reduced loop
> gain will result in more spurious FM within the loop
bandwidth,
> resulting in
> a degraded close-in spectrum.
I have monitored the spectra of 144MHz
stations for 10 years or so.
The width of the transmitted carrier is never a problem on 144
when
you look at it from the DX point of view. All of this can create
sideband
noise that is a big problem to the near neighbours.
> Acceptable figures for transmitter
linearity, in amateur terms, are still
> based on 1950's requirements! While we can just about get
away with
> intermods of -20dB wrt pep (which seems not to be untypical
of 100W
> solid-state PAs on 12V) when using unprocessed ssb, even a
few dB
> of speech
> processing starts to make an increased
demand on transmitter linearity.
I am not so sure the problem lies here. The
power spectra of the
SSB signals of all the rigs I tested were very ugly indeed.
"Absolutely unacceptable" is the proper wording.
I do not think this is because of amplifier linearity. I think it
is
because there is no speech processing. The peak power is simply
limited by the ALC. The ALC amplifier is simply an AM modulator
and it is used to amplitude modulate the narrowband SSB signal
in such a way that the peak amplitude stays within limit.
To do this, the "AM modulator" has to be fed with a
signal of
a rather wide bandwidth - otherwise peaks can not be compressed!
The modulation sidebands then extend 25 kHz on either side:(
I do not know for sure, but as I understood the explanation of
the SSB controls, this is the way SSB is done nowadays. I do
know for sure that this is the way it is done in the FT221. I
also
know for sure that there is a dramatic improvement easily
obtained in the FT221 - just by moving a wire one can feed the
SSB generator with the amplitude limited audio intended for the
FM modulator. This makes the ALC unnecessary and the improvement
on signal quality is dramatic.
> Linearising amateur transmitters could
result in much cleaner
> sounding bands - particularly when signals are strong.
I am really very sceptic here. It would be
nice with narrower signals,
but it is even more important to avoid noise at larger frequency
separations. With the current state of the art, adding
linearisation
circuits (noise generators) will be a bad idea unless the
spectrum
puritu is checked at several frequency separations. The
measurements
I have done check Tx noise at 5, 10, 20, 100, 200 and 500 kHz.
An improvement at 10 and 20 kHz must not degrade the performance
at 100kHz and above.
> The same would be true if we spent
more time, trying to
> reduce transmitter noise outputs, but to do that we need a
standard. The
> current RSGB VHF Contests Committee attempt, dating from the
1970s, is
> meaningless!
OK. I agree 100%. I am doing my best to do
something about it:)
> I didn't mean to write this rant when
I started! It's also moved some way
> from the original topic. However, it's clear that the bell
and whistles on
> modern transceivers haven't really led to improved
perfomance. I certainly
> wouldn't spend my money on a modern 'all-band' transceiver
if I wanted to
> use it for serious operation, at vhf/uhf. They seem
strictly for Mr & Mrs
> Mouse.
Yes - but that is because we amateurs do not
appreciate the efforts
made now and then. A good radio does not get it's qualities held
out
properly by the reviewers.
> SM5BSZ's Linrad software and hardware
may well be pointing the way to a
> better future. It would be hard to better the concept for
it's intended
> purpose , but it's very application-specific - at the
moment.
I think this is unfair. Linrad is not
primarily for EME - everybody
seems to think so because that is from where I come;)
Linrad should be most useful on crowded HF bands in difficult
interference
situations. I am not interested myself so it will take some time
before
HF people discover how to use this tool.
> My inclination would be to design very
good single conversion transceivers
> for the individual bands, probably operating under PC
control.
> and to use IF
> sampling DSP following a phase-linear crystal filter. They
would only need
> to tune in relatively wide frequency steps (like the DSP-10)
with
> interpolation taking place in the DSP. For a number of
reasons I wouldn't
> use a zero, or near-zero IF topology.
A simple solution, but you will have problems
if you try to open up the
bandwidth. There is no need at all for a phase-linear filter. It
is trivial
to correct the phase non-linearities in the computer.
73
Leif / SM5BSZ
5) From:
Leif Åsbrink Sm5BSZ
> To: Steve Thompson <steve@eltac.co.uk>;
<vhf-dx-discuss@blacksheep.org>
> Sent: 09 June 2003 20:56
> Subject: RE: VHFDX: LA6LCA remote controlled 2m transceiver
>
>
> Hi Steve,
>
> > I don't mean this to appear critical, but is there any
point in having such a
> > high IP3? Doesn't phase noise (from even the best rigs)
get you
> > long before the intermods?
>
> If you are careful, you can get noise sidebands at
-145dBc/Hz from
> conventional amateur equipment at 100kHz separation. This is
the
> performance of the LO of TM255E for example. (The tx is not
of
> this class due to some design error)
>
> In case you are interested in EME (like I am) you want the
lowest
> possible noise figure which means that you have a mast
mounted
> preamplifier that places the noise floor from the antenna at
> a power level somewhere around -150dBm to allow for a 4dB
noise
> figure at the receiver input.
> ------------------
> I'm with you up to this point
>
> >The antenna noise should be 20dB
> above the noise floor of the receiver.
> ---------------------
> Sorry, but I don't understand where the 20dB comes from
here.
Oooh! This was a remark trying to explain from where the -150dBm
comes. (It should have been -150dBm/Hz)
Sky temp is 290K = -174 dBm/Hz.
Preamp gain is 24 dB so temp at preamp output is -150dBm/Hz.
The NF of the second stage, the receiver itself is 4dB which
means that the noise floor is at -170dBm/Hz, that is why 24 dB
gain was choosen to place the antenna noise 20 dB above
-170dBm/Hz (=-150 dBm/Hz)
With this arrangement, 1% of the noise at the receiver output
will come from the second stage while 99% will come from the
preamp. The 1% noise contribution will degrade the system noise
figure by 0.043 dB. If you choose a "typical" gain
figure of
18dB rather than 24 dB, 3.8% of the noise will come from the
second stage while 96.2% will come from the preamp.This will
degrade the system noise figure by 0.16dB, something which
most EME'ers happily accept because the IP3 of the system
will be 6dB better. The 0.12 dB S/N improvement one can get
here is not meaningless and I argue that one can get advantages
of this size at several places and they sum up to a substantial
improvement.
73
Leif / SM5BSZ
6) G8GSQ
Steve wrote:
OK, Leif - I understand and follow your comments. I come to an
answer from my
own calculations, and one I always come back to whenever I
revisit them. I'm
often wrong, and I'll be very glad if someone can correct me
here.
Having written my tome, I'd like to add a preface - I really hope
this
doesn't come over as me trying to be a smartarse, or trying to
get into a
pissing contest, neither is the case.
Steve
> > ---------------------
> > Sorry, but I don't understand where the 20dB comes from
here.
>
> Oooh! This was a remark trying to explain from where the
-150dBm
> comes. (It should have been -150dBm/Hz)
>
> Sky temp is 290K = -174 dBm/Hz.
> Preamp gain is 24 dB so temp at preamp output is -150dBm/Hz.
The reference point usually used is when the noise floor degrades
by 3dB -
for the sake of discussion, lets assume that all of that
degradation comes
from one external transmitter. The noise at the preamp output is
-147dBm/Hz,
that's -150 from the sky noise and -150 from the transmitter.
Allowing for
the preamp gain, the noise from the transmitter is -174dBm/Hz at
the preamp
input.
As you said earlier, a good transmitter will have noise sidebands
of
-145dBc/Hz, so a signal level of -29dBm at the preamp input will
cause 3dB
noise floor degradation. If we consider the IMD produced by 2 x
-29dBm, then
in SSB bandwidth, then -140dBm is 0dB S/N, which equates to IP3
of +26dBm. At
100Hz, -154dBm IMD gives +33.5dBm IP3 - +40 IP3 only becomes
valuable at 5Hz
bandwidth. Clearly, this is just in the realms of the most recent
developments in digital techniques but it's rarely made clear
that it's a
waste of effort for the average 'dxer' to strive for such
improvements.
Personally, I think that the figures above are seriously
optimistic for most
people. Noise in your own receiver isn't going to be negligible,
and it's
highly unlikely that all other signals you encounter are going to
as clean as
the example assumes. The noise floor increase might well be due
to
contributions from several other transmitters - if there's two in
the example
above, the -29dBm goes to -32, and the IP3s drop accordingly. If
there's two
signals and your own rig has the same -145dBc/Hz LO noise, then
the -29 drops
to -34 and so on. If any of the rigs have poor noise, then
everything goes
out of the window. For example, if any one of them has noise only
10dB higher
than the assumed figure, then the IP3 values you need to maintain
your
dynamic range drop from +33 to about +16dBm at 100Hz bandwidth.
If the phase
noise of any one is 15dB higher, then the useful IP3 drops to
about +8. Given
the quality of most of the equipment out there, I reckon the
phase noise gets
to you long before the IMD.
Finally, I did some tests this morning, and I think the 3dB noise
floor
degradation might be a convenient but not realistic value to use.
These
numbers are for my dud old ears - your values might be different.
I set up a
just readable keyed signal on a set (worked out to be -6dB S/N)
and then
wound up a noise generator until it was unreadable. For my ears,
it's a noise
floor increase of 1.5dB, meaning that the transmitter signals are
a further
4dB lower than examples above.
7) SM5BSZ wrote:
Hi Steve,
> OK, Leif - I understand and follow your comments. I come to
an
> answer from my own calculations, and one I always come back
to
> whenever I revisit them. I'm often wrong, and I'll be very
glad
> if someone can correct me here.
>
> Having written my tome, I'd like to add a preface - I really
hope this
> doesn't come over as me trying to be a smartarse, or trying
to get into a
> pissing contest, neither is the case.
Oooh! I know I am an extremist - but I think it is justified.
A good EME station would use 8 Yagis which would cost something
like 1000$ extra compared to a 4-yagi array. Probably a lot more
if you include the cost for bigger rotors, towers stacking frames
and so on. Just make it 300$ per dB for the sake of argument.
This means one would pay 30$ for a 0.1 dB improvement. Take it
just as a rough estimate. Maybe others on the list who have
really erected big antennas would know this better and come up
with a more realistic value.
Another thing is that various problems with neighbours might
close
the route to improvement by a bigger antenna.
> > > ---------------------
> > > Sorry, but I don't understand where the 20dB comes
from here.
> >
> > Oooh! This was a remark trying to explain from where
the -150dBm
> > comes. (It should have been -150dBm/Hz)
> >
> > Sky temp is 290K = -174 dBm/Hz.
> > Preamp gain is 24 dB so temp at preamp output is
-150dBm/Hz.
> The reference point usually used is when the noise floor
degrades
> by 3dB - for the sake of discussion, lets assume that all of
that
> degradation comes from one external transmitter. The noise
at
> the preamp output is -147dBm/Hz, that's -150 from the sky
noise
> and -150 from the transmitter.
OK.
> Allowing for the preamp gain, the
noise from the transmitter
> is -174dBm/Hz at the preamp input.
>
> As you said earlier, a good transmitter will have noise
sidebands of
> -145dBc/Hz, so a signal level of -29dBm at the preamp input
will
> cause 3dB noise floor degradation. If we consider the IMD
produced by
> 2 x -29dBm, then in SSB bandwidth, then -140dBm is 0dB S/N,
which
> equates to IP3 of +26dBm.
Yes. I agree 100%, but this is at the preamplifier input. At the
preamplifier output IP3 should be 24 dB higher and I was
discussing
the "VHF receiver", the unit we have in the schack, not
the mast
mounted preamplifier......
With your numbers, an IP3 of 40dBm would be reasonable to strive
for
at the input of the main receiver.
> At 100Hz, -154dBm IMD gives +33.5dBm
IP3 - +40 IP3 only becomes
> valuable at 5Hz bandwidth.
When operating CW one uses the "builtin filter" in the
brains
and for discussions like this a bandwidth of 50Hz is reasonable.
It is just not possible to make IP3 anywhere near what is
required
with these assumptions.
> Clearly, this is just in the realms of
the most recent
> developments in digital techniques but it's rarely made
> clear that it's a waste of effort for the average 'dxer'
> to strive for such improvements.
Sure. The average dxer does not care to listen for EME signals
even though they are easily hear on a HB9CV antenna if proper
equipment is used.
The discussion I made was with a reference to EME and it is
intended for the advanced dxer who looks for EME, IONO and
other really weak signals.
> Personally, I think that the figures
above are seriously
> optimistic for most people. Noise in your own receiver
> isn't going to be negligible, and it's highly unlikely
> that all other signals you encounter are going to
> as clean as the example assumes.
Now I think you twist the argument in an unfair way.
In case the noise from your own receiver is not
negligible, it is just because you did not put
sufficient gain in the preamplifier. If you really
want minimum noise floor, it is trivial to increase the
preamp gain and to get a noise figure of 0.3dB or less.
The problem is you can not do it because your own
receiver does not have the IP3 required!!!!
But this was the thing we started from. Is there any
reason to improve IP3 by use of methods like those
suggested by LA6LCA and I argue that the answer is
a clear yes for the advanced dxer but I fully agree
it is no for the average dxer.
> The noise floor increase might well be
due to contributions
> from several other transmitters - if there's two in the
> example above, the -29dBm goes to -32, and the IP3s drop
> accordingly. If there's two signals and your own rig has
> the same -145dBc/Hz LO noise, then the -29 drops to -34
> and so on. If any of the rigs have poor noise, then
everything
> goes out of the window. For example, if any one of them has
> noise only 10dB higher than the assumed figure, then the
> IP3 values you need to maintain your dynamic range drop from
> +33 to about +16dBm at 100Hz bandwidth.
Yes, but 16dBm at the antenna is +40dBm at the main receiver
input.
> If the phase noise of any one is 15dB
higher, then the useful
> IP3 drops to about +8.
Yes. The +40dBm IP3 is surely not needed always. There will be
situations where reciprocal mixing dominates. Unfortunately
there is nothing I can do about it in my own equipment but
I do not take that as an argument to do nothing about the third
order qrm which can be removed by use of better equipment.
> Given the
quality of most of the equipment out there, I
> reckon the phase noise gets to you long before the IMD.
Well, at a frequency separation of 20 kHz you may
be right,
but for IM3 we have to consider signals at larger frequency
separations also.
Here are some numbers:
Rig
noise@100kHz noise@500kHz
(dBc/Hz) (dBc/Hz)
TM255E(rx)
144.5
151.3
TM255E(tx)
125.8
141.7
IC706(rx)
125.8
141.3
IC706(tx)
125.7
134.7
IC970H(rx)
141.7
147.7
IC970H(tx)
132.3
134.1
IC821H(rx)
128.0
131.2
IC821H(tx)
130.7
132.9
FT817(rx)
132.6
135.3
FT817(tx)
130.7
132.9
TS850+TV144(rx)
138.8
148.8
TS850+TV144(tx)
134.1
137.6
FT847(rx)
131.9
146.9
FT847(tx)
130.2
141.1
FT100(rx)
129.3
143.8
FT100(tx)
129.7
124.3
TR9130mod(rx)
147.8
147.8
TR9130mod(tx)
141.7
147.5
If you have a friendly relation to your ham
neighbours, 145dBc/Hz
is perfectly realistic with standard commercial rigs. It is
obvious
that the synthesizer is good enough in several of the above rigs,
the modification needed to bring the tx to the quality of
the rx should be simple enough.
A little better than 145dBc/Hz is not difficult at
all if you are prepared
to build something yourself.
> Finally, I did some tests this
morning, and I think the 3dB noise floor
> degradation might be a convenient but not realistic value to
use. These
> numbers are for my dud old ears - your values might be
different.
> I set up a
> just readable keyed signal on a set (worked out to be -6dB
S/N) and then
> wound up a noise generator until it was unreadable. For my
ears,
> it's a noise
> floor increase of 1.5dB, meaning that the transmitter
signals are
> a further
> 4dB lower than examples above.
Let me suggest you do the test as follows:
First set the signal level so you copy 80% of random text
correctly. Then degrade S/N by 1dB and see what happens to
your error rate. CW speed and message type has to be unchanged.
You may be surprised:)
73
Leif / SM5BSZ
8) G8GSQ Steve writes:
On Tuesday 10 June 2003 15:36, Leif
Åsbrink wrote:
> Hi Steve,
>
> > OK, Leif - I understand and follow your comments. I
come to an
> > answer from my own calculations, and one I always come
back to
> > whenever I revisit them. I'm often wrong, and I'll be
very glad
> > if someone can correct me here.
> >
> > Having written my tome, I'd like to add a preface - I
really hope this
> > doesn't come over as me trying to be a smartarse, or
trying to get into a
> > pissing contest, neither is the case.
>
> Oooh! I know I am an extremist - but I think it is
justified.
Of course - and it makes sure that life is
always interesting.
> > > > ---------------------
> > > > Sorry, but I don't understand where the 20dB
comes from here.
> > >
> > > Oooh! This was a remark trying to explain from
where the -150dBm
> > > comes. (It should have been -150dBm/Hz)
> > >
> > > Sky temp is 290K = -174 dBm/Hz.
> > > Preamp gain is 24 dB so temp at preamp output is
-150dBm/Hz.
> >
> > The reference point usually used is when the noise
floor degrades
> > by 3dB - for the sake of discussion, lets assume that
all of that
> > degradation comes from one external transmitter. The
noise at
> > the preamp output is -147dBm/Hz, that's -150 from the
sky noise
> > and -150 from the transmitter.
>
> OK.
>
> > Allowing for the preamp gain, the noise from the
transmitter
> > is -174dBm/Hz at the preamp input.
> >
> > As you said earlier, a good transmitter will have noise
sidebands of
> > -145dBc/Hz, so a signal level of -29dBm at the preamp
input will
> > cause 3dB noise floor degradation. If we consider the
IMD produced by
> > 2 x -29dBm, then in SSB bandwidth, then -140dBm is 0dB
S/N, which
> > equates to IP3 of +26dBm.
>
> Yes. I agree 100%, but this is at the preamplifier input. At
the
> preamplifier output IP3 should be 24 dB higher and I was
discussing
> the "VHF receiver", the unit we have in the
schack, not the mast
> mounted preamplifier......
Yes - understood, I think we have been
cross purposes to some extent. Like I
said, I'm often wrong......
I think we mostly agree - I'm just lazy in not pursuing
improvements if I
think I'm unlikely to see the benefit in practice. It's been fun
forcing the
grey cells into action.
Cheers,
Steve
9) Graham F5VHX:
I think trying to herd the rig reviewers towards more relevant
testing is a
good thing and this conversation is both interesting and
educational.
It is an unfortunate fact however that little, if any, progress
will be made
with respect to off the shelf equipment. The traditional
designers/suppliers
of amateur equipment, in most cases being aware of their
shortcomings and in
many cases really wishing to improve things, have to operate
within a
business model that just doesn't allow it.
The highly compromised DC to light band 'one size fits all' box
is here to
stay, until ultimately it will die too, as the hobby does. Most
of the
amateur equipment suppliers are already looking for the exit, not
the
stairway to yet more development time/expense.
This hobby is heading back to the future, ie. we are already
seeing the
re-emergence of a new layer of specialty small volume kit/rig
manufacturers
which are pandering to what used to be a 'niche' market (the
folks that know
what they want and understand what's in the box) as this is
rapidly becoming
THE market. I'm not even sure how long they will be able to keep
it up
either because the components supply market is changing as well,
they may
simply not be able to find the parts needed to build the gear and
if they
do, it is becoming increasingly hard to find parts that amateurs
can mount
for a kit format sale.
Everyone else that sustained the critical market mass necessary
for onward
development, at the amateur level, has gone away to play space
invaders,
re-load Windows 4 times a week and all just to fiddle with the
colour of
their desktop, and are spending 100 quid a month on
texting/talking about
nothing on their mobiles instead of talking about nothing on the
bands...
and at the professional level, have shifted to exploit cell
infrastructure
for voice and data transport. Without the private mobile radio or
private
vhf/uhf data trunking backbones, then the suppliers simply cannot
warrant/sustain 'ham' development in ANY form.
I'm not trying to be negative, this is just modern, commercial
fact and
industry trend.
Graham F5VHX
10) Leif, do you have the relevant noise figures for
each of the rigs ?? that
would also be useful to see.
When you have a chance try an IC202.
Graham F5VHX
---
Hi Graham,
> Leif, do you have the relevant noise
figures for each of the rigs ?? that
> would also be useful to see.
Yes (of course) and I have closer frequency
separations,
power spectra in SSB mode (very ugly) as well as IP3.
I am working on the presentation of all of it.
> When you have a chance try an IC202.
Oooh! It is an old friend since many years:)
I have one myself, but it is useless for my purpose,
I have modified it so it is much better than the
original.....
I will most probably collect some
more data at the RS03
meeting in August 21 to 24.
This is a major effort from my side, I do it because I
think new standard procedures are absolutely necessary
to avoid really bad problems caused by the introduction of
computers closer to the antenna. There are all sorts of
problems that could make practical performance unacceptable
for units that would show very good results if they were
tested with today's test procedures. This is at least the
case with QST, I do not have the details about how other
magazines do their tests, but I hope I will get some info
through DUBUS. I am writing a rather large article on the
subject for Dubus:)
At the moment I am not interested in the
performance data
as such. I am looking for consistency. First priority
for the measurement procedure has to be relevance.
Reproducibility has to come second. I just hate the ISO9000
philosophy - it is a tool to make something become identical
every time you do it. Not bad in itself of course, but not
an acceptable excuse to choose a deficient technical solution
because it may be reproduced at a narrower tolerances.
I do not want to stir up too much noise, but on the other hand
I need some assistance.
For example. Is there anyone on this list who knows how
compression is made in today's SSB transmitters? It seems to
me one uses the ALC. Is it really so?
73
Leif / SM5BSZ
11) From: "Chris
Bartram G4DGU" <chris@chris-bartram.co.uk>
Sent: Wednesday, June 11, 2003 3:02 PM
Dear Lief
Although we have some differences of approach, which might cloud
our basic
attitudes, I don't think we're that far apart..
>The effort required to do something about this unfortunate
>situation does
not seem utopic to me:)
>Last weekand I made measurements on several transceivers at
>the
>Scandinavian VHF meeting. It was not unexpected to find that
>most
>units had what I would call obvious design errors causing
poor
>spectral purity. At the moment I spend most of my time trying
>to do
>something about it, the ultimate goal is to have the
measurement
>standards changed in amateur reviews. The reason I came into
>this
>problem is that when the standard tests are performed on DSP
>systems,
>the resulting performance data may become utterly misleading.
>The measurement is reproducible and standardized - but it is
>not relevant. The measured property of the test object is not
>the thing that the measurement is supposed to quantify.
>Example: BDR as measured by ARRL is not blocking dynamic
range as it is
>defined in text. The weakest signal a receiver
>can receive without significant degradation due to a strong
>signal simultaneously present. ARRL just measures blocking
>which is something else (and not relevant).
As you say, Leif, the problem is one of inflexibility. New ways
of doing
things are now available, but the standards by which equipment is
assessed
are firmly rooted in the days of the FT101 and FT221.
My worry is that the improvements to VHF/UHF receiver/transmitter
performance which some of the bigger manufacturers of amateur
radio made in
the late 1980s and 90s are being swept away by the current mania
for
one-size-fits-all solutions, and that some important (to me)
aspects of
performance are conveniently swept under the carpet.
>If you drift by less than 50Hz per minute, your signal is
good enough for
144MHz EME CW. Surely JT44 will not work, but there are several
EME
stations who have much worse stability than this.Stability is
fine (of
course) but not really a problem with modern rigs.
That's not true, Leif. Even for conventional CW the stability of
many modern
rigs seems less than adequately stable for serious operation at
VHF and
above. I find it intensely annoying to have to chase people
around. In my
previous amateur radio life, possibly because my ears and brain
were
younger, I found narrow filters less than useful, and I could
copy EME cw
signals happily in ssb bandwidths. However, since
coming-back on the air,
I've started to find very narrow filters helpful. Last weekend, I
had to
chase one well-known EME station over 100Hz in a 2min over!
Medium term (1
min) stabilities of around a part in 10e8 (1.4Hz on 2m) are
easily available
from an unovened crystal oscillator if it isn't subjected to
constant
thermal cycling.
>I have monitored the spectra of 144MHz stations for 10 years
or so.
The width of the transmitted carrier is never a problem on 144
when
you look at it from the DX point of view. All of this can create
sideband
noise that is a big problem to the near neighbours.
Three years ago, when CW was the only really viable DX mode, that
would have
been true. Transmitter adjacent channel noise needs to be
considered
separately from the close-in spectrum, as the mechanisms are not
quite the
same.
The close-in problem has been highlighted by JT44, and even now,
hasn't been
appreciated by a lot of people. I emphatically don't want to get
into a
stupid CW/JT44 argument. I use, and like, both modes, and I'll
choose the
appropriate mode for the QSO I'm trying to make. I'm not
dogmatic, and feel
that for serious DXing access to all modes is desirable. Hell,
I'll even
work sporadic-E DX on FM if necessary! If other people want to
use other
modes exclusively, that's fine by me, providing that they don't
try to tell
me what to do, or produce spurious technical arguments trying to
prove that
one mode is better than another!
JT44 requires good frequency stability - an order of magnitude
better than
CW. Many modern transceivers are at best marginal in this
respect. In the
course of quite a few hours operating off the Moon using JT44, it
has become
clear that not only do some signals drift a lot less than the
majority, and
thus provide cleaner copy, but even amongst the stations
transmitting stable
signals, some are better than others. The most logical reason for
this is
that the width of the transmitted spectral line, due to some
residual FM, is
approaching, or slightly exceeding the filter (bin) bandwidth
used within
the DSP software.
This leads to questions about the appropriate frequency stability
and
close-in spectral purity required for both current and
medium-term future
modulation modes. Even now, there are improved modulation
techniques being
trialed, and JT44 will soon become obsolete. The limiting factor,
will be
the the characteristics of the propagation medium. Without going
into any
real detail, partly because there's not a lot of information
easily
available, probably the minimum decorrelation bandwidths
typical stations
will see will be over space-earth paths, although Worked All
Solar Planets
is probably a few decades away. The EME path on 2m/50MHz is
probably next
with reported minimum decorrelation bandwidths on 144 of a bit
less than a
Hz. Terrestrial propagation modes often introduce quite severe
decorrelation: just yesterday I saw in excess of 400Hz spreading
on a 2m
iono/tropo backscatter signal.
From a bit of thinking, it's not a great leap to see that we
probably need
to aim for a spectral line which is less than 0.5Hz wide at -3dB
if we are
to be limited by the minimum expected signal path degradation.
Put another
way, there's very little point in going for receiver bandwidths
of less
than about 1Hz. JT44 effectively uses filters of just under 6Hz
BW, and I
suspect that's no accident.
>The power spectra of the SSB signals of all the rigs I tested
were very
ugly indeed.
>I do not think this is because of amplifier linearity. I
think it is
>because there is no speech processing. The peak power is
simply
>limited by the ALC. The ALC amplifier is simply an AM
modulator
>and it is used to amplitude modulate the narrowband SSB
signal
>in such a way that the peak amplitude stays within limit.
>To do this, the "AM modulator" has to be fed with a
signal of
>a rather wide bandwidth - otherwise peaks can not be
compressed!
>The modulation sidebands then extend 25 kHz on either side:(
>I do not know for sure, but as I understood the explanation
of
>the SSB controls, this is the way SSB is done nowadays. I do
>know for sure that this is the way it is done in the FT221. I
also
>know for sure that there is a dramatic improvement easily
>obtained in the FT221 - just by moving a wire one can feed
the
>SSB generator with the amplitude limited audio intended for
the
>FM modulator. This makes the ALC unnecessary and the
improvement
>on signal quality is dramatic.
I think that some of the new transmitters implement RF speech
clipping in
DSP.
> Linearising amateur transmitters could result in much
cleaner
> sounding bands - particularly when signals are strong.
>>I am really very sceptic here. It would be nice with
narrower signals,
>>but it is even more important to avoid noise at larger
frequency
>>separations. With the current state of the art, adding
linearisation
>>circuits (noise generators) will be a bad idea unless the
spectrum
>>puritu is checked at several frequency separations. The
measurements
>>I have done check Tx noise at 5, 10, 20, 100, 200 and 500
kHz.
>>An improvement at 10 and 20 kHz must not degrade the
performance
>>at 100kHz and above.
If you have a gain control system with unlimited bandwidth (OK, I
do know
that you can't make a control system with unlimited bandwidth!!)
it is
indistinguishable from a limiter. All linear transmitters need to
be
designed with some form of power limiting. This shouldn't be the
function of
the PA, and ALC loops around the PA shouldn't be expected to
provide speech
processing. We agree here!! Using an audio limiter followed by a
low-pass
filter isn't a bad solution, although it will result in some
audio
distortion if it's taken too far. There are better techniques.
However,
Leif, the linearity of the amplifier strip driving the antenna is
_very_
important in an ssb transmitter.
We get away with relatively poor linearity in amateur ssb
transmitters
because the power density of speech in the area where the
amplifier
linearity is poor (approaching saturation) is low. As soon as the
power
density in that area is increased, by speech processing, or when
the
transmitter is used as an upconverter for non-constant envelope
data, the
energy in adjacent channels, generated by intermods stemming from
the
non-linearity, grows at alarming rates. Nowadays this gets called
by the
(ugly) name 'spectral regrowth'. As an aside, the extreme
difficulty of
containing spectral regrowth from transmitters using wideband
CDMA
modulation schemes, without compromising efficiency unduly, (or
requiring a
10GW power station in your back pocket!) is one of the many
reasons that
so-called 3G mobile phone systems have been slow to take-off.
Spectral regrowth is a problem - listen to any band in a contest!
I feel
that it's a more serious problem, in practice, than transmitter
noise,
because it becomes a nusiance at possibly only -40dB. There are
techniques
of transmitter linearisation which could be designed-in to modern
amateur
transceivers for a relatively small amount of money. There
shouldn't be a
noise penalty. The 'big three' know all about this, but won't do
anything,
unless they feel that there's a demand. -60dB 3rd orders in a 2m
SSB
transmitter is entirely practicable, without backing-off the
power output.
>> I didn't mean to write this rant when I started! It's
also moved some way
>> from the original topic. However, it's clear that the
bell and whistles
on
>> modern transceivers haven't really led to improved
perfomance. I
certainly
>> wouldn't spend my money on a modern 'all-band'
transceiver if I wanted to
>> use it for serious operation, at vhf/uhf. They
seem strictly for Mr &
Mrs
>> Mouse.
>Yes - but that is because we amateurs do not appreciate the
efforts
>made now and then. A good radio does not get it's qualities
held out
>properly by the reviewers.
One of the problems is that reviewers are writing for the amateur
radio
general public. Given the way in which standards within the hobby
are being
erroded, that public is tending to become even less aware of
technical
issues than it was when I was involved in the amateur radio
business.
>> SM5BSZ's Linrad software and hardware may well be
pointing the way to a
>> better future. It would be hard to better the concept
for it's intended
>> purpose , but it's very application-specific - at the
moment.
>I think this is unfair. Linrad is not primarily for EME -
everybody
>seems to think so because that is from where I come;)
>Linrad should be most useful on crowded HF bands in difficult
interference
>situations. I am not interested myself so it will take some
time before
>HF people discover how to use this tool.
Leif, I don't think I am being unfair. In its current hardware
implementation Linrad is primarily aimed at the CW EME community.
I did say
that it would be hard to better the concept for its intended
purpose, and I
stand by that! What you have done - effectively singlehandedly -
is a
considerable achievement. I believe that the integration of radio
hardware
into PC systems is the way forward for serious operation, and I'm
slowly
moving down that road myself.
>> My inclination would be to design very good single
conversion
transceivers
>> for the individual bands, probably operating under PC
control.
>> and to use IF
>> sampling DSP following a phase-linear crystal filter.
They would only
need
>> to tune in relatively wide frequency steps (like the
DSP-10) with
>> interpolation taking place in the DSP. For a number of
reasons I wouldn't
> >use a zero, or near-zero IF topology.
>A simple solution, but you will have problems if you try to
open up the
>bandwidth. There is no need at all for a phase-linear filter.
It is trivial
>to correct the phase non-linearities in the computer.
The choice of IF bandwidth would be a function of the performance
of the
sampling A/D converter. A nominally phase linear filter would
still need the
phase non-linearities correcting.... My choice of a phase-linear
filter has
more to do with limiting the Q of the crystal resonators in the
transition
regions, and thus the potential voltage across individual
crystals, their
mechanical deformation, and the potential for filter intermods,
than with
phase characteristic.
73
Chris
GW4DGU
12)
From: "Leif Åsbrink SM5BSZ"
Sent: Thursday, June 12, 2003 3:40 AM
Subject: RE: VHFDX: Receiver Linearity - and more
Hi Chris,
> Although we have some differences of approach, which might
cloud our basic
> attitudes, I don't think we're that far apart..
I agree 100%
> As you say, Leif, the problem is one
of inflexibility. New ways of doing
> things are now available, but the standards by which
equipment is assessed
> are firmly rooted in the days of the FT101 and FT221.
Hmm, I would say they are rooted in the vacuum tube era..........
> My worry is that the improvements to
VHF/UHF receiver/transmitter
> performance which some of the bigger manufacturers of
amateur
> radio made in
> the late 1980s and 90s are being swept away by the current
mania for
> one-size-fits-all solutions, and that some important (to me)
aspects of
> performance are conveniently swept under the carpet.
I have the same worry, but there is an additional problem
that is even
worse. Did you see Ulrich Rodhes article in QEX. He advocates
dual
loop AGC. Essentially a S/N controlled attenuator at the antenna
input
which is tailored to make S/N never go above 40dB. Measurements
done at high signal levels will show fantastic performance for a
really
bad receiver which is practically useless. There is also a
problem
when the A/D converter moves closer to the antenna so the A/D
will
see more bandwidth. There is a need for a major revision now!
> > If you drift by less than 50Hz
per minute, your signal is good enough for
> > 144MHz EME CW. Surely JT44 will not work, but there are
several EME
> > stations who have much worse stability than
this.Stability is fine (of
> > course) but not really a problem with modern rigs.
>
> That's not true, Leif. Even for conventional CW the
stability of
> many modern rigs seems less than adequately stable for
serious
> operation at VHF and above. I find it intensely annoying to
have
> to chase people around. In my previous amateur radio life,
possibly
> because my ears and brain were younger, I found narrow
filters
> less than useful, and I could copy EME cw signals happily in
> ssb bandwidths. However, since coming-back on the air,
I've
> started to find very narrow filters helpful. Last weekend, I
had to
> chase one well-known EME station over 100Hz in a 2min over!
Medium term (1
> min) stabilities of around a part in 10e8 (1.4Hz on 2m) are
> easily available from an unovened crystal oscillator if it
isn't
> subjected to constant thermal cycling.
I am not saying I think you are wrong, it is just a matter of
priority. If someone choses to use an unstable transmitter he
makes
his own signal less easy to copy for some stations. All of those
who
do not use narrow filters will have no problem and those who
install
Linrad will have no problem at all because the computer will make
sure the tuning is always optimum. Correcting the unstability
design error is fine - but compared to the silly design errors
that makes it impossible for anyone in the surroundings
to hear anything because of the noise emissions it is not
really a problem:)
> >I have monitored the spectra of
144MHz stations for 10 years or so.
> The width of the transmitted carrier is never a problem on
144 when
> you look at it from the DX point of view. All of this can
create sideband
> noise that is a big problem to the near neighbours.
>
> Three years ago, when CW was the only really viable DX mode,
that would have
> been true. Transmitter adjacent channel noise needs to be
considered
> separately from the close-in spectrum, as the mechanisms are
not quite the same.
Oooh, I was refering to CW only. JT44 needs better
equipment, there
is a large number of EME'ers who do not have the stability
required.
> The close-in problem has been
highlighted by JT44, and even now,
> hasn't been appreciated by a lot of people.
Absolutely. When Joe adds coherent averaging the
requirements will
be much higher - but there is a substantial improvement one can
get out of that!
> JT44 requires good frequency stability
- an order of magnitude better than
> CW. Many modern transceivers are at best marginal in this
respect. In the
> course of quite a few hours operating off the Moon using
JT44, it has become
> clear that not only do some signals drift a lot less than
the majority, and
> thus provide cleaner copy, but even amongst the stations
transmitting stable
> signals, some are better than others. The most logical
reason for this is
> that the width of the transmitted spectral line, due to some
residual FM, is
> approaching, or slightly exceeding the filter (bin)
bandwidth used within
> the DSP software.
Well, I have real data on this. If you have a look at
http://antennspecialisten.se/~sm5bsz/arrl2001/index.htm
you will find recordings over a long time where more than a
hundred 144MHz
stations can be analyzed for spectral width. These are recordings
with
the huge array of SM5FRH so all signals are strong. It is easy to
check
which stations had adequate stability for JT44 in 2001. You are
absolutely
right in that a large number would need to do something to their
rigs to
become really sucessful on JT44.
> This leads to questions about the
appropriate frequency stability and
> close-in spectral purity required for both current and
medium-term future
> modulation modes. Even now, there are improved modulation
techniques being
> trialed, and JT44 will soon become obsolete. The limiting
factor, will be
> the the characteristics of the propagation medium. Without
going into any
> real detail, partly because there's not a lot of information
easily
> available, probably the minimum decorrelation
bandwidths typical stations
> will see will be over space-earth paths, although Worked All
Solar Planets
> is probably a few decades away. The EME path on 2m/50MHz is
probably next
> with reported minimum decorrelation bandwidths on 144 of a
bit less than a
> Hz. Terrestrial propagation modes often introduce quite
severe
> decorrelation: just yesterday I saw in excess of 400Hz
spreading on a 2m
> iono/tropo backscatter signal.
You are absolutely right - but now the subject is moving
away pretty far from
arguments whether better IP3 is a good idea or not;)
> From a bit of thinking, it's not a
great leap to see that we probably need
> to aim for a spectral line which is less than 0.5Hz wide at
-3dB if we are
> to be limited by the minimum expected signal path
degradation. Put another
> way, there's very little point in going for receiver
bandwidths of less
> than about 1Hz. JT44 effectively uses filters of just under
6Hz BW, and I
> suspect that's no accident.
Well, have a look at:
http://antennspecialisten.se/~sm5bsz/arrl2001/sm5frh/sm5frh.htm
here you can see that a bandwidth down to about 0.1Hz is
meaningful
on 144 MHz. It will be possible to decode JT44 at 0.1Hz bandwidth
by coherent processing, the sync tone will be like a carrier
which
one can find in a narrow filter. Once the phase is established,
which
is possible in qsb peaks, one has the info to do coherent
averaging
to dig the keying info out of the noise.
> I think that some of the new
transmitters implement RF speech clipping in
> DSP.
Would be nice if they did.
> All linear transmitters need to be
> designed with some form of power limiting. This shouldn't be
the
> function of
> the PA, and ALC loops around the PA shouldn't be expected to
> provide speech
> processing. We agree here!! Using an audio limiter followed
by a low-pass
> filter isn't a bad solution, although it will result in some
audio
> distortion if it's taken too far. There are better
techniques. However,
> Leif, the linearity of the amplifier strip driving the
antenna is _very_
> important in an ssb transmitter.
Well, I once modified the FT225. This rig, in it's
original shape
sounded like all the rigs I have tested now. Really not good!!!
Just by moving a single wire all the splatter from the
transmitter
could be removed. I am not talking about a few dB improvement,
it was really a big difference. The ALC was used for speech
processing,
the mod was to take the audio from the audio limiter. (the one
for FM)
I know there are many articles written about the superiority of
an RF
clipper over an audio clipper. They are all misleading. The fact
that
there is a big difference when the input is a sinewave does not
mean that
a voice signal is distorted. If the input signal is pulses, there
is no
difference and the voice is much more a pulse train than a
sinewave.
There is no disadvantage of an RF clipper (ALC if you wish) but
it has to
be followed by a filter with SSB bandwidth! Not very practical so
it is
absolutely a reasonable soultion to make the clipping at audio.
> We get away with relatively poor
linearity in amateur ssb transmitters
> because the power density of speech in the area where the
amplifier
> linearity is poor (approaching saturation) is low. As soon
as the power
> density in that area is increased, by speech processing, or
when the
> transmitter is used as an upconverter for non-constant
envelope data, the
> energy in adjacent channels, generated by intermods stemming
from the
> non-linearity, grows at alarming rates.
I am not so sure. Poor linearity is mainly third order but
the interference
seems to me to have a much higher order.
> Nowadays this gets called by the
> (ugly) name 'spectral regrowth'. As an aside, the extreme
difficulty of
> containing spectral regrowth from transmitters using
wideband CDMA
> modulation schemes, without compromising efficiency unduly,
(or
> requiring a
> 10GW power station in your back pocket!) is one of the many
reasons that
> so-called 3G mobile phone systems have been slow to
take-off.
>
> Spectral regrowth is a problem - listen to any band in a
contest! I feel
> that it's a more serious problem, in practice, than
transmitter noise,
> because it becomes a nusiance at possibly only -40dB.
Yes - this is exactly what the spectra of the tested
transceivers show.
I do not think it is caused by poor linearity, I think it is
caused
by a misuse of the ALC. A simple test would be to feed the
microphone input
with a properly processed audio signal and turn the mic volume
for the output
level to become a few tenths of a dB below where ALC starts
acting.
> There are techniques
> of transmitter linearisation which could be designed-in to
modern amateur
> transceivers for a relatively small amount of money. There
shouldn't be a
> noise penalty. The 'big three' know all about this, but
won't do anything,
> unless they feel that there's a demand. -60dB 3rd orders in
a 2m SSB
> transmitter is entirely practicable, without backing-off the
power output.
OK. But I am not so sure there is any need for it. The
fundamental
frequency of the human voice is not high at all. I think third
order
products will essentially fall within the passband. I think SSB
is much easier
than the modern digital modes.
> One of the problems is that reviewers
are writing for the amateur radio
> general public. Given the way in which standards within the
hobby are being
> erroded, that public is tending to become even less aware of
technical
> issues than it was when I was involved in the amateur radio
business.
This is exactly the problem that worries me. Something can
be done
about it and I am prepared to try to do it.
> Leif, I don't think I am being unfair.
In its current hardware
> implementation Linrad is primarily aimed at the CW EME
community.
This is not true at all. Linrad has no hardware. Linrad is a
computer program that runs well with for example the TimeMachine,
a simple HF direct conversion radio that anyone can buy easily.
Linrad can be used to remove interference on HF bands that can
not be removed by any conventional receiver. At the moment
Linrad is good for CW and SSB but you can work any digital
mode that you can work with a standard SSB receiver if you
connect
a second computer to the Linrad loudspeaker output (yes, it is
silly
but sometimes it is the only way to get rid of the interference)
The WSE converters that I work with are not specifically designed
for Linrad. You may equally well use them together with any
other software package. The design is not for 144MHz EME, the
design
is for all amateur bands. I have a prototype for conversion from
HF
to 70 MHz on the bench but since I am alone on all of this it
is a slow process and it will not be available very soon. I hope
to bring the first prototype to the RS03 meeting in August to
have a look at HF spectra of all the transceivers people bring
to the meeting.
> The choice of IF bandwidth would be a
function of the performance of the
> sampling A/D converter. A nominally phase linear filter
would still need the
> phase non-linearities correcting.... My choice of a
phase-linear filter has
> more to do with limiting the Q of the crystal resonators in
the transition
> regions, and thus the potential voltage across individual
crystals, their
> mechanical deformation, and the potential for filter
intermods, than with
> phase characteristic.
OK. I never studied intermodulation in crystal filters,
but I know it
is not a simple thing.
I made the WSE boxes the way they are because I wanted to get
maximum
performance out of the computer. It is not trivial if you sample
at
IF. I did not want to go into LCA programming to implement
sophisticated filters that would be needed to feed the computer
with about 100 kHz bandwidth without oversampling. If I had had
the resources I would have gone for an A/D at 60 to 100 MHz
sampling
speed and some digital data decimation. One could feed anything
into it
with medium performance, but one could convert it to an excellent
high performance receiver just by adding a preamplifier with a
suitable
filter, one for each band of interest. It is beyond my resources,
but
it is the hardware for which Linrad is designed.
Chris, I do not really think we differ in opinion at all once
we talk about the same thing:)
73
Leif / SM5BSZ
13. From: "Leif Aasbrink
SM5BSZ, Sent: Friday, June 13, 2003 2:13 PM
Hi Chris, Steve and All,
> On Thursday 12 June 2003 23:54, Chris Bartram wrote:
> snip
> >Transmitter/receiver performance is very important to
the sort of
> > people who'll be reading this, and many of the modern
boxes really don't
> > have terribly good performance..........
> A very british way of phrasing it. I reckon much of the
amateur
> equipment is
> poor, or just bl**dy awful.
Here I agree with Steve......
I think it is because the ALC is often used for speech
processing.
As I remember from the FT225RD where it was possible to change
easily. Sending a suitably processed signal intended for FM into
the SSB generator generated splatter levels that were equal to
the CW noise sideband levels. In the original shape, with the ALC
as speech processor this rig had an ugly spectrum indeed:(
The measurement is done the following way: The Linrad screen is
used
as a spectrum analyzer. The SSB signal peak level is recorded
from
the A/S saturation margin and/or the Linrad S-meter peak level
reading.
The level at which the entire spectrum is 20 kHz wide is recorded
and
it is converted to dBc/Hz by a comparision with the CW carrier
measurements. Finally the peak-to-average power is extracted from
the S-meter readings and subtracted. For an ideal SSB
transmitter,
the values should be equal to the CW values. In CW the same
transceivers
range from 100 to 138 dBc/Hz (also at 10kHz)
Except for my IC202 which is modified for CW only (I do not use
SSB)
the excess splatter ranges from 39dB (TM255E and TR9130) to 4 dB
(FT817) This is really not a fair comparison at all because
interference
is no better if this ratio is good because of poor phase noise
performance
but the comparison is interesting because it illustrates what the
measurement means.
I have started to process the data from Gavelstad. It is at:
http://antennspecialisten.se/~sm5bsz/dynrange/gavelstad/gav.htm
Chris wrote:
> I'd like to very breifly amplify my reasons for believing
that
> linearisation of amateur ssb transmitters would be useful.
Snip.
I do not agree at all. Not because what you say is incorrect in
any way
but because the standard of today's transmitters is far below the
level where modern technology like this will be needed.
It would be like a quantum leap if manufacturers corrected
obvious
design errors - and I do not think it would cost anything at all
in most cases.
Then, in the distant future, linearisation would allow a second
major improvement, but introducing it now to produce better
two-tone tests in QST would not improve the situation on the
bands at all unless something is done about the sideband noise.
Many rigs have a "noise generator" inserted between the
SSB
generator and the antenna:(
There is a large unbalance between rx and tx performance as you
will see from the Gavelstad site when all the data is in place.
73,
Leif / SM5BSZ
14. From: "Steve Thompson" <g8gsq@qsl.net> Sent: Saturday,
June 14, 2003 9:02 AM
Hi Chris, Steve and All,
> On Thursday 12 June 2003 23:54, Chris Bartram wrote:
> snip
> >Transmitter/receiver performance is very important to
the sort of
> > people who'll be reading this, and many of the modern
boxes really don't
> > have terribly good performance..........
> A very british way of phrasing it. I reckon much of the
amateur
> equipment is
> poor, or just bl**dy awful.
Here I agree with Steve......
I think it is because the ALC is often used for speech
processing.
As I remember from the FT225RD where it was possible to change
easily. Sending a suitably processed signal intended for FM into
the SSB generator generated splatter levels that were equal to
the CW noise sideband levels. In the original shape, with the ALC
as speech processor this rig had an ugly spectrum indeed:(
-------------------
I've handled relatively few rigs, but I'ne not seen one which
deliberately
uses ALC for speech compression. I have, on the other hand, seen
lots of
very poor ALC systems. ALC shuts the stable door after the horse
has gone,
and it's misused by both manufacturers and operators. Problems
I've seen:
All 'normal' ALC systems allow the transmitter to be overdriven
for as long
as it takes them to act, producing splatter. The overdrive and
distortion
can be great if the manufacturer is too lazy to properly control
the
transmitter gain, and relies on ALC to reduce it. Where a gain
control is
provided, operators are reluctant to reduce it until the ALC
doesn't
operate - they might lose 1% output!
ALC applied badly in the wrong place can cause the controlled
stage to
become non-linear, increasing rather than decreasing distortion.
In some sets (FT990 for example, I think) the ALC is used for
output power
control. If the output is turned down to, say, 20W then the ALC
reference is
reduced but the transmit strip gain is not and the PA is still
overdriven to
150W until the ALC operates. If you've turned the power down to
suit the
drive level of an amplifier, the result can be horrible, if not
destructive.
Metering is important. It's often said that the TS830 is cleaner
with speech
processing on. It isn't - it's just that the meter is really slow
to respond
so the short peaks in normal speech don't give any ALC reading
until the
drive is turned up a long way so the overdrive into the PA is
very high. If
you set the ALC to just operate with a prolonged yell, then speak
normally,
you still get full peak output but the signal is really clean.
With
processing on, the peaks are reduced and made longer duration, so
they show
on the ALC meter (and it's still cleaner than most).
Steve
See:
SM 5 BSZ - RX Dynamic range (and TX)
measurements at Gavelstad June 2003
(June 13 2003) http://antennspecialisten.se/~sm5bsz/dynrange/gavelstad/gav.htm
