FT-990 vs TS-850
Uri Blumenthal
> Ken..I own the 850..The reliability is better on the 990...
I beg to differ! I'm quite pleased with the reliability of
my FT-990 <knock-knock on the wood! :-> 'cause so far it
didn't give me a single problem, nor to anybody of my
friends, who seeing my 990 chose to buy similar rig
for themselves.
> The scaf filter is no great schucks...
??? I use/like this filter A LOT, on all manual modes.
> Difficult decision..
> But Id trade my 850 for a 990 in a minute because of overall
> quality. (also if 990 uses 24 volt finals..TX is better)
(:-)
--
Regards,
Uri. u...@watson.ibm.com scifi!angmar!uri
------------
<Disclaimer>
Fred Lloyd [Phoenix SE]
>
>I've narrowed my choice of a new rig to either the Yaesu FT-990 or the
>Kenwood TS-850. My interests right now are both ragchewing and DXing using
>both CW and SSB.
>
>I've read the reviews, etc., and would like any input you have about the
>performance of these rigs. If you have had a chance to use both of these
>rigs, a comparison would be great!
>
As many on this newsgroup know, I'm a big TS-850 fan and have owned
more than one - my first being one of the first to arrive into the
local store.
Last weekend, at Phoenix's largest annual swap meet, I saw about
three or four TS-850's for sale, all for around $1250.00 each.
But, I didn't see any FT-990's for sale... Hmmmmm.
I still like my 850, but lot'sa folks really like the 990.
-fred
[ Fred Lloyd, AA7BQ Fred....@west.sun.com ]
[ Sun Microsystems, Systems Engineer ]
[ Phoenix, AZ (602) 224-3517 ]
hamilton on BIX
>I don't think one should pick a radio just on looks. I have an
>850S-AT with the optional 1.8 SSB filter and enjoy a lot of low band
>work (160 and 75m) I tried a friend's Yeasu 890AT for two weeks,
>side by side, using a Daiwa switch in reverse and did a lot of
>*listening* (as opposed to *seeing*) switching rapidly from one
>radio to the next. The Kenwood 850 is a lot "noisier", i.e., the
>human voice is not clearly separated from the background noise,
>unless you cut back considerably on the RF gain. With the Yeasu, it
>is seldom necessary to play with the RF gain.
>A friend has both the Yeasu 757 and a Kenwood 440AT and he confirms
>the same findings. The Yeasu is more pleasant to listen to and most
>of us do a lot more listening than talking. My next rig is likely to
>be a Yeasu.
>The above impression is about SSB reception only. I have not tried
>anything else really. ...
I have an FT-990 that I bought about 2 months ago after a *lot* of
time spent soliciting opinions, reading reviews (if you're thinking
of buying a radio and haven't yet bought the two volumes of collected
QST reviews that the ARRL offers, you should!), and sitting in front
of the units I was considering at the local Ham Radio Outlet.
I'm _very_ pleased with the 990. I haven't tried transmitting anything
yet (I just took and passed my Novice, Tech, General and Advanced tests
about 5 weeks ago and am still waiting for my ticket) but have spent
a lot of time listening. I'll confirm that the the 990 is a very
pleasant radio to listen to. On CW, it's terrific. The built-in
audio filters allow you to zero right in on a signal.
Regards,
Doug Hamilton hami...@bix.com Ph 508-358-5715
Hamilton Laboratories, 13 Old Farm Road, Wayland, MA 01778-3117
Kok Chen
>In article <2m4sff$4...@doc.cc.utexas.edu>, kre...@doc.cc.utexas.edu (Bob Nagy) writes:
>> Ken..I own the 850..The reliability is better on the 990...
>I beg to differ! I'm quite pleased with the reliability of
>my FT-990 <knock-knock on the wood! :-> 'cause so far it
>didn't give me a single problem, nor to anybody of my
>friends, who seeing my 990 chose to buy similar rig
>for themselves.
Don't get too excited :-), you both said the same thing!
I also run an FT-990, for over a year now, I think. Not a single
problem so far.
I use the Yaesu side by side with an Ten-Tec Omni V and actually
prefer the FT-990 over the Omni for all modes except CW. If you
know the Omni, that is saying a lot. Very sweet rig.
73,
Kok Chen, AA6TY kc...@apple.com
Apple Computer, Inc.
NADO000
>>didn't give me a single problem, nor to anybody of my
>>friends, who seeing my 990 chose to buy similar rig
I don't think one should pick a radio just on looks. I have an
850S-AT with the optional 1.8 SSB filter and enjoy a lot of low band
work (160 and 75m) I tried a friend's Yeasu 890AT for two weeks,
side by side, using a Daiwa switch in reverse and did a lot of
*listening* (as opposed to *seeing*) switching rapidly from one
radio to the next. The Kenwood 850 is a lot "noisier", i.e., the
human voice is not clearly separated from the background noise,
unless you cut back considerably on the RF gain. With the Yeasu, it
is seldom necessary to play with the RF gain.
A friend has both the Yeasu 757 and a Kenwood 440AT and he confirms
the same findings. The Yeasu is more pleasant to listen to and most
of us do a lot more listening than talking. My next rig is likely to
be a Yeasu.
The above impression is about SSB reception only. I have not tried
anything else really. To be fair, the 850SAT is a little better at
avoiding strong QRM, but in practice this was not very important,
through the 2 weeks I tried both rigs. The 850SAT is also a quicker
rig to utilize and its tuner is faster. My 2 year-old rig has not
had a single problem since I bought it new.
Luis Nadeau VE9LN
>.
Kok Chen
>In article <2mfkd6$8...@apple.com> kc...@apple.com (Kok Chen) writes:
>>>didn't give me a single problem, nor to anybody of my
>>>friends, who seeing my 990 chose to buy similar rig
> ^^^^^^^^^^
>I don't think one should pick a radio just on looks.
QRZ, QRZ. That ain't me who said that!
I also own a Ten-Tec rig. Have you ever seen a Ten-Tec rig that
looks good? :-) :-). Moreover, I drive a Saab and I get taunted
all the time for uglifying the parking lot.
Vote for best looking rig: the National HRO-500.
Scott Turner
Referring to the choice between a Yaesu FT990 and a Kenwood TS850:
: A friend has both the Yeasu 757 and a Kenwood 440AT and he confirms
: the same findings. The Yeasu is more pleasant to listen to and most
: of us do a lot more listening than talking. My next rig is likely to
: be a Yeasu.
This seems to be a popular pair of radios. There've been several
postings over the last months asking for comparisons between these rigs.
I've also narrowed my choices down to these two if (when?) I ever
replace my Drake rigs. One thing that almost never seems to get
mentioned however...
How do these rigs sound on the air! At the risk of starting a flame
war, I'll venture the opinion that, in general, I find Kenwoods putting
out among the best sounding signals on the air. I've got to admit that
I enjoy hearing the reports of superior audio quality that my Drakes
seem to get on sideband, and I'd hate to give that up. I like putting
out a good pleasing signal! If it weren't for that factor, I'd probably
lean towards the Yaesu. As it is, I'm not sure.
Any thoughts or comments on the transmitting audio qualities of the
major brands on sideband, particularly these two radios?
Scott Turner N0VRF sc...@hpisla.LVLD.HP.COM
David Stockton
: In article <19MAR94.18...@UNBVM1.CSD.UNB.CA>, NADO000 <NA...@UNB.CA> writes:
: > >>didn't give me a single problem, nor to anybody of my
: > >>friends, who seeing my 990 chose to buy similar rig
: > I don't think one should pick a radio just on looks.
: Come on, don't you really understand, that "seeing" in the
: context meant they tested it in all the modes they wanted?
: If it was a joke - I apologize.
: (:-).
Hmm this takes me back a couple of years, the first TS850 I saw was
on a stand at a rally. There was a large crowd around it, fighting to
spin its knobs and poke its buttons. It seemed rather futile, really, as
none of the stands had any electrical power, and the set was
consequently unpowered. They must have really liked the feel of those
knobs because some people were offering to buy it there and then,
untried. One individual in my hearing offered the stall holder a premium
over the list price just to get that one, said to be the first in the
area. To fully understand the meaning of this, readers in the US must
note that in the UK, list price is strictly selling price, although you
might get them to throw in an accessory worth a few percent of the deal.
UK list price is about double the US prices seen in QST.
I was somewhat stunned by this, and have watched for the effect
recurring, which it did at another rally when the TS50 showed up.
I have witnessed people buying radios by appearance !
I just can't decide whether it is a joke or not, although P T Barnum
would have found it funny....
Cheers
David GM4ZNX
Jay Kesterson K0GU
: How do these rigs sound on the air! At the risk of starting a flame
: war, I'll venture the opinion that, in general, I find Kenwoods putting
: out among the best sounding signals on the air.
As I remember from the 850 review in QST sometime back they didn't like
the 850 at near full output on SSB. They claimed it got a bit dirty (wide)
I think. Maybe someone with the article handy can give the exact quote.
73, Jay K0GU ja...@fc.hp.com
Ignacy Misztal
1. Passband characteristics of the AF chain (including the microphone),
2. IMD of the PA,
3. Type of processing, audio or RF?
I am wondering why the QST reviews do not mention the type of processing,
which has a large effect on signal quality. Signals with audio processing
have higher content of AF harmonics, and are subsequently less efficient
(3db?) and more difficult to tune. All cheaper rigs such as IC 725-737,
FT 747-757, TS 430-450 use AF processing. IC 751-, FT990-, TS 830-
use RF processing. I am not sure about the rest.
Ignacy Misztal, NO9E, SP8FWB
University of Illinois
ign...@uiuc.edu
Ignacy Misztal
>ign...@ux2.cso.uiuc.edu (Ignacy Misztal) writes:
>>It seems that the SSB quality depends on 3 factors:
>>1. Passband characteristics of the AF chain (including the microphone),
>>2. IMD of the PA,
>>3. Type of processing, audio or RF?
>You are forgetting probably the most important factor, the voice characteristics
>of the operator. I recently sold a radio to a friend of mine. During the
>five years I had it, I consistantly got unsolicited comments on how GOOD
>my audio sounded. My friend is always getting complaints on his.
>I have heard him on a TS520, TS430, and now the Yaesu FT980 (not 990)
>His voice sounds like, well like it is being compressed.
>I have a resonating, full "FM" sounding voice. He doesn't.
>Voices are like faces, everyone would rather look at a pretty face
>than an ugly one (unfortunately, his face is prettier than mine hi hi),
>and when you hear a good-sounding voice, you ususally will compliment it.
>I know the purist (real ones and those who think they are) will take exception
>to this, but most commercially available radios have similar TX charactersitics.
>Yes, RF processors are superior than most AF ones, but in the 32 years I have
>been hamming, I have found that voice characteristics determine who gets
>unsolicited compliments.
>my $.02 worth
>Mike WA5CMI
>Extra Class since 1973
>5-band WAS, WAZ, DXCC (303 cfmd)
>Ragchewer
>Electrical Engineer, CONVEX Computer Corp.
>The opinions are mine alone, not my employer's
I agree but the reply is not quite on the topic. Surely a good driver
with Geo Metro may be faster than a poor driver with Ferrari, but it does
not mean that the GEO is better.
Ignacy, NO9E
Bill Turner
<cowart.764364068@neptune>
Organization: Eskimo North (206) For-Ever
In article <cowart.764364068@neptune>,
changes locations and starts getting poor audio complaints,
one might want to be sure there is no "RF in the mike" syndrome
present. Major RF levels will be obvious, but low levels
might not. SSB is especially sneaky since the distortion
only occurs on voice peaks, not the whole signal.
73 es gl
Bill Turner, W7LZP
w...@eskimo.com
Ignacy Misztal
>harmonics. Aren't there digital signal processing algorithms that
>could prevent this effect? Even before DSP, didn't people use split
>band audio processing to reduce the content of harmonics?
>>--
>Zack Lau KH6CP/1 2 way QRP WAS
> 8 States on 10 GHz
>Internet: zl...@arrl.org 10 grids on 2304 MHz
Cheap AF processors use AF clippers. DSP-based processors are not only
novelties now, but they are more xepnsive to built than RF processors.
Why AF clippers are worse than RF (IF) clippers? Consider a 500Hz
tone test. With AF processor you will get extra 1000,1500,2000,2500
Hz tones. With RF (SSB and DSB) processor 500Hz will be the only
output. Please note that some older rigs have "implicit" RF
processors. For instance, SWAN 500 has 7360, a beam deflection tube,
as a DSB modulator. By clipping peaks, it acts with the following XTAL
filter as a DSP processor.
--------------------------------------------------------------------------
Ignacy Misztal Ham radio: NO9E, SP8FWB
Internet: ign...@uiuc.edu Bitnet: ign...@uiucvmd.bitnet
University Of Illinois 1207 W. Gregory Dr., Urbana, IL 61801, USA
Zack Lau (KH6CP)
:
: Cheap AF processors use AF clippers. DSP-based processors are not only
: Why AF clippers are worse than RF (IF) clippers? Consider a 500Hz
: tone test. With AF processor you will get extra 1000,1500,2000,2500
: Hz tones. With RF (SSB and DSB) processor 500Hz will be the only
: output. Please note that some older rigs have "implicit" RF
: processors. For instance, SWAN 500 has 7360, a beam deflection tube,
: as a DSB modulator. By clipping peaks, it acts with the following XTAL
: filter as a DSP processor.
If you clip an ideal DSB waveform (1 kHz modulation), aren't there
two tones spaced 1 kHz apart that could generate IMD products at
1.5 kHz and 1.5 kHz (receiver output)? What if you had a significant
amount of carrier leakthrough that was cleaned up by the crystal
filter? Couldn't this give you extra tones at 1, 1.5, 2, and 2.5
kHz (at the receiver)?
Alan Bloom
: Ignacy Misztal (ign...@ux2.cso.uiuc.edu) wrote:
: :
: : .... Please note that some older rigs have "implicit" RF
: : processors. For instance, SWAN 500 has 7360, a beam deflection tube,
: : as a DSB modulator. By clipping peaks, it acts with the following XTAL
: : filter as a DSP processor.
: If you clip an ideal DSB waveform (1 kHz modulation), aren't there
: two tones spaced 1 kHz apart that could generate IMD products at
: 1.5 kHz and 1.5 kHz (receiver output)? ...
Clipping a full-carrier DSB (AM) signal is equivalent to clipping the
positive peaks of the baseband audio. I believe that symmetrical
clipping of a suppressed-carrier DSB signal is equivalent to clipping
both + and - peaks of the baseband audio. If you want to get the full
benefits of RF speech clipping, you have to do it on the SSB signal.
AL N1AL
Gary Coffman
>Gary Coffman (ga...@ke4zv.atl.ga.us) wrote:
>
>: Phfffft! The phase flatness through the audio phase shift networks
>: used in amateur phasing SSB rigs was much worse than any phase
>: distortion in a filter rig. The audio phasing network had to cover
>: octaves while the crystal filter only has to work over a tiny fraction
>: of an octave.
>
>This is *wrong*
>
>SSB crystal filters are designed for steep skirts for good
>shape factors. This means that without any equalizing networks
>(which normally double the complexity and send the cost through
>the roof), the phase response at the passband edges are *terrible*
>The fact that the center frequency of the crystal filter is much
>higher just means that the Q of the parts has to be that much
>better. The mathematics of the phase and amplitude response
>tradeoffs are unchanged-- the tradeoffs are identical for a
>3 kHz audio filter and a 3 kHz SSB filter (assuming ideal
>parts--with real parts its easier at audio...)
Apples and oranges. The phasing SSB exciter is using an audio
*phase shift network*, the filter exciter is using a RF filter.
Now the AF phasing network may be considered a sort of filter,
but that's not it's designed purpose, and for sure it's not a
3 kHz bandpass response. Instead it has to maintain a constant
90 degree phase shift across multiple octaves. That's tougher.
Gary
--
Gary Coffman KE4ZV | You make it, | gatech!wa4mei!ke4zv!gary
Destructive Testing Systems | we break it. | uunet!rsiatl!ke4zv!gary
534 Shannon Way | Guaranteed! | emory!kd4nc!ke4zv!gary
Lawrenceville, GA 30244 | |
Alan Bloom
sound better than filter-type generators because phasing exciters have
flatter amplitude and delay response. Gary Coffman disputed that. Rather
than respond to Gary's long replies in detail, I'll just summarize how
phasing-type SSB exciters work:
_______ _____________ I ________
Audio | Audio | | Phase-shift |----------- Mixer ->| |
Input --| Filter|-->| Network | Q | | Summer |--> SSB Out
|_______| |_____________|--- Mixer --------->|________|
| |
_______________ | ____|____
| RF Oscillator |--+->| +90 deg |
|_______________| |_________|
I and Q are two audio outputs with a constant phase difference between
them of 90 degrees. The input filter limits the audio frequency response
to the range of the phase-shift network. The "+90 deg" box can be switched
to -90 degrees to get the opposite sideband. (The output of each mixer is
a DSB signal.)
The audio phase shift network is the interesting (read difficult) part
of the system. It must maintain a 90 degree phase difference and
excellent amplitude matching between the two outputs over something like
a 10:1 frequency range (300 Hz - 3000 Hz). It generally does that by
causing each of the two outputs to have a constantly-rising phase shift
versus frequency characteristic, like thus:
/ /
Phase Shift / /
/ /
720 deg / /
/ /
/ /
540 deg / /
/ /
/ /
360 deg / /
I / /
/ / Q
180 deg / /
/ /
/ /
0 deg ___/_/
| | | | |
300 600 1200 2400 4800 Hz
Frequency
I may have gotten the scaling off a bit, but the principle is right:
Both channels have constantly-changing phase shifts, but the difference
is always 90 degrees. Note that the frequency scale is logarithmic.
If phase were linear with frequency, then that would equal constant
delay. Since that's not true, there is some variation in group delay
with frequency, but it is a nice smooth curve that has little affect
on audio quality. (As opposed to the crystal filter used in a filter-
type SSB generator which has "bumpy" group delay, expecially at the
high and low band edges.)
The design of the two channels' phase-shift networks is such that
any errors in linearity occur in different places. That means that
you can't make it work properly unless both channels have nice
linear phase versus log(frequency). The same goes for amplitude.
I suppose you could design a diabolical phase-shift network that
had unflat (but matched) frequency response in the two channels,
but why would you do that?
The input audio filter can also add to amplitude or delay distortion.
However, it's not hard to design the filter to minimize the problem.
You don't need the sharp cutoff of a crystal filter designed for
receiving applications because any spurious below 30 or 40 dB down
will be covered up by the transmitter power amplifier's splatter
anyway. Also, audio filters are easier to build accurately than
crystal filters because of the lower Q and lower frequency.
The conclusion: Phasing-type SSB generators have flatter group delay
and amplitude than filter-type generators. You really can hear the
difference in the on-the air signal, in my experience.
Gary Coffman (ga...@ke4zv.atl.ga.us) wrote:
: In article <1994Mar26....@arrl.org> zl...@arrl.org (Zack Lau (KH6CP)) writes:
: >SSB crystal filters are designed for steep skirts for good
: >shape factors. This means that without any equalizing networks
: >(which normally double the complexity and send the cost through
: >the roof), the phase response at the passband edges are *terrible*
: >The fact that the center frequency of the crystal filter is much
: >higher just means that the Q of the parts has to be that much
: >better. The mathematics of the phase and amplitude response
: >tradeoffs are unchanged-- the tradeoffs are identical for a
: >3 kHz audio filter and a 3 kHz SSB filter (assuming ideal
: >parts--with real parts its easier at audio...)
: Apples and oranges. The phasing SSB exciter is using an audio
: *phase shift network*, the filter exciter is using a RF filter.
I think Zack was referring to the input audio filter.
: Now the AF phasing network may be considered a sort of filter,
: but that's not it's designed purpose, and for sure it's not a
: 3 kHz bandpass response. Instead it has to maintain a constant
: 90 degree phase shift across multiple octaves. That's tougher.
But the hard part is getting the ampltude and phase matching
to within a fraction of a dB or degree. As explained above, if you
do that, the overall amplitude and delay response versus frequency
will be quite good.
AL N1AL
Richard Karlquist
Phasing type SSB exciters produce a higher fidelity output
than filter type exciters *using filters of the type typically
found in ham equipment*. If you get high quality filters like
the Lumda FDM telecom filters, you get do just as well with
a filter type exciter. These filters will run you about $125 each.
In any event, if the receiver is a transceiver, and it uses
the same filter for receive and transmit, then all the nasty
ripples you avoided with a phasing type transmitter will
be reintroduced at the receiver. So you really need a phasing
transmitter and phasing receiver to get "hi-fi" audio. Or
use Lumda filters at both ends.
(Lumda is a small outfit that took over the FDM xtal filter
market when the big boys pulled out after T1 replaced FDM for
99% of the telecom market. FDM is still used for phone calls
to Alaska and Hawaii. It's SSB but doesn't sound like ham SSB.)
Rick Karlquist N6RK
rka...@scd.hp.com
Zack Lau (KH6CP)
: In article <1994Mar26....@arrl.org> zl...@arrl.org (Zack Lau (KH6CP)) writes:
: >SSB crystal filters are designed for steep skirts for good
: >shape factors. This means that without any equalizing networks
: >(which normally double the complexity and send the cost through
: >the roof), the phase response at the passband edges are *terrible*
: >The fact that the center frequency of the crystal filter is much
: >higher just means that the Q of the parts has to be that much
: >better. The mathematics of the phase and amplitude response
: >tradeoffs are unchanged-- the tradeoffs are identical for a
: >3 kHz audio filter and a 3 kHz SSB filter (assuming ideal
: >parts--with real parts its easier at audio...)
: Apples and oranges. The phasing SSB exciter is using an audio
: *phase shift network*, the filter exciter is using a RF filter.
: Now the AF phasing network may be considered a sort of filter,
: but that's not it's designed purpose, and for sure it's not a
: 3 kHz bandpass response. Instead it has to maintain a constant
: 90 degree phase shift across multiple octaves. That's tougher.
Actually, what I was writing about was Gary's misconception that
phase distortion is somehow much easier to deal with if you
move the center frequency higher. Its actually tougher--just try
and build a crystal frequency with good phase characteristics
and a good shape factor. (Or, try and buy one...) Of course, it
is true that you need an audio filter for a phasing exciter, as
there are limits to how broad you can make the phase shift network.
Fortunately, there is no requirement to transmit 60 Hz hum with
perfect fidelity.
I would agree that it isn't necessary for a phasing rig to have low
phase and amplitude distortion--I'm sure that someone could work
*really* hard and come up with one that sounded awful and still
managed to reject the opposite sideband.
The dark side of DSP? :-)
But, in practice, the easiest way to make one to work well is
to just go ahead and design for low distortion.
FWIW, one of the fanatical AM types showed off his phasing
receiver at Deerfield NH a few years ago... Guess he didn't
notice the distortion Gary is worried about. Come to think
of it, I don't recall hearing complaints about the Sony
2010's audio quality, which also uses audio phase shift
networks. (go through the archives of the shortwave newsgroup?)
Tom Bruhns
(concerning the generation of quadrature phase audio signals)
: But the hard part is getting the ampltude and phase matching
: to within a fraction of a dB or degree. As explained above, if you
: do that, the overall amplitude and delay response versus frequency
: will be quite good.
So, just how good is a practical network?
In some recent ARRL pubs (e.g., ARRL Handbook, 1987, pg 18-9) there's
a "matrix" type of phase shift network that claims to be able to give
good performance with loose-tolerance parts. I put this network into
Spice last nite, and thought some might be interested in the
results. Below are four columns plus a frequency column. The
second column is phase of one output channel. The third is that
phase with a constant 235 degrees and a ramp at -.066 degrees/Hz taken out.
The fourth is the magnitude frequency response for that channel.
The fifth is the magnitude frequency response with an additional
5-pole Butterworth low-pass at 3kHz and 5-pole Butterworth high-pass
at 320Hz, to limit the audio passband to the 300Hz-3kHz voice band.
(I hope tabs in the table won't make it unreadable at too many sites.)
phase filtered
freq phase error mag mag
Hz degrees degrees dB dB
200 -193.5 54.6 2.2 -8.4
230 -203.8 46.4 1.9 -6.1
264 -214.0 38.4 1.6 -4.0
303 -224.4 30.6 1.3 -2.3
348 -234.9 23.1 1.0 -1.2
400 -245.6 15.8 0.7 -.5
459 -256.5 8.8 .5 -.1
527 -267.6 2.3 .3 0
606 -278.8 -3.8 .2 0
696 -290.2 -9.2 .1 0
800 -301.6 -13.8 0 0
919 -313.1 -17.4 0 -.1
1056 -324.5 -19.9 0 -.1
1213 -335.9 -20.9 0 0
1393 -347.1 -20.2 .1 0
1600 -358.4 -17.8 .2 0
1838 -369.5 -13.2 .4 0
2111 -380.4 -6.1 .6 -.1
2425 -391.1 3.9 .8 -.5
2786 -401.7 17.2 1.1 -1.2
3200 -411.8 34.3 1.3 -2.4
3676 -421.5 56.1 1.6 -4.1
Note: I didn't calc the phase error including the Butterworth
filters, which are NOT linear phase, but better than the Chebyschev
typically approximated in a crystal filter.
The exact meaning of the "phase error" column is certainly open
to discussion, but to me it's obvious that it's a very low error,
and quite smooth with frequency. And both the magnitude
columns show extremely good flatness in the voice band:
particularly the one including the out-of-band audio filters.
I didn't include columns about the phase and amplitude difference
between the channels, but it's a very close match in both.
(I hope to follow this with another followup about actually
achieving 90 degree phase differentials with a finite number of
poles and zeros, but that will have to wait for now.)
John Welch
> In another thread, I claimed that phasing-type single-sideband generators
> sound better than filter-type generators because phasing exciters have
> flatter amplitude and delay response. Gary Coffman disputed that. Rather
> than respond to Gary's long replies in detail, I'll just summarize how
> phasing-type SSB exciters work:
> _______ _____________ I ________
> Audio | Audio | | Phase-shift |----------- Mixer ->| |
> Input --| Filter|-->| Network | Q | | Summer |--> SSB Out
> |_______| |_____________|--- Mixer --------->|________|
> | |
> _______________ | ____|____
> | RF Oscillator |--+->| +90 deg |
> |_______________| |_________|
>
> I and Q are two audio outputs with a constant phase difference between
> them of 90 degrees. The input filter limits the audio frequency response
> to the range of the phase-shift network. The "+90 deg" box can be switched
> to -90 degrees to get the opposite sideband. (The output of each mixer is
> a DSB signal.)
>
> The audio phase shift network is the interesting (read difficult) part
> of the system. It must maintain a 90 degree phase difference and
> excellent amplitude matching between the two outputs over something like
> a 10:1 frequency range (300 Hz - 3000 Hz). It generally does that by
> causing each of the two outputs to have a constantly-rising phase shift
> versus frequency characteristic, like thus:
<much deleted>
So, the audio phase shift is the only 'interesting' part...
How, pray tell, can one having only the usual ham test gear (scope,
probably, dmm, maybe power supply) make the RF phase shift be 90
degrees and the same amplitude at, say 12MHz?
My dual-trace scope is not perfectly calibrated, so that's
out. Generating 48MHz and using flip-flops to get 12MHz in quadrature
doesn't work well in reality (theory is great, but unless your 48MHz
signal is *exactly* 50% duty cycle it has a *strong*component at just
under half of 48MHz, usually near 22MHz. Flip-flops, like all
non-linear devices, are very good as mixers, and the 22MHz mixes with
the 12MHz to make some *interesting* spurs. Filtering out these spurs
usually trashes the 90 degree and equal amplitude you got in the first
place, leaving you back at square one.)(yes, i do know about this.
i've tried it. repeatedly. i KNOW phasing sounds better, and i
WANTED it to work *sigh*). Generate it in quadrature with a dual DDS
and two DACs? Then you must filter the DACs outputs through two
different filters, introducing slightly different phase and amplitude
errors.
I wanted it to work. Really. I've listened to DC receivers
and to crystal filtered super-hets and the difference is amazing.
However, I also want to be able to build a receiver and actually *use*
it, not spend eternity designing the 'perfect' one.
One other interesting thing re: phasing vs filtering: you'll
need *some* additional filtering to do a transmitter anyway (for SSB
at least). This filter will cost you $$, and if you already have to
spend the $$ why not use it for the receiver too? That rather neatly
explains why darned few commercial ham rigs use phasing any more.
--
While (its_not_working()) John Welch, N9JZW
mess_with_it(); j...@seastar.org
Gary Coffman
>Gary Coffman (ga...@ke4zv.atl.ga.us) wrote:
>
>: Apples and oranges. The phasing SSB exciter is using an audio
>: *phase shift network*, the filter exciter is using a RF filter.
>: Now the AF phasing network may be considered a sort of filter,
>: but that's not it's designed purpose, and for sure it's not a
>: 3 kHz bandpass response. Instead it has to maintain a constant
>: 90 degree phase shift across multiple octaves. That's tougher.
>
>Actually, what I was writing about was Gary's misconception that
>phase distortion is somehow much easier to deal with if you
>move the center frequency higher. Its actually tougher--just try
>and build a crystal frequency with good phase characteristics
>and a good shape factor. (Or, try and buy one...) Of course, it
>is true that you need an audio filter for a phasing exciter, as
>there are limits to how broad you can make the phase shift network.
>Fortunately, there is no requirement to transmit 60 Hz hum with
>perfect fidelity.
>
>I would agree that it isn't necessary for a phasing rig to have low
>phase and amplitude distortion--I'm sure that someone could work
>*really* hard and come up with one that sounded awful and still
>managed to reject the opposite sideband.
> The dark side of DSP? :-)
>But, in practice, the easiest way to make one to work well is
>to just go ahead and design for low distortion.
Well if you look at the table Tom posted, you'll see that even a
matrix network audio phase shifter (published in the ARRL Handbook)
has lousy phase response at the edges too, and a simple first order
network is worse. The Dome networks and B&W networks used in older
designs were even worse. Now compare that to the phase response of
a Collins mechanical filter. Except at the *edges* it's phase response
is flatter. And as you noted, we can cut off the edges with a pre-filter
in either case. Bill Orr notes in his Radio Handbook that while 60 db
opposite sideband rejection is easy with a filter, it's difficult to do
as well as 40 db with a phasing network because of balance problems,
especially near the edges, which shows up as *distortion product aliases
in the passband*.
>FWIW, one of the fanatical AM types showed off his phasing
>receiver at Deerfield NH a few years ago... Guess he didn't
>notice the distortion Gary is worried about. Come to think
>of it, I don't recall hearing complaints about the Sony
>2010's audio quality, which also uses audio phase shift
>networks. (go through the archives of the shortwave newsgroup?)
Better still consult the Hi Fi magazines. The Sony 2010, and
a few other AM receivers, have been *panned* for their poor
implementation of synchronous detection. Differential phase
distortion is a hot topic with the high end folks now, probably
because they've licked almost all the other problems. In
rec.radio.shortwave the 2010 was panned because it's synchronous
detector isn't really synchronous. It's actually a form of ISB
instead of correlating upper and lower sidebands as a true sync
detector does.
Gary Coffman
Now this chart illustrates the problem I've been talking about. As
we can see, the difference in delay with frequency is quite marked.
Sure the phase delay increases *smoothly* with frequency delta, but
the magnitude of the error rapidly climbs with increasing frequency
delta. This is our old friend click-boom. From 600 Hz to 2400 Hz delay
decreases by almost 20%, or about 0.2 ms. That much delay difference
is clearly audible.
If we look at Tom Bruhns' chart for a more complex matrix network:
phase filtered
>freq phase error mag mag
>Hz degrees degrees dB dB
>200 -193.5 54.6 2.2 -8.4
>400 -245.6 15.8 0.7 -.5
>527 -267.6 2.3 .3 0
>606 -278.8 -3.8 .2 0
>696 -290.2 -9.2 .1 0
>800 -301.6 -13.8 0 0
>1056 -324.5 -19.9 0 -.1
>1213 -335.9 -20.9 0 0
>1600 -358.4 -17.8 .2 0
>2111 -380.4 -6.1 .6 -.1
>2425 -391.1 3.9 .8 -.5
>2786 -401.7 17.2 1.1 -1.2
>3676 -421.5 56.1 1.6 -4.1
Now this is much better. The ends are horrible of course, but in the
region 600-2400 Hz there is only a delay delta of 0.014 ms. That's
hardly audible at all to someone with *good* ears. I'd note that this
matrix phase shift network is considerably more complex than typical
networks found in older phasing type equipment. And as Richard Karlquist
noted, that's all for naught anyway unless the IF filtering of the
receiver has correspondingly flat phase delay. Good RF filters can be
as flat away from their edges as the matrix above, and RF filters work
both ways.
Zack Lau (KH6CP)
:
: >FWIW, one of the fanatical AM types showed off his phasing
: >receiver at Deerfield NH a few years ago... Guess he didn't
: >notice the distortion Gary is worried about. Come to think
: >of it, I don't recall hearing complaints about the Sony
: >2010's audio quality, which also uses audio phase shift
: >networks. (go through the archives of the shortwave newsgroup?)
: Better still consult the Hi Fi magazines. The Sony 2010, and
: a few other AM receivers, have been *panned* for their poor
: implementation of synchronous detection. Differential phase
: distortion is a hot topic with the high end folks now, probably
: because they've licked almost all the other problems. In
: rec.radio.shortwave the 2010 was panned because it's synchronous
: detector isn't really synchronous. It's actually a form of ISB
: instead of correlating upper and lower sidebands as a true sync
: detector does.
So what?
The point is, does the audio phase shift networks used in the 2010
cause a noticeable degradation in audio quality as perceived by
the users of the radio? And, since we are primarily talking
about SSB, as opposed to AM, there is *no* benefit to having
a detector that can correlate the upper and lower sidebands--we
only have one sideband to work with on receive.
Gary Coffman
> <much deleted>
>
> So, the audio phase shift is the only 'interesting' part...
>How, pray tell, can one having only the usual ham test gear (scope,
>probably, dmm, maybe power supply) make the RF phase shift be 90
>degrees and the same amplitude at, say 12MHz?
This is the easy part. There are three main ways to do it. The most
straight forward way is a quarterwave transmission line, distributed
or lumped. A second way is to use two loosely coupled LC tanks arranged
so they're in quadrature phase. A third way is to use a RL lag network
in one leg and a RC lead network in the other. Each will give 45 degrees
of phase shift when R=L=C with a net shift between the two of 90 degrees.
Since the RF is a pure tone at one frequency, getting an accurate phase
shift is fairly trivial.
Alan Bloom
: zl...@arrl.org (Zack Lau (KH6CP)) writes:
: >If you clip an ideal DSB waveform (1 kHz modulation), aren't there
: >two tones spaced 1 kHz apart that could generate IMD products at
: >1.5 kHz and 1.5 kHz (receiver output)? What if you had a significant
: >amount of carrier leakthrough that was cleaned up by the crystal
: >filter? Couldn't this give you extra tones at 1, 1.5, 2, and 2.5
: >kHz (at the receiver)?
: Wait... I am completely confused by Zack's arguments.
: A DSB signal that has a 1 kc modulation consists of two "carriers"
: spaced 2 kc apart, not 1 kc, no? (Imagine AM with 1 kc modulation.
: Now take away the carrier.)
Right. If you assume Zack meant .5 kHz modulation, then it makes sense:
The two sidebands are carrier + and - .5 kHz and the third-order
IMD products are carrier + and - 1.5 kHz. So the recovered audio has
a third harmonic, but no second harmonic. So symmetrical clipping of
a DSB waveform is the same as symmetrical clipping of the baseband
audio in that it generates odd-order products, but not even-order.
AL N1AL
David Hough
of 1800Hz carriers which are 90 degrees out of phase, and fairly easy to
generate a couple of 10.7MHz carriers which are 90 degrees out of phase, and
the rest is reasonably straightforward without any expensive bits. SBL1 mixers
are cheap, so the fact that you need four shouldn't be prohibitive.
Dave
--
*****************************************************************************
* G4WRW @ GB7WRW.#41.GBR.EU AX25 * Start at the beginning. Go on *
* da...@llondel.demon.co.uk Internet * until the end. Then stop. *
* g4...@g4wrw.ampr.org Amprnet * (the king to the white rabbit) *
*****************************************************************************
Alan Bloom
: So, the audio phase shift is the only 'interesting' part...
: How, pray tell, can one having only the usual ham test gear (scope,
: probably, dmm, maybe power supply) make the RF phase shift be 90
: degrees and the same amplitude at, say 12MHz?
Easy: Just feed in an audio tone to the mic input and adjust the
RF amplitude and phase until the unwanted sideband disappears.
: One other interesting thing re: phasing vs filtering: you'll
: need *some* additional filtering to do a transmitter anyway (for SSB
: at least).
Not really, unless you want to do RF clipping. And in that case, you
still have saved one of the two required crystal filters.
: This filter will cost you $$, and if you already have to
: spend the $$ why not use it for the receiver too? That rather neatly
: explains why darned few commercial ham rigs use phasing any more.
Or turn it the other way around: Since the receiver portion of a
transceiver already has a crystal filter, why not use it for the
transmitter as well? I think that's the reason you don't find
phasing-type SSB generators in commercial transceivers.
AL N1AL
Richard Karlquist
> My dual-trace scope is not perfectly calibrated, so that's
>out. Generating 48MHz and using flip-flops to get 12MHz in quadrature
>doesn't work well in reality (theory is great, but unless your 48MHz
>signal is *exactly* 50% duty cycle it has a *strong*component at just
>under half of 48MHz, usually near 22MHz. Flip-flops, like all
No, you get a spur at exactly 24 MHz, the second harmonic. If you
drive the mixer differentially from Q and Qbar, I believe this
second harmonic will cancel out for all practical purposes.
Many popular mixers (SBL-1 etc) have floating LO windings that make
this easy to do.
Also, the propagation delay doesn't have to be zero, it only has to
be identical between the I and Q outputs, which if they are flip
flops on the same chip will be very close (easily under a nanosecond
for FACT logic, 100 psec. for ECLIPSE logic.)
The tough part isn't the audio 90 degree shift or the RF 90 degree shift,
it's getting amplitude and phase matched mixers.
Rick Karlquist N6RK
rka...@scd.hp.com
Alan Bloom
: delta. This is our old friend click-boom. ...
Other people besided Gary may be confused by this, so I'll post an
explanation.
The graph above plots phase, not delay. A constant delay results in
a constantly-rising phase plot. For example, a 1 millisecond delay
is 36 degrees at 100 Hz, 360 degrees at 1000 Hz, 3600 degrees at
10,000 Hz, etc.
Constant delay does not cause the "click-boom" that Gary mentioned.
Think about it: You could record a voice on a tape recorder and play
it back DAYS later without any "click boom." It's only when the
delay is different at different frequencies (that is, the phase versus
frequency plot is not a straight line) that there is a potential problem.
While the plot above looks like a straight line, it really isn't because
of the logarithmic x-axis. However, as the chart that Tom Bruhns posted of
a typical phase-shift network shows, it really isn't too bad. His chart
shows that between 400 and 2786 Hz, the maximum phase error from a straight
line varies smoothly between +17.2 to -20.9 degrees, which is far better
than you would get with a typical transceiver-type crystal filter.
I expect most of the delay variation would come from the audio filter
(that comes before the phase-shift network). Such a filter can be
much flatter than a typical crystal filter for the three reasons I
mentioned in a previous posting: (1) Doesn't need as sharp a cutoff
as a receiver-type filter, (2) Can use a filter type with inherently
flatter delay (non-Chebyshev), and (3) Pole Q and placement are much
easier to control at audio frequencies than at RF.
AL N1AL
Tom Bruhns
: In another thread, I claimed that phasing-type single-sideband generators
: sound better than filter-type generators because phasing exciters have
: flatter amplitude and delay response.
Yesterday, I posted results of a simulation of a quadrature audio network
outlined in the ARRL Handbook. One thing to note is that the phase ramp
is as Alan suggests in his posting. The "frequency" column is in 1/5
octave steps, and the delta phase between entries is practically the same
across the band. So the phase looks like a straight line on a log-
frequency scale.
But what if it was a design goal to come up with a quadrature phase
network with flat frequency response and linear phase? Would that be
possible? I think so, and I offer comments below in support of that
idea. If I don't get to it, maybe someone else can check out the
suggestion at the end to see if it really can work.
Soapbox: I'm treating this as a design problem, not something to wave
my arms about. I start by summarizing some things that are likely to
be relevant that could be tools to lead to a good solution. There is
an assumption that flat response and linear phase will lead to something
that people will think "sounds good." That may be wrong, but so far there
seems to be good agreement about it. My purpose here is to simply
explore ways to achieve this in a phasing system. This does NOT say that
the same thing can't be done with filtering at carrier frequencies.
Observations:
0. The relative phase and amplitude response of a system can be
completely determined by the positions of the poles and zeros
representing the system; there is no need to commit to a
particular physical implementation until the desired pole and
zero positions are determined.
1. If a pole at -x+jy is balanced by a zero at +x+jy, the frequency
response from that pair is exactly flat. The phase response,
d(phi)/df, is exactly twice as much for such a pair as it is
for the pole alone, at every frequency.
2. Techniques exist for putting poles at any interesting place on
the left half of the s-plane, and for putting zeros at any
interesting place on the s-plane. For example, for audio work,
state-variable filter blocks can do this.
3. If one channel of a quadrature audio network has a phase ramp
vs frequency d(phi)/df = x, then the other channel must have
the same d(phi)/df over the band of interest to maintain a
constand phase difference between the channels. Since d(phi)/df
determines group delay, the two channels will have identical
group delays.
4. Linear phase is characterized by d(phi)/df invariant with frequency.
5. A strict delay has zero phase shift at zero frequency. Thus to
achieve quadrature phase, at least one channel will be characterized
by a phase shift in addition to any delay, even if d(phi)/df is
invariant over the band of interest.
6. None of the above precludes having a constant d(phi)/df in each
channel with a constant 90 degree phase difference between channels
and a flat frequency response in the band of interest. However, it
also doesn't tell you how to achieve this--it only provides some
hints.
7. A maximally-flat-delay (MFD) filter achieves a very constant d(phi)/df
over its passband.
8. A MFD _lowpass_ filter has nearly constant d(phi)/df through zero
frequency, and zero phase shift at zero frequency. It represents a
simple delay.
9. A MFD _bandpass_ filter has nearly constant d(phi)/df over its
passband, but not outside. The result can be a phase shift relative
to a MFD lowpass in the shared passband. THIS IS A KEY TO A POSSIBLE
ANSWER TO THE ORIGINAL QUESTION!
10. It _should_ be possible to find a pair of MFD filters with overlapped
passbands (either two bandpass filters or a lowpass and a bandpass)
which have equal d(phi)/df and a 90 (or 45--see paragraph 11) degree
phase difference between the channels in the shared passband.
Actually finding such a pair is the exercise remaining to be done!
Any volunteers?
11. MFD filters are notorious for non-flat amplitude response. This can
be taken care of by invoking paragraph 1: put a zero in for each pole,
which will double d(phi)/df and make the frequency response dead flat.
This is why we only need 45 degrees between the MFD filters used as
prototypes as suggested in paragraph 10.
12. Expect that it will take fairly high order filters to accomplish this
over the 300Hz-3kHz band. Tenth order wouldn't surprise me, to get
0.1 degree matching. The amplitude matching will depend only on
how accurately the implementation can place the poles and zeros at
the desired positions.
As a simple feasibility check, I asked a curvefitter to fit to a couple
linear phase ramps with constant amplitude, offset by 90 degrees, in the
range from 300Hz to 3.5kHz, and it didn't have any trouble getting
within a tenth dB and under a degree, with about 14 poles and 14 zeros
for each. None of the poles was particularly high Q: max about 8.
Zack Lau (KH6CP)
: :
: : Cheap AF processors use AF clippers. DSP-based processors are not only
: : novelties now, but they are more xepnsive to built than RF processors.
: : Why AF clippers are worse than RF (IF) clippers? Consider a 500Hz
: : tone test. With AF processor you will get extra 1000,1500,2000,2500
: : Hz tones. With RF (SSB and DSB) processor 500Hz will be the only
: : output. Please note that some older rigs have "implicit" RF
: : processors. For instance, SWAN 500 has 7360, a beam deflection tube,
: : as a DSB modulator. By clipping peaks, it acts with the following XTAL
: : filter as a DSP processor.
: :
: If you clip an ideal DSB waveform (1 kHz modulation), aren't there
Oops, this should be 500 Hz modulation, not 1 KHz modulation in the
parenthesis.
: two tones spaced 1 kHz apart that could generate IMD products at
: 1.5 kHz and 1.5 kHz (receiver output)? What if you had a significant
: amount of carrier leakthrough that was cleaned up by the crystal
: filter? Couldn't this give you extra tones at 1, 1.5, 2, and 2.5
: kHz (at the receiver)?
What this means is that having two crystal filters, one before and
after the clipper, results in less distortion on your transmitted
signal. The extra filter can also be useful on receive, since IF
amplifiers typically reintroduce noise on the unwanted sideband
detected by most product detectors. It isn't difficult to obtain
matched sets of crystal filters, if you have enough $$$.
Gary Coffman
>
>The tough part isn't the audio 90 degree shift or the RF 90 degree shift,
>it's getting amplitude and phase matched mixers.
That's easy, 7360s are almost ideal devices for this.
Gary Coffman
>Gary Coffman (ga...@ke4zv.atl.ga.us) wrote:
>: Now this chart illustrates the problem I've been talking about. As
>: we can see, the difference in delay with frequency is quite marked.
>: Sure the phase delay increases *smoothly* with frequency delta, but
>: the magnitude of the error rapidly climbs with increasing frequency
>: delta. This is our old friend click-boom. ...
>
>Other people besided Gary may be confused by this, so I'll post an
>explanation.
I'm not confused. I calculated the delays based on the graph you posted.
Delay equals the reciprocal of frequency times the total phase delay in
degrees divided by 360.
>The graph above plots phase, not delay. A constant delay results in
>a constantly-rising phase plot. For example, a 1 millisecond delay
>is 36 degrees at 100 Hz, 360 degrees at 1000 Hz, 3600 degrees at
>10,000 Hz, etc.
Yeah, but that isn't what your graph showed.
>While the plot above looks like a straight line, it really isn't because
>of the logarithmic x-axis.
Bingo! I check plotted it on semi-log paper then replotted on a linear
graph from which I calculated my differential delay numbers.
>However, as the chart that Tom Bruhns posted of
>a typical phase-shift network shows, it really isn't too bad. His chart
>shows that between 400 and 2786 Hz, the maximum phase error from a straight
>line varies smoothly between +17.2 to -20.9 degrees, which is far better
>than you would get with a typical transceiver-type crystal filter.
As I commented, his table looked much better than your graph, and I
calculated differential delays based on it too that were nearly 10
times smaller. It's just that I've seen phase plots for crystal and
mechanical filters which, *away from the edges*, where Tom's table
looked bad too, looked at least as good as Tom's phase shifter. And
note also that the network Tom modelled is not "typical", it is
considerably more complex than the traditional Dome based networks.
They tend to really suck in the differential delay department because
they're based on the same simple semi-log response as your graph. I
don't think he included an AF pre-filter in the table either.
I'll readily agree that the receiver type filters used in many ham
rigs for SSB transmit signal generation, suck wind. But that's a
different issue. Good filter designs are available, as witness
filters we use for VSB video, and in certain telco FDM equipment
that have a maximally flat phase response in the passband. We simply
can't tolerate differential delay in video systems, yet we use filters
instead of phasing to generate VSB signals. It's not a matter of economy,
it's what works best.
John Seboldt
>In any event, if the receiver is a transceiver, and it uses
>the same filter for receive and transmit, then all the nasty
>ripples you avoided with a phasing type transmitter will
>be reintroduced at the receiver. So you really need a phasing
>transmitter and phasing receiver to get "hi-fi" audio. Or
Like the Campbell R2 receiver, Jan. 1993 QST! For whatever combination of
reasons, it IS a clean sounding unit. I'll bet his companion phasing SSB
exciter sounds just as good.
John K0JD
Alan Bloom
: Why not use a Weaver (Third Method) exciter? It is easy to generate a couple
: generate a couple of 10.7MHz carriers which are 90 degrees out of phase, and
: the rest is reasonably straightforward without any expensive bits. SBL1 mixers
: are cheap, so the fact that you need four shouldn't be prohibitive.
For some reason, the "third method" of SSB generation invented by Weaver has
never caught on. Perhaps part of the reason is the fact that the suppressed
carrier comes out right in the middle of the audio passband. Even with
40 dB of carrier suppression (typical with diode balanced mixers), people
might find it objectionable because of the AGC action of typical SSB
receivers. (Which would make the carrier "pop up" during speech pauses.)
I have often thought, though, that the Weaver method would be well-suited
to implementation in a DSP, since you can get mathematically perfect
carrier suppression.
AL N1AL
David Hough
>For some reason, the "third method" of SSB generation invented by Weaver has
>never caught on. Perhaps part of the reason is the fact that the suppressed
>carrier comes out right in the middle of the audio passband. Even with
>40 dB of carrier suppression (typical with diode balanced mixers), people
>might find it objectionable because of the AGC action of typical SSB
>receivers. (Which would make the carrier "pop up" during speech pauses.)
>
own passband, instead of clobbering adjacent channels like the other
methods do. Hmmmmm.
>I have often thought, though, that the Weaver method would be well-suited
>to implementation in a DSP, since you can get mathematically perfect
>carrier suppression.
>
Not sure if it is the most efficient way though, and you are limited in your
output frequency - or were you only thinking of doing the audio stages in the
DSP and converting the 'mixed' output back to analogue for injection into
the RF mixers?
Gary Coffman
>never caught on. Perhaps part of the reason is the fact that the suppressed
>carrier comes out right in the middle of the audio passband. Even with
>40 dB of carrier suppression (typical with diode balanced mixers), people
>might find it objectionable because of the AGC action of typical SSB
>receivers. (Which would make the carrier "pop up" during speech pauses.)
>
>I have often thought, though, that the Weaver method would be well-suited
>to implementation in a DSP, since you can get mathematically perfect
>carrier suppression.
Better still, make *use* of the carrier to implement ACSSB compandoring.
Do the final notching at the receiver. Naturally audio quality would
suffer with a notch at 1800 Hz, but that won't hurt communications
effectiveness much.
Tom Bruhns
: Now this is much better. The ends are horrible of course, but in the
: region 600-2400 Hz there is only a delay delta of 0.014 ms. That's
: hardly audible at all to someone with *good* ears.
For some points in my table, I get the following approximate group
delays:
Freq, Hz Group Delay, ms
200 1.09
400 0.57
800 0.31
1600 0.15
3200 0.07
Dunno how Gary got 14 microseconds; you'd have to have better-than-
golden ears to hear that, from reports I've seen. Anyway, the numbers
in the table above are from
group-delay (seconds) = d(phase [radians])/d(frequecy [radians/sec])
As expected, they go essentially inversely with frequency.
: I'd note that this
: matrix phase shift network is considerably more complex than typical
: networks found in older phasing type equipment. And as Richard Karlquist
I recall having the values used in the B&W phase shift network around
somewhere, but couldn't find them. I wanted to put that into Spice
originally, cuz it would have been a lot simpler than that "matrix"
network. Can someone supply the values? I'd be happy to run them
for comparison.
David Stockton
: Why not use a Weaver (Third Method) exciter? It is easy to generate a couple
: generate a couple of 10.7MHz carriers which are 90 degrees out of phase, and
: the rest is reasonably straightforward without any expensive bits. SBL1 mixers
: are cheap, so the fact that you need four shouldn't be prohibitive.
: Dave
: --
This avoids the need for broadband (multi-octave) phase shifters but
still leaves the need for precise amplitude matching to get accurate
cancellation of the unwanted sideband. The required amplitude and phase
matching to get comparable suppression to a reasonable quality filter
exciter are both severe. You can adjust to get best cancellation, but
this still needs it to be stable and for all frequencies to cancel at
the same position of the adjuster.
An attractive compromise is to use a phasing source (polyphase
network, weaver or whatever) to get modest suppression of the unwanted
sideband, the clipper section of the RF speech processor, and finally a
wide-ish lower than usual filter. We get the sum of the suppression
factors of the two systems, the transmitted audio has benefitted from
passing through a much lower Q filter than would be needed by a simple
filter type exciter.
I think this debate is nearing its best-before date, ADCs to digitise
speech are widely available and cheap. DSP devices capable of
implementing an SSB modulator with "RF" speech processor are available,
but still a bit pricy yet. DACs to give an IF output with plenty of
dynamic range are also available and getting cheaper, especially if a
low IF is used. A complete system has the promise of being cheaper for
manufacturers than a single crystal filter, and will also handle lots of
other modes.
Remember how VFOs were dropped the moment synthesisers became cheaper
than a dial and gearbox ? and how only a few people seemed to care
that those synthesisers were so dirty ? With a bit of luck this
change might be done better.....
David GM4ZNX
Jon Bloom (KE3Z)
: I have often thought, though, that the Weaver method would be well-suited
: to implementation in a DSP, since you can get mathematically perfect
: carrier suppression.
See "A Weaver Method SSB Modulator Using DSP," September, 1993 QEX,
by Carlos M. Puig, KJ6ST, and "A Simple SSB Receiver Using a Digital
Down Converter," March, 1994 QEX, by Peter Traneus Anderson, KC1HR
for examples of DSP-based Weaver-method SSB generation and detection.
--
Jon Bloom KE3Z jbl...@arrl.org
Alan Bloom
> In article <CnrLx...@srgenprp.sr.hp.com> al...@sr.hp.com (Alan Bloom) writes:
> >I have often thought, though, that the Weaver method would be well-suited
> >to implementation in a DSP, since you can get mathematically perfect
> >carrier suppression.
> Not sure if it is the most efficient way though, and you are limited in your
> output frequency - or were you only thinking of doing the audio stages in the
> DSP and converting the 'mixed' output back to analogue for injection into
> the RF mixers?
I figured you could do it at a low "RF" frequency of a few 10's of kHz
and then heterodyne it up to the radio's IF frequency.
David Stockton (dst...@hpqmoca.sqf.hp.com) wrote:
: David Hough (da...@llondel.demon.co.uk) wrote:
: : Why not use a Weaver (Third Method) exciter? ...
: This avoids the need for broadband (multi-octave) phase shifters but
: still leaves the need for precise amplitude matching to get accurate
: cancellation of the unwanted sideband. The required amplitude and phase
: matching to get comparable suppression to a reasonable quality filter
: exciter are both severe. You can adjust to get best cancellation, but
: this still needs it to be stable and for all frequencies to cancel at
: the same position of the adjuster.
Except that the "unwanted sideband" does not fall into an adjacent
channel as it does with the "first" and "second" methods of SSB generation.
Since the unwanted sideband folds back into the wanted sideband, poor
rejection shows up as audio distortion, rather than adjacent-channel
interference. A Weaver SSB generator should achieve around 40 dB
suppression which amounts to around 1% distortion. I think that's
rather better than most SSB rigs currently achieve.
: An attractive compromise is to use a phasing source (polyphase
: network, weaver or whatever) to get modest suppression of the unwanted
: sideband, the clipper section of the RF speech processor, and finally a
: wide-ish lower than usual filter. We get the sum of the suppression
: factors of the two systems,
... less the amount of clipping used.
If the phasing generator has 30 dB suppression, there is 20 dB of clipping,
and the post-clipping filter has 30 dB suppression, you get a total of
40 dB suppression, not 60. However I agree it shouldn't be hard to get
such a system to be plenty good enough.
AL N1AL
Wayne Covington
> As a simple feasibility check, I asked a curvefitter to fit to a couple
> linear phase ramps with constant amplitude, offset by 90 degrees, in the
> range from 300Hz to 3.5kHz, and it didn't have any trouble getting
> within a tenth dB and under a degree, with about 14 poles and 14 zeros
> for each. None of the poles was particularly high Q: max about 8.
Another interesting case is to start with a conventional elliptic function
bandpass response, then proceed to the two networks with flat group delay
and 90 degree phase difference, keeping the nice elliptic magnitude response.
The finite jw-axis zeros may well wreak havoc -- with the number of poles
and zeros (for the same overall tolerances on amplitude and phase errors as
you have above) increasing significantly.
Wayne
KD0EA
nb: As you mentioned, nowadays the right way to go would be DSP.
Tom Bruhns
: Another interesting case is to start with a conventional elliptic function
: bandpass response, then proceed to the two networks with flat group delay
: and 90 degree phase difference, keeping the nice elliptic magnitude response.
: The finite jw-axis zeros may well wreak havoc -- with the number of poles
: and zeros (for the same overall tolerances on amplitude and phase errors as
: you have above) increasing significantly.
If you look at this a little differently, it's easy to see that the number
of poles & zeros shouldn't be significantly affected. Come up with a
pair of filters for quadrature phase that you are happy with for
amplitude and phase matching. Add the same zeros and/or poles to
both. Then the amplitude and phase matching will be unchanged. However,
it should be easier to put the frequency shaping outside the quadrature
phase network, since it can then be guaranteed to be identical for both
channels. Leave the quadrature network all-pass; if you wish, shape its
absolute phase to compensate the frequency-shaping filter. At least, that
is how I'd approach it if I were constrained to do it analog.
Robert J. Kelley
>I have often thought, though, that the Weaver method would be well-suited
>to implementation in a DSP, since you can get mathematically perfect
>carrier suppression.
>AL N1AL
Isn't this only true (mathematically perfect carrier suppression) if you
happen to use perfect "brick wall" filters after the two Weaver mixers?
Practically speaking, very good Hilbert transformers for the phasing method
can be implemented with DSP's nowadays. My question is, which of the two
methods uses the least amount of DSP power for the same level of performance?
Wayne Covington
>Wayne Covington (wa...@fc.hp.com) wrote:
I think the key phrase is "shape its absolute phase to compensate the
frequency-shaping filter." I didn't explain the situation I had in mind
very well. Let me try again.
Suppose the system has been realized with a conventional all-pole bandpass
filter such as Chebychev or Butterworth for the amplitude shaping, followed
by all-pass networks to flatten the system's group delay and get the
90-degree phase difference. The 90-degree phase difference and flatness of
group delay are just within certain tolerances.
Now you decide to improve the amplitude response (better shape factor) by
changing the bandpass filter to the elliptic version, with the same number
of poles but additional jw-axis zeros. You try to readjust the all-pass
networks to restore the flat group delay and the 90-degree phase difference
to within the original tolerances.
My conjecture is that this cannot be done without adding more all-pass
pole-zero pairs. If the group delay is within tolerance, the 90-degree
phase difference isn't, or vice-versa.
Wayne
Alan Bloom
: al...@sr.hp.com (Alan Bloom) writes:
: >I have often thought, though, that the Weaver method would be well-suited
: >to implementation in a DSP, since you can get mathematically perfect
: >carrier suppression.
: >AL N1AL
: Isn't this only true (mathematically perfect carrier suppression) if you
: happen to use perfect "brick wall" filters after the two Weaver mixers?
No, the carrier suppression is perfect (except for round-off error). The
imperfect filtering would, however, affect the unwanted sideband suppression.
: Practically speaking, very good Hilbert transformers for the phasing method
: can be implemented with DSP's nowadays. My question is, which of the two
: methods uses the least amount of DSP power for the same level of performance?
The Weaver method only requires a couple multiply operations per sample
to generate the audio 90 degree phase shift. I'm not familiar with
Hilbert transform approximation algorithms, but I gotta believe they
are more complicated than that.
AL N1AL
Alan Bloom
: to...@lsid.hp.com wrote:
: Suppose the system has been realized with a conventional all-pole bandpass
: filter such as Chebychev or Butterworth for the amplitude shaping, followed
: by all-pass networks to flatten the system's group delay and get the
: 90-degree phase difference. The 90-degree phase difference and flatness of
: group delay are just within certain tolerances.
: Now you decide to improve the amplitude response (better shape factor) by
: changing the bandpass filter to the elliptic version, with the same number
: of poles but additional jw-axis zeros. You try to readjust the all-pass
: networks to restore the flat group delay and the 90-degree phase difference
: to within the original tolerances.
: My conjecture is that this cannot be done without adding more all-pass
: pole-zero pairs. If the group delay is within tolerance, the 90-degree
: phase difference isn't, or vice-versa.
So long as the additional filtering is done to both channels identically,
the phase and amplitude matching between the two channels is not afffected.
AL N1AL
Wayne Covington
> the phase and amplitude matching between the two channels is not afffected.
> AL N1AL
Umm... true. But the overall group delay is affected. There is an offline
discussion going on on this. The essence of it is that the finite zeros out
of the passband by themselves don't affect the phase (certainly true) but
the somewhat different pole positions have an effect of to-be-determined
significance. Stand by.
Wayne