THE FACTS ABOUT eSSB AND IMD
THE FACTS ABOUT eSSB AND IMD
In various mediums (on the air, forums, reflectors, etc.), the
subject of eSSB comes up periodically. You often see the same
people posting the same tired words about how it's wasteful, rude,
sounds bad, etc., etc., etc... However, there are a few out there
that claim that eSSB creates terrible IMD Products that can be
heard up and down the bands. I will not go into what odd-order IMD
products are and how they are created here -- that's beyond the
scope of this. Suffice to say that they are products created when
signals are amplified, and are worse when the amplifier (especially
a non-Class A amp) is driven into compression. For our purposes,
when you begin to get ALC deflection, you are approaching that
level and IMD products shoot up rapidly. Also, bear in mind that
rigs are designed differently, and ALC is implemented in different
ways. Also keep in mind that rigs can be mis-aligned with terrible
results (i.e. twisting adjustments to get 150W out of 100W rigs).
Using poorly designed amplifiers (like the type often found on the
CB band) with improper bias or poor bias regulation also severely
degrades IMD performance. And all too common these days,
over-driving and over-compression are ways to make your neighbors
on the bands unhappy. While it is true that a wider signal (of ANY
type) is also wider IMD-wise, it can be shown that properly done,
eSSB is no more offensive on the bands than standard SSB. There are
a few things to consider: How is this measured? Under what
conditions? What drive level? What output level? There are about as
many opinions about this as there are people to measure it! I
borrowed an HP 8591E spectrum analyzer for my experimentation. It
is said that a picture is worth a thousand words, and that's the
beauty of a spectrum analyzer! You can infer that you are fairly
clean by using an oscilloscope, an RF sampler, and a detector, but
if you really want to know what's going on, you can't beat a
spectrum analyzer! Another way is to use a second receiver
(preferably the high-quality commercial grade type) and tune
off-channel. And by "off-channel", I mean away from the occupied BW
of the intended signal, but not so far away that you're beyond the
IMD products. A few kHz should do it. Also, knowing the exact BW of
the receiver is imperative. It must be selective enough to reject
the main envelope, and still be able to listen the intended BW
effectively. You want to put the receiver's BW next to the signal
being produced but not overlap the BW, like THIS. Ideally, you
would have things set up so the signal under test is attenuated
enough to not swamp the test receiver. If the input to the test
receiver can be set so that the S-Meter is about an S9, things
should be okay. Now, it is well known that S-Meters are quite
inaccurate. The purpose is not to get absolute values, but an idea
of how the transmitter is doing. Also, it is a fair representation
of a real world scenario (minus typical band noise, which would
mask much of the IMD products anyway, unless signals are extremely
strong between the transmitter & receiver) and something many
Amateurs would be able to repeat themselves. John-NU9N has done
some experimentation like this (it can be found HERE). He also has
an MP3 of it. It's worth repeating, though: The BEST WAY to measure
this is a spectrum analyzer!
Now to describe what I have done... I took a quick look at my
rig typical of how I operate it (about 80-90W PEP output). I
thought, "Wow! This is actually better than I thought!". Probably
because of all the rhetoric I've been hearing about this. So I
decided to run the tests at the rig's max power level, and even
mis-tuned the ALC to allow the rig to put out more power than it is
supposed to be capable of. The TS-950SDX is specified as a 150W
rig, which it does easily on all bands. The final amplifier uses a
pair of MRF-150 TMOS FETs, which are actually capable of 150W
EACH!! As an example of this, the Ameritron ALS-600 uses four (4)
of these for 600W PEP output. Therefore, a pair at 150W (or even
more, as I've found) should be pretty clean. I was able to get in
excess of 200W out of this rig by disabling the ALC. Don't worry, I
re-aligned it after I was finished with these tests. The image to
the left is what I found as a first pass (click it to see a larger
image). It is my rig being driven to max output with white noise,
with a saved trace of eSSB overlaid. As can be seen, things don't
look too bad. One thing I'd like to point out is that a statement
commonly made is "eSSB is much worse than white noise at producing
IMD". As can be seen from the test I did, that is not the case at
all. White noise is worse than eSSB. That has been found each and
every time I did the comparison, regardless of power level or ALC
deflection. Here are some examples of White Noise and eSSB alone. I
had previously mentioned that I had seen this before when I wanted
to just verify what the rig was doing, and was told that I was flat
wrong. See for yourself.... There's one more thing I'd like you to
notice here. When you look at my eSSB spectra plot made using my PC
& sound card, the signal is about 4.2 kHz wide, going from the
-6db point on the high side, to the equal level on the low side.
The spectrum analyzer pretty much confirms that! It's good to see
some consistency here! The occupied BW shows 99% of all the power
within 4.9 kHz. Please notice that the analyzer I borrowed did NOT
have HP's Narrow BW option. Therefore, you must subtract the 300 Hz
IF BW of the analyzer from the result, giving us 4.6 kHz. That's
pretty good as well. Anytime you see the Occupied BW measurement,
you must subtract the analyzer's IF BW from the value shown.
I also ran a test comparing my eSSB signal to a "standard" SSB
signal (at left), which in the case of my rig is ~2.7 kHz wide. You
can see that the eSSB signal is wider (which is of course the
idea!), but it isn't significantly more offensive than the 2.7k
signal. Both of these were done with speech. I'd also like to show
some other images for reference. I was sent a couple links from Jim
(Jan)-SM2EKM when the subject of this came up. He has traces done
on the Yaesu FT-1000 and the Collins KWM-2. I have the same images
on my site HERE and HERE, respectively. The KWM-2 is known for
being very clean, so it could be used as a very good performing
reference. The point of showing these are twofold. One is that it's
a secondary source to show that what I'm seeing is roughly what has
been found elsewhere. And two, my rig is at least on par with a
different one, even in eSSB mode. I'd like you to notice the noise
level at the edges of the screen referenced to the highest level of
the test signals on all of these traces. Also, most of the images I
have are done at 20 kHz, edge to edge, like Jim's are. I do have
some that were done at 25 kHz, but that's because I hadn't paid
close attention to that at first. What I want to point out is that
my rig is ~60-65dB down from the peak at +/- 10 kHz, just like
Jim's FT-1000 (remember, that's at max output -- I usually run much
less power). The KWM-2 is ~65-70dB down, clearly the better
performer. Also notice that he doesn't have the signal centered in
the analyzer's display, so one side is a bit better than the other.
I attempted to center the signal in the display, evening out the
sides.What I set out to do was to prove or disprove some of the
claims made buy a few out there that eSSB is substantially worse at
producing IMD than standard SSB. The claim was made without any
real measurements (in other words, pure speculation!), so I wanted
to see for myself! It is clear that the IMD claim is being used
falsely or at least blown way out of proportion, and there are much
worse signals out there that AREN'T eSSB!! Grossly overdriven rigs
and amplifiers produce far more IMD than a well controlled eSSB
signal. A vast majority of eSSB users drive their rigs so that
there is NO ALC DEFLECTION, into amplifiers that are capable of
producing much more than legal limit, even when being driven with
less than the rig's maximum power output. In my case, I drive a
Henry 3K-A which is capable of about 2.5 kW PEP output due to its
hefty power supply (almost 4 kV @ nearly 1A). This is what I was
referring to earlier about only running about 80-90W -- that's all
I need to get legal limit. This is pretty typical. Most run an 8877
or pair of 3cx800s, which do 1500W out with well under 100W drive,
as well as no ALC. So is eSSB wider than standard SSB? Of course!
That's why it sounds better (if you have a receiver capable of
listening to it all)! Is eSSB substantially worse than standard
SSB? I don't think so. Especially when you consider that the rigs
are driven to much less than their maximum output power
capabilities, and without showing any ALC deflection. Judge for
yourself.To see all images referenced, go HERE.
MY THOUGHTS ON AMATEUR RADIO AND eSSB*
I use (or have used) a variety of rigs, from "boat anchors"
(vintage AM rigs) to the latest & greatest from Icom, Kenwood,
Yaesu, and Ten-Tec. As I have said above, my main HF rig at home is
the TS-950SDX. I have modified it so the receive is fairly flat
from ~DC to about 6k on SSB (a BIG thanks to WZ5Q for all the help
identifying the required parts that needed to be changed!!). Of
course, I can still narrow it to below 2.5 kHz if necessary, but I
rarely have the need. There is some roll-off above 3 kHz, so it's
not perfectly "flat", but it's quite good, and much better than
stock. The transmit is good from around 20 Hz to about 4.2 kHz. My
processed TX audio effectively uses approximately 40 Hz to about
4.5 kHz, as you can see from the above link. And as with my
receive, I can narrow the TX up as much as necessary. All of the
above applies to SSB. I also tinker with CW and PSK (as well as
other soundcard Digi-Wigi modes), and this rig can narrow down very
tightly. I do NOT have any optional filtering installed since it
comes stock with 500 Hz, 2.7 kHz, and 6 kHz filters in both the
8.83 MHz and 455 kHz IFs. I can choose whichever filter I want
independent of mode, and when both IFs have a filter of the same
BW, it yields a slightly narrower BW than each by themselves (a
benefit of cascading filters). I have not found a need for any
optional filters, especially since I can adjust whatever filtering
I have inline at the first and second IFs and vary the upper and
lower slopes. For maximum BW, the 8.83 filter can be bypassed
completely via the front panel, and I have installed a piece of
coax across one of the 455 kHz optional filter slots to bypass that
one as well. That effectively lets the DSP handle all the RX
filtering, so that the recovered audio contains as much BW as
possible (again, that's approximately 0-6 kHz max on SSB) with no
analog filter distortion.If you haven't figured it out by now, I'm
into high quality SSB audio also known as "Hi-Fi" SSB or eSSB. Many
feel this facet of Amateur Radio is wasteful or in extreme cases
very rude or even illegal. I can only say that who gets to decide
this? The FCC's Part 97 makes no specific reference to allowed BW
for SSB or AM, so it's up to us to decide as long as there is
enough available BW open to support it. eSSB is not for everyone --
I would certainly never claim that to be the case. Just like
contesting, QRP, DX'ing, CW, Satellites, or any other mode isn't
for everyone, neither is eSSB! It most often takes up less space
than AM, and definitely less space than a dirty, overdriven 2.4 kHz
SSB signal. So what's the problem, then? For some reason, this mode
has elicited literally hostile actions by more than a few people. I
have been in QSO many times and have had to put up with rude
comments, intentional QRM, disgusting bodily function noises, etc.
just because of how I sound! Now how silly does that sound to you?
In reality, not too many can even receive eSSB anyway! Even if you
have a rig with wide enough filtering, you have to take a couple
simple steps to hear it all. First, forget using the internal
speaker, and in most cases, even an external speaker plugged
directly into the rig. Usually, the best place to pull the receive
audio is out of the accessory jack usually found on the back of the
rig. Of course, you can't drive a speaker with that, so you have to
use some kind of amplifier and external speaker. I feed mine
directly to my PC then on to a dbx preamplifier which drives an
Alesis 150W amplifier. It drives a pair of modified Bose 301s and a
JBL professional sub-woofer located under my desk. Not only does
this work very well for eSSB, but also for listening to MP3s or
streamed radio stations, watching movies, and gaming.The bottom
line is that many folks seem to judge something that they haven't
really heard! Others appear to not even want to hear it. I have
pointed a few people to sites with lots of eSSB samples and they
have come back either claiming to not hear much difference (oh come
ON!!) or, they claim that it sounds terrible because of one thing
or another. They have either made their mind up and are unwilling
to change their opinion, or are just plain deaf! I honestly believe
that a lot of people tune to an eSSB QSO with a rig that has no
chance of hearing the entire transmitted signal accurately (or it
simply isn't set up to hear it), and decides that eSSB does nothing
and just wastes space. If you have decent speakers on your PC, or
at least half decent headphones, go to John-NU9N's or Mike-WZ5Q's
websites and listen to the clips there. John has the clips
organized by BW, so you can hear the subtle but important
differences between less than 3 kHz SSB, 6 kHz SSB, and everything
in between. Mike also has a variety of clips of all different types
of sounds. eSSB is about sounding good, learning, experimenting,
and most importantly, HAVING FUN! After all, besides bona fide
emergency communications, isn't that why we do what we do?
* The term eSSB is owned and copyrighted by John Anning,
NU9N.
WHY PROCESS OUR SIGNALS INTO eSSB (or at all)?
Well, that's a pretty big can of worms, isn't it? A good place
to start is understanding how speech works, how available bandwidth
affects it, and what happens when the listener hears it. A Google
search on "Speech Intelligibility" turns up many hits. Another good
source of information is the Polycom Whitepaper on the subject.
Here are a few more:
http://www.assta.org/sst/2004/proceedings/papers/sst2004-103.pdf
http://www.utdallas.edu/~loizou/thesis/kalyan_ms_thesis.pdf
http://www.meyersound.com/support/papers/speech/section2.htm What's
important to understand is that the mantra "SSB at 2.4kHz (or less)
has been good enough for 50 years, why change?" is what a lot of
the controversy is about. I think the above references answer that
question in great detail, but the short answer is that very narrow
bandwidth causes holes in our speech when it contains energy above
or below the transmission equipment's bandwidth limits. The high
frequency parts are important for clarity and articulation. They
basically allow us to determine the difference between the hard
consonants within words. When they are missing, our brains engage
and attempt to fill in the missing parts based on context. The
lower the upper bandwidth limit is in frequency, the more
ambiguities there are and the more work our brains have to do. In
the simplest forms, it really isn't too much of an issue because
casual contacts heard on the Amateur Bands in a typical day just
don't require all that much accuracy, except for maybe call signs
which are usually given phonetically, anyway. Let's face it, it's
easy to tell the difference between Yaesu, Icom, Kenwood, or
Ten-Tec and Vertical, Yagi, or Dipole, etc. You get the idea. The
low end adds warmth and personality to our voices, and while not
really critical to understanding, adds comfort and realism.Long
periods of ragchewing can be fatiguing because of the totally
random subjects and the duration. That's where eSSB really shines!
If you listen to my audio clip, you can easily tell the difference
between the P in "PR-30" and T in "TS-950". Also, the S in "SDX" is
clearly an S and not an F. Try that with a 300-2.4k band pass and
you'll hear a lot of missing parts. My old call was "KS4KF" and I
can tell you first hand, "normal" bandwidths just don't contain
enough information to determine what I said. Obviously, calls are
basically random groups of letters and numbers, and since there is
no valid context, most folks heard it as "KS4KS" or "KF4KF" -- I
have heard both. Probably because our brains like repetition. This
is clearly why phonetics exist. But we don't talk that way in daily
person to person speech, so it's unnatural to have to do that
constantly (especially with THAT call!!).Another phenomenon is our
natural speech rolls off as frequency increases. On the Amateur HF
Bands, it's easy for those parts of speech to decrease below the
ambient noise levels if nothing is done to compensate. I made a
spectral plot of my voice into a flat studio condenser mic. Look at
it HERE. What you see there is a lot of energy in the low frequency
area up to about 700 Hz. After that, it sort of stair-steps down
until you get to 10 kHz then it rapidly rolls off. An interesting
thing is happening with my voice around 5.8 kHz -- there is a
fairly deep "V" starting around 4 kHz, coming back up around 7.5
kHz. I have seen other natural voice plots, and most have something
similar, but are different in each individual case. What that tells
me is that even if I had a rig capable of passing frequencies up to
6 kHz, it would do me little good even with an EQ. Granted, in very
good signal conditions it would be possible to hear differences,
but normally I would think the difference between 4 and 6 kHz
transmitters in my case would be nil. These speech dynamics
(frequency and level variations) are very difficult if not
impossible for the electronics in our radios to handle accurately.
Enter voice processing.There are several things that need to be
done to the natural voice to make it more "friendly" to our radios
and the medium (HF spectrum in this case) in which they operate.
There is some of that done already in terms of the microphones,
internal mic preamps, and other circuits in the rigs. However, it
is rarely if ever perfectly tuned to each of our voices. Like I
mentioned above, our natural voices have different characteristics,
and while the rigs certainly work as they are, they are likely not
anywhere near optimal. The first thing that we can do is equalize
or "EQ" our voices to lower some of the boominess in the midrange,
and increase some of the high end of our speech up to the highest
frequency the radio will pass. This brings us closer to something
the radio can handle. Additionally, we can tailor it for good
signal ragchewing or for "DX style" high noise, high interference
conditions. Also, we can use compression to decrease the difference
between the low and high volume parts of speech. When adjusted
properly, it allows for a more consistent level to feed to the rig
by reducing the peaks and increasing the valleys which means more
overall talk power. Related to this is peak limiting which keeps
loud spikes from entering the rig causing clipping and distortion.
Also related is noise gating or downward expansion. Gating simply
blocks anything below a certain level from going to the next stage.
When done properly and in the right environment, it can be fairly
transparent. However, gating can chop the first few milliseconds
off words or sentences and can be annoying. Therefore, I prefer
downward expansion. Sort of opposite to compression, it takes
everything below a certain level and reduces it even further.
Background noise is reduced to almost nothing and other ambient
noises are also reduced in level bringing your voice literally out
of the noise.These steps so far go a long way toward making your
voice clean, clear, and easily heard by the distant end. However,
doing all that can make your voice sound flat and unnatural. There
are still a couple things we can do with processing to fix this.
One is to add harmonics, the other is some type of effect to give
your voice a more life-like sound. What we're trying to do is
eliminate the effect that can be heard if you've ever gone into a
special sound-proof room called an anechoic chamber. These are
typically found in acoustic labs that basically eliminate any kind
of room reflections so that whatever is under test is the only
source of acoustic energy. Walking into one of these and closing
the door is an eerie feeling. You hear nothing reflecting off the
walls or floor and even your own breathing is only heard from
inside your head. Your speech sounds unnatural because we're used
to hearing ourselves both internally, and from external room
reflections. Although that's an extreme case, there is an element
of this when we've done the processing above. Also, harmonic
processing can fill in the holes (like the one I have from 4.2 to
7.5 kHz). However, both of these must be done carefully so that the
only way to tell it's there is to turn it off. Too much room effect
can be distracting and too many harmonics can sound harsh or over
processed. Either of these cases are not what we're after. The idea
is to increase our intelligibility, not reduce it!Okay, so once all
these pieces are wired together, there is the daunting task of
getting everything adjusted properly. This can be a never ending
process as you become more and more critical and aware of how you
sound. This is the reason the non-eSSB'er happens across a group
and hears a lot of back & forth about turn this adjustment up
or that one down, etc. Also, recording & playback are heard
frequently in eSSB circles. All that can turn the non-eSSB'er off
at best or at worst for some reason prompts a few to intentionally
QRM or jam the QSO. I don't know that there's any way to fix that,
but I hope that through learning and awareness, eSSB can achieve
more of an acceptance.So to answer the question at the top of this
piece, it's partially because our radios aren't meant to handle the
dynamic range of natural speech, and the built-in generic
processing isn't really good enough to do exactly what we need at
all times. And we can go a long way towards creating much more
intelligible signals when conditions are difficult (which is not
necessarily "eSSB" as John-NU9N defines it, but a definite
improvement over stock, one-size-fits-all audio for those
conditions), as well as a much more warm, personal sound when
conditions are good. The latter is generally the "eSSB" as defined
by John, where the "e" stands for "Extended" not enhanced. He
explains the differences HERE. So my hope is to proliferate the
idea that eSSB is about good audio, good Amateur practice, and
above all, adds to Amateur Radio positively.
