[RTTY] DXP38 vs Soundcard stuff

[Kok Chen]

AA5AU wrote:
> "I think those tha[t] "shelve" their HAL products are foolish."

Not saying it didn't work for you, Don, but I had a DXP-38 loaner which 
I used for a couple of months.

What follows below is the comment I had sent back to HAL.

BTW, when I wrote it, I didn't know that the HAL ST-8000 came with the 
same AGC mechanism and advantage which I commented on re the Timewave.  
And to be fair to HAL, the DSP-4100's (similar to the slow-but-accurate 
tuning mode on the DXP-38) tuning mechanism is probably perfectly 
adequate for the target customer, which is probably a professional 
station on an assigned frequency. Also, my current tuning program is no 
longer uses gray bars but crossed bananas (more CPU cycles to burn and 
lower sampling latency with the new fangled computers, HI HI).

Remember, this is just one man's opinion, and I am not even a decent 
op, just look at my contest scores, HI HI.  Being an engineer, I had 
tried to be brutally objective (at least I tried not to do a CQ 
Magazine type "review") in the face of the very kind offer from HAL.  
And I have to say, with my various interactions with them is that their 
support is superb (there are guys on the staff who gets on the RTTY 
contests).

BTW, I still have two screenshots which shows the DXP-38 making 
mincemeat out of the standalone KAM Plus :-).

Another caveat is that I did not compare the DXP-38 with a software 
modem like MMTTY or RITTY.  A DSP modem today is just an A/D converter 
followed by DSP code; exactly what a computer with a sound card does.  
I am not saying that MMTTY has better demodulation algorithms than the 
DXP-38, since I have never used it.  All I am saying is that with 
today's state of art, as the Timewave unit showed, you can do better in 
the DSP domain than the DXP-38, and software is just software.

73
Chen, W7AY

--------------------------------------------- (sissors) 
--------------------------------------------------------------
Test Conditions
---------------

The other modems which I have in the shack are

1) a modified KAM-Plus
2) a stock KAM'98
3) a HAL DSP-4100
4) a Timewave DSP-599zx

The KAM Plus is modified by increasing the Q of the
switched-capacitor tank circuit.

The Timewave is used both as a DSP filter ahead of the
KAM-Plus, and also as a Remodulator for the KAM.

For my daily use, I use a standalone software
program which displays subbands of an FFT as gray
levels.  There are two markers beneath the bars at
the Mark and Space positions.  The sampling buffers
are kept to the size of the FFT so there is as little
delay as possible from the sampled signals to the
display (more on this later when describing the
DXP-38's tuning display).  It is extremely accurate
and adequately fast.  The accuracy stems from using
gray levels for the spectral bands, unlike the LEDs
on a KAM, for example.

The control program which I used was a homebrew program
which I developed for the Macintosh.  A different control
program should not change any of the observations
of the DXP-38 below.

Initial Observations
--------------------

No problem is setting up at all.  I did not have to even
open the documentation package.  I simply took the DSP-4100
setup and transferred it over to the DXP-38 setup.

I have a switchable audio rail in the shack in RCA-connector
form.  Since I only put the DXP-38 through demodulation paces
(I never transmitted through it), that was all I had to do to
connect it to the receiver (actually, to the switchable audio
rail in the shack).

This audio rail system allows me to feed the A receiver output
from my FT-1000MP to one modem, and the B receiver output of the
FT-1000MP to a second modem.  I can reverse A/B at either the
audio level with the switch, or at the RF level with the FT-1000MP.
Finally, I can feed receiver A to both modems.

The RS-232 had to go through an adaptor (the DSP-4100 uses
a 25 pin D-sub) and with that, the DSP-4100 program came up and
initialized the DXP-38 fine.  There was nothing else I had to do
to operate the DXP-38.

My usual operating mode is to set the system up so that S7 and
above signal (rig RF AGC operational) would saturate the A/D
converters of a modem slightly.  I have found that the distortion
on the loud signals were not enough to cause any harm (the DXP-38
had no problem at all with the loud signals) and the weaker signals
has more A/D dynamic range to work with.


Tuning
------

I hate to say it, but for my style of operating, the tuning
meter is all but useless.  I tried it for a few minutes, and never
looked at the tuning display again.

The Mark/space mode of the tuning meter did not provide enough
accuracy to tune a weak signal.  And the Delta-F mode is too
slow to be any use.  I suspect the Delta-F mode is the same
as a similar mode in the DSP-4100 (a delta frequency was returned
as a status inquiry), which I also found not very practical.

One of the problem with DSP modem tuning meters is the lag time.
Very often, they sample a large buffer before processing the
buffer.  This long buffer length translates into a hysteresis
when tuning the receiver.  If the buffer length can be shortened
for the tuning meter, the Delta-F mode might actually be useful.
But as it is, unless you are willing to tune at the rate of a
few Hz per second, the Delta-F mode indicator has too much
hysteresis.  You really need an update time that is -much- faster
than one per tenth of a second to be useful for assisting the
manual tuning of a receiver.

Demod Observations
------------------

The first time I used the DXP-38 extensively was in the 2002 ARRL
RTTY Roundup.  As is my usual practice in S&P operation in
contests, I would tune a signal using a primary modem.  If the
station (W1XX) I have tuned in is busy during an exchange, I would
swap VFO, placing the station in question on a second modem.  I then
proceed to tune until I see this station, now on the second modem,
free up.

In this manner, I find that I can often find a second station
(W2YY) while waiting for the first station to be free again.

With two stations tuned in on separate modems, I wait for the
first station to become free (for example, W1XX).  I would then
switch to work W1XX, and then immediately have W2YY already
tuned in.  If W2YY is free, I work him, otherwise, I keep
him parked, while tuning with the VFO that was tuned to W1XX.

For the contest, I had the DSP-599zx as a remodulating front end
to the KAM acting as the primary modem, and used the DXP-38 as the
secondary modem.  At less busy moments, I would also connect both
modems to the same signal to see what the prints are like.

I have done the same thing during the WPX RTTY contest.

Outside of the contests, I use the same two modems during split
DX operation.

I would tune a station in using the primary modem, then swap VFOs
so the second modem is now monitoring the DX, while the primary modem
is used to scan for the QSX of the DX.  Once I find the frequency the
DX is listening to, I can print both ends of the QSO.  Using the
switch on the audio rail, I can also switch between which modem is
listening to the DX, but I always transmitted through the KAM.

For signals over S3 to S5 on the S meter (about where the FT-1000MP's
AGC kicks in), I can usually not find a significant difference, except
the KAM'98, between all my modems analog or digital.  The standalone
KAM'98 is the poorest performer of the bunch.  The DXP-38 performs as 
well
as the others, which is to say perfect copy most of the time, modulo
QRM, or a noise static now and then.

My Portland QTH is very quiet, and in the evenings, I can copy
signals in and across the Pacific when they do not even light
a single LED on the S meter.

The DXP-38 actually performed very well with weak and clean
signals.  It is about on par with the Timewave/KAM combination,
with the Timewave in remodulation mode.  It beats the Timewave/KAM
with the Timewave in RTTY Filter mode, and definitely beats the
standalone KAM-Plus.

However, the Timewave Remodulator comes out ahead with FSK
which exhibits transpolar or equatorial flutter.  In a certain
case with a fluttery FO5, the Timewave provided much better print
than the DXP-38.  The same with signals through the pole (most
of Europe) when there is significant flutter.

How much better print?  Well, I probably would not have lost
a single contact if I only had the DXP-38, and no Timewave on hand.

That is to say that when the DXP-38 starts misprinting badly, the
Timewave would also be misprinting.  Just that I can get a slightly
cleaner print from the Timewave.

I can pull more callsigns cleanly with the Timewave.  If I can print
a DX coming back to me on the Timewave, I am pretty sure I can
eventually print him on the DXP-38, if the DX repeats my call a
few more times.

This is more than I can say about analog modems.  In many
critical cases, the analog modems I own would just keep
printing garbage no matter how long I listen to the DX.

Conclusions
-----------

Although the DXP-38 is superior to a standalone KAM Plus and
KAM'98, I have to give a slight edge to the Timewave DSP-599zx
over the DXP-38 when it comes to weak and fluttery signals.

In fact, I do not notice any observed difference in performance
between the DXP-38 and my HAL DSP-4100, which has been the secondary
modem I have been using for a while now.

The tuning mechanism on the DXP-38, however, is something I cannot
recommend to anyone who is either a serious DX'er or a serious
contester.

Unless the DSP tuning mechanism can be made both accurate and fast,
I would recommend that with any future HAL product that you use analog
methods as the tuning mechanism in a DSP modem.  The analog
filters can also supply cross-ellipse output for the more mature
RTTY operator.  A faster standalone DSP filter for tuning that is
not encumbered by the algorithms of FSK decoding can also be used.

I would also recommend that HAL provides a "VU meter" readout
(through the RS-232 port, if there is no front panel alphanumeric
display) together with as AF level pot in the modems.  My believe
is that very few hams are using their DSP modems to full potential
because the AF levels are not optimally adjusted.

I have used the AF AGC system in the Timewave to good advantage,
and perhaps HAL might want to look at that approach too.  It
is quite useful for small signals, effectively increasing the
dynamic range of the A/D system, but I often have problems
(understandably) with the first few characters of a very strong
signal.

Another suggestion is to design the DSP matched filters to optimize
for the case where (S+N)/N is low and there is a lot of polar
flutter (Kalman and Wiener filters all have different optimal
designs depending not just upon the signal spectrum, but also
the noise spectrum and magnitude).  Such a modem would not be
optimal for the clean, strong signal case, but those are the cases
where even suboptimal demodulators would provide perfect print anyway.
Perhaps HAL is already doing this, if not, it is food for thought.

Finally, since we are talking DSP modems, and not fixed analog
hardware, it is possible to have a few FIR designs, each optimized
for a different S/N or flutter characteristic, and chosen based
on some heuristic on the characteristics of the received signal.