[Lf] DSP info

[Frank Gentges]

LFers,

I had a nice chat with Chuck Ripple and he will come down to see us 
either Friday night or Saturday (or both) at the cottage.  We got into a 
discussion on DSP receivers and he had received an email that explains 
why DSP receivers sound better than analog receivers.  (Chuck had 
observed this phenomena and wondered why.)  

I think the explanation makes some really good points.  We need to be
aware of these so we don't break our DSPs.  I think a convolutional filter
would be even better than an FIR filter so Bob's work should be good here. 

Frank K0BRA

---------- Forwarded message ----------
Date: Mon, 25 Jan 1999 16:45:00 +0100
From: KD Elektronik GmbH <KuD-BS at t-online.de>
To: Chuck Rippel <crippel at erols.com>
Subject: text to numero uno

Dear mr. Rippel,
thank you for sending the copy of the text to numero uno.
Here I will only comment on the text and write about the other items
later in a separate e-mail.

>From your text I understand that you confirm that DSP-receivers sound
different from analog receivers and that the readability of weak signals
is better. But you cannot quite pinpoint the reason for the better
quality. Maybe I can. This is going to be a somewhat longer explanation
and if I tell you something that you already know, excuse me for that. I
am sending you this for the preparation of the demonstration and I want
you to tell the people the right things.

Comparison of DSP-Receivers and analog (conventional) receivers:

There are two reasons for the better audio- or signal-quality of the
DSP-receivers:
one is the properties of the bandpass-filters and the second is the
properties of the demodulator or downconverter.

1. Bandpass filters
The bandpass-filters used in analog receivers are either crystal or
mechanical filters. Both filters suffer from phase distortion, the more
the steeper the skirts are. This means that the delay time of different
frequencies in the passband is not the same. The time or phase
relationship of the frequency components of a signal is lost or at least
distorted. This can easily be observed with digital signals like fast cw
or RTTY. The pulses are severely rounded or even can get pointy. Or this
can be seen by reveiving fax pictures. Due to the phase distortion the
vertical lines get fuzzy of are doubled. This does happen with audio
signals too, but the human ear cannot detect the phase error, but the
sound and readability are affected. There are very expensive receivers,
e.g. from Rohde u. Schwarz, which have quite elaborate phase
compensation networks to compensate the phase distortion, but these
receivers are very rare.

The bandpass filters in the DSP-receivers are of the type FIR. These
filters are strictly phaselinear, which means that the delay time for
all frequencies in the passband is the same. Often the expression
phaselinear is used, although many people do not know what it means. It
means that the phase increases in a linaer function with the frequency.
If the factor is correct, the delay time is constant. That the
phaselinearity of the filters is mathematically exact linear is very
important for the signal quality. I have always stressed this in my
brochures and publications, but the reviewers do not pay attention or
they do not know why this is so important. You can reread the review
from Radio Netherland (there is a link in our homepage). They too write
a lot about the special sound and do not know the reason. Some reviewers
even write that the sound is somewhat artificial. The contrary is
correct. The sound is more natural with a DSP-receiver than with an
analog receiver, but they have never heard it before. The absense of
phase distortion can again best be seen by receiving digital signals and
looking at the signals on a scope or by looking at fax pictures. And the
digital filters do not ring. You can receive fast cw or RTTY with a very
narrow filter, which is not possible with analog filters.
There is no analog counterpart for the FIR-filters. The can not be built
in the analog technology. Thus these filters and their performance is
really something new in the art of communication.
It is important too, that the filters in the front-end of the receiver
or the first i.f. do not cause phase distortions. Therefore are we using
a pretty wide crystal filter in the 1. i.f. of 15 kHz bandwidth.

2. Demodulators
All demodulators are mixers or multipliers. The frequency conversion is
mathematically a multiplication.
The simple diode demodulator for AM uses the nonlinearity for mixing the
carrier with the sidebands. This is the wanted signal. But the sideband
frequencies multiply with each other too. Every frequency in one
sideband generates a signal with all other frequencies which are present
in the passband. This leads to an almost unlimited number of unwanted
signals. These are smaller because the sideband frequencies are smaller
than the carrier, but they are there. Therefore the diode demodulator
has a distortion factor of 3 to 5 %. or more. The situation is a bit
better with sync detectors and product detectors (product =
multiplication), because the added carrier is much stronger than the
signal and so the spurious signals are relatively smaller. Basically
there is no difference. It can not be prevented, that the signal
components multiply with each other.

This is completely different with the digital multiplication. As said
before, any frequency conversion is a multiplication of two frequencies.
If two frequencies are multiplied in the digital representation, only
this is performed and nothing else. A multiplication of the signal
components does not happen. So when the signal is downconverted in the
DSP, the resulting signal is as clean as it was. There are of course
different algorithms for the demodulation of am and ssb or other
signals. But common for all is that they do not cause a distortion like
the diode demodulator or product-detector. Basically the demodulator
algorithms are free of distortion, except maybe the resolution. In a
16-bit system the resolution is 65,000 and in a 32-bit system it is 4.3
billion bits or steps. In the KWZ 30 we use double precision math, which
is 32-bit. So the resolutiuon error is not a big deal. It can be said
that the digital downconversion and the demodulation does not cause a
detectable distortion.

The properties of both the filters and the downconverters/demodulators
were unknown before and contribute to the special and exceptional signal
quality of the DSP-receivers. A real DSP-receiver is something
completely different than a conventional receiver with an added Timewave
filter. Do, not mention that thing again.
I think that this is enough about this matter and I hope that it gives
you the information that you have missed to understand the differences
between a DSP-receiver and an analog receiver. If you need more
information about this or have any questions, please let me know.


Remote keyboard:
The remote keyboard looks the same as the built-in keyboard and it does
the same. It has a cable of about 6 ft. We have added one feature: with
the keys 0 and Enter you can step the received frequency up and down one
tuning increment for fine tuning. If you have chosen a step of 50 Hz,
you can step the frequency up and down in steps of 50 Hz (0=down,
EN=up). This function works also with the built in keyboard, but is not
needed.
The remote keyboard is available at the price of DEM 232.-- plus
shipping.

Best regards
Hans-J. Kneisner