GC_ET_Phased-B.html

Nyack glene77is Glen Ellis K4KKQ KK4LPG Audio Filter HW-8 HW8 HeathKit Vintage Heathkit+Fan+Club QRP / CW Weak-Signal Ham+Radio Hi-Per-Mite HiPerMite

**************************************************************************************************************************************

Signal Phase-Delay Analysis
for Ham Radio C.W. operations ,
by Glen Ellis, August 27, ,2015, presented on Research Gate . com :

( Documentation for the current project to display signal-phase-shifts / delays )

*******************************************************************************************************************************

Abstract:
For Amateur Radio C.W. (MorseCode) operations.
Series of analog filters approaching DSP quality.
Phased-Filters and Parallel-Channel-Filters provide exceptional results.
Applications could include in-lab analog instrument filters
( proto-boarded and no PC required).

Rationale:

My goal is to better understand the concept of phase-delay.
Then build guidelines for designing filters
such that the phase-delay and group-delay distortions
can be controlled in narrow-band C.W. Amateur Radio operations.

Given that a C.W. narrow-band audio signal has no odd harmonics,
it may be that Phase-Delay and InterModulation-Distortion are the only effects for study.

For this study, the standard signal is f(700),

and odd harmonics for this are 233Hz and 2100Hz,

which are far outside the narrow bandpass of 100Hz.

Thus, the Narrow-Band C.W. audio signal has no odd harmonics,
and the turn-on and turn-off shapes are of a sine-wave shape
( not a square-pulse shape ). This is termed "rise-time".
It can be shown , via an O'scope, that the leading and trailing edges
of a C.W. audio signal have the characteristic sine-shape .
The author's measurements in SPICE of the rise-time
for a given 700Hz square-wave signal,
passing through a Q=10 band-pass filter is about 1.4 mS. .

Phase-Delay and Group-Delay are well defined in WIKI,
as they are related to HiFi wide-band music audio.
In Wiki, neither concepts are developed
relating to Narrow-Band Continuous-Wave Morse-Code type signals.
Narrow-Bandwidth C.W. signals are about 100Hz wide
[ depending on the relation of BW=(Baud * K ) + Delta-F()
which can apply to all digital pulsed signals. ].
This is an average of measurements/calculations for C.W. speeds
varying from 10 wpm to 40 wmp, that is from 30 Hz to 120 Hz band-width,
containing the central signal and sideband IMD passing through the filters. .
The author adopts this 100Hz band-width as typical and usable for study.

In the real analog world, square wave pulses are hard to generate,
always having a measurable rise/fall time, over-shoot, etc.,
and never being truly square.
Square wave pusles of audio always require strong Odd sub-and-supra Harmonics as a 'package'
in order to begin producing the 'square'ness of the pulse wave form.
When passed through narrow bandpass filters ( here read 100Hz BandWidth or Q=7 )
the C.W. audio signal will have a sine-shaped leading/trailing edge ( No Squareness ).

Given that this study is with C.W. Narrow Band
the author has adopted the "triple-signal",
comprised of 600Hz and 700Hz and 750Hz,
which is all that will pass through from earlier bandpass filter stages.
This can be described as a data signal and adjacent background noise signals.

The GOAL is to see if usuable data can be provided by SPICE,
such that some GUIDELINES could be formed for band-pass filter design.

If the Triple-Signal is analogous to a Single-Audio-Signal ,
then Phase-Delay and Inter-Modulation Distortion
may be a proper study in C.W. audio signals.

It can be shown that the Triple-Signal,
passing through a narrow band-pass filter,
has Phase-Shifting and Intermodulation Distortion
sufficient to alter a C.W. pulsed signal
without reference to Group-Delay concepts.

"Is it possible to measure Group Delay, using SPICE ? "

Available from: https://www.researchgate.net/post/Is_it_possible_to_measure_Group_Delay_using_SPICE
[accessed Aug 26, 2015].

**********************************************************************************************

In Active Audio Filters, "Q" = F() / BW.
“Q” can also be used as an indicator of the level of Filter Activity
required to do the specified filtering to acquire the Band-Width.
.
Filter Activity at any level will generate Phase Delay,
by causing Phase Shifts at each frequency [ based on T= ( 2 pi f R C ) ].
These Phase-Shifts can affect the quality of CW signal passed through an audio filter.
Ham CW operators call this interference “Ring” .
.
"Ring" is often reported by "ham radio" operators
and various filters have been designed claiming to have avoided the problem.
The author proposes that "Ring" is often confused with the CW signal combining
with the amplified background band noise within the narrow passband.
This narrow passband amplified audio is the "Whistle" or "Tunnel" effect
.
This is more accurately described as the "white noise narrow filter effect"
which always occurs when combining with the phase-shifted signals
on the edge of the filtered passband.
Further Low Pass filtering can remove the 3rd harmonics responsible.
.
When the Filter passband is narrow enough ( say 100 Hz)
to match the CW passband ( say 100 Hz)
then phase-shifted smearing may be heard
as the Phase-Delay frequences are forced into the CW signal
causing cancelaton of the 3rd harmonics.
.
Few CW operators have heard true "Ring" or self-oscilation of an active filter.
Active Filter self-oscillation is not covered in this study.
.

The Phase-Shifted smearing changes a crisp "dit" to a soft "Thith" sound.
At CW speeds of 30wpm the smeared edges of the CW signal may become uncopyable.
At CW speeds of 10wpm the smeared edges of the CW signal may be un-noticed.
The phase-shifted smearing changes a crisp "dit" to a soft "Thith" sound.
Some measurements are following.
.
In HiFi wide-band music,
the wide spread of frequencies may be dispersed across 20 KHz,
and Group-Delay is a noted and corrected phenomenom.

In C.W. narrow-band communications,
the narrow spread of frequencies may be dispersed across 1 KHz or filtered down to 100 Hz,
and Phase-Delay is caused by the natural phase shifting during RCL filter action on a 100 Hz band-width.

Phase Delay is caused by the RC network in the filter itself.
Once the CW signal has passed through a narrow filter, odd harmonics are filtered out,
and the required 1/3 and 3rd harmonics required to form a square pulse C.W. signal are not present.
The primary odd harmonics for f(700) would be f(233) and f(2100), which are attenuated by the active filter.
What remains are a smaller number of similar frequency signals.
An example is 10 CW signals ( dispersed across 1 KHz ) entering
and then only 3 CW signals exiting the filter ( dispersed across 100 Hz ).

In the Signal-Phase-Delay experiment which follows, using a triple-signal set of 650Hz and 700Hz and 750Hz,
the author has tried to show that the narrow C.W. bandpass limites the available frequencies in the group
to only 100 Hz wide, and is so narrow that the Phase-Delay amongst these signals controls
the distortion from Phase-Delay.

***********************************************************************************
Harmonic structure of a CW audio signal defined:

The CW signal is defined by the odd harmonics
on the leading/trailing edges and middle of each "dit-dah".
The harmonics have a naturally occuring Phase Shift during filter action.
Without the odd harmonics the Square-Wave pulse shape will not be formed,
and the CW signal will be a Sine Wave with defined time periods.

**********************************************************************************************************

*************************************************************************
The author's Bread-Board set up for testing circuits,
used in real-time on the radio bands , and with signal generators.

***********************************************************************************************
Summry of the authors approach to designing basic MFB Phase-Shift-free Audio Filters.

Group Delay is present to some degree in all audio filters,
Group Delay is fundamentally a Phase Shift that naturally occurs in all R/L/C filters.

This approach has been to sequence multiple Butterworth inverting filter stages
such that the successive Skirt BandPass pulls signal from the 'inside' of the prior stage, as Q increases ;
ie, the succesive Skirt Attenuation attenuates the previous Phase Shifted Frequences.

This has been done by increasing the “Q” of each successive stage
which progressively descreases the Band-Width,
and cuts off the edges of the prior band-pass.
The idea is to cut off the prior skirt from -3dB down to the -9dB area, successively .

In the AFX design,
the author's best design is based around calculated Q = 2.0, 2.5, 3.0, 5.0
which is measured Q = 2.0, 3.5, 4.5, 7.0, with the final "Q" measuring 7.0 cumulative.
In the AFX design, the author's goal has been to have the successive AFX stages
centered on the same Frequency f() Center,
with each stage generating successively more “Q”,
with successively Narrower BandWidths,
in order to attenuate the Prior Stage's Phase Shift Effect.

The CW "dit-dah" is not a pure sine wave,
but has harmonically defined leading/trailing edges.

The CW signal is defined by these harmonics
on the leading/trailing edges of each "dit-dah".
The harmonics have a naturally occuring Phase Shift during filter action.
This Phase Shift (for each frequency) is measured
as a Time Shift with reference to the original signal.
This phase shift on the leading / trailing edges
is a distortion of the original C.W. pulsed audio signal.

(af-f19) example of OverAll-Group-Delay-Effect-in-Low-Pass-Filter

NOTE: Below: the GREEN original signal
NOTE: Below: the RED Phase-Delayed signal is the Filter V(out).

aimg-Square&Ring_Sound-Westhost-com_af-f16.gif

Graphic source: AFX_GroupDelay&AttenuationEffect_Sound-Westhost-com_af-f19

********************************************************************

aimg-a-group-delay-animated

NOTE: Above figure, Various Signals in the center of the pass-band
are passing through at a faster rate.
Various Signals lower and higher than the center of the pass-band
are passing at slower rates.
This demonstrates the Phase-Shifting in a "Group of Signals".

When the signals are HiFi music (wide band = 20 to 20K Hz )
then this Group-Delay ( low edge to high edge ) is destructive.

When the signals are C.W. (narrow band = 650 to 750 Hz ) ,
then the Phase-Delay results in less distortion of the C.W. characteristics.
:

*********************************************************************************************************************
*********************************************************************************************************************

Where does the Filtered Energy go ? Up in Smoke ???

The sharply shaped high frequency Filtered Energy
is phase-shifted into a wider low frequency output signal.

The Filtered Energy is Phase Shifted into the existing energy mass
and produces a wider spread on filtered signal pulse,
with the filtered energy showing up in the base-band noise.

***************************************************************************************

This is a description of measuring Phase-Delay using Spice and a 'Triple-Signal' to simulate "Group Delay"

"Experiments in tracking Signal-Phase-Delay"

**************************************************************************************
Signal-Phase-Delay
*** Triple-Signal injected, f(650), f(700), f(750).
*** Signal phase delays are observed in the #2 (middle) Q=10 MFB filter.
*** Mixing of signals and new peaks / nulls are observable, relative to the Triple-Signals interaction.
*** This interaction among the Triple-Signals is InterModulation Distortion.

*** First, a little background experiment to show how the Triple-Signal mixes. The AF10 ( Q=10 ) filter response.

*** AF10 filters all the signals producing this Bode Plot.

Below: Transient Plot shows Phase Delays and inter-actions.
Triple-Signals are phase-shifting across time.
Note: AF10 forces phase shifting and produces a symetrical signal.
Note: AF10 (blue) responds to the beat frequencies with peaks and nulls in its output.

************************************************************************
*** Now the actual Research Gate experiment
************************************************************************

Method:
Measurements were visual and recorded by hand.
In ngSPICE Transient tests, it can be observed, by visual measurement,
that the three signals have initial wavefronts that pass through the RC filter network,
at Differing Rates to a Specified Time Point.
Observaton was for each signal to rise to .707V of the initial 1.0V level.
That gives a signal-phase-delay measurement of the most elementary form.

I was initially interested in the sub / supra harmonics of 3rd order ,
which shape the turn-on / turn-off characteristics of my audio signal,
which would be 233Hz and 2100Hz,
but my circuit very effectively attenuated these down -48dB.
There are no practical odd harmonics to cause a square pulse shaped C.W. audio waveform.
Thus, square shaped audio waveform becomes a sine shaped pulse.

For that reason,
this little experiment could NOT show the square wave turn-on and turn-off
as it might actually be distorted by the group-delay.
It only shows elementary signal-phase-delays
which must be interpreted for some use.

It was understood, from the start,
that if the 3rd order harmonic signals were removed from the initial signal,
then the final signal could be only a sine-shape turn-on / turn-off.
This is, in practice , the C.W. audio signal that is displayed by the author's O'scope.
.
*** Summary:
*** triple signals injected, f(650), f(700), f(750).
*** Signal phase delays are observed in the #2 (middle) Q=10 MFB filter.
*** Mixing of signals and new peaks / nulls are observable, relative to the triple signals dispersion.
*** Mixing of signals generate beat-frequency signals, and are InterModulation Distortions (IMD).
***
*** Read the Legend Chart, then track the Triple-Signal trace, then track the NetF10 trace.
*** Signal phases are mixing, and delayed, referenced to the incoming triple signal.
*** Triple-Signals are blue/orange/green, starting as one signal at time=0, dispersing over time.
*** Faint Blue (Absolute) trace is an inverting sum of the triple signals, and shows immediate mixing of the Triple-Signals.
*** Black trace is Filter(Q=10) and is slow to respond to triple signals , peaks, then gives a null.
*** (The signal delay through Filter(Q=10) indicates the latency charge-up time for the filter to function.)
***
***
*** Initial analysis is that the triple signals are being smeared due to Signal-Phase-Delay
*** because they represent the total of the available CW audio waveform, and they are displaced in time.
*** Observe that at 10.5mS, the Q=10 signal responds to the mix of signals with a Null.
*** Observe that the Q=10 trace null lags the Absolute trace by 4 mS ( latency through Filter).
*** Plans are to measure Triple-Signal and NetF10 plotted out to 24mS and up to 100mS.
*** Will be to trace the repetition of the Null and relate it to the mix / un-mix of the triple-signals.
*** Will be plotting f(700) against NetF10, to more clearly define that elementary relation.
*** Will be plotting Triple-Signal against NetF10, to more clearly define that basic Phase-Shifting relation.
***

*****************************************************************************************************************
*** Test to show the Single-Signal f(700) into AF10 (Q=10) MFB BandPass filter.

*** Orange is the Q=10 signal requiring aprox. 6mS to charge the filter to .707.

*****************************************************************************************************************
*** Test to show the Triple-Signal combining and passing through a Q=10 Band-Pass Filter.

*** Observe the Triple-Signal mixing, creating beat-frequency signals, with peaks and nulls.
*** Observe the 'beat frequency' (IMD) distortions on the Red AF10.

*** Observe the Triple-Signal mixing and producing 'beat frequency' distortions on the Red AF10.
All Signal Rise/Fall characterists ar those of a Sine-Wave turn on/off pulse.
*** Observe the 'beat frequency' distortions on the Red AF10..

******************************************************************************************

General Summary :

[Jul 27, 2015]

Measurements of Phase-Delay on the Triple-Signal
do not produce measurements which link Signal-Phase-Delay with Group-Delay ,
in the field of C.W. communication audio narrow pass-band signal.
At least not directly.

(1) It is found that the prior stage filtering attenuates
the 1/3 and 3rd odd harmonic of the 700Hz target signal, down -36dB,
such that no C.W. audio square-wave pulse characteristic are practically measured.
(2) It is found that the Triple-Signal (650,700,750)
mixes with itself, causing 'beat frequency' InterModulation Distortion.
Signal phase shift on each of the three signals
is in the order of 50uS after 12mS time, Q=10,
Phase-Shift depends on Q of the filter
with Q=3 being less Phase-Shift and Q=10 being more Phase-Shift.

HiFi wideband music audio and C.W. communications audio
are two very different scenarious.
Group delay is widely reported and corrected in HiFi audio.
Group delay is casually reported in C.W. communications,
with no experimental evidence.

This experiment shows, in detail ,
the Signal-Phase-Delay of 4mS , produced by the Q10 filter itself,
and in an additive pattern in multi-stage audio filters, depending on the Q of the stages.

This experiment shows, in detail, the important 6mS rise/fall time
of the C.W. audio pulse, and the sine-shaped characteristic.

Of concern to a radio C.W. operator, working at speeds of 10 words per minute (wpm),
( where 'dit'=60mS, 'dah'=180ms ),
this phase-delay is only 10% of the shortest 'dit' signal
and only 3% of the 'dah' signal.
(1) In practice, at 10wmp this delay is very tolerable,
since the sine-shaped pulse is long enough to be detected by ear. .
(2) In practice, at 40wmp this delay is very destructive,
since the sine-shaped pulse is indistinct to the ear. .

For example some real-time measurements:
Given 10wpm and 'dit' as 120mS, then 6mS rise and 6mS fall takes away 10% of the 'dit.
Given 20wpm and 'dit' as 60mS, then 6mS rise and 6mS fall takes away 20% of the 'dit'.
Given 40wpm and 'dit' as 30mS, then 6mS rise and 6mS fall takes away 40% of the 'dit'.

Results here are that signal-phase-delay ( alone )
can distort the rise/fall times of a C.W. audio signal,
and can 'smother' a really fast CW signal.

Results here are that Signal-Phase-Delay
does not indicate Group-Delay,
and Signal-Phase-Delay is adequate to cause distortion.

Further investigation needs to be done
to provide more understanding of this phenomenom

Is it possible to measure Group Delay, using SPICE ?
Not exactly,
but by using Signal-Phase-Delay as analogy "YES".

Results at this stage are that Signal-Phase-Delay
does not indicate Group-Delay exactly.

But, Triple-Signal Phase-Delay
is adequate to cause a phase distortion (IMD)
which is analogous to Group-Delay.

Further observation leads to these considerations :
(0) Filtering practically removes sub/supra Harmonics from the f(700) signal.
(1) the Triple-Signals are the only and basic signal for use.
(2) the Triple-Signals are representative of C.W. signals with 100Hz bandwidth.
(3) the Phase-Delay for the Triple-Signals do cause distortion (InterModulation distortion).
(4) the distortion is analogous to Group-Delay distortion.

Further investigation and observation is in process
to provide more understanding of this phenomenom

*** Practical Summary of Results :
********* Too much filtering ( see below ) will deteriorate the C.W. intelligibility at higher data rates.

******************************************************************************************************************

Group Delay is present to some degree in all audio filters, Group Delay is fundamentally a Phase Shift that naturally occurs in all R/L/C filters.

Group Delay is present to some degree in all audio filters,
Group Delay is fundamentally a Phase Shift that naturally occurs in all R/L/C filters.