| Author |
Message |
Patrick Turner
Guest
|
 Posted:
Sun Oct 23, 2005 9:37 pm Post subject:
Multiplex Decoder questions. |
|
|
Hi all,
I built a stereo FM multiplex decoder some years ago and the schematic
is at
http://www.turneraudio.com.au/htmlwebpgs02/schemampxdecoder.htm
If you find all of what I am saying here or at the above address all
too hard to follow, could you be so kind as to simply
direct me to a URL the where the exact specifications are listed for the
standard
stereo transmitted signal with all relevant voltage levels of
L+R signal levels and 38kHz double sideband levels.
The principle of my decoder involved filtering out the L-R DSB signal
from the main composite
signal, and applying that to one input of an LTP, then applying a
synchronised
38kHz signal to the other input to get two oppositely phased output
38khz AM waves modulated
with the L-R signal so that when the +ve and -ve going peaks in these
signals from the LTP are diode detected you get a L-R and -L+R audio
signal.
These two signals are matrixed in a resistance divider
and the L+R signal is applied to the centre of the divider you get
L and R signals.
Ripple frequency in the diode detectors is 76kHz, and the idea was that
this
can better define the higher audio frequencies.
Anyway, the the output of the decoder didn't give very good separation
which became quite poor by 15kHz, and noise and spuriae were a problem.
I am now trying to improve on the decoder yet still use vacuum tubes.
I would like to know if it is possible to decode the stereo signal more
simply
and without the phase altering filters in the following way :-
1.Use a voltage follower to buffer the main composite signal from the
ratio detector.
2.Take an output from this using a tuned filter to extract the 19kHz to
use to synchronise the 38khz oscillator.
3.After the buffer remove the 19kHz with a notch filter,
and add one more voltage follower buffer no2.
4.Have a resistance divider between the oscillator output and buffer
no2,
and the output from the junction of the two resistor buffer
will contain L+R, and the AM wave form with the L-R modulation envelope.
5.If the modulation level of L-R is what I think it is, then
the positive going peaks of the 38khz AM wave will be one channel,
and the negative going peaks will be the other channel.
So if you had a 1khz sine wave for one channel, and no modulation for
the other channel
then the AM wave would have 1kHz modulation on the positive peaks and
no modulation on the negative peaks.
6.The signal from the divider would contain about 4Vrms of carrier, with
modulation levels
around 0.5Vrms, and this signal would need to be buffered by yet another
voltage follower no3 before being applied to a pair of diode-capacitance
detectors, and followed by a de-emphasis and filter network to remove
38khz ripple.
I have also seen some general descriptions of the use of symetrical
differential amps used in the chip decoders which could be done using
vacuum tubes but unfortunately the explanation offered on the pdfs I
have seen
is so damned confusing and inadequate that I still cannot follow
how the circuits work, and what all the wave forms are and voltage
levels.
The chip based decoder I have in a 1980 Audio Reflex tuner clearly
outperforms
my existing tubed decoder technically, although the tubed type did
sound ok..
The schematic of mine has been on the website for 4 years yet nobody
except myself has tried to improve on the design which I think
could easily be improved.
Most tubed decoders suffered from an inadequate amount of hardware to do
the job
due to accountants controlling purse strings and ppl trying to avoid
plagerizing someone else's designs, or designs becoming unecessarily
complex in order to be
unique and to claim some superiority, something I have not heard from
many stereo tubed decoders; the mono
signal often sounds clearer and better, just as the measurements would
suggest.
I am only intending to use any information for my own use.
Regards,
Patrick Turner.
|
|
| Back to top |
|
 |
Robert McLean
Guest
|
| Posted:
Mon Oct 24, 2005 8:06 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435BBB27.172664E5@turneraudio.com.au...
| Quote: | Hi all,
I built a stereo FM multiplex decoder some years ago and the schematic
is at
http://www.turneraudio.com.au/htmlwebpgs02/schemampxdecoder.htm
If you find all of what I am saying here or at the above address all
too hard to follow, could you be so kind as to simply
direct me to a URL the where the exact specifications are listed for the
standard
stereo transmitted signal with all relevant voltage levels of
L+R signal levels and 38kHz double sideband levels.
The principle of my decoder involved filtering out the L-R DSB signal
from the main composite
signal, and applying that to one input of an LTP, then applying a
synchronised
38kHz signal to the other input to get two oppositely phased output
38khz AM waves modulated
with the L-R signal so that when the +ve and -ve going peaks in these
signals from the LTP are diode detected you get a L-R and -L+R audio
signal.
These two signals are matrixed in a resistance divider
and the L+R signal is applied to the centre of the divider you get
L and R signals.
Ripple frequency in the diode detectors is 76kHz, and the idea was that
this
can better define the higher audio frequencies.
Anyway, the the output of the decoder didn't give very good separation
which became quite poor by 15kHz, and noise and spuriae were a problem.
I am now trying to improve on the decoder yet still use vacuum tubes.
I would like to know if it is possible to decode the stereo signal more
simply
and without the phase altering filters in the following way :-
1.Use a voltage follower to buffer the main composite signal from the
ratio detector.
2.Take an output from this using a tuned filter to extract the 19kHz to
use to synchronise the 38khz oscillator.
3.After the buffer remove the 19kHz with a notch filter,
and add one more voltage follower buffer no2.
4.Have a resistance divider between the oscillator output and buffer
no2,
and the output from the junction of the two resistor buffer
will contain L+R, and the AM wave form with the L-R modulation envelope.
5.If the modulation level of L-R is what I think it is, then
the positive going peaks of the 38khz AM wave will be one channel,
and the negative going peaks will be the other channel.
So if you had a 1khz sine wave for one channel, and no modulation for
the other channel
then the AM wave would have 1kHz modulation on the positive peaks and
no modulation on the negative peaks.
6.The signal from the divider would contain about 4Vrms of carrier, with
modulation levels
around 0.5Vrms, and this signal would need to be buffered by yet another
voltage follower no3 before being applied to a pair of diode-capacitance
detectors, and followed by a de-emphasis and filter network to remove
38khz ripple.
I have also seen some general descriptions of the use of symetrical
differential amps used in the chip decoders which could be done using
vacuum tubes but unfortunately the explanation offered on the pdfs I
have seen
is so damned confusing and inadequate that I still cannot follow
how the circuits work, and what all the wave forms are and voltage
levels.
The chip based decoder I have in a 1980 Audio Reflex tuner clearly
outperforms
my existing tubed decoder technically, although the tubed type did
sound ok..
The schematic of mine has been on the website for 4 years yet nobody
except myself has tried to improve on the design which I think
could easily be improved.
Most tubed decoders suffered from an inadequate amount of hardware to do
the job
due to accountants controlling purse strings and ppl trying to avoid
plagerizing someone else's designs, or designs becoming unecessarily
complex in order to be
unique and to claim some superiority, something I have not heard from
many stereo tubed decoders; the mono
signal often sounds clearer and better, just as the measurements would
suggest.
I am only intending to use any information for my own use.
Regards,
Patrick Turner.
|
You probably know everything on this page already, but just in case try
http://transmitters.tripod.com/stereo.htm
It is mostly SPICE stuff which may not be of interest, but buried away near
the end is the fact that the 19 KHz pilot signal is 10% of the carrier
level. The absolute values of voltages obviously depends on the gains of
your circuit, but in terms of proportions you have carrier at say 4 volts,
19Khz tone at .4 volts, and signal at 0 to 2 volt depending on degree of
modulation.
Your point number 5 is correct. A chopped version of the Left channel
signal is on top, chopped Right channel signal is on bottom, and this means
that you can use a positve peak detector for one, and a negative peak
detector for the other. It also means you can extract the signal by
switching back in forth in time at a 38Khz rate. Add these two methods to
the method of extracting upper and lower sidebands and combining in a matrix
and you have 3 categories of circuit that I am aware of.
However I do not understand step 4 of your procedure. As I read it you have
a resistor divider going from your 38K osc output to your 19Khz signal. I
dont see the point.
The circuits I have seen that begin like yours does then go on and do the
following : Take the 38Khz and apply it to the primary of a transformer.
The secondary is center tapped. To that center tap you apply the input
signal minus the 19Khz, ie the signal from your step 3. Then across the
secondary is a ring of diodes arranged such that one output gets fed the
center tap signal when the 38Khz is one polarity, and the other output gets
the center tap signal the other half cycle of 38K. This is the same as left
and right. Then apply the appropriate filtering.
You said you did not get very good seperation in your circuit, but how much
did you get ? From what I gather from chip specs 30dB is what you get from
a mediocre one. 35 dB is what my tube type Sansui 250 receiver spec sheet
claims.
Some time ago I did a survey of all the demux circuits I could find from old
radio schematics, and the RCA tube book has one, Eico and Heathkit have some
circuits and so on. I mainly just wanted to see how they worked, but also
wanted to see if a discrete component version, tube or solid state, could be
tweaked to outperform an IC version. I did SPICE simulations of them, and of
various combinations and permutations of them in attempts to "improve" them.
Now I realize SPICE is not the same as real life, but I basically I
concluded that nothing I could come up with would beat any of the better IC
chips. |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Tue Oct 25, 2005 10:24 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
Robert McLean wrote:
| Quote: | "Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435BBB27.172664E5@turneraudio.com.au...
Hi all,
I built a stereo FM multiplex decoder some years ago and the schematic
is at
http://www.turneraudio.com.au/htmlwebpgs02/schemampxdecoder.htm
If you find all of what I am saying here or at the above address all
too hard to follow, could you be so kind as to simply
direct me to a URL the where the exact specifications are listed for the
standard
stereo transmitted signal with all relevant voltage levels of
L+R signal levels and 38kHz double sideband levels.
The principle of my decoder involved filtering out the L-R DSB signal
from the main composite
signal, and applying that to one input of an LTP, then applying a
synchronised
38kHz signal to the other input to get two oppositely phased output
38khz AM waves modulated
with the L-R signal so that when the +ve and -ve going peaks in these
signals from the LTP are diode detected you get a L-R and -L+R audio
signal.
These two signals are matrixed in a resistance divider
and the L+R signal is applied to the centre of the divider you get
L and R signals.
Ripple frequency in the diode detectors is 76kHz, and the idea was that
this
can better define the higher audio frequencies.
Anyway, the the output of the decoder didn't give very good separation
which became quite poor by 15kHz, and noise and spuriae were a problem.
I am now trying to improve on the decoder yet still use vacuum tubes.
I would like to know if it is possible to decode the stereo signal more
simply
and without the phase altering filters in the following way :-
1.Use a voltage follower to buffer the main composite signal from the
ratio detector.
2.Take an output from this using a tuned filter to extract the 19kHz to
use to synchronise the 38khz oscillator.
3.After the buffer remove the 19kHz with a notch filter,
and add one more voltage follower buffer no2.
4.Have a resistance divider between the oscillator output and buffer
no2,
and the output from the junction of the two resistor buffer
will contain L+R, and the AM wave form with the L-R modulation envelope.
5.If the modulation level of L-R is what I think it is, then
the positive going peaks of the 38khz AM wave will be one channel,
and the negative going peaks will be the other channel.
So if you had a 1khz sine wave for one channel, and no modulation for
the other channel
then the AM wave would have 1kHz modulation on the positive peaks and
no modulation on the negative peaks.
6.The signal from the divider would contain about 4Vrms of carrier, with
modulation levels
around 0.5Vrms, and this signal would need to be buffered by yet another
voltage follower no3 before being applied to a pair of diode-capacitance
detectors, and followed by a de-emphasis and filter network to remove
38khz ripple.
I have also seen some general descriptions of the use of symetrical
differential amps used in the chip decoders which could be done using
vacuum tubes but unfortunately the explanation offered on the pdfs I
have seen
is so damned confusing and inadequate that I still cannot follow
how the circuits work, and what all the wave forms are and voltage
levels.
The chip based decoder I have in a 1980 Audio Reflex tuner clearly
outperforms
my existing tubed decoder technically, although the tubed type did
sound ok..
The schematic of mine has been on the website for 4 years yet nobody
except myself has tried to improve on the design which I think
could easily be improved.
Most tubed decoders suffered from an inadequate amount of hardware to do
the job
due to accountants controlling purse strings and ppl trying to avoid
plagerizing someone else's designs, or designs becoming unecessarily
complex in order to be
unique and to claim some superiority, something I have not heard from
many stereo tubed decoders; the mono
signal often sounds clearer and better, just as the measurements would
suggest.
I am only intending to use any information for my own use.
Regards,
Patrick Turner.
You probably know everything on this page already, but just in case try
http://transmitters.tripod.com/stereo.htm
|
I checked all that out again.
I also had another look at the composite signal from my BA1404
based FM test transmitter.
Using a 1kHz test tone to modulate one channel only,
the L-R DSB modulation is equal in amplitude to the L+R mono signal.
It must be thus for all existing decoders to work.
| Quote: | It is mostly SPICE stuff which may not be of interest, but buried away near
the end is the fact that the 19 KHz pilot signal is 10% of the carrier
level. The absolute values of voltages obviously depends on the gains of
your circuit, but in terms of proportions you have carrier at say 4 volts,
19Khz tone at .4 volts, and signal at 0 to 2 volt depending on degree of
modulation.
|
Well, its not quite like that.
Max F deviation allowed is +/- 75kHz at 100Mhz.
So the pilot tone is at a constant voltage that produces 1/10 of the max
deviation,
ie, a voltage that causes 7.5kHz deviation.
It means that there would be plenty of programme material that would have
less audio amplitude than the pilot tone, a good reason the reduce it
with a notch filter; some tuners just don't bother....
In my BA1404 tester, the 19khz pilot is a damned square wave because the guy i
bought a transmitter
module from took the cheap and nasty option of not having a filter to
make the pilot tone a sine wave.
The app notes on BA1404 tell you about this as well as at the site
http://transmitters.tripod.com/stereo.htm
| Quote: |
Your point number 5 is correct. A chopped version of the Left channel
signal is on top, chopped Right channel signal is on bottom, and this means
that you can use a positve peak detector for one, and a negative peak
detector for the other. It also means you can extract the signal by
switching back in forth in time at a 38Khz rate. Add these two methods to
the method of extracting upper and lower sidebands and combining in a matrix
and you have 3 categories of circuit that I am aware of.
|
Switching would be a pain with tubes, and you have you looked at the
wave form for a 15khz sine wave expressed by a 38 khz carrier?
Its a mess.
| Quote: |
However I do not understand step 4 of your procedure. As I read it you have
a resistor divider going from your 38K osc output to your 19Khz signal. I
dont see the point.
|
Its a way of adding the carrier to the main composite signal after removing the
19khz pilot.
But I have worked a better way; it'd be far easier to foget the R divider.
The buffered composite signal comes at low impedance from a cathode follower,
and an output winding on the 38khz oscillator tranny can simply
be connected at one end to the buffer output, and you have
no voltage losses, so that if the L+R plus DSB is 2Vrms, the adding 5Vrms of
oscillator signal is about right to apply to the high impedance input of
another buffer to drive the detection diodes for L and R.
But the ripple F of the detectors is at 38khz, and I see no way of having
it at 76kHz unless full wave detection is done but I have no idea if
that would work while there is the L+R signal present.
| Quote: |
The circuits I have seen that begin like yours does then go on and do the
following : Take the 38Khz and apply it to the primary of a transformer.
The secondary is center tapped. To that center tap you apply the input
signal minus the 19Khz, ie the signal from your step 3. Then across the
secondary is a ring of diodes arranged such that one output gets fed the
center tap signal when the 38Khz is one polarity, and the other output gets
the center tap signal the other half cycle of 38K. This is the same as left
and right. Then apply the appropriate filtering.
You said you did not get very good seperation in your circuit, but how much
did you get ? From what I gather from chip specs 30dB is what you get from
a mediocre one. 35 dB is what my tube type Sansui 250 receiver spec sheet
claims.
|
The chip gives about 40+ dB of separation.
Trouble is with many decoders is that the phaseshifts with filters and cause
incomplete
cancelation in the R matrixes, so that although 0ver 35db sep is possible at
200Hz, it fall to only
12dB at 10k if you are lucky.
I am still fidlin round to see if I can improve on matters.
| Quote: |
Some time ago I did a survey of all the demux circuits I could find from old
radio schematics, and the RCA tube book has one, Eico and Heathkit have some
circuits and so on. I mainly just wanted to see how they worked, but also
wanted to see if a discrete component version, tube or solid state, could be
tweaked to outperform an IC version. I did SPICE simulations of them, and of
various combinations and permutations of them in attempts to "improve" them.
Now I realize SPICE is not the same as real life, but I basically I
concluded that nothing I could come up with would beat any of the better IC
chips.
|
That tells you that if you were to use the same techniques as inside the chips
then
perhaps you would get the same outcome, ie,
better snr, and lower thd and better separtation to a higher frequency.
But what is done in the chips?
Nobody sems to explain their workings for dummies.
And remember much of the chip internals are CCS, regulators, buffers etc.
One would think that synchronous detection of the L-R signal from the
reconstructed 38kHz AM wave would be possible, and matrixing would then be
easier, filters better etc.
But the chip maker high priests don't really want ppl to know what's in there.
Some have been around for 30 years; you'd think they wouldn't need to keep
secrets.
But they just give the app notes and spec and block diagram maybe, none
of the real details are properly explained.
The majority of the early cheap 3 transistor or 4 tube decoders fitted
to tuners were worse sounding in stereo than mono; sure you did have something
called stereo,
but it wasn't marvellously good.
Patrick Turner. |
|
| Back to top |
|
|
Chris Hornbeck
Guest
|
| Posted:
Wed Oct 26, 2005 4:42 am Post subject:
Re: Multiplex Decoder questions. |
|
|
On Tue, 25 Oct 2005 17:24:08 GMT, Patrick Turner
<info@turneraudio.com.au> wrote:
| Quote: | But what is done in the chips?
|
Bandwidth. You can do the same, but it's harder.
| Quote: | One would think that synchronous detection of the L-R signal from the
reconstructed 38kHz AM wave would be possible, and matrixing would then be
easier, filters better etc.
|
The subcarried information is *always* synchronously
detected. So the regenerated carrier is very much larger
than any possible sidebands; this should never approach
being an issue.
Still not sure what's bugging you about your current
results. Distortion in stereo reception is predominantly
from the IF filtering. And separation
of 20 dB midband is fine for most folks, practically
speaking. Did I miss the actual issue? (Likely; sorry.)
Good fortune,
Chris Hornbeck
Gen. Miller, Gen. Sanchez, Donald Rumsfeld, President Bush. |
|
| Back to top |
|
|
Robert McLean
Guest
|
| Posted:
Wed Oct 26, 2005 5:12 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435E6918.769F40FE@turneraudio.com.au...
| Quote: |
Robert McLean wrote:
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435BBB27.172664E5@turneraudio.com.au...
Hi all,
I built a stereo FM multiplex decoder some years ago and the schematic
is at
http://www.turneraudio.com.au/htmlwebpgs02/schemampxdecoder.htm
If you find all of what I am saying here or at the above address all
too hard to follow, could you be so kind as to simply
direct me to a URL the where the exact specifications are listed for
the
standard
stereo transmitted signal with all relevant voltage levels of
L+R signal levels and 38kHz double sideband levels.
The principle of my decoder involved filtering out the L-R DSB signal
from the main composite
signal, and applying that to one input of an LTP, then applying a
synchronised
38kHz signal to the other input to get two oppositely phased output
38khz AM waves modulated
with the L-R signal so that when the +ve and -ve going peaks in these
signals from the LTP are diode detected you get a L-R and -L+R audio
signal.
These two signals are matrixed in a resistance divider
and the L+R signal is applied to the centre of the divider you get
L and R signals.
Ripple frequency in the diode detectors is 76kHz, and the idea was that
this
can better define the higher audio frequencies.
Anyway, the the output of the decoder didn't give very good separation
which became quite poor by 15kHz, and noise and spuriae were a problem.
I am now trying to improve on the decoder yet still use vacuum tubes.
I would like to know if it is possible to decode the stereo signal more
simply
and without the phase altering filters in the following way :-
1.Use a voltage follower to buffer the main composite signal from the
ratio detector.
2.Take an output from this using a tuned filter to extract the 19kHz to
use to synchronise the 38khz oscillator.
3.After the buffer remove the 19kHz with a notch filter,
and add one more voltage follower buffer no2.
4.Have a resistance divider between the oscillator output and buffer
no2,
and the output from the junction of the two resistor buffer
will contain L+R, and the AM wave form with the L-R modulation
envelope.
5.If the modulation level of L-R is what I think it is, then
the positive going peaks of the 38khz AM wave will be one channel,
and the negative going peaks will be the other channel.
So if you had a 1khz sine wave for one channel, and no modulation for
the other channel
then the AM wave would have 1kHz modulation on the positive peaks and
no modulation on the negative peaks.
6.The signal from the divider would contain about 4Vrms of carrier,
with
modulation levels
around 0.5Vrms, and this signal would need to be buffered by yet
another
voltage follower no3 before being applied to a pair of
diode-capacitance
detectors, and followed by a de-emphasis and filter network to remove
38khz ripple.
I have also seen some general descriptions of the use of symetrical
differential amps used in the chip decoders which could be done using
vacuum tubes but unfortunately the explanation offered on the pdfs I
have seen
is so damned confusing and inadequate that I still cannot follow
how the circuits work, and what all the wave forms are and voltage
levels.
The chip based decoder I have in a 1980 Audio Reflex tuner clearly
outperforms
my existing tubed decoder technically, although the tubed type did
sound ok..
The schematic of mine has been on the website for 4 years yet nobody
except myself has tried to improve on the design which I think
could easily be improved.
Most tubed decoders suffered from an inadequate amount of hardware to
do
the job
due to accountants controlling purse strings and ppl trying to avoid
plagerizing someone else's designs, or designs becoming unecessarily
complex in order to be
unique and to claim some superiority, something I have not heard from
many stereo tubed decoders; the mono
signal often sounds clearer and better, just as the measurements would
suggest.
I am only intending to use any information for my own use.
Regards,
Patrick Turner.
You probably know everything on this page already, but just in case try
http://transmitters.tripod.com/stereo.htm
I checked all that out again.
I also had another look at the composite signal from my BA1404
based FM test transmitter.
Using a 1kHz test tone to modulate one channel only,
the L-R DSB modulation is equal in amplitude to the L+R mono signal.
It must be thus for all existing decoders to work.
It is mostly SPICE stuff which may not be of interest, but buried away
near
the end is the fact that the 19 KHz pilot signal is 10% of the carrier
level. The absolute values of voltages obviously depends on the gains of
your circuit, but in terms of proportions you have carrier at say 4
volts,
19Khz tone at .4 volts, and signal at 0 to 2 volt depending on degree of
modulation.
Well, its not quite like that.
Max F deviation allowed is +/- 75kHz at 100Mhz.
So the pilot tone is at a constant voltage that produces 1/10 of the max
deviation,
ie, a voltage that causes 7.5kHz deviation.
It means that there would be plenty of programme material that would have
less audio amplitude than the pilot tone, a good reason the reduce it
with a notch filter; some tuners just don't bother....
In my BA1404 tester, the 19khz pilot is a damned square wave because the
guy i
bought a transmitter
module from took the cheap and nasty option of not having a filter to
make the pilot tone a sine wave.
The app notes on BA1404 tell you about this as well as at the site
http://transmitters.tripod.com/stereo.htm
Your point number 5 is correct. A chopped version of the Left channel
signal is on top, chopped Right channel signal is on bottom, and this
means
that you can use a positve peak detector for one, and a negative peak
detector for the other. It also means you can extract the signal by
switching back in forth in time at a 38Khz rate. Add these two methods
to
the method of extracting upper and lower sidebands and combining in a
matrix
and you have 3 categories of circuit that I am aware of.
Switching would be a pain with tubes, and you have you looked at the
wave form for a 15khz sine wave expressed by a 38 khz carrier?
Its a mess.
However I do not understand step 4 of your procedure. As I read it you
have
a resistor divider going from your 38K osc output to your 19Khz signal.
I
dont see the point.
Its a way of adding the carrier to the main composite signal after
removing the
19khz pilot.
But I have worked a better way; it'd be far easier to foget the R divider.
The buffered composite signal comes at low impedance from a cathode
follower,
and an output winding on the 38khz oscillator tranny can simply
be connected at one end to the buffer output, and you have
no voltage losses, so that if the L+R plus DSB is 2Vrms, the adding 5Vrms
of
oscillator signal is about right to apply to the high impedance input of
another buffer to drive the detection diodes for L and R.
But the ripple F of the detectors is at 38khz, and I see no way of having
it at 76kHz unless full wave detection is done but I have no idea if
that would work while there is the L+R signal present.
The circuits I have seen that begin like yours does then go on and do the
following : Take the 38Khz and apply it to the primary of a transformer.
The secondary is center tapped. To that center tap you apply the input
signal minus the 19Khz, ie the signal from your step 3. Then across the
secondary is a ring of diodes arranged such that one output gets fed the
center tap signal when the 38Khz is one polarity, and the other output
gets
the center tap signal the other half cycle of 38K. This is the same as
left
and right. Then apply the appropriate filtering.
You said you did not get very good seperation in your circuit, but how
much
did you get ? From what I gather from chip specs 30dB is what you get
from
a mediocre one. 35 dB is what my tube type Sansui 250 receiver spec
sheet
claims.
The chip gives about 40+ dB of separation.
Trouble is with many decoders is that the phaseshifts with filters and
cause
incomplete
cancelation in the R matrixes, so that although 0ver 35db sep is possible
at
200Hz, it fall to only
12dB at 10k if you are lucky.
I am still fidlin round to see if I can improve on matters.
|
with regard to separation at higher frequencies, the specs for the HH Scott
310E tuner claim 34dB @400 Hz, 30 dB @ 3KHz, and 28dB @ 12KHz. So it seems
that it is possible to do much better than the run of the mill decoders.
Now, in this case it is done with a fairly elaborate looking circuit using
10 tube sections, not counting 4 more tubes for driving relays, and 12
diodes. I only recently came across this, and I have not really figured out
all its details, but you can see it for yourself at
http://www.mcmlv.org/Archive/HiFi/Scott310E.pdf , if you have not already
done so. You may find some ideas for your fiddling in it.
| Quote: |
Some time ago I did a survey of all the demux circuits I could find from
old
radio schematics, and the RCA tube book has one, Eico and Heathkit have
some
circuits and so on. I mainly just wanted to see how they worked, but
also
wanted to see if a discrete component version, tube or solid state, could
be
tweaked to outperform an IC version. I did SPICE simulations of them, and
of
various combinations and permutations of them in attempts to "improve"
them.
Now I realize SPICE is not the same as real life, but I basically I
concluded that nothing I could come up with would beat any of the better
IC
chips.
That tells you that if you were to use the same techniques as inside the
chips
then
perhaps you would get the same outcome, ie,
better snr, and lower thd and better separtation to a higher frequency.
But what is done in the chips?
Nobody sems to explain their workings for dummies.
And remember much of the chip internals are CCS, regulators, buffers etc.
One would think that synchronous detection of the L-R signal from the
reconstructed 38kHz AM wave would be possible, and matrixing would then be
easier, filters better etc.
But the chip maker high priests don't really want ppl to know what's in
there.
Some have been around for 30 years; you'd think they wouldn't need to keep
secrets.
But they just give the app notes and spec and block diagram maybe, none
of the real details are properly explained.
The majority of the early cheap 3 transistor or 4 tube decoders fitted
to tuners were worse sounding in stereo than mono; sure you did have
something
called stereo,
but it wasn't marvellously good.
Patrick Turner.
|
|
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Wed Oct 26, 2005 5:33 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
Robert McLean wrote:
| Quote: | "Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435E6918.769F40FE@turneraudio.com.au...
Robert McLean wrote:
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435BBB27.172664E5@turneraudio.com.au...
Hi all,
|
snip,
| Quote: |
I am still fidlin round to see if I can improve on matters.
with regard to separation at higher frequencies, the specs for the HH Scott
310E tuner claim 34dB @400 Hz, 30 dB @ 3KHz, and 28dB @ 12KHz. So it seems
that it is possible to do much better than the run of the mill decoders.
|
Those specs are good.
| Quote: |
Now, in this case it is done with a fairly elaborate looking circuit using
10 tube sections, not counting 4 more tubes for driving relays, and 12
diodes. I only recently came across this, and I have not really figured out
all its details, but you can see it for yourself at
http://www.mcmlv.org/Archive/HiFi/Scott310E.pdf , if you have not already
done so. You may find some ideas for your fiddling in it.
|
Yeah, the Scott is complex.
I went right through it and decided to go more in my own direction, but maybe
the filtering
ideas are worth a look...
Patrick Turner. |
|
| Back to top |
|
|
John Byrns
Guest
|
| Posted:
Wed Oct 26, 2005 9:25 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
In article <%oK7f.6045$ki7.356040@news20.bellglobal.com>, "Robert McLean"
<robert.mclean1@sympatico.ca> wrote:
| Quote: | "Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435E6918.769F40FE@turneraudio.com.au...
The chip gives about 40+ dB of separation.
Trouble is with many decoders is that the phaseshifts with filters and
cause
incomplete
cancelation in the R matrixes, so that although 0ver 35db sep is possible
at
200Hz, it fall to only
12dB at 10k if you are lucky.
I am still fidlin round to see if I can improve on matters.
with regard to separation at higher frequencies, the specs for the HH Scott
310E tuner claim 34dB @400 Hz, 30 dB @ 3KHz, and 28dB @ 12KHz. So it seems
that it is possible to do much better than the run of the mill decoders.
Now, in this case it is done with a fairly elaborate looking circuit using
10 tube sections, not counting 4 more tubes for driving relays, and 12
diodes. I only recently came across this, and I have not really figured out
all its details, but you can see it for yourself at
http://www.mcmlv.org/Archive/HiFi/Scott310E.pdf , if you have not already
done so. You may find some ideas for your fiddling in it.
|
A slight correction, only 8 of the tube sections are part of the actual
multiplex decoder in the 310E, the other 6 tube sections drive the
relays. Of the 8 tube sections in the actual multiplex decoder, 4
sections are used to amplify the composite signal, separate the 19 kHz
pilot, and regenerate the 38 kHz subcarrrier, the other 4 tube sections
are used in the left and right audio output amplifiers, which include the
auxiliary matrix circuit. A total of 10 semiconductor diodes are used in
the actual multiplex decoder, 2 in the 19/38 kHz doubler circuit, and 8 in
the switching bridges. 5 more semiconductor diodes are used in the relay
driver circuits.
The H.H.Scott 4310 circuit is similar and can be found here:
http://users.rcn.com/jbyrns/pics/4310-p1.jpg
http://users.rcn.com/jbyrns/pics/4310-p2.jpg
The 4310 circuit ups the tube section count in the actual multiplex
decoder to 10, excluding the switching diodes, by adding 2 more tube
sections to the left and right audio amplifier stages upping the total
number of tube sections used in the audio stages to 6, while the other 4
sections are used in the subcarrier processing section as with the 310E.
The 4310 replaces the 8 semiconductor diodes the 310E uses in the
switching bridge with 8 vacuum diodes, to give the 4310 that nice tube
sound. Counting the 8 vacuum diodes in the switching bridges, the 4310
uses 18 tube sections in the actual multiplex decoder, including the audio
stages. The 4310 also uses vacuum diodes in the relay driver circuits.
The relay driver circuits in the 4310 are somewhat more complex than in
the 310E because the 4310 has diversity reception capability.
Regards,
John Byrns
Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Thu Oct 27, 2005 4:42 am Post subject:
Re: Multiplex Decoder questions. |
|
|
John Byrns wrote:
| Quote: | In article <%oK7f.6045$ki7.356040@news20.bellglobal.com>, "Robert McLean"
robert.mclean1@sympatico.ca> wrote:
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:435E6918.769F40FE@turneraudio.com.au...
The chip gives about 40+ dB of separation.
Trouble is with many decoders is that the phaseshifts with filters and
cause
incomplete
cancelation in the R matrixes, so that although 0ver 35db sep is possible
at
200Hz, it fall to only
12dB at 10k if you are lucky.
I am still fidlin round to see if I can improve on matters.
with regard to separation at higher frequencies, the specs for the HH Scott
310E tuner claim 34dB @400 Hz, 30 dB @ 3KHz, and 28dB @ 12KHz. So it seems
that it is possible to do much better than the run of the mill decoders.
Now, in this case it is done with a fairly elaborate looking circuit using
10 tube sections, not counting 4 more tubes for driving relays, and 12
diodes. I only recently came across this, and I have not really figured out
all its details, but you can see it for yourself at
http://www.mcmlv.org/Archive/HiFi/Scott310E.pdf , if you have not already
done so. You may find some ideas for your fiddling in it.
A slight correction, only 8 of the tube sections are part of the actual
multiplex decoder in the 310E, the other 6 tube sections drive the
relays. Of the 8 tube sections in the actual multiplex decoder, 4
sections are used to amplify the composite signal, separate the 19 kHz
pilot, and regenerate the 38 kHz subcarrrier, the other 4 tube sections
are used in the left and right audio output amplifiers, which include the
auxiliary matrix circuit. A total of 10 semiconductor diodes are used in
the actual multiplex decoder, 2 in the 19/38 kHz doubler circuit, and 8 in
the switching bridges. 5 more semiconductor diodes are used in the relay
driver circuits.
The H.H.Scott 4310 circuit is similar and can be found here:
http://users.rcn.com/jbyrns/pics/4310-p1.jpg
http://users.rcn.com/jbyrns/pics/4310-p2.jpg
The 4310 circuit ups the tube section count in the actual multiplex
decoder to 10, excluding the switching diodes, by adding 2 more tube
sections to the left and right audio amplifier stages upping the total
number of tube sections used in the audio stages to 6, while the other 4
sections are used in the subcarrier processing section as with the 310E.
The 4310 replaces the 8 semiconductor diodes the 310E uses in the
switching bridge with 8 vacuum diodes, to give the 4310 that nice tube
sound. Counting the 8 vacuum diodes in the switching bridges, the 4310
uses 18 tube sections in the actual multiplex decoder, including the audio
stages. The 4310 also uses vacuum diodes in the relay driver circuits.
The relay driver circuits in the 4310 are somewhat more complex than in
the 310E because the 4310 has diversity reception capability.
Regards,
John Byrns
Surf my web pages at, http://users.rcn.com/jbyrns/
|
Well John, the schematics are more complex than I recall when i first had a look
some years back.
Is there a block diagram giving the general manner in which stereo decoding is
done with the
shapes of the filters used in each block element and with the wave forms?
Or do we have to re-draw the whole lot out so its many basic blocks can more
easily
be seen and understood?
Is there a general description of the circuit which ordinary ppl like myself
can understand?
Do you HTF the darn thing works?
Patrick Turner. |
|
| Back to top |
|
|
Robert McLean
Guest
|
| Posted:
Thu Oct 27, 2005 6:55 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
"John Byrns" <jbyrns@rcn.com> wrote in message
news:jbyrns-2610051125450001@216-80-74-164.d.enteract.com...
Thanks. More circuitry to digest. |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Thu Oct 27, 2005 9:42 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
Robert McLean wrote:
The 38khz oscillator I dreamed up a few years ago
dropped one of its diodes, and that created some 19kHz that
leaked through to the output, so I rebuilt the whole oscillator.
It did have a 19kHz triode amp ( 1/2 6U8A) with selective filter in
front of it,
and that fed the other pentode 1/2 as a synchronised 19khz oscillator
with a pair of diodes to double the F
into a tuned 38kHz tranny.
one of those diodes went strange as germanium does.
But now I have the same triode amp to amp up the 19kHz piloy being
received, then
the more usual doubler diodes and then a tuned 38khz tranny loading the
pentode.
I will have to make an untuned tranny for the triode input stage because
after the mod the
input stage oscilated with a tuned grid and tuned plate circuit, and the
following
38kHz oscillator wouldn't lock onto the diode pulses from the first
stage, so far more easy oscillation
than I had before, but not synchronised to the input 19khz pilot.
With every mod there is always the unexpected problems of things working
not quite as expected,
especially with oscillator circuits that one wants to synchronise with a
small input signal.
With both stages trying to oscillate, they went a bit LF unstable, and
the oscilations
undulated if the 19khz signal from a station was too low.
So I have to get the thresholds right and range of lock in F and the
stability right.
I have a lot to read about on decoders, but not one that
has diode detectors to give a detected L-R and -L+R signal using 76kHz,
which seems to me to be only possible in the scheme I have been using.
This means filtering the L+R signal out of the composite signal from the
ratio detector
and then adding the 38kHz carrier, so a pair of symetrical AM waves of
opposite
phase can then charge diodes at a 76khz rate making filtering out the
ripple much easier,
but the filtering prevents the phase tracking one must have to get good
separation at HF.
Many decoders do have filtering of signals before the matrixing, so
good separation at 15khz is impossible.
I revised the BA1404 circuit with stereo audio input signals so that the
19kHz isn't a square wave but a sine wave.
This meant adding an LC circuit to the board and a couple or other R&C
parts,
and when i tested it i got the 19khz sine wave but still rather
distorted with what looks like
about 20% 5th harmonic, and nobody on the web who promoted this add on
circuit said that this would occur. So like all the experts one sees
saying how it ought to be done,
one has to expect their advice isn't always full or perfect.
But when i retested the stereo FM generator using an old solid state
Audio Reflex tuner
a I got very nice results, with slightly less noise and more separation,
so the mod didn't upset the BA1404 chip operation.
I may finish up installing a spare 38kHz transformer to act as a drive
element
for the 4 diode balanced demodulator, and settling for the 38khz ripple,
but then I think the shape of that is more benign than the usual saw
tooth one gets with
a normal diode charging a cap with R discharging it, so perhaps
filtering after the diodes
won't be too bad.
I don't think may people understand MPX decoders here. Its just that one
step too hard,
like anything to do with TVs. Such devices as FM radios and TVs and
videos all have
routine techniques which are quite difficult to understand if you don't
deal with them everyday.
One never sees posts about TV sets here.
One decoder circuit I have looks so very simple with a
tube that has deflector plates so the cathode current can be swung to
one or the other of two anodes
at the 38kHz rate.
But the tubes for these have not been made for 40 years, and won't be
again;
maybe there is a way of using a twin triode and a couple of diodes to
act as the beam switcher,
but at the end of the day i doubt the stereo outcome is any more
listenable than a Scott.
The Scott circuits have the refinement of muting and threshold action
on tuning; they are civilised to use, now howls of noise between
stations etc.
So considerable parts of the circuit are devoted to such bells and
whistles.
But there is considerable filtering going on, and I need to study rather
a lot more.
I somehow think that despite John Byne's avid interest in the venerable
Scott
machines, he may not be totally familiar with exactly how the MPX
section works.
There is no spread sheet I know with all the typical wave forms at all
parts of the circuit
so its easy to visualize what takes place.
Patrick Turner. |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Sun Oct 30, 2005 9:58 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
I am offering further explanation of matters I raised below.
Patrick Turner wrote:
| Quote: | Robert McLean wrote:
"John Byrns" <jbyrns@rcn.com> wrote in message
news:jbyrns-2610051125450001@216-80-74-164.d.enteract.com...
The H.H.Scott 4310 circuit is similar and can be found here:
http://users.rcn.com/jbyrns/pics/4310-p1.jpg
http://users.rcn.com/jbyrns/pics/4310-p2.jpg
Thanks. More circuitry to digest.
The 38khz oscillator I dreamed up a few years ago
dropped one of its diodes, and that created some 19kHz that
leaked through to the output, so I rebuilt the whole oscillator.
It did have a 19kHz triode amp ( 1/2 6U8A) with selective filter in
front of it,
and that fed the other pentode 1/2 as a synchronised 19khz oscillator
with a pair of diodes to double the F
into a tuned 38kHz tranny.
one of those diodes went strange as germanium does.
But now I have the same triode amp to amp up the 19kHz piloy being
received, then
the more usual doubler diodes and then a tuned 38khz tranny loading the
pentode.
I will have to make an untuned tranny for the triode input stage because
after the mod the
input stage oscilated with a tuned grid and tuned plate circuit, and the
following
38kHz oscillator wouldn't lock onto the diode pulses from the first
stage, so far more easy oscillation
than I had before, but not synchronised to the input 19khz pilot.
With every mod there is always the unexpected problems of things working
not quite as expected,
especially with oscillator circuits that one wants to synchronise with a
small input signal.
With both stages trying to oscillate, they went a bit LF unstable, and
the oscilations
undulated if the 19khz signal from a station was too low.
So I have to get the thresholds right and range of lock in F and the
stability right.
|
I got a suitable new tranny wound to use with the 19khz amplifier prior
to conversion to 38kHz by means of fullwave diode rectification
of the amplified 19khz pilot tone.
It is a solenoid winding of about 30mm long on a
piece of 10mm dia ferrite rod.
there are 3 x 1,000 turn windings of 0.2mm dia wire each so they can be
arranged S-P-S.
The P is taken to some B+, then its other end to a collector circuit of a
pair
of darlington connected transistors. No point using a triode to do the
donkey work of picking out and amplifying the pilot tone, when transistors
have far more gain.
The tranny gives to give a 1:2 step up ratio, with the secondary having a
grounded CT, and then
each end has diodes to give negative waves of rectification into a resistoe
of around 470k,
which will then bias a 12AU7 twin triode oscillator similar to that shown in
the Scott circuits.
With some fiddling, this arrangement should be more stable and work a lot
better than a
6U8A triode-pentode so that a harmonic free 38khz tone is produced
that is not affected by audio modulation leakage, and the phase relationship
between the
pilot tone and the re-constructed carrier wave of 38khz should be more
accurate.
| Quote: |
I have a lot to read about on decoders, but not one that
has diode detectors to give a detected L-R and -L+R signal using 76kHz,
which seems to me to be only possible in the scheme I have been using.
This means filtering the L+R signal out of the composite signal from the
ratio detector
and then adding the 38kHz carrier, so a pair of symetrical AM waves of
opposite
phase can then charge diodes at a 76khz rate making filtering out the
ripple much easier,
but the filtering prevents the phase tracking one must have to get good
separation at HF.
Many decoders do have filtering of signals before the matrixing, so
good separation at 15khz is impossible.
|
I searched to find some graphs of the phase response of the modulation
envelope shape
in an AM wave form but there was nothing except incomprehensible maths.
I set up a tuned LC filter tuned to 38kHz with a low Q and with 10k
between a low Z sig gene source and the LC I was able to plot the F response
of the
filter.
Its -3db points were at 23kHz and 53kHz.
At the tuned F the phase shift between the input signal at 38 kHz and the
output of the filter
was 0d.
There was about +45d lead in the phase at 23kHz, and a lag of 45d at 53kHz.
So I wondered how the filter affected the phase of the modulation envelope.
Where one has a 38khz carrier modulated by say 15kHz sine wave, there are
actually
3 frequencies present, the carrier at 38khz, and the lower "sideband" of
23kHz,
and upper sideband of 53kHz.
So what would be the phase relationship of the envelope shape at 15kHz of
modulation?
So I set up a 38khz wave and amplitude modulated it with a variable F
oscillator,
then fed it into the tuned LC filter with a dual trace CRO to monitor
the modulation from the AF oscillator and the 38khz modulated wave at the
output of the filter.
Sure enough the shape of the envelope is a lagging one of about -50d with
15kHz of AM.
At low AF there is negligible phase shift, but it increases according to
whatever the filter
does be it tuned BP filter or HP filter.
So OK in my original MPX decoder design I have filtered out the "double
sideband signal",
DSB from the main composite signal from the ratio detector and added
that to an LTP with 38khz carrier generated by a locked oscilator.
The resulting two AM envelopes in at the LTP anodes will produce
delayed AF -L+R and +L-R signals because of the the HPF action
used to separate the DSB from the comp signal.
So the addition of the L+R signal also filtered out by a LPF from the comp
signal must have as near as possible a phase response of the
delays in the modulation recoved by the diode detectors working from the
LTP.
This seems doable since the modulation recovered from the 38khz AM signal
is delayed, not phase advanced, and thus the addition and subtraction in the
resistor matrix will give reasonable separation, because the L+R signal
at the output of the LPF is also delayed..
| Quote: |
I revised the BA1404 circuit with stereo audio input signals so that the
19kHz isn't a square wave but a sine wave.
This meant adding an LC circuit to the board and a couple or other R&C
parts,
and when i tested it i got the 19khz sine wave but still rather
distorted with what looks like
about 20% 5th harmonic, and nobody on the web who promoted this add on
circuit said that this would occur. So like all the experts one sees
saying how it ought to be done,
one has to expect their advice isn't always full or perfect.
But when i retested the stereo FM generator using an old solid state
Audio Reflex tuner
a I got very nice results, with slightly less noise and more separation,
so the mod didn't upset the BA1404 chip operation.
I may finish up installing a spare 38kHz transformer to act as a drive
element
for the 4 diode balanced demodulator, and settling for the 38khz ripple,
but then I think the shape of that is more benign than the usual saw
tooth one gets with
a normal diode charging a cap with R discharging it, so perhaps
filtering after the diodes
won't be too bad.
I don't think may people understand MPX decoders here. Its just that one
step too hard,
like anything to do with TVs. Such devices as FM radios and TVs and
videos all have
routine techniques which are quite difficult to understand if you don't
deal with them everyday.
One never sees posts about TV sets here.
One decoder circuit I have looks so very simple with a
tube that has deflector plates so the cathode current can be swung to
one or the other of two anodes
at the 38kHz rate.
But the tubes for these have not been made for 40 years, and won't be
again;
maybe there is a way of using a twin triode and a couple of diodes to
act as the beam switcher,
but at the end of the day i doubt the stereo outcome is any more
listenable than a Scott.
The Scott circuits have the refinement of muting and threshold action
on tuning; they are civilised to use, now howls of noise between
stations etc.
So considerable parts of the circuit are devoted to such bells and
whistles.
But there is considerable filtering going on, and I need to study rather
a lot more.
I somehow think that despite John Byne's avid interest in the venerable
Scott
machines, he may not be totally familiar with exactly how the MPX
section works.
There is no spread sheet I know with all the typical wave forms at all
parts of the circuit
so its easy to visualize what takes place.
|
From what I have studied after re-reading what John Byrns sent to me
about 6 years ago it appears the Scott is a "switcher" decoder.
The 353 unit for which I have a schematic uses a balanced network of
resistors and
8 solid state diodes and a balanced output from a locked ( synchronised to
the 19khz pilot tone )
38khz oscilator and applies the composite signal to the demodulator diode
circuit
that switches the currents at a rate of 38khz thus
producing the L and R signal in a train of 38khz steps.
The 19khz pilot and switching artifacts are filtered away
after the "matrixing" is done, but considerable phase delay of the recovered
final audio
signals is caused by the final filtering. De-emphasis is also finally
added to the filtering which counters the emphasis and phase advance to
all of the audio enetering the encoder at the FM transmitter.
I think I will try my idea once again before resorting to doing a version of
the
Scott method which looks so far to be a very nice unit
with very good specs, so it should sound well, but if i can
at least get 25db of separation up to 10khz with low thd and noise i will be
happy.
Patrick Turner. |
|
| Back to top |
|
|
Robert McLean
Guest
|
| Posted:
Mon Oct 31, 2005 7:18 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
"Patrick Turner" <info@turneraudio.com.au> wrote in message
news:4364EC63.DDAA5119@turneraudio.com.au...
| Quote: | I am offering further explanation of matters I raised below.
|
.... some discussion snipped ... ( or should I say notch filtered considering
the subject matter ? )
| Quote: | There is no spread sheet I know with all the typical wave forms at all
parts of the circuit
so its easy to visualize what takes place.
|
I have been working on such a thing for some time now ( ie months )based on
the Spice models I have developed for several demux circuits, both from
actual examples, and also by using ideal circuit building blocks. I may
have something I can post in a week or two.
| Quote: | From what I have studied after re-reading what John Byrns sent to me
about 6 years ago it appears the Scott is a "switcher" decoder.
The 353 unit for which I have a schematic uses a balanced network of
resistors and
8 solid state diodes and a balanced output from a locked ( synchronised to
the 19khz pilot tone )
38khz oscilator and applies the composite signal to the demodulator diode
circuit
that switches the currents at a rate of 38khz thus
producing the L and R signal in a train of 38khz steps.
The 19khz pilot and switching artifacts are filtered away
after the "matrixing" is done, but considerable phase delay of the
recovered
final audio
signals is caused by the final filtering. De-emphasis is also finally
added to the filtering which counters the emphasis and phase advance to
all of the audio enetering the encoder at the FM transmitter.
|
Of all the dozens of circuit diagrams I looked at, all but 2 were of the
switcher type. The exceptions were the one in the RCA tube manual, and an
EICO one. Of the switchers only the Scott used the 8 diodes, ie dual bridge
arrangement. The rest use 4 diodes in a ring that accomplishes the same
thing. This circuit is most common probably because it is simplest. |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Mon Oct 31, 2005 8:37 pm Post subject:
Re: Multiplex Decoder questions. |
|
|
Robert McLean wrote:
| Quote: | "Patrick Turner" <info@turneraudio.com.au> wrote in message
news:4364EC63.DDAA5119@turneraudio.com.au...
I am offering further explanation of matters I raised below.
... some discussion snipped ... ( or should I say notch filtered considering
the subject matter ? )
There is no spread sheet I know with all the typical wave forms at all
parts of the circuit
so its easy to visualize what takes place.
I have been working on such a thing for some time now ( ie months )based on
the Spice models I have developed for several demux circuits, both from
actual examples, and also by using ideal circuit building blocks. I may
have something I can post in a week or two.
From what I have studied after re-reading what John Byrns sent to me
about 6 years ago it appears the Scott is a "switcher" decoder.
The 353 unit for which I have a schematic uses a balanced network of
resistors and
8 solid state diodes and a balanced output from a locked ( synchronised to
the 19khz pilot tone )
38khz oscilator and applies the composite signal to the demodulator diode
circuit
that switches the currents at a rate of 38khz thus
producing the L and R signal in a train of 38khz steps.
The 19khz pilot and switching artifacts are filtered away
after the "matrixing" is done, but considerable phase delay of the
recovered
final audio
signals is caused by the final filtering. De-emphasis is also finally
added to the filtering which counters the emphasis and phase advance to
all of the audio enetering the encoder at the FM transmitter.
Of all the dozens of circuit diagrams I looked at, all but 2 were of the
switcher type. The exceptions were the one in the RCA tube manual, and an
EICO one. Of the switchers only the Scott used the 8 diodes, ie dual bridge
arrangement. The rest use 4 diodes in a ring that accomplishes the same
thing. This circuit is most common probably because it is simplest.
|
I tested a tranny I wound for the 19khz detector and it works fine
and can have its primary tuned, but sec untuned to avoid the phase shift
of a pair of tuned coils.
The recipe I gave in an earlier post was self resonant at about 35kHz,
but with 1,400pF added across the 1,000 turn primary the coil became
resonant at 19khz, and with a tuned impedance of 90k, a very
nice load for a bjt or triode to drive.
I also have retained the two 19kHz LC tuned circuits to filter the 19kHz out of
the
composite signal prior to feeding to the 19khz detector amp stage because
originally I found that without this filtering the 38khz oscillator output tone
was slightly
amplitude modulated by some of the audio leaking through, and this
would cause serious distortions if not controlled.
But after giving thought to the idea of pre filtering the
composite signal to get the DSB and L+R separated, I have realised
that some L+R will get past the DSB and prevent good separation
in the circuit I have been using no matter what i do.
So I have gone back to the idea of a matrix with 4 diodes, driven into two
points
by the balanced output of a floating secondary off the 38khz oscillator.
The CT of the sec is fed a signal from a CF supplying the composite signal but
with the 19kHz filtered out.
The other two ports of the matrix are fed into CRC filters to get the L and R
outputs.
At this time I think it best to at least filter out the pilot before sending
the composite signal to the diode matrix since adding a considerable voltage of
19khz
into the diodes + C can cause intermodulation.
Anyway, I will try this before anything else.
From what i can see the wave forms at the caps that are charged by diodes is
more like a stepped wave form of a digital signal before it is filtered.
The amount of 38khz switching noise should be a lot less than a normal diode
detector where
you have diodes charging caps from a pulse in one direction only and then
depending
on the rate of the DC discharge of the cap to be about the same
as the rise time of the highest audio F sine wave.
This means ripple voltage is large; its ok in an AM radio where the carrier F =
455kHz,
and is easy to filter the ripple away without affecting the audio, but in an MPX
unit
the audio band is just under the subcarrier band.
Anyway I think the balanced diode demodulator will work as well as if I had the
ripple F = 76kHz as in my original experimental design.
It has occurred to me that one may be able to synchronise a sub carrier
to the 19kHz pilot of say 152 kHz which is 8 times the pilot tone F.
I don't know if such a scheme could be made to work so that by heterodyning
the reconstructed 38khz carrier with L-R modulation with a mixer tube such as a
6BE6
which is oscillating at 152khz to transfer the shape of the AF modulation to the
higher difference frequency of 152kHz - 38kHz, or 114kHz.
One may have to use a balanced pair of mixer tubes.
Then the frequency of diode detection would be 114kHz, and much easier to
filter away switching artifacts after the detection.
The original RCA circuit is a little primitive imho.
I have moved away from trying to build some
suitable filters before matrixing because they are just too hard to get right
and
with complementary phase shifts, then deal with the detection phase shift as
well plus the
incomplete filtering between the top of the AF band and the bottom of the
subcarrier band.
For the 38khz oscilator, all that's needed ois a 12AU7 with common cathode R,
and a an LC tuned winding to feed each anode, and from a1 a signal is fed to g2,
via CR coupling, and g1 is fed the raw 38khz doubled signal of the pilot.
It would be quite ok to use all bjts in this whole 19khz amp and
locked oscillator stage because the stage does not produce a signal that is
listened to.
It only provides the 38kHz carrier that was removed at the transmitter to
allow the L+R signal to be the majority signal to cause the F deviation.
So it does not matter what means are used to reconstruct a 38kHz carrier
as long as it is fairly pure and is very stable regardless of
what signals may be in the tuner or PS.
The R+diode marix does load the oscillator if it isn't buffered with some
followers.
but a triode oscillator would have low output impedance compared to the
collector output impedance if bjts were used so buffering here isn't a bad idea
to help control creation of intermodulation products.
Patrick Turner. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in