I have a Leak Troughline III with its own (horrible) decoder so I've been wondering about a better decoder...
Some background. To ensure reverse compatibility with mono receivers, the stereo difference signal (L-R) is transmitted using double sideband suppressed carrier modulation on a 38kHz subcarrier. DSBSC requires the original carrier to be regenerated for demodulation, so a 19kHz pilot tone is transmitted between the 30Hz-15kHz (L+R) audio and the DSBSC (L-R) signal. The usual way to demodulate DSBSC is to apply the regenerated carrier and the DSBSC modulated signal to a four-quadrant multiplier and low-pass filter the output to leave just the original modulating signal. Regenerating the carrier correctly is crucial, hence the preponderance of phase-locked loops.
I believe that there's another way of doing the job. One can apply the entire multiplex signal to a switch driven by regenerated carrier. One output of the switch is L and the other R. Obviously, the output needs some serious low-pass filtering.
What I'm thinking of is a CMOS logic/valve hybrid approach. I had some 4.864MHz crystals made (cheaper than you'd think) so that the crystal oscillator can be divided down by a CMOS 28 synchronous divider to give a 19kHz square wave that can be compared with the 19kHz pilot using an EXOR gate and an RC integrator driving a varicap diode across the crystal. That's the phase-locked loop oscillator.
A 38kHz square wave can be picked off the divider and phase split by an invertor to give a differential output and amplified by a 7N7 differential pair to produce a rather large 38kHz differential square wave.
The switch will be a 6AR8 sheet beam valve. Essentially, this is a pentode with two anodes and beam deflection plates. The rather large 38kHz square wave from the 7N7 will drive the beam deflection plates. The stereo multiplex signal will be applied to the control grid and the demodulated stereo will appear at the two anodes.
Now, I know (from looking at the tuner thread) that there are people out there who know more about receiver design than I do, so are there any pitfalls I ought to consider?
Some background. To ensure reverse compatibility with mono receivers, the stereo difference signal (L-R) is transmitted using double sideband suppressed carrier modulation on a 38kHz subcarrier. DSBSC requires the original carrier to be regenerated for demodulation, so a 19kHz pilot tone is transmitted between the 30Hz-15kHz (L+R) audio and the DSBSC (L-R) signal. The usual way to demodulate DSBSC is to apply the regenerated carrier and the DSBSC modulated signal to a four-quadrant multiplier and low-pass filter the output to leave just the original modulating signal. Regenerating the carrier correctly is crucial, hence the preponderance of phase-locked loops.
I believe that there's another way of doing the job. One can apply the entire multiplex signal to a switch driven by regenerated carrier. One output of the switch is L and the other R. Obviously, the output needs some serious low-pass filtering.
What I'm thinking of is a CMOS logic/valve hybrid approach. I had some 4.864MHz crystals made (cheaper than you'd think) so that the crystal oscillator can be divided down by a CMOS 28 synchronous divider to give a 19kHz square wave that can be compared with the 19kHz pilot using an EXOR gate and an RC integrator driving a varicap diode across the crystal. That's the phase-locked loop oscillator.
A 38kHz square wave can be picked off the divider and phase split by an invertor to give a differential output and amplified by a 7N7 differential pair to produce a rather large 38kHz differential square wave.
The switch will be a 6AR8 sheet beam valve. Essentially, this is a pentode with two anodes and beam deflection plates. The rather large 38kHz square wave from the 7N7 will drive the beam deflection plates. The stereo multiplex signal will be applied to the control grid and the demodulated stereo will appear at the two anodes.
Now, I know (from looking at the tuner thread) that there are people out there who know more about receiver design than I do, so are there any pitfalls I ought to consider?
Get yourself one of these on EBay:
Fit all the digital goodies into an early 1960's box -- (you should see my digital attenuator in my PAS preamp)
Fit all the digital goodies into an early 1960's box -- (you should see my digital attenuator in my PAS preamp)
An externally hosted image should be here but it was not working when we last tested it.
EC8010 Great minds think alike see this:
http://www.hi-fiworld.co.uk/hfw/featureshtml/timdeparadecoder.html
The trend to phase locked decoders was started in 1969 ish with the publication in Wireless World of the Portus and Heywood decoder
This prompted the bods at Motorola to come up with the MC1310 integrated circuit decoder
http://www.hi-fiworld.co.uk/hfw/featureshtml/timdeparadecoder.html
The trend to phase locked decoders was started in 1969 ish with the publication in Wireless World of the Portus and Heywood decoder
This prompted the bods at Motorola to come up with the MC1310 integrated circuit decoder
forr said:
I don't know. Am I? Obviously, I could use a pair of 6AR8 and modify the duty cycle of the signal feeding the deflection plates so that the signal is sampled. That might slightly ease the filtering at the expense of reducing output.
The Tim de P decoder is completely different (although rather elegant).
I can't see why this wouldn't work. Upon rethinking this I'm assuming that both the L+R baseband and the double side band supressed carrior L-R are fed to the grid of the 6AR8 demodulator? I suspect delays in the IF strip would have to be consistent across the IF bandwidth for this to work ok though. I'd say its worth trying..
IIRC Isn't this time domain de-multiplexing? If so this is the basis of all Scott decoders in existance during the tube era, and they worked quite well. They used (double) balanced diode demodulators to do the job. My favorite outboard decoder was the Scott 335..
IIRC Isn't this time domain de-multiplexing? If so this is the basis of all Scott decoders in existance during the tube era, and they worked quite well. They used (double) balanced diode demodulators to do the job. My favorite outboard decoder was the Scott 335..
kevinkr said:
Definitely the place to start, but I've no idea how you'd simulate the 6AR8.
Well, you do what Norm Koren did when he made up some SPICE models -- instead of going to "first principles" you take the data and form a series of equations which acurately represent the properties of the tube -- great if you have some youngster from MIT to do this for you --
Here's a link to Eric Barbour on the 6AR8 -- http://www.cgs.synth.net/tube/beam.html
It is indeed the JLH stereo decoder. The ETI article confirmed my understanding of the principle and added a few useful practical points. Obviously, the phase of the 38kHz switching waveform is critical. JLH used an adjustable all-pass network to achieve optimum phase, but a pre-loadable counter would allow my scheme to achieve correct phase and maintain it without drift. JLH used a 3rd order filter to remove the switching waveform - that should be easy enough to do, and a 6C45 cathode follower would allow a low enough output impedance to make an LC filter with air-cored inductors. The remaining problem is to make the 6AR8 operate linearly, and that will require experimentation...
Hi EC8010,
I may say that I built the JLH tuner from a Hart kit around 1985 and was very disappointed by the stereo separation and the lack of treble. I spent hours trying to fix it, I've a Philips FM multiplex generator to help me, but I failed to detect what was wrong.
I may try to connect the PLL demodulator output which is very quiet to a more standard stereo decoder but I currently live very happy with a vintage lovely little Sansui TU555 which I had carfeully aligned and which is a clear winner for my ears in a comparison involving twelve tuners.
However this does not mean at all that there is something wrong with the sample and hold approach and would be glad to know the results you will arrive at, hoping they will be better than mine.
I may say that I built the JLH tuner from a Hart kit around 1985 and was very disappointed by the stereo separation and the lack of treble. I spent hours trying to fix it, I've a Philips FM multiplex generator to help me, but I failed to detect what was wrong.
I may try to connect the PLL demodulator output which is very quiet to a more standard stereo decoder but I currently live very happy with a vintage lovely little Sansui TU555 which I had carfeully aligned and which is a clear winner for my ears in a comparison involving twelve tuners.
However this does not mean at all that there is something wrong with the sample and hold approach and would be glad to know the results you will arrive at, hoping they will be better than mine.
Hello forr, I've just been for a brisk walk in the woods and I suspect I know what was wrong with the JLH decoder. The signal is being sampled at 38kHz and once the sampled signal approaches the sampling frequency, you get aperture loss. I'd have to look it up, but I suspect it might be as much as 2dB or even 3dB at 15kHz. The problem with aperture loss is that it's phaseless, so you can't compensate for it with a simple RC network, but need to use a misterminated delay line.
The other problem I would have to solve with the 6AR8 approach is the switching spikes. Looking in from the anodes, the output problem looks like a pair of pentodes each being switched on and off at a 19kHz rate. Assuming a typical drop across the anode load resistor of 100V, switching the pentode off guarantees 100V spikes. The linearity of pentodes isn't amazing, so we need to keep the wanted signal small. Filtering away those spikes whilst leaving the audio intact won't be trivial.
One possible solution might be a pair of 6AR8 as cathode followers using the beam plates as a means of switching each valve on and off. If the cathode follower has its cathode sourced by a semiconductor CCS from a negative supply, the switching spikes could be reduced from 100V to perhaps 20V. Because the linearity of the cathode follower is much better, the audio could swing to 40Vpk-pk without appreciable distortion, thus greatly improving the ratio between audio and switching spike amplitude. I need to think some more on this...
The problems really are very similar to a non-oversampling DAC...
The other problem I would have to solve with the 6AR8 approach is the switching spikes. Looking in from the anodes, the output problem looks like a pair of pentodes each being switched on and off at a 19kHz rate. Assuming a typical drop across the anode load resistor of 100V, switching the pentode off guarantees 100V spikes. The linearity of pentodes isn't amazing, so we need to keep the wanted signal small. Filtering away those spikes whilst leaving the audio intact won't be trivial.
One possible solution might be a pair of 6AR8 as cathode followers using the beam plates as a means of switching each valve on and off. If the cathode follower has its cathode sourced by a semiconductor CCS from a negative supply, the switching spikes could be reduced from 100V to perhaps 20V. Because the linearity of the cathode follower is much better, the audio could swing to 40Vpk-pk without appreciable distortion, thus greatly improving the ratio between audio and switching spike amplitude. I need to think some more on this...
The problems really are very similar to a non-oversampling DAC...
Hi EC8010,
Thanks for your comments about the JLH decoder. My idea that there was something sub-optimal is now confirmed.
I am aware of an other sample and hold decoder, it was used in a commercial product by Harman Kardon but I do not have the schematics.
Some quick Google search gave the following results :
patent 4550424 at
http://www.pat2pdf.org/
An IC :
w.st.com/stonline/products/literature/ds/12326.pdf
regards
Thanks for your comments about the JLH decoder. My idea that there was something sub-optimal is now confirmed.
I am aware of an other sample and hold decoder, it was used in a commercial product by Harman Kardon but I do not have the schematics.
Some quick Google search gave the following results :
patent 4550424 at
http://www.pat2pdf.org/
An IC :
w.st.com/stonline/products/literature/ds/12326.pdf
regards