Early Color Television
RCA CPA Prototype Restoration
Now that most of the schematic has been traced, I will begin the long process of trying to get the chassis working.
April 4, 2012
Today I installed a bunch of new electrolytics. I just put them in parallel with the old ones for now. Nick will probably want to invest in custom replacements at some point.
I built a power supply: 2 - B+ supplies, about 400 volts each, a 6.3 v 6 amp transformer, and a couple of lower current 6.3 volt transformers.
I started fired it up. I have over 100 volts p-p at the grid of the hor. output tube. With the 6BQ6 HV regulator out, I have only 10 kv of HV, but 750v of boost. The width control changes the hor. output tube screen grid voltage from about 170 to 350. Changing it has no effect on the HV, which seems strange to me.
I would think that the HV would be around 20 kv, but the doorknob capacitor looks too small for that. Tomorrow I'll take the assembly apart so I can look at the doorknob and see if it has markings that would indicate its rating.
When I connect the HV regulator tube the HV drops to 6 kv. I put a bias supply on the grid of the regulator tube and applied up to -75 v. No effect on the 6 kv. Seems like with -75 on the grid the tube would not conduct and the HV would rise to 10 kv.
April 5, 2012
I've made some significant progress, but have a mystery. The yoke had a wire broken on it. Fixing it resulted in 17 kv of HV. I discovered the yoke problem by unplugging it. The HV didn't change. I then measured the resistance and one half was open. Removed the rivets that held the plastic cover on the rear and saw arcing at one of the terminals where the heavy wire to the plug is soldered. I removed the fine wire to the winding and re-soldered it to the terminal.
Now the mystery. The doorknob capacitor on the HV line looks to me to be at most 15 kv. It is only slightly bigger than the ones used in black and white sets. I measured it and it is 500 pf, but there are no markings on it. The negative end is returned to the focus line (about 2 kv), so at most the HV of this chassis would be 17 kv.
In reading the material about CPA there was reference to defects in the picture being visible with brighter CRTs. We know that the model 1 used 20kv. It looks to me that this set was designed for 15kv, to mask the defects, for demo purposes?
I'm not convinced that 17 kv is the design voltage. Bob Galanter has a 20 kv doorknob that is almost the same size as the one in the chassis, so I think that the chassis may be designed for 20 kv (tying the negative lead of the doorknob to the 2 kv focus voltage gives a maximum of 22 kv using a 20 kv doorknob). I may still have a fault.
The regulator doesn't work, though I know exactly how it is supposed to. I will spend some time today trying to get it working.
Also, I'll insert a signal into the crystal socket and determine what frequency the oscillator/buffer amp is tuned to.
April 6, 2012
I now have 21 kv. My power supply has two sections: 330 v. and 410 v. I had been using the 330 v sections for the tests. I moved to the 410 v section and the HV came up. The doorknob didn't explode, so it must be a 20 kv one. It is the same exact dimensions as the one Bob Galanter has that is marked 20 kv.
When I connect the 6BQ6 regulator, the voltage drops to 14 kv, and the HV control does nothing. I've changed tubes (the 6BQ6 and the 12AT7 that feed it) with no change. I've also checked (substituted) the .0005/6kv capacitor and the 5 180K resistors between the flyback windings. There are a few more parts in the circuit I haven't checked yet.
I connected a signal generator to the crystal socket and a scope to the CPA transformer. The reference oscillator and amplifier are peaked at exactly 3.58 mHz.
Connected the filament supply for the tuner/if/video sections and hooked up a signal. Got nice clean video at the output of the video amp.
Verified that I have chroma signal of some sort at the inputs of the demodulators. Tomorrow I'll sweep the chroma amp (really a cathode follower) to see what the bandpass is.
The video input to the burst amp is missing. I found a remnant of a wire at the cathode of the video cathode follower and tried connecting that point to the grid of the burst amp. Got a burst pulse out of it. I also tried connecting the burst amp input to other possible video sources and got almost identical burst output with several locations, so I'm going to assume for now that the cathode of the video cathode follower is the proper source.
Had two electrolytics explode. I checked the rest to see if they were warm and found one more to replace.
April 7, 2012
Today was a frustrating day with the prototype. I decided to work on the AGC and sync.
The AGC circuit doesn't work at all, and the control marked AGC has no effect on anything. The horizontal pulse is on V19 pin 2, and video is on pin 7. All the parts are good. The circuit is so simple there isn't much more to do, though I'm not sure I understand how it works. I will look at some RCA black and white sets from this era to see if they used the same circuit. Maybe that will give me a clue. It is possible that a part or two was clipped out.
Same thing with the sync. I have sync pulses at V17 pin 6, but nothing at V18 pin 6. The control labeled Sync has no effect. In the process I discovered that something is wrong with either my documentation or the negative power divider. Based on the resistors in the divider, I should have about -10v at B-2, but I actually have -3.
April 8, 2012
Shows why it is good to put something aside when you reach an impasse. Today I discovered that the AGC and sync tubes were missing, though the tube shields were in place.
I put in the AGC tube, but still nothing. Then I discovered a bad 4 mfd electrolytic I had missed. Replaced it, and then had some effect when I changed the AGC control. I attached a bias supply to the AGC line and determined that I needed about -3 volts on the line for an undistorted video waveform with the RF input I was supplying. The AGC control was in its full counterclockwise position when the AGC voltage was at its maximum negative (about -1 volt). Because I suspect that I need a power supply with about 400 volts output compared to the 330 I now have, I experimented with the values of the resistors to B+ and to ground connected to the AGC control. I discovered that if I replaced the 10k resistor to ground with a 2.2k, the AGC voltage could be adjusted to -3, and it changed with signal level changes. I decided to stop at that point and declare success until I get the proper power supply.
Installing the sync tube made the sync work. I now have nice pulses at the vertical and horizontal oscillators. The horizontal oscillator locks nicely. I haven't done any work on the vertical section yet, but I have a good vertical pulse on the output of the integrator.
I then used a capacitor to bridge the place where I think the delay line went (from the cathode of V9A to the grid of V9B). I have video at the output of the 3 matrix tubes.
Here is what I still need to do:
1. Get the vertical section going
2. Install a delay line from a more modern set and see if it passes video without degradation or too much loss.
3. Plug a 3.58 crystal into the oscillator and see if I can get it to work.
4. Sweep the chroma amp to see its response
5. Connect the cut wires in the demodulator circuits and install the missing parts
Then I'll see if I get anything like color out of the thing with a NTSC input.
April 9, 2012
Got the vertical section working - bad tube plus I had knocked loose a wire while documenting it.
AGC still isn't right so I'm using a bias supply for now.
I swept the chroma bandpass amp. It passes approximately 1.9 to 4.3 mHz (3 db down points).
More evidence the modifications were done by a non-RCA person. I wanted to make sure that the local oscillator was on frequency before I did any more work in the color section. I connected by spectrum analyzer and measured 102.0 mHz for a channel 3 input. It should be 107.0. I then put the analyzer on the output of the IF section and found that the video is at 40.75 and the audio at 36.25. The response is sloped about 10 db. This suggests that someone twiddled with the IF adjustments and the LO.
Next task is to do an alignment. Fortunately the IF is identical to a 1952 black and white set and I have the Sams.
April 11, 2012
Today I did a video IF alignment. It went well, almost like the Sams said it should. I now have an essentially flat response from 45.0 mHz to 42.25 mHz, down 3 db at 45.75 mHz. This is what the RCA black and white response is supposed to be. I didn't have time to check it through the tuner, which I'll do tomorrow.
April 14, 2012
The low video was due to a bad detector diode. I replaced it and did the alignment over again in case there was a difference (there was). I now have a good curve and about 4 volts of video out of the detector. Video waveform looks good.
The AGC still doesn't work, but I when I replaced the diode I put it in backwards first. The AGC did work then. I will explore whether the video input to the AGC tube was connected somewhere else (the input of the video output tube instead of the output of that tube?) with inverted polarity.
April 15, 2012
Today is the last time I'll be able to work on the chassis for a couple of weeks.
When I connected the sweep to the tuner input I had a horrible response. I decided to do the sweep again, but this time insert the signal through a floating tube shield over the osc/mixer. The RCA black and white set instructions have you insert the signal directly into the first IF transformer. The prototype has a slightly different connection from the tuner to the IF.
I redid the alignment that way (it was way off) and then looked at it through the tuner. With a few tweaks on the tuner, the response looks good. There is something intermittent in the tuner that shifts the oscillator about 2 mHz when you wiggle it, so I'll have to figure out what that is.
Now I don't have enough video output, so I need to go back to the alignment and be careful that I not only have a good response curve but also sufficient gain. Very frustrating.
The good news, though, is that before I redid the alignment I had about 4 volts out of the detector (the Sams for the black and white set says 6), and the AGC worked nicely. I put back in the original 10K resistor in the AGC control pot circuit. So I'm confident that when I get the alignment right the AGC will work.
May 16, 2012
I spent some time yesterday stuffing 4 of the electrolytics. The main advantage was that it cleared up space near the reference
oscillator so I could work on it. I have a loose connection somewhere in the oscillator, which I'll track down at some point.
I connected the missing bias to R315 near the CPA transformer. This provides about -3 volts to G3 of the two demodulators. By adjusting C310 (reference osc. amp plate) I was able to get demodulated color, in sync, out of both demodulators. Waveform is very noisy and I don't think I have nearly enough output. I think this may be due to 3.58 output from the osc/amp being much too low.
Something is still not right with the AGC. Yesterday I could get it to be stable by paralleling a 1 meg resistor across the 2-5.6 meg
ones in the AGC tube circuit. Today it doesn't work. I don't think it is temperature sensitive. For now I'm using a bias supply for the AGC voltage.
One of the problems with working on this set is that parts have been
clipped out, so when something doesn't work I have that added
May 17, 2012
Today was one step forward, two backward. I think I got the AGC working by reducing the 340 supply to about 300. However, just as I was making final measurements I banged on the chassis to clear up an intermittent problem in the tuner (I'll deal with it tomorrow), I knocked a loose nut (left over from the stuffing the electrolytics) into a position where it shorted out the 310 v line, frying the two centering pots. Amazingly, they still have continuity, but the cases have some sort of short to the wipers. For now I've just isolated them from the chassis until I can find out what is wrong.
Now I have a problem with the sync, which, of course, affects the AGC because the horizontal sweep is not locked.
May 19, 2012
I have never worked on a set that has had so many faults.
Today I worked on instability in the horizontal circuit. First I found that the hor. locking range trimmer had broken down and had about 400 k DC resistance across it. I didn't have a spare, so I modified a larger value one by removing one of the leaves. Close enough, I think.
That helped, but still not right. I started replacing micas, with no change in performance. In the process I discovered that the wire from R437 and C426 to -65v was missing. I replaced it and now the horizontal looks good (for now). The hor. output tube was getting its DC bias from the bad locking range trimmer. I''ll check tomorrow to see what effect having the proper bias has on HV.
Then I noticed that the video had a lot of white noise on it. A check of cathode voltages in the IF section revealed that one 6CB6 was bad. I replaced it, which increased the IF gain (AGC voltage went more negative), but still noisy video.
As I moved my hand around to top of the tuner the noise would get better and worse. I discovered that there is something wrong in the balun. There are 4 pins on the antenna plug, like all early RCA sets. Two are the 300 ohm input and I'm not sure what to do with the other two. I noticed that if I grounded one of the spare pins the noise disappeared. Why this suddenly happened (no noise yesterday) I don't know. I'll work on that tomorrow.
Another mystery is C7A. It is rated at 450v, but I measure about 525 across it. The boost voltage is about 575. R435 is 1/2 watt, so I think the voltage drop across it is normal. Is the boost voltage too high? If so, why? Is there another explanation?
May 20, 2012
I wired the RF input properly, and now the video waveform has very little noise.
AGC and sync stability at last. Everything looks good.
I connected the demodulator outputs to the video matrix tubes and have color waveforms on the R, G, and B CRT grids. Lots of work to do on them, though. Not good clean waveforms, and not the right colors.
I randomly picked a delay line from the ones the John Folsom brought to the convention. Without a line I have 60 v pp video out of V9a. With the delay line with a 1500 ohm termination at the grid of V9b I have 50v, a good waveform, and I can see the delay on the scope.
I then turned on a monochrome stairstep video pattern and compared the output of the video generator with the output of the video detector. Strange results. The first half of the waveform looks good, but the second half (more white) is clipped out of the detector. I first suspected that the detector diode was bad (I had replaced it earlier), but then saw that the waveform changed if I reduced the RF input to the tuner. I then measured the AGC voltage going to the tuner and found that it was 0. There was a solder bridge on a terminal strip that shorted the tuner AGC line to ground.
After I removed it, the waveform became very rounded, as if the video had no high frequency response. I think I need to repeat the IF and tuner alignment now that I have the right AGC voltages on both the IF and tuner.
It really feels good when things come together. AGC and sync are still stable. I found that I had accidentally disconnected C123 when fixing the solder bridge. Putting it back cured the poor IF response (don't ask me how). That left me with the clipping of the whiter areas of the waveform. This had me stumped for a while. Then I realized the obvious. I was putting too much video into the modulator. It has a 75 ohm termination switch, which was off. Turning it on cured the problem.
Now the really exciting news. Connecting the grid return of the hor. output tube to -65v increased the HV to 22 kv, and the regulator now adjusts from 18 to 20kv. The set was designed for 20 kv after all.
I also finished the repairs of the hor and vert centering controls and put the original micas back in the hor. oscillator.
Next I will play with the color circuits some more. Guys, what amplitudes would you expect for the chroma (G1 of V20 and V21) and ref oscillator (G3 of V20 and V21). Also, what DC bias would you expect on these grids? The bias on G1 is about -200mv, and that is the way the chassis was wired when I got it. Seems sort of low to me.
The wire from R315 to the - supply was cut. If it is connected to the -65v line I get about -3 volts on G3. The RCA patent shows a bias of -3v on these grids.
This photo shows the output of the two demodulators (top is R-Y). The waveforms are just to show that we are detecting two phases. I've made no attempt to get them right.
This photo shows the waveform on the red grid of the CRT. Again, just to demonstrate that we are actually seeing demodulated color. This is taken with the chroma control and the R-Y gain control all the way up and the contrast control set to produce about 60 v pp.
A couple of observations. First, L307, 308 and related components at V20 (and the corresponding stuff at V21) are a low pass filter. Though the schematic doesn't show it, the inductors are variable. I removed R338 because its effect is to bypass the filter and put a lot of 3.58 in the output. I wonder why it was there?
Second, there is nothing to filter the 3.58 out of the luminance channel. It is present all the way to the CRT. In order to take the picture Colorbars on R grid I put a .001 capacitor across the output of the delay line to attenuate the 3.58.
Today I took apart the tuner, cleaned the contacts, and replaced the missing LO adjusment screw for channel 3.
I took the 3.58 trap that is next to the delay line in my donor CT-100 chassis and installed it in a similar location in the prototype. It does a good job of reducing the 3.58 mHz signal in the luminance channel.
Tomorrow I will make an extension cable for the 15GP22. I will use a set with a good 15G in it, install the prototype chassis yoke, and hook it up. With luck, I will be able to see some sort of color on it.
Couldn't make an extension cable, so I made a mount for a 15G using a mask donated by Nick Williams and a shield salvaged from one of my Dage 15 inch monitors.
With high hopes I connected the chassis and this is what I saw:
Without the vertical sweep
The white vertical lines are some sort of foldover in the horizontal sweep.
May 26, 2012
Here is what I tried today:
- added an additional 500 pf across convergence element to see if garbage was getting in there.
- added an additional 500 pf on the ultor
- shorted the vertical yoke winding to make sure nothing was deflecting the beams through the yoke.
- cleaned the CRT base pins
- put a large electrolytic on the plate of the damper tube to make sure no garbage there
- installed the purity coil
- disconnected the HV regulator
Nothing made a difference, though the foldover I had is now gone. I have no idea why. The three scanning lines converge in the center of the screen, but then diverge on the edges, one being more or less horizontal, one sloping one direction, one in the other. Could I have the CRT oriented wrong? Could that cause it? Something to check.
Something is arcing on the HV or convergence lines on the chassis. I can hear it, but can't isolate it.
I when measuring the focus voltage at the CRT I noticed that it is erratic. It is stable coming out of the focus pot. I will track this down later today. Maybe there is a problem with something in the dynamic convergence circuits (though the tubes are out) that is causing the erratic focus voltage and the convergence problem?
Here are screen shots of the rasters. Notice that the blue is almost normal, while the red and green tilt substantially.
I have no idea what is going on. In talking to John Folsom we agreed that I should try 1) degaussing, 2) removing the metal shield to see if it is a problem, and 3) trying a different 15G.
Sort of sad for first images on the screen from the prototype chassis.
May 27, 2012
R456 in the focus circuit went bad, which is what was causing the erratic focus voltage.
I started by putting a 2 ohm resistor in series with the hor. yoke winding and looking at the waveform across it. There is some ringing, but not enough to cause the problem.
I then removed the 6 foot yoke extension cable and plugged the yoke directly into the chassis. Much less ringing, but the tilted rasters remained.
Then I put on the CT-100 yoke, and used short leads to connect it to the chassis. The picture below shows the 3 rasters with the vertical off - notice the nice, flat lines, all converged. The crosshatch of all 3 rasters shows fairly good convergence.
Without vertical sweep
All 3 rasters
So the problems were 1) the yoke extension cable and 2) the original yoke.
There are three possible scenarios with the original yoke: 1) it is bad, 2) it is incompatible with the 15G, or 3) there are brass adjustment pieces that attach to the front of the yoke. Installing these and adjusting them may solve the problem. I'll play with that when I have time.
I have insufficient contrast, which I don't understand. I have over 100 volts of video on the grids, which should be plenty. I didn't have time to measure the cathode voltage, which could be the problem.
May 28, 2012
Today's Progress (and the last report for a couple of months).
I'm happy to end this for now with some major successes.
First, I got the original yoke to work. I reversed pin 5 and 6 on the yoke plug and the raster is perfect now. At some time in the past someone must have reversed those connections. Maybe an engineer in the 50s spent days trying to get it to work, just like I did. In any case, the old yoke works nicely, with almost no ringing. Purity is horrible, but the picture below shows the convergence.
Before I tried the original yoke, I cranked up the color and took some photos (with the CT-100 yoke). The pictures below show color! First time since the 50s, most likely.
Obviously a long way to go - I still can't get decent purity with the original yoke and purity coil; the colors are all wrong, and there are retrace lines, etc., etc., etc.
When I return to the project later this summer I will clean things up and get the best pictures I can with NTSC.
September 7, 2012
I'm back to spending some time on the CPA chassis. Today I made additional attempts to get decent purity, with limited luck. I am using a 15GP22 and the prototype yoke. I tried both the prototype purity coil and a CT-100 purity coil, both operated on an external power supply with which I could adjust the current through the coil and the polarity. I was never able to get a good red over the whole screen. At best, I got red over the bottom 2/3 and green at the top.
I think that the only solution is to use the CT-100 yoke, which is designed for the 15G. If you remember from months ago, that yoke works, but has some significant ringing. I will experiment with adding capacitance to the yoke circuit to see if I can improve the ringing.
If anyone has ideas, please let me know.
I also played with the monochrome balance and discovered that the background controls have almost no effect. I will explore that more this afternoon.
My hope is to get reasonable purity and reduce the ringing. Then I will move back to the color demods and try to get the proper waveform out of them. I think I am close on that, and I'll have a working NTSC receiver soon.
Next step will be to implement CPA. I'll build a multivibrator based on the RCA patents, with no field one detection. At some point I'll need a CPA signal source (Darryl?)
September 8, 2012
Today I put the CT-100 yoke back on. I added a 270 pf capacitor across one half of the horizontal winding (that is what the CT-100 schematic shows) and the ringing went away. I now have a nice raster. With that yoke and the CT-100 purity coil I was able to get decent but not perfect purity. More careful adjustment will probably fix that. Convergence is OK but not right. The control is an one extreme for the best convergence. There are four possibilities that I can think of. 1) there are differences between the prototype tube and the 15G in convergence voltage required, 2) Convergence voltages are off, due to resistor drift in the prototype chassis, 3) HV, which is now 16kv, should really be 20kv, indicating remaining problems in the prototype chassis HV supply. 4) we are missing the horizontal dynamic convergence circuit. Maybe that throws off the static convergence.
In any case, it is good enough to put that part aside and go back to the color demods.
September 9, 2012
Today I messed with the R-Y and B-Y demods. After playing with all the adjustments in the 3.58 osc circuit, I finally concluded that I could either get a proper R-Y waveform, or what appeared to be an inverted B-Y waveform, but I couldn't get them simultaneously because they required different settings of the trimmers.
I decided to remove the B-Y demod from pin 1 of the CPA transformer and move it to pin 3. After making that change I am able to get a good R-Y waveform out of the R-Y demod, but the waveform out of the B-Y demod is inverted. No amount of tweaking the trimmers will change the polarity.
Now, for the exciting news. By moving V20 (R-Y) to pin 4 and V21 (B-Y) to pin 3 I now have proper waveforms and a real (though poor) NTSC color pattern on the screen. The green bar is compressed because I turned the width down to get all the bars on the screen and that screwed up the linearity.
September 10, 2012.
I spent a few minutes with the width, background and gain adjustments, and here is what the screen looks like now:
September 14, 2012.
I took a break to rebuilt the power supply. I had a chassis with the 370v supply and one filament transformer, and a CT-100 transformer and voltage doubler "air wired" on the bench. I had an old aluminum chassis which I used to consolidate everything.
On it are:
The CT-100 transformer and voltage doubler to produce the 360 v plus and 80 v negative lines. This transformer also provides the 6.3 v for the CRT.
An old transformer from who knows where with a silicon diode full wave rectifier to provide the 370 volt line
A big filament transformer to provide 6.3 for the chassis.
I also experimented some more with the purity coil. I had to add a resistor to the purity circuit to get enough current through the CT-100 coil to get decent purity.
I also measured the trimmer capacitor in the oscillator circuit so I can order the 3.89 crystal, which I plan to do today.
September 15, 2012
Today I worked on high voltage and focus. If you remember back to when I first got the chassis working I was only able to get about 18kv out of it. Now, for some reason, I have almost 25kv with the regulator disconnected. Maybe the flyback had moisture in it that has worked itself out? I have no other explanation.
With the regulator connected I can adjust the HV from 17 to about 18.5 kv. By reducing the screen voltage on the 6BQ6 to 60 the range changes to 18 to 20 kv. For now I've done that. It is possible that the B+ line that feeds the 6BQ6 screen is supposed to be 60v, since the voltage dividers for the B+ have been hacked.
I have too high a focus voltage. I jumpered one of the 2.7 meg resistors from the bottom of the focus pot and that brought the focus in line. Either the 2.7 meg resistors changed value (didn't check) or there is a difference in focus voltage between the prototype tube and the 15G.
Here is a screen shot showing where I am with convergence. We have no horizontal dynamic convergence circuit, and I haven't messed with the vertical dynamic convergence.
September 17, 2012
Today I worked some more on the color demods.
The trimmer across the crystal was hanging in midair, making it difficult to tune. I installed a mounting bolt, and was finally able to adjust C313, C310, C312 and C318 to produce a non-inverted B-Y waveform on the output of the demod with its input on pin 1 of the CPA transformer. I connected the R-Y demod input to pin 4 of the CPA transformer and was able to adjust C321 for the proper R-Y waveform.
I was then able to get the proper waveforms out of the R, G and B matrix tubes by adjusting all the pots.
So, I think the circuits are working the way they are supposed to.
I have ordered the 3.89 crystal and Darryl is working on the converter.
November 4, 2012
Finally got a chance to get back to the CPA prototype.
First, I spent a couple of hours trying to get Cliff's field 1 detector to work. No luck. I am confident I am feeding it with the right waveforms, but no matter what I do I get 3 sharp pulses out of it at the beginning of the VBI of both fields. So I've given up and have gone back to plan B: a divide by 2 multivibrator triggered by the vertical sweep, with a pushbutton to manually select the CPA phase.
I built one using a schematic I found on the internet. Unfortunately, no component values, so I had to experiment. I got it to trigger properly, but when the diode switches were connected the waveform was very rounded. That might not be a problem, but because I have 3 triode sections to work with, I added an amplifier with a much lower impedance output to drive the diodes.
Take a look at the schematic. It works perfectly, with a nice square wave output. I can see the B-Y demod output being switched at the 30 Hz rate.
Tomorrow I'll turn on the CPA switching on Darryl's converter and see what it looks like.
November 5, 2012
CPA at last! This afternoon I saw CPA on Nick's prototype set, the first time since 1952 that anyone had seen it. The problem was the setting of R352, the diode bias adjustment. I had to eliminate R351 to allow R352 to adjust to the right point for the diodes to switch.
The color balance is off, so the colors aren't exactly right, but they are the same as when I turn off CPA in Darryl's converter.
November 17, 2012
The Wizard of Oz on CPA:
There is still much to do, especially on convergence. Though the vertical dynamic convergence circuit is there and should work, parts are missing in the horizontal circuit. We will have to experiment to find what will work.
Today I plan to build a metal frame out of angle iron to hold the chassis, tube and power supply so that it will be easier to make adjustments. I will also install a better 15GP22 - the tube I'm using has a weak red gun.
Then I'll start working on the convergence. There is also work to do on the focus voltage - the pot is at one extreme, either due to resistor aging or to the fact that we are using a different CRT than the original prototype.
November 23, 2012
Here is the frame I made. It makes it easy to adjust stuff and keeps the CRT and yoke stable.
Convergence: I got the vertical dynamic circuit to work, at least the controls produce waveforms that can be adjusted. I installed the horizontal phasing coil from the CT-100 and wired the horizontal dynamic circuit like I think it was originally. I get a waveform that will change in amplitude (by adjusting the pot in the cathode of the hor. output tube), but the phasing coil doesn't change the phase. I suspect it is the wrong inductance.
Pot in the cathode of the hor. output tube: This was a strange condition. It is 100 ohms, but if you rotate the shaft, with nothing connected to the wiper, the resistance would change to about 90 ohms at about the midpoint. The wirewound element was warped at the midpoint, allowing a part of the wiper that normally misses the element to contact it, shorting out a bunch of turns.
Drifting color balance: I had noticed that the green and blue would get brighter as the set stayed on for a while. The culprit was leaky coupling capacitors from the 6BL7s to the CRT.
Red matrix amplifier: The waveform out of the red amplifier was very distorted. I noticed that someone had added a .05 cap in series with a 10 k resistor from the 6BL7 plate to the cap that drives the red CRT grid. When I removed them, the waveform disappeared. I took apart L207, thinking that the inductor labeled as 165 uh was open, but found it to be OK. I put it back together, re-installed it, and the waveform was there. There must have been something intermittent in the inductor, and the .05 and 10k were a crude fix to get some red video. I know I didn't fix whatever was wrong, and it may return.
Delay line: I had significant distortion of the video coming out of the delay line, which I determined was because my circuit wasn't matched to the impedance of the line. Adding a 2.2k resistor in series with the input fixed it.
November 24, 2012
Worked some more on convergence. I can see that the hor. dynamic circuit will work if I get the right inductor. Vertical doesn't have a great deal of effect, so I need to see what is wrong. Even without the dynamic convergence circuits working the convergence isn't bad.
While playing with the color balance I discovered that I had a weak 6BL7 in the blue matrix. Replacing it made the black and white setup much better. The picture now looks pretty decent from a DVD.
The vertical started shrinking while I was working on the set. The problem was C407, which started leaking wax and getting hot. The reason is that the voltage at that point is about 650 v, and about 700 at the point marked Boost in the schematic. C407 is a 600 v capacitor. Even stranger is that C7A, in the horizontal oscillator (10k to boost) is rated at 450v. This is a can electrolytic which looks to be original, so it isn't likely that someone put in the wrong replacement. I have two 450 v caps in series now, but how could this have worked? Could the boost have originally been 500v rather than the present 700? How is it possible that it is so far off now? Anyone have any ideas?
Perhaps related to this is that there is foldover in the middle of the screen with the original components. I had to reduce the hor. output screen grid resistor to 2500 ohms to get rid of it. Also, the waveform on the grid of the hor. output tube is not clean - rather than it being a relatively straight ramp there is a dip about 1/10 of the way up. Nothing visible in the picture, which is strange, but that may be because it is within the blanking inverval or just outside it and not visible because of overscan. There is a wavy pattern visible when the scene is dark.
Seems like this set has an endless list of problems.
Another oddity. There is no vertical retrace circuit, and retrace lines are visible if the brightness is turned up only slightly above optimum.
November 26, 2012
I did some more exploration of the glitch in the horizontal waveform. Changing to the original yoke made no difference, and nothing else helped. I did notice that with the new 450 v supply for the hor. osc. that adjusting the hor. drive trimmer made the wavy line disappear off the side. Also, I removed the 2500 ohm resistor I'd put on the hor. out. screen grid, and the width was fine. So it appears that the horizontal section is working well,even with the waveform glitch, so I won't look further.
I removed the delay line and discovered that the ringing was caused by it. There is a very slight ghost about 3/16" inch following sharp lines which I will look for later. I tried a number of delay lines and series resistors until I got one to work with no ghosting. The picture looks much improved.
I now have an intermittent high voltage sizzle which occurs with the brightness turned up. HV is normal, so something else is going on.
Next I'll go back to the dynamic convergence. I need to find an adjustable inductor that is similar to the one in the attached spec sheet. I don't have anything to measure inductance with, so I thought that paralleling the coil with a capacitor and putting a signal general and scope on it would be the easiest way. Anyone have a better solution?
November 27, 2012
Today I solved one of the mysteries I've been putting off tackling, quite by accident. I've been able to get a balanced waveform out of the R-Y demod. By that I mean that the two phases produce the same amplitude outputs. But when the R-Y outputs are balance, the B-Y are not. This doesn't appear visible in the picture because I believe the two levels are averaged and the rate is quick enough not to be visible. But it always bothered me.
Today I accidentally turned the diode bias pot and the B-Y suddenly became balanced. I must have been on the edge with the switching voltage and one diode wasn't completely conducting.
The picture is looking really good. Black and white looks good, as does the color.
Everything appears to be stable now.
The last remaining issue is the dynamic convergence. John Folsom is sending me 3 inductors, one of which should work in the horizontal circuit.
The vertical circuit produces a waveform which changes when I adjust the two pots. However, I can't see much change on the CRT. The picture does shrink about 1/4" when the pots are at one extreme, but the convergence doesn't appear to change. I've changed the caps and checked the resistors, and tried a new tube.
November 28, 2012
I modified by metal frame to make adjustment of the yoke easier so I can get better purity.
I have noticed that the horizontal is still flaky for a few minutes as things warm up (10 minutes or so). I started by installing C459, a .22 from the -100v line to ground in the hor. osc. It used to be there, but I removed it in an attempt to solve an earlier problem. When I re-installed it the horizontal went completely unstable and wouldn't sync. This is not right, obviously. This is only a bypass cap on the -100 line, which has many other bypass caps in other parts of the chassis. I will go back to that later.
Then I spent some time on the vertical dynamic convergence circuit. I have only about 10 v pp on the plate, which is not enough to do anything for convergence. Since I don't know how the circuit works or whether parts are missing, all I can do is check parts, which I've done twice.
After I got home I looked at the circuit from a 1951 RCA prototype (courtesy of Ed Reitan). It also takes its signal from the cathode of the vert. osc. circuit. My problem is that there is no signal there in our chassis. C412 completely squashes any AC. I noticed that the corresponding capacitor in the 1951 circuit is 5 mfd (compared to 250 mfd in my chassis). I wonder if someone put the wrong value cap in. Tomorrow I'll try a 5 mfd and see if I get a parabolic waveform on the cathode.
The color still looks great. I can see some minor flicker in the picture occasionally with a DVD.
November 30, 2012
The inductors arrived from John Folsom. The 130-250 uh one works great. Attached are the horizontal dynamic convergence waveforms at minimum and maximum inductance. Though I haven't done a convergence yet I can see that it should work.
The vertical circuit remains a mystery. Here are some more observations:
I have a 10v pp sawtooth waveform at the top of R418. There is no AC, but about +330v dc, at the bottoms of R417 and R418.
With a 20 mfd cap on the cathode of V26B I have a 4 v pp parabolic waveform at the top of R417. With a 250 mfd cap on the cathode I have no AC there, as expected.
There is no significant difference in the waveform at the plate of V27A with either value of cap. This makes sense, since the parabolic waveform is mixed with the sawtooth at C414 through a 100k resistor while R418 is 7500 ohms. Almost no parabolic signal reaches C414.
So, as the circuit is wired now, the presence or absence of the parabolic signal is irrelevant. Since the set had a 250 mfd cap in it, it seems likely that no parabolic signal was obtained from V26B.
So how did this circuit work? As Terry pointed out, R417 could vary the DC voltage on something, since at the bottom is about -80v and at the top about +330v. But R417 (through R419) is AC coupled into V27A.
John Folsom's Admiral 15 inch set uses only the vertical sweep sawtooth to derive a convergence waveform. When he gets it working again he will send some waveforms and voltages. That might provide a clue.
December 1, 2012
Success! After the realization I had (last email) that the parabolic signal didn't do anything, I started thinking about what could create the proper convergence waveform. What I had was a sawtooth, sloped one way. I figured that if I put a sawtooth sloped the other way on the grid of V27A I would have something with a dip in the middle. I removed the wire that went from R417 to the cathode of V26B and connected the 100k "resistor to nowhere" through a .1 to the plate of V26B, where I had a sawtooth of high amplitude of the opposite polarity.
The result is a convergence waveform of about 800 v pp which can be adjusted both in phase and amplitude.
This works so well I think it must be how it was done originally. The mystery is how the wire from the cathode of V26B to R417 got there.
December 2, 2012
Here is the convergence. It is very good, and can probably be better with more care. Please ignore the color balance and purity. I'll work on that next. Also, the focus is actually very good. The problem is in the photography:
You will notice a ghost to the right of the vertical bars. This is my next challenge - I'll probably have to do another IF alignment. I have posted the latest revisions to the schematics - I've found a bunch of errors, and all the circuit changes I've done are there. The links are at the top of this page.
A portion of the screen, enlarged, to show the ghosting. There are two possible sources: IF alignment and problems in the video amplifiers.
December 4, 2012
Today I discovered the source of the main ringing - a resistor and inductor at the grid of the video amplifier. Removing the inductor eliminates the ringing. I have no idea why - maybe there are parts missing there? In any case, most of the defects in the picture are now gone. There is still a minor ghost about 1/4 inch away from sharp lines. Here are pictures taken today. The photography isn't perfect, and the picture on the screen is much sharper than the photos. Also, the color balance was done very quickly, and can be much better.
I thought I'd solved the mystery of why shorting out the inductor from the grid of the video amp to ground greatly reduced the ringing. The RCA 21 inch b &w set that was the model for the IF strip in the CPA chassis also uses the same video amp. The Sams shows a 6800 ohm resistor in parallel with the inductor, which is missing in the CPA chassis. I added it, but the ringing was still pronounced (the first "ring" was over 10% of the value of the pulse). The Sams also showed a 3300 ohm resistor in parallel with the inductor from the detector diode to the video amp. I added that resistor, and replaced the short across the inductor to ground. The ringing disappeared (I can't see it with the pulse full screen on the scope). So, still a mystery, but I'm not going any further. While making these changes I compared the amplitude of the burst to make sure that they weren't attenuating high frequencies, and found that the burst stayed the same.
Now that the picture is so much better flicker is visible at times. I've never been satisfied with the symmetry of the two phases coming out of the R-Y demod. This is because the switching affects the B-Y symmetry, and I had to make a compromise, resulting in neither waveform having the same shape in both phases. This contributes to the flicker.
In addition, the LC circuit (L303/C312) on the secondary of the torroid transformer that drives the CPA transformer doesn't do anything, which tells me that the circuit was not wired that way originally.
So I experimented with different ways of connecting the input to the B-Y demod which would eliminate the interaction between the B-Y and R-Y adjustments for symmetry.
I discovered that if I moved the B-Y input to directly on the torroid output, before the 3300 ohm resistor (R314) I could adjust the B-Y symmetry using either L301 or L302, then adjust the R-Y symmetry using C318, C321 and C323 without any interaction. I can now get almost perfect symmetry on both. Attached is a revised schematic to show what I did.
had a 33K 2 watt carbon resistor go bad by shorting (2 ohms). I've never had that happen before. Fried a choke in the power supply. I will install some fuses.
I also had an intermittent in the video show up. Cleared up before I could find it. I suspect this chassis is going to continue to have things go bad.
On the positive side, I replaced the micas in the horizontal oscillator. Before the waveform adjustment didn't work right, though I could get it to be correct, I couldn't tilt it both directions like all the other ones I've done. Now it works normally. No apparent difference in the operation of the set.
I also reduced the CRT screen grid voltages to 320 from 350, which is the design maximum for the 15G.
Finally, I chased down a shading effect that was visible with a black scene. It was a vertical bar about 1/8 the width of the screen on the right side in the blue and green, but not the red. It was coming from the B-Y demod, which was producing a pulse even with the chroma control all the way down. The cause was the pulse from the flyback, which was much too wide and was extending past the hor. blanking interval into the video portion of the line. I added a resistor to ground after R444 and R445 to make a discriminator with C435. The pulse is now much sharper and stays within the blanking interval. That fixed the shading.
Here are two photos showing the wrong phase:
Here are some pictures showing my next problem. Yesterday I mentioned that I reduced the width of the pulse from the flyback. The picture on the left shows the b-y (top) and r-y demod outputs before I narrowed the pulse. You can see the slight difference between the two fields in both demod outputs. You can also see a negative going pulse in the b-y just following the burst (visible in r-y) followed by a distortion of the waveform of the white color bar. The picture on the right shows the same waveforms with the pulse narrowed. The waveform during the white bar is fixed, but the pulse is still there.
The pulse goes through the blue matrix tube, but, since it is negative, doesn't do anything to the picture. However, the b-y is inverted in the blue matrix tube and fed to the green matrix tube (along with some of the r-y). The pulse is therefore positive as it hits the green CRT grid. Though it is during the horizontal blanking period, it appears as a wide green bar on the screen (picture 0011). This bar moves as you turn the hor. hold control but is always visible in the picture during dark scenes.
So, this is the latest demon I'm fighting. The better the picture gets the more of this subtle stuff is visible.
Other demons include an intermittent flashing in the color and video (like a loose connection) that comes and goes, a strange green/blue shading condition in black and white that may be purity adjustment or the lack of a field neutralizing coil, and today for the first time I saw a pulsing in the picture that looked like it might be bad filtering in the AGC line but which disappeared before I could find it.
At John Folsom's suggestion, I rebuilt the metal rack that holds the chassis and CRT so that the chassis is on its side. That way it doesn't get as hot underneath, I can work on it without tipping it over, and when it is put on display people can see both the top and bottom of the chassis.
I worked some more with the green retrace shading bar. By accident I discovered that it goes away with a particular setting of the burst and osc. adjustments. So, I guess that is fixed.
Just as I was finishing up there was a snap, a hiss, and smoke. Something has died, in the area of the centering/width controls. A seemingly never ending project.
I'm finished for now. Attached are a couple of screen shots, though the picture looks better on the screen. I didn't pay much attention to convergence when I took them.
The smoke was coming from the horizontal centering control. A bolt had fallen next to it and shorted it to ground. Amazing it survived.
The intermittent hissing noise from the HV returned, and changed as I rotated the frame. I used piece of plastic hose (John Folsom's suggestion) as a single stethoscope to locate its source, and found it came from the plexiglass where the lead to the CRT attaches. I cleaned the plexiglass with alcohol using Q tips, and the noise disappeared. I suspect it might return.
The chassis will have to wait for the spring and the expertise of Pete Deksnis and others to do a good convergence and setup.
A couple more observations. I noticed that the front control shaft lengths are all different. I think that means that this chassis was never put in a cabinet. Also, the chassis isn't square. It is almost 1 inch shorter on the tuner side than on the HV cage side.
December 21, 2012
At Ed Reitan's request, I took some photos of color bars with CPA on and off to see if cancellation of phase errors could be seen between the white and yellow bars. Here is a cover and one page from an article from Electronics magazine describing this cancellation (courtesy of Ed Reitan):
I took 4 shots of each of 2 conditions, using a shutter speed of 1/15 second, and selected the best photo of each condition to post here:
CPA off, closeup of white/yellow transition
CPA on, closeup of white/yellow transition
It appears that quadrature phase cancellation is taking place with CPA.
January 6, 2013
George Lemaster wrote us, with information on how RCA accomplished field one detection with a single tube:
|I've been doing more research on RCA and CPA and ran across some schematics that may help with the field detector in the CPA circuit. The sketch of the field detector was used in "Experimental Model #1" CPA sets. At least three experimental sets, based on 16 in. B&W model TC166 chassis, were built about May 1951 and these used three sampling color detectors at 120 degree phasing.
After some initial confusion about the polarity of the vertical pulse, the operation of the circuit was understood. Positive going pulses from the horizontal and vertical sweep are combined and fed to a diode clipper. At field one, the horizontal pulse coincides with the vertical, resulting in increased amplitude at the diode. At field two, the pulses don't coincide. The clipper level is set so that only the higher (field one) pulse reaches the grid of the multivibrator.
Over the next couple of weeks Cliff Benham breadboarded the circuit and connected it to a RCA 630 chassis. With the help of John Folsom, he was able to get it to work and document the pulse levels that worked best with the circuit.
January 19, 2013
I arrived back in Columbus for a visit and built the circuit, first on a breadboard, then on the prototype chassis. Some changes were needed to connect the circuit to the H and V signals in the prototype chassis, but it worked basically as Cliff had described.
The prototype now locks automatically to the proper CPA phase.
I am convinced that this circuit (or one very similar) was what was originally in the prototype chassis. The evidence is this:
1) Though there are 3 available triode sections on the chassis where the switching could have been implemented, one of these sections had never been used (nothing had been soldered to the tube socket). This means that RCA implemented the switching using a single dual triode.
2) There was a 5-50 pf trimmer capacitor on a terminal strip in the vertical sweep section. Such a trimmer could not have been used in the vertical sweep because of its low value. The trimmer does, however, work perfectly to adjust the level of the V pulse to the CPA switcher (C3435 on the Color With CPA schematic).
3) There is an emply lug on the terminal strip next to the horizontal pulse line from the flyback. This is where I obtained the H pulse for the CPA switcher. There is evidence that the lug once had components attached to it (R360 and R361 on the schematic).
I believe we are to the end of the road. All of the mysteries of the CPA prototype have now been uncovered.
Februay 5, 2013
My next step with the CPA chassis is to recheck the alignment.
The IF/tuner should be down 6 db at the visual carrier frequency, then flat to about 3.5 mHz, then down 6 db at 3.89mHz. After considerable confusion I was able to get the alignment right. I also swept the video amplifier, delay line, and video cathode follower.
Then I turned to the chroma. Documents provided by Ed Reitan showed that the chroma passband should be flat from about 2.9 to 4.2 mHz. No matter how I adjusted the three inductors in the chroma bandpass filter I was not able to get that response. I then looked at a document that George Lemaster sent and found that the schematic for the filter in the Model 1 CPA prototype was identical to the one in our chassis. This filter was designed to be flat from about 2 to 5.4 mHz. I calculated what components would be required in the filter for 2.9 to 4.2 and they were radically different from the ones actually there.
After playing with different components I decided that changing the filter would defeat the purpose of restoring this chassis. The idea is to see what CPA looked like in 1951, so I decided to keep the original filter design and simply align it for as close to 2.9 to 4.2 as I could. The best I could do was about 2.2 to 4.2. Here is the response:
Then I swept the R-Y and B-7 demodulators, which according the Ed's documentation should be flat to about 1 mHz. The R-Y demod was, in fact, flat to 1 mHz, then rolled off. The B-Y demod, however, was flat to only 500 kHz. Was this changed when the chassis was converted to NTSC, or was it originally that way?
February 6, 2013
I decided to align both demodulators for 1 mHz bandpass. This is almost certainly the way it was originally - the modification to the B-Y demod may have been done when they changed the chassis to NTSC. Here is the response I got on both demods (the marker is 1 mHz):
I also cleaned up some loose ends. The vertical centering control was falling apart, and would arc to the chassis occasionally. I replaced it. Also, during the alignment process smoke started coming from the isolation transformer in the filament of the damper. One of the two micas from the filament to ground shorted. I replaced all three in that circuit. I am ready to move the chassis onto the museum floor and do a complete adjustment - purity, convergence, black and white balance, and color balance.