pilot: Increase Video DC-restoration from 31pct to 100pct
pilot: Increase Video DC-restoration from 31pct to 100pct
Fellow Radiophiles,
DC-restoration of the video signal is the process that fixes the black level seen in the CRT (Cathode Ray Tube), independent of scene content. For example, when watching ice skating, the scene is mostly white on the ice, but is much darker if the background changes to the audience. The exposure of the skater in the scene does not vary in transmission, but in a set without DC-Restoration, which is to say, the video signal is AC-coupled on its way to the CRT, it will have very different exposure levels on the skater, depending on light vs dark surroundings.
Click any graphs or images to enlarge them in a separate window.
LTspice Simulation with Approximated NTSC Signal
I made a reasonable approximation of an NTSC video signal for the purposes of simulating the DC-restoration of an AC-Coupled video signal in LTspice. The graph on the right is zoomed in around the vertical sync pulse. I did not include the 3.58MHz color burst signal or the black level just before and just after the horizontal and vertical pulses. These black level sections blank the CRT beam while it returns from vertical and horizontal sweeps.
If you are closely familiar with the NTSC signal, you will see that I also did not include the extra equalization horizontal sync pulses before, during and after the vertical sync pulse.
The important thing is to get the correct duty cycle for the vertical and horizontal pulses. The horizontal and vertical sync pulses are active high only 7.8% and 1.2% respectively, which makes a net 9.0% sync duty cycle.
The amplitude of the signal was scaled to 20Vp-p of video and 6Vp-p for the sync pulses. These amplitudes are typical for the video signal at the the output of the 6BA6 video amplifier in the Pilot TV-37, which drives the cathode of the CRT through an AC-coupling cap. See the circuit excerpt below.
Nearly Perfect DC-Restoration
With DC restoration, peak white and peak black levels stay fixed, regardless of scene content. This is typically accomplished by detecting the peaks of the sync pulses to clamp them at a fixed level.
Clamping is a peak detection process that involves storage and pushes an AC-Signal to one side of a fixed DC voltage. There is no loss of AC information.
Clipping simply clips a waveform above or below a certain level.
The simplest circuit for this is perhaps a simple diode with a grounded cathode and an input AC-Coupling capacitor driving a resistive load in parallel with the diode. The DC-restored video is at the "Video_Dio_Restore" terminal. This simple circuit compares the input "VIDEO" signal to AC-Coupled video at "Video_AC" and to a DC-restored video at "Video_Dio_Restore". The "Video_ACEnvelope" terminal also follows the positive envelope of the AC-coupled video.
The following plot shows 1.2s of video, where the scene starts out black for 0.2s, then goes white for 0.5s and returns to black for the remaining 0.5s.
The solid black lines are the zero reference levels for the three traces.
The red trace is the input video signal with +6V sync pulses and -20V peak white video.
The magenta trace is the AC-Coupled version of the input video. Note how the AC-coupled white level first steps negative at 0.2s by -20V, but then drifts to only -5V, making the white scene much darker. When the input video level steps back up to black by +20V at 0.7s, it takes about 0.3s for the proper black level to be reached. This illustrates a scene variation from white ice to a black background.
The cyan trace shows the video signal after DC-restoration at the "Video_Dio_Restore" terminal with the simple R-C-Diode clamping circuit. The cyan trace is nearly identical to the red trace. This is close to ideal DC-restoration.
This circuit has a finite level of efficiency. One issue is 0.7V of voltage drop for a Silicon diode, but that can be overcome with a simple offset (brightness) correction. The input AC coupling cap gets charged by the positive sync peaks, and must loose very little charge between sync pulses. It is imperative that the AC-coupling cap charges much faster through the diode and source impedance than it discharges through the load resistor. Something like a 100/1 charge to discharge current ratio would be acceptable. If the load resistor is 1MΩ, then the source impedance and diode impedance would have to be less than 10kΩ if the sync pulses had a 50% duty cycle. The actual charge/discharge behavior is calculated below by taking the actual 9% V&H sync pulse duty cycle into account.
After the charge/discharge current ratio is satisfied, the time constant of the load resistor and cap also has to be long enough so that less than 1% of the charge is lost between horizontal and vertical sync pulses.
DC-Restoration in the Pilot TV-37
Early post-war B&W TVs often had no DC-restoration, or had very incomplete DC-restoration, as is the case with The Pilot TV-37 from 1949.
I have been looking into the DC restoration circuit in the Pilot TV37, as shown in the schematic excerpt. The video path from the 6BA6 video amp plate to the CRT-cathode is AC-coupled with a 0.25uF cap. The DC restoration voltage is fed into the CRT-grid by the cathode of the 6AU6 DC-restorer and sync amp.
This pentode also delivers amplified sync at its plate. This is a clever dual use of a single pentode. During the positive-going AC-coupled sync pulses at the grid, the plate turns on hard and causes the plate to swing hard negative below ground by 60V. Keep in mind that the 6AU6 operates between ground and Bminus=-99V. During this period, most of the 5mA peak cathode current actually flows to the screen grid g2 because the plate current is limited by the 220kΩ load resistor to 100V/220kΩ=450uA.
During a sync pulse, the 6AU6 pentode works as a triode follower with μ=35 with the screen grid g2 serving as the plate and the cathode follows the positive peaks of the grid voltage to keep the 5uF load capacitor charged up with the positive envelope of the AC-coupled video signal. With 85V between the cathode the screen g2 and μ=35, the 6AU6 cuts off with just -95V/35=2.4V. This limits the minimum required sync amplitude at the 6AU6 grid to at least 2.5Vp-p for reliable sync operation.
After the positive sync peaks, the 6AU6 grid goes hard negative by at least the -6V of the sync pulse for black video and by -26V for white video. The cathode also turns off like a diode, so that the charge is retained by the 5uF reservoir capacitor as the positive envelope of the video.
The CRT beam current and thus brightness, responds to the voltage difference between its grid (Wehnelt cylinder) and cathode. So the final step of the DC restoration occurs inside the CRT as the positive DC envelope of the AC-coupled video is fed to the CRT grid voltage to replace the missing DC component in the AC-coupled cathode.
The large 5uF value and the 22kΩ cathode load resistor have a 110ms time constant while the 6AU6 is cut off. The V and H sync pulses have a total 9% duty cycle with respect to the video content. The peak charging current during a white scene is 5mA and the discharge current is 0.5mA through the 22kΩ load resistor.
The charge path for the 5uF reservoir cap is through the internal 6AU6 cathode resistance and the external series 270Ω resistor. The total cathode charging resistance in simulation is 1450Ω.
Original DC-restoration Clamp Efficiency Limited to 31%
The charging time constant including the 270Ω cathode resistor is 7.25ms. This means that only 18.7% of the charge into the 5uF cap is accumulated during the 9% sync duration in the 1/60s field period. This time is 1/60*9%=1.5ms during the active H+V sync pulses in one 1/60s field sweep.
The discharging time constant through the 22kΩ load is 110ms. This means that some 12.9% of the charge accumulated on the 5uF cap is lost during the 1/60*91%=15ms between H&V sync pulses in one 1/60s field sweep.
During the 110ms discharge time constant, the capacitor voltage charge/discharge stabilizes to a constant value over several 1/60s field sweeps.
The net efficiency of the DC restoration is 100%-12.9%/18.7%=31.1%. This calculation is strongly dependent on how well modeled the 6AU6 cathode resistance is. This resistance is non-linear and depends on the cathode current. The cathode resistance in the linear calculations below is the value that makes the calculated DC-Recovery efficiency match the simulated circuit result.
With only 31% of DC recovery, there is still a very noticeable change in the brightness level of a particular item in the scene as the background changes from dark to light.
DC-Restoration improved to 100%
The efficiency can be improved by either reducing the charging time constant or increasing the discharging time constant. The charging time constant is limited by the internal 6AU6 cathode resistance, so it can't be improved. I doubled the discharge time constant by doubling the 22kΩ load resistor to 44kΩ. Repeating the calculation, we get 64% of DC recovery. Much better, but still not 100%. One possible reason for the original 22kΩ value choice may have been to guaranty full plate saturation at the 6AU6 for a reliable full sync swing at the plate. But, I found no sync swing degradation with the 44kΩ cathode load.
DC recovery circuits are typically done in the same path that delivers the video signal to the CRT. In the Pilot TV-37, the AC-coupled video drives the CRT-cathode and the video DC-recovery voltage is fed to the CRT-grid. This means that if I were to add a modest DC recovery circuit to the CRT-cathode circuit, I might be able to make up the missing 36% of the DC-recovery voltage.
I added a 1N34 diode to the CRT-cathode circuit, like the diode type in the existing video detector shown in the schematic above, in parallel with the 180kΩ resistor passing the DC brightness bias current to the CRT-cathode. This diode was able to accumulate enough DC recovered voltage on the increased 1.25uF cathode AC-coupling cap in the video path to the CRT-cathode, that I was able to obtain close to 100% net DC restoration, with 64% DC recovered in the CRT-grid path and 36% DC-recovered in the CRT-cathode path.
I also increased the AC-coupling cap into the CRT cathode from 0.25uF to 1.25uF. The cathode charge accumulation cap was also cut in half from 5uF to 2.5uF. This improves the time constant match between the CRT-cathode and CRT-grid paths. This makes for a faster and cleaner response to scene brightness changes, as the CRT-grid should always track the black level in the CRT-cathode video.
Having the two separate paths available for AC-video and DC-restoration is what made it possible to improve the DC restoration to 100% with such a simple fix.
Simulated DC-Restoration
This LTspice schematic was used to simulate DC restoration. The ideal sources and circuit elements on the lower left generate an approximation of the NTSC video signal with a 0.5s white scene followed by a 0.5s black scene. The 6AU6 shown was modelled in KorenTube.lib and simulates the Sync Amp and DC restoration stage of the TV37.
These plots show the simulated DC-restoration results comparing the ideal case on the left, the original circuit in the midddle and the modified improved circuit, including the 1N34, on the right.
The first plot on the left shows results using the simple R-C-Diode circuit shown above. Ideally, the thin blue trace with the recovered DC level should match the shape of the top side envelope of the magenta AC-coupled video waveform. In this the case, the DC-recovered level is nearly ideal in the cyan trace, because it looks almost exactly like the original input video in the red trace.
The plot in the middle is a simulation of the original circuit and it is clear that the amplitude of the thin blue trace with the recovered DC level has a much lower amplitude than the top side envelope of the magenta AC-coupled video trace. The cyan trace is the difference between the blue and magenta traces. Ideally, it would have the same shape as the red input video trace. In this case The black level on the right end of the wave did move up 31% off where it should have been to match the red trace. The peak-peak swing of the blue trace is simply too small at 31% for a full DC restoration.
There is also a time constant mismatch between the magenta and blue traces in the middle plot. This makes the cyan trace lumpy. It overshoots quickly following the fast AC-coupled video, then it is brought down slowly by the undersized DC-recovered level.
The plot on the right is the simulation after I incorporated my DC-recovery improvement in the CRT-grid circuit and the additional 1N34 diode in the CRT-Cathode circuit for a net 100% recovery. There is also a change in the CRT-Cathode and CRT-Grid time constants to make them match more closely. This time constant match results in clean transitions between dark and light scenes in the final DC-restored video inside the CRT.
The magenta AC-coupled CRT trace on the right plot already shows 32% DC-recovery from the new 1N34 diode in the cathode circuit, as compared to the magenta curves on the two plots to the left, that are purely AC-coupled. Now the thin blue trace with the improved, but incomplete DC-recovery into the CRT-grid, is combined in the CRT with the partially DC-recovered magenta trace at the CRT-cathode to achieve 100% DC-recovery in the cyan trace.
Simulated Ideal vs. Original vs. Improved DC Restoration
The following plot simulates a direct comparison of ideal 100% DC-recovery in the thin black line overlaying the magenta trace, the original 31% DC-recovery in the thin black trace overlaying the AC-coupled video in the red trace and and the improved 68% DC-recovery in the thin black grace overlaying the 32% partially DC-recovered video in the cyan trance.
Measured results
The following measured results, for the improved 100%DC-restoration, confirm the simulation results.
The following scope photos show the video sisgnal for horizontal line number 24 at the CRT-Cathode and the DC-restored level at the CRT-Grid. The HP54601b oscilloscope has a TV sync mode where you can select the line number to view. The CRT-Cathode waveforms float up and down with the scene brightness, because they are AC-coupled. The CRT-Grid is supplied with the recovered DC level that follows the black level at the top of the CRT-Cathode video waveform.
The black level in the video in the first photo matches the DC grid voltage perfectly. This is the ideal case. In the the following three photos, the black level of the video on both sides of the horizontal sync pulses still match the DC level fed to the grid quite well.
The brightness of the CRT is set by the difference between its cathode and grid. The difference between the grid and cathode waveforms at the black level on either side of the H-sync pulses is zero or nearly zero in all 4 photos. This means that the DC-Restoration is very close to 100%.
You will notice that the amplitude of the sync pulses is suppressed from about 8Vp-p with a black scene down to 6Vp-p with a white image. This is due to the loading of the additional 1N34 diode I wired into the CRT-Cathode circuit to partially restored the video in the CRT-cathode path.
NTSC Test Pattern Generation
A side comment on test pattern generation: Instead of buying a classic NTSC generator, I simply used a small cheap set-top DVR with an NTSC CH3 output. I put all the test patterns that are are freely available on the web in a USB stick plugged into the DVR. You can even view the classic American test pattern, the Indian head. The DVR also works as an ATSC to NTSC converter for antenna reception off the air in the USA.
For a full report on the earlier restoration and analysis of this set go to Pilot TV37 Repair and Operational Details.
Conclusion
The possibility of injecting incomplete DC-restoration in two parallel video paths to the CRT-grid and CRT-cathode is what made it possible to reach a net 100% DC-restoration without an optimized dedicated DC-restoration circuit.
With DC-restoration improved from 31% to 100%, the TV-37 is much more enjoyable to watch with correct brightness for any scene, regardless of scene content.
Regards,
-Joe
Attachments:To thank the Author because you find the post helpful or well done.