Mod a CRT to increase its TVL

[LukeEvansSimon]

Multi-sync arcade CRTs are "the missing link" between low TVL consumer CRT TVs and high TVL professional video monitors.

With the goal of understanding the engineering differences between a high TVL CRT and a low TVL CRT, I've been comparing cathode amplifier designs from consumer CRT TVs, high resolution multi-sync arcade CRTs, professional video monitor CRTs, and PC monitor CRTs. I have found a few interesting clues that I will point out using specific CRT case studies below. In the following case studies, I will refer to the maximum TVL input test pattern that a cathode amplifier can amplify without distortion as "Amp TVL".

Consumer CRT TV vs. Multi-sync Arcade CRT
Consumer CRT TVs and high resolution multi-sync arcade CRTs use the same consumer grade CRT tubes, but the high resolution multi-sync arcade CRTs are capable of supporting higher resolutions than the consumer grade CRTs. Why? It comes down to two things: (i) negative charging the G1 anode, and (ii) cathode amplifier bandwidth. What is VERY interesting is that negative charging the G1 anode helps the cathode amplifier achieve a higher bandwidth, because the cathode amplifier can be operated at a lower voltage when the G1 anode has a negative charge! Here are the neckboard designs of a consumer CRT TV and a high resolution multi-sync arcade CRT that both use the same tube:

R=Description&Model Number&Tube&G1 Anode Voltage&Amp Supply Voltage&Amp Design&Amp TVL R=ElBartoMe TV&Sony KV-E3431D&???&0 V&200 V&Cascode + Push-pull&likely around 480 R=JVC D-Series&JVC AV-36D502&A90AEJ15X01 36-inch consumer grade curved slot mask&0 V&217 V&Common Emitter&likely less than 480 R=Multi-sync Arcade CRT&Weiya M3138F-82&A90AEJ15X01 36-inch consumer grade curved slot mask&-70 V&150 V&Cascode + Push-pull&960

Consumer TV: JVC D-Series
Here is the schematic for the JVC D-Series neckboard. The schematic is confusing because at first glance it looks like it uses a cascode amp design, but you can see the footnotes for the schematic state that 3 of the NPN transistors are not mounted. So it really uses a very simple common emitter amplifier design. The NPN transistor that is used in the common emitter amp is the 2SC4544, and its datasheet states that its bandwidth product at 10 volts from collector to emitter is 70 mhz. However, the amplifier is running at approximately 217 volts from collector to emitter, so the realized bandwidth will be much lower as the amplifier's application varies cathode voltage by around 135 Vpp as opposed to just 10 Vpp.


Multi-sync Arcade CRT: Weiya
So now take that same picture tube (A90AEJ15X01), but hook it up to the Weiya's neckboard and higher resolutions can be achieved. But why? First note that the Weiya's flyback has a dedicated pin and the chassis has a dedicated rectifier diode, smoothing capacitor, and potentiometer for setting G1 anode voltage. This is equivalent to ElBartoMe's version of the mod, where he adds a new winding to the flyback. The Weiya's flyback just has this extra winding hidden inside the flyback.

At first I thought the Weiya chassis was using negative charging of G1 to only control the spot size. While it is used to control the spot size, it is also used by this chassis to allow for the cathode amplifiers to be run at a lower voltage! Inside the CRT's electron gun, the relative voltage difference between cathodes and G1 ends up being roughly the same as found in the JVC D-Series. Running the cathode amps at a lower voltage is an important point to note, and something we will see over and over again in high resolution CRT neckboard designs. When displaying 240p content, both the JVC and the Weiya have very thick blank scanlines in between the illuminated scanlines, so they both appear to have a nice small spot size, but the Weiya has a much higher bandwidth amplifier design.


Weiya vs ElBartoMe's Sony
The amplifier design in the Weiya is the same cascode + push-pull design used in ElBartoMe's Sony KV-E3431D, but many of the transistors seem to have a higher bandwidth, AND the Weiya chassis is running its amplifier at a lower voltage of 150 V versus ElBartoMe's Sony KV-E3431D, which is running its amp at 200 V. Here is a comparison of the datasheet specs of the transistors in the Weiya's amplifier and ElBartoMe's Sony KV-E3431D:
R=Chassis&Stage&Type&Configuration&Original Part&Original Bandwidth R=Weiya M3138F-82&Cascode&NPN&Common Emitter&PH2369&500 mhz R=Sony KV-E3431D&Cascode&NPN&Common Emitter&2SC2412&180 mhz R=Weiya M3138F-82&Cascode&NPN&Common Base&2SC3790&[i][b]150 mhz[/b][/i] R=Sony KV-E3431D&Cascode&NPN&Common Base&BF871&[i][b]60 mhz[/b][/i] R=Weiya M3138F-82&Push Pull&NPN&Common Collector&BF422&60 mhz R=Sony KV-E3431D&Push Pull&NPN&Common Collector&BF871&60 mhz R=Weiya M3138F-82&Push Pull&PNP&Common Collector&BF423&60 mhz R=Sony KV-E3431D&Push Pull&PNP&Common Collector&2SA1091&60 mhz

As I mentioned before, the NPN transistor used in the "Common Base" amplifier configuration is likely the most bandwidth intensive point in this amplifier design because it handles more voltage displacement than any of the other transistors in this amplifier ensemble. The Weiya's amplifier circuit uses a 150 mhz transistor for this component, versus the Sony's amp circuit that uses a 60 mhz transistor. This suggests the importance of upgrading the Sony's common base transistor, but as we will see in the next case study, a lower cathode voltage may make the biggest difference.

The most important take away from this comparison to the arcade CRT is that THE PICTURE TUBE ITSELF IS NOT THE SOLE DETERMINING FACTOR IN RESOLUTION! Neckboard design is a very big determining factor in a CRT's resolution!

Comparisons to a Pro Video Monitor CRT
In this case study, we compare and contrast the neckboard design of ElBartoMe's consumer grade Sony KV-E3431D with a professional video monitor that uses a professional grade "M-class" CRT. Again, the goal is to see what neckboard design differences enable higher resolutions on a CRT.
R=Description&Model Number&Tube&G1 Anode Voltage&Amp Supply Voltage&Amp Design&Amp TVL R=ElBartoMe TV&Sony KV-E3431D&???&0 V&200 V&Cascode + Push-pull&likely around 480 R=Pro Video Monitor CRT&Panasonic BT-H1390Y& M34KNZ092X 13-inch pro grade curved dot mask&-70 V&145 V&Cascode + Push-pull&750 R=Multi-sync Arcade CRT&Weiya M3138F-82&A90AEJ15X01 36-inch consumer grade curved slot mask&-70 V&150 V&Cascode + Push-pull&960

The Panasonic pro video monitor runs its amplifiers at an even lower voltage than the arcade CRT and much lower than the consumer grade Sony CRT, but all of these 3 CRTs use a very similar cascode + push-pull amplifier design. So let's compare the transistor datasheet specs to see if that is how the pro monitor achieves 750 TVL:
R=Chassis&Stage&Type&Configuration&Original Part&Original Bandwidth R=Panasonic BT-H1390Y&Cascode&NPN&Common Emitter&2SC4502&300 mhz R=Sony KV-E3431D&Cascode&NPN&Common Emitter&2SC2412&180 mhz R=Panasonic BT-H1390Y&Cascode&NPN&Common Base& 2SC4544&[i][b]70 mhz[/b][/i] R=Sony KV-E3431D&Cascode&NPN&Common Base&BF871&[i][b]60 mhz[/b][/i] R=Panasonic BT-H1390Y&Push Pull&NPN&Common Collector&2SC3334&100 mhz R=Sony KV-E3431D&Push Pull&NPN&Common Collector&BF871&60 mhz R=Panasonic BT-H1390Y&Push Pull&PNP&Common Collector&2SA1321&80 mhz R=Sony KV-E3431D&Push Pull&PNP&Common Collector&2SA1091&60 mhz

As I mentioned before, the NPN transistor used in the "Common Base" amplifier configuration is likely the most bandwidth intensive point in this amplifier design because it handles more voltage displacement than any of the other transistors in this amplifier ensemble. But wait a second, the Panasonic's part is spec'ed at 70 mhz versus the Sony's part that is spec'ed at 60 mhz. That doesn't seem to be enough of a difference to explain how the Panasonic's amplifier correctly amplifies a 750 TVL test pattern. So how does it work? The lower amplifier supply voltage is a clue. The negatively charged G1 voltage is a clue. The schematic for the Panasonic's neckboard holds the smoking gun for achieving higher amplifier bandwidth:


The Panasonic is running the cathodes at around 50 volts peak to peak! Consumer CRT TVs run the cathodes at around 110 to 200 volts peak to peak! As we saw above in the case study of the JVC D-Series, it is running its cathodes at around 135 volts peak to peak! So that brings this case study to an important CRT neckboard modification idea.

Run Both Cathodes & G1 At Lower Voltage To Achieve Higher TVL
So another option for modifying a consumer grade CRT TV's cathode amplifiers to make it correctly amplify high TVL content, is to run the amplifiers at a lower voltage and leverage an even MORE negative G1 anode voltage to restore cutoff. Lowering the amplifier voltage involves lowering the gain of the amplifier, and the voltage differential between the cathodes and the G1 anode should not be lower than before, as that would increase spot size and cause issues with cutoff, but instead the cathode to G1 voltage will be higher than before the mod, due to even more negatively charged G1 anode. Since modifying your TV to negatively charge the G1 also has the benefit of creating a smaller spot size, simply lowering the voltage of the cathode amplifiers may be an easier path to higher TVL than trying to achieve it by upgrading transistors. Of course, upgrading the transistors may also help.

One final case study shows that lowering the amplifier voltage may be as simple as replacing one resistor per cathode. This brings me to the LM2422, which is a monolithic IC that contains analog video amplifiers for R, G, and B cathodes in a single package. This IC is designed to be used in HDTV CRTs that support up to 1080i input signals. 1080i has a horizontal resolution of 1920, which translates to a TVL of (1920 * 9/16) = 1080 TVL. In other words, the LM2422 has very high bandwidth! Here is the LM2422's datasheet's schematic for the amplifier design inside the monolithic IC:


This is the same cascode + push-pull amplifier design as used in the other CRTs, with two differences. First the LM2422 does not use diodes for biasing the push-pull stage, but instead it uses the transistor Q6. I don't think this helps improve bandwidth, but instead decreases the push-pull cross-over distortion. The second difference is the PNP common collector amplifier Q5, which may help increase bandwidth by increasing input impedance of the amplifier ensemble. Here is the datasheet's explanation of the circuit:


This case study suggests that the absolute value of the amplifier gain can be decreased in a cascode + push-pull circuit by changing the value of the R1 resistor or the value of R1. I am going to simulate in LTSPice to double-check.

[LukeEvansSimon]

Since 15khz consumer CRT TVs lack the cathode amplifier bandwidth to properly amplify a high TVL signal, I have been exploring ways to modify 15khz consumer CRT TVs to increase their cathode amplifier bandwidth so that the G1 charging mod's full potential can be realized. OnSemi is the only company that still manufactures the type of high voltage transistor (the KSC3503DS ) needed to drive CRT cathode amplifier circuits, and they publish a free PSpice model for the transistor KSC3503DS, so I played around with LTSPice simulations of a few cathode amplifier circuit designs that use the KSC3503DS to drive the CRT cathodes. Here are what the simulations showed.

Common Emitter Amplifier
This is a very common neckboard design for late generation 15khz consumer CRT TVs, and even with the very high bandwidth KSC3503DS, the achieved bandwidth is around 3mhz @ -3dB. The JVC D-Series uses a neckboard with the common emitter amplifier design and it uses a much less capable transistor. This amplifier design can only amplify an approximately 312 TVL pattern and beyond that, the white lines will start to become increasingly grey and the black lines will become increasingly grey. This brings up a clear problem with the TVL metric for analog CRTs. This amplifier could output a 400 TVL test pattern. The white lines would be light grey and the black lines would be dark grey, but you could still count distinct alternating lines, but technically they are no longer white and black lines. At what point of white and black correctness does it still count as achieving that TVL? The measure I am using here is -3dB, but the TVL standard does not mention this as the threshold.


Cascode Amplifier
This was a very common consumer CRT neckboard design in the early days of transistor amplifiers, when transistors were less capable than later years. This amplifier design uses two transistors to compensate for the Miller effect that robs the common emitter amplifier from its bandwidth. At -3dB, this amplifier can drive up to 468 TVL. Again, a higher TVL pattern would still be discernible at -10dB, but the white lines would be light grey and the black lines would be dark grey. The TVL metric is ambiguous in this regard.


Cascode Amplifier with Emitter Follower Output
This amplifier design is the same as the cascode, but adds a third transistor in an emitter follower configuration as the output stage of the amplifier. This emitter follower acts as a current amplifier, not a voltage amplifier, and it allows the CRT's electron gun to vary the current it draws from the amplifier circuit, with less negative impact to the bandwidth of the circuit. This amplifier design can amplify a 958 TVL pattern at -3dB.

[LukeEvansSimon]

The OnSemi KSC3503DS transistor is quite impressive. I did more simulations with it to see how hard it can be driven, and it turns out that even in a simple common emitter amplifier CRT neckboard circuit design, it can amplify a 1000 TVL test patterns at -0.5dB and 1910 TVL test patterns at -3dB. Below is the simulation that shows how to achieve this using a CRT neckboard design taken from the JVC D-Series. While the amplifier circuit architecture matches that of the JVC D-Series, the high voltage transistor is significantly upgraded and some of the key resistor and capacitor values are taken from the pro video monitor Panasonic BT-H1390Y, with some tuning of their values. I explain the theory behind why this works below. The important thing to note is that this outlines a simple way to increase JVC D-Series neckboard bandwidth by upgrading a few resistors and transistors. Many other low-end to mid-range consumer CRT TVs from the 1990s and early 2000s use the same common emitter neckboard design, so they can also be easily modded using the approach outlined below:



Effectively Driving Capacitive Loads
The electron gun is simulated in terms of an impedance load RLoad and a capacitive load CLoad. RLoad simulates the cathode ray current that draws the picture on the CRT, but it is the capacitive load CLoad that causes the most challenges for the amplifier. Inside the electron gun, the cathodes are positioned very close to the G1 anode, and this creates a parasitic capacitance of around 10pF. As the video signal's frequency increases, the amplifiers gain degenerates because this capacitive load starts to act like a short to ground. This can be addressed by decreasing the amplifier's output impedance, which historically was accomplished by either adding a push-pull output stage to the amplifier circuit or by adding an emitter follower output stage... or both!

However, transistor manufacturing capabilities have improved considerably since the late 1990s, and for this reason, the KSC3503DS is a very capable transistor, and a more simple solution is now possible: decrease the amplifier's output impedance by decreasing the impedance of R354 and the impedance of R351. Lowering the impedance of R351 increases the current that flows across the KSC3503DS, but the KSC3503DS can handle it, if you clip on a little heatsink to it. Clip on heatsinks for transistor amps are very common in professional video CRT neckboards and multi-sync arcade CRT neckboards. In fact, the values for these two resistors is very close to the values used in the pro video CRT: Panasonic BT-H1390Y. In addition, lowering the impedance of R351 slightly decreases the voltage gain of the amplifier, but it dramatically increases the switching speed of the amplifier.

Improving High Frequency Roll Off
C3 and R2 are crucial for optimizing the high frequency roll off response of the common emitter amplifier. R13 acts as a negative feedback, which is necessary to stabilize the amplifier by suppressing DC voltage gain and preventing thermal runaway, but in order to prevent gain degeneration at higher frequencies, the bypass capacitor C3 and and corresponding bypass resistor R2 are added (reference). The value of C3 needs to be tuned to correspond to the high frequency roll off point, and R2 needs to be tuned to large enough to provide amplifier stability at high frequency but not too high impedance as that would degenerate the gain too much at high frequencies.

[ElBartoME]

Nice work Luke! That is a lot of useful information for the future. My transistor came in the mail some days ago and I hope I find enough time soon to exchange the transistors on the E3431D.

In the meantime I implemented the mod into some other monitors.

Sony KV-29X5D

Before

Spoiler

After

Spoiler

As you can see the tube of the 29X5D was a bit tired and blurry at high contrast. The mod was able to make the picture much sharper. Standard G1 voltage is 0V. I changed it to -130V. Unforunately on this one you need to scratch away traces. If someone needs more information let me know.

Here an imgur album where you can see the modification: https://imgur.com/a/XW2QMpt


PVM-2730 (same as PVM-2530)

Before

Spoiler

After

Spoiler

The 2730 was already pretty good before the mod. The difference after the mod is subtle, but the scanlines are definitely a bit thinner now. Modding the 2730 was basically exactly the same as with the A3411D. I made a lengthy post for that TV some pages ago. G1 was standard at around +30V and I changed it to -120V.



I removed R705 completely. I exchanged C703 with a 200V 10µF electrolytic cap and reversed the polarity. Otherwise the cap will explode if you put -120V across it. Otherwise I use the same PCB I made for the A3411D. Only difference is I used an electrolytic cap instead of a film cap as I ran out of them. Works just as well.

Album with more images: https://imgur.com/a/AXc12R8

[LukeEvansSimon]

...

More before and afters is great. The vertical resolution is visibly improved as the horizontal fields are overlapping less.

By the way, I have found that some CRTs have a flyback pin that outputs an inverted form of the spiking voltage wave form that all of the other pins output. It looks like this pin can be used to generate the negative voltage rail, as opposed to adding the extra winding to the flyback. Here is the reference waveforms from the JVC AV-32D303:


Note that pin 6 has an inverted wave form of 250 Vpp. The JVC D-Series uses a single diode to rectify this voltage coming off of this pin and then passes it into a smoothing capacitor to create a 29V rail for the vertical deflection IC's input voltage rail. This suggests that most of the 250 Vpp is below chassis ground, and could therefore be used to also generate roughly -220 volt rail? I need to oscope it this weekend to see if this makes sense. As long as a current limiting resistor is used to limit the current draw during powering on, the flyback coil that drives pin 6 should be able to drive both the vertical deflection IC and charging the capacitor that charges the G1.

All of my JVC CRT TVs have this inverted pin, and all of my consumer Panasonic CRT TVs also have it. Have you checked your Sony CRTs to see if an existing flyback pin can be used?

[ElBartoME]

All of my JVC CRT TVs have this inverted pin, and all of my consumer Panasonic CRT TVs also have it. Have you checked your Sony CRTs to see if an existing flyback pin can be used?

It would be great to get the negative supply without using the additional winding but there are no waveforms for the flyback pins in the Sony service manual. The only thing I find is a -15V winding coming out of the flyback.

[LukeEvansSimon]

Have you oscoped them all to see if any of them are inverted? The JVC chassis has the flyback pins with a non-inverted waveform roughly 90% above chassis ground and 10% below chassis ground. The pin with the inverted waveform is the opposite with most of the waveform below chassis ground.

[Blacksheep]

This brings up a clear problem with the TVL metric for analog CRTs. This amplifier could output a 400 TVL test pattern. The white lines would be light grey and the black lines would be dark grey, but you could still count distinct alternating lines, but technically they are no longer white and black lines. At what point of white and black correctness does it still count as achieving that TVL? The measure I am using here is -3dB, but the TVL standard does not mention this as the threshold.

The signal (generator) is supposed to use black voltage level and white voltage level for the lines, and the TVL count as long as the individual lines are still discernable/countable on the screen to be measured, hence:

the maximum number of alternating light and dark vertical lines that can be resolved per picture height

The definition is clear in that regard, no ambiguity at all. Otherwise, interesting stuff, keep it coming!

[LukeEvansSimon]

This brings up a clear problem with the TVL metric for analog CRTs. This amplifier could output a 400 TVL test pattern. The white lines would be light grey and the black lines would be dark grey, but you could still count distinct alternating lines, but technically they are no longer white and black lines. At what point of white and black correctness does it still count as achieving that TVL? The measure I am using here is -3dB, but the TVL standard does not mention this as the threshold.

The signal (generator) is supposed to use black voltage level and white voltage level for the lines, and the TVL count as long as the individual lines are still discernable/countable on the screen to be measured, hence:

the maximum number of alternating light and dark vertical lines that can be resolved per picture height

The definition is clear in that regard, no ambiguity at all. Otherwise, interesting stuff, keep it coming!

Do you have a reference for your claims? The only definitions that I have read mention discernible alternating dark and light lines. I am skeptical it would be specified in terms of electron gun voltages. It would make more sense to use nits, lumens, lux, or what scientists use photo flux density to define white and black. I could calibrate white and black voltage levels to crush whites and blacks and the reach a higher TVL that way.

I honestly think TVL is a metric that was more used for marketing purposes than for engineering purposes.

[Ikaruga11]

All of this stuff is completely over my head, but I am still amazed nonetheless at the progress and research/effort being put into this. Keep up the amazing work!

[LukeEvansSimon]

All of this stuff is completely over my head, but I am still amazed nonetheless at the progress and research/effort being put into this. Keep up the amazing work!

CRT gaming is life

More cathode amplifier simulations, this time with the exact same R13 and R351 resistor values as the Panasonic BT-H1390Y, but using the modern KSC3503D transistor in a common emitter amplifier design. More current across the transistor, so more heat dissipation, but bandwidth is improved with ~1000 TVL at +0.01dB.

[ElBartoME]

Have you oscoped them all to see if any of them are inverted?

No, I haven't. I'll take a look when I replace the transistors.


That common emitter amplifier is interesting. Can you plot the output of the transistor? How big is the voltage swing and at what voltage levels is it operating?

[LukeEvansSimon]

Have you oscoped them all to see if any of them are inverted?

No, I haven't. I'll take a look when I replace the transistors.


That common emitter amplifier is interesting. Can you plot the output of the transistor? How big is the voltage swing and at what voltage levels is it operating?

The amplified output is roughly 85 volts peak to peak. The JVC D-Series that this simulation is calibrated for uses 100vpp. Assuming black levels are calibrated correctly, this decrease in vpp will decrease the light emission for the whites and bright color ranges. However, since the decrease in resistors R354 and R351 allows for increased cathode currents at this lower voltage, it is not clear if the peak cathode emissions will be decreased or increased. I need to get the soldering tools out to experiment in real life and not in simulation. However, LTSpice simulation is useful for getting close to the right values for the passive components.

After doing these simulations, I have come to believe that for your CRT, you will get terrible results if you don’t calibrate the circuit by adjusting a few resistors and capacitors.

[Blacksheep]

Do you have a reference for your claims? The only definitions that I have read mention discernible alternating dark and light lines. I am skeptical it would be specified in terms of electron gun voltages. It would make more sense to use nits, lumens, lux, or what scientists use photo flux density to define white and black. I could calibrate white and black voltage levels to crush whites and blacks and the reach a higher TVL that way.

I honestly think TVL is a metric that was more used for marketing purposes than for engineering purposes.

You misunderstood. Even though the signal (generator) is supposed to use black voltage level and white voltage level for the lines, the TVL definition is not concerned with voltages—neither are they part of the definition nor did I claim that, thus I owe no reference for that whatsoever. At the end of the day what matters is what is seen on screen, and that is what TVL measures.

I think it's undisputed and I see no reason to challenge that the TVL count as achieved as long as the individual lines are still discernable/countable on screen, and I hope it's self-explanatory that you definitely will not be able to count more individual lines by decreasing voltage difference between the parts of the signal that represent the lines from the start, that is, in the signal generator. Best case scenario you will be able to keep the frequency of alternation and thus count the same number of lines (although since by definition the lines are already gray with minimal difference when displaying maximum achievable TVL pattern—if that is not the case then you should increase alternation frequency, which, the less difference between the lines from the start, the sooner a ceiling of number of countable lines will be hit), and by tendency, you will have to decrease frequency of alternation until the lines become countable again and therefore count less lines.

[LukeEvansSimon]

Do you have a reference for your claims? The only definitions that I have read mention discernible alternating dark and light lines. I am skeptical it would be specified in terms of electron gun voltages. It would make more sense to use nits, lumens, lux, or what scientists use photo flux density to define white and black. I could calibrate white and black voltage levels to crush whites and blacks and the reach a higher TVL that way.

I honestly think TVL is a metric that was more used for marketing purposes than for engineering purposes.

You misunderstood. Even though the signal (generator) is supposed to use black voltage level and white voltage level for the lines, the TVL definition is not concerned with voltages—neither are they part of the definition nor did I claim that, thus I owe no reference for that whatsoever. At the end of the day what matters is what is seen on screen, and that is what TVL measures.

I think it's undisputed and I see no reason to challenge that the TVL count as achieved as long as the individual lines are still discernable/countable on screen, and I hope it's self-explanatory that you definitely will not be able to count more individual lines by decreasing voltage difference between the parts of the signal that represent the lines from the start, that is, in the signal generator. Best case scenario you will be able to keep the frequency of alternation and thus count the same number of lines (although since by definition the lines are already gray with minimal difference when displaying maximum achievable TVL pattern—if that is not the case then you should increase alternation frequency, which, the less difference between the lines from the start, the sooner a ceiling of number of countable lines will be hit), and by tendency, you will have to decrease frequency of alternation until the lines become countable again and therefore count less lines.

My comments are in the context of CRT neckboard high voltage amplification of the raw RGB signals that drive the cathodes in the CRT's electron gun, not the input signals, which operate at such a low voltage that high bandwidth amplifier design is trivial. The high voltage amplifiers start to decrease their gain for very high TVL signals, which causes alternating white and black vertical lines to become alternating light grey and dark grey vertical lines. In the extreme, the light grey and dark grey become the same lightness of grey and the lines blur together, not due to phosphor pitch and not due to spot size, but due to a lack of amplifier bandwidth.

If the TVL metric only requires "discernible" lines of alternating light and dark color, then there is huge amounts of ambiguity. Discernibility will vary depending on the person doing the discerning. A better metric would define the nits of black and the nits of white, and therefore if the amplifier can't push true blacks and true whites at high frequency, it doesn't count. The HDTV industry is doing a better job here with things like HDR standards that use standard measures of luminance (such as "nits").

Regardless, I've found an inexpensive way to upgrade the neckboard amps in my Panasonic and JVC consumer CRT TVs, which use a simple common emitter amplifier design. I am sure ElBartoMe will engineer a bandwidth upgrade to the more complicated amplifier architecture used by Sony's CRTs (cascode with a push-pull output stage). There is still a good chance that the pre-amp bandwidth in consumer CRTs is not high enough, but I think it is feasible to engineer an inexpensive upgrade to pre-amp bandwidth by just copying the circuit design from late generation multi-sync arcade CRTs. I will eventually get to that when I finish testing the cathode amp upgrade mod.

That leaves phosphor pitch as the only limiting factor that has no feasible inexpensive solution, but as I mentioned early in this thread, most consumer CRTs use a horizontal resolution that is half the resolution that can be achieved with the phosphor pitch. So removing all of these other resolution bottlenecks should result in a measurable improvement to both horizontal and vertical resolution in most consumer CRTs. There will also be the added benefit that G1 anode charging and upgraded cathode amps will allow for squeezing more hours out of a single CRT tube. No longer will people need to discard a CRT because its cathode tips have dulled and the image sharpness is lost. The upgraded cathode and G1 anode circuitry will allow for maintaining a tight cathode ray.

[Josh128]

If the TVL metric only requires "discernible" lines of alternating light and dark color, then there is huge amounts of ambiguity.

He is not only arguing that, but also argued earlier that the lines dont have to be equal in width. So if you take all that together, you could measure TVL by using alternating primary colors (remember, all you need is "light and dark") and thus it is mathematically possible to get a maximum 3X increase in TVL for a given pitch aperture grille using his interpretation vs allowing only black and white lines for the measurement. I think thats a ridiculous interpretation of the definition of TVL, but thats what the man said.

Code: Select all

                                                                     RGB    RGB    RGB    RGB 
TVL using the widely accepted interpretation:                         1      2      3      4

TVL using BlackSheeps interpretation:                              123     456    789   10 11 12

[Derf]

I love seeing this thread continue to be active! That's fantastic a built-in inverted voltage pin was found, that will make it a lot easier for many sets.

This brings up a clear problem with the TVL metric for analog CRTs. This amplifier could output a 400 TVL test pattern. The white lines would be light grey and the black lines would be dark grey, but you could still count distinct alternating lines, but technically they are no longer white and black lines. At what point of white and black correctness does it still count as achieving that TVL? The measure I am using here is -3dB, but the TVL standard does not mention this as the threshold.

Any interest in adding "authored ISO 90007" to your CV? Nothing says dedication to the craft like proposing new standards to 13 years out-of-production hardware!
I think you're definitely correct that it's a marketing metric, not an engineering one, and someone can get really creative with the measurements. Whether someone does actually bother to propose an industry standard or not, I think the current CRT community should create a replacement for the interpretive TVL metric that is objectively based on values (voltage, lumens, nits, etc) and clearly defined parameters (equal size columns, each column comprising of a RGB triad).

[Blacksheep]

My comments are in the context of CRT neckboard high voltage amplification of the raw RGB signals that drive the cathodes in the CRT's electron gun, not the input signals, which operate at such a low voltage that high bandwidth amplifier design is trivial. The high voltage amplifiers start to decrease their gain for very high TVL signals, which causes alternating white and black vertical lines to become alternating light grey and dark grey vertical lines. In the extreme, the light grey and dark grey become the same lightness of grey and the lines blur together, not due to phosphor pitch and not due to spot size, but due to a lack of amplifier bandwidth.

Exactly, and it does not matter. What I said still holds true. If you reduce voltage difference from the start, that is, in the signal generator, you will reduce voltage difference everywhere down the path, and you will also reduce the resulting difference in luminance this way. In fact if you reduce voltage difference at any point in the system, achievable TVL will by tendency be reduced as well. It's not possible to increase achievable TVL by decreasing voltage difference anywhere. This should be obvious. It's exactly those bottlenecks that you are trying to upgrade here; they decrease achievable TVL, they don't increase it.

If the TVL metric only requires "discernible" lines of alternating light and dark color, then there is huge amounts of ambiguity. Discernibility will vary depending on the person doing the discerning. A better metric would define the nits of black and the nits of white, and therefore if the amplifier can't push true blacks and true whites at high frequency, it doesn't count. The HDTV industry is doing a better job here with things like HDR standards that use standard measures of luminance (such as "nits").

TVL seems to allow for some wriggle room, and it would indeed be great if there could be some other metric for measuring and describing sharpness of CRT displays in an objective way. But luminance (which candela per square metre, aka "nit" is a unit of) is not it. It's very odd that you propose this. Luminance does not say a thing about sharpness. If you want to measure (improved) sharpness, you need a unit that measures that, not something that measures luminance. This should also be obvious.

Maybe number of horizontal and vertical phosphor triads is better, but only as an upper limit/bottleneck, other factors can reduce achievable sharpness before as we've seen.

He is not only arguing that, but also argued earlier that the lines dont have to be equal in width. So if you take all that together, you could measure TVL by using alternating primary colors (remember, all you need is "light and dark") and thus it is mathematically possible to get a maximum 3X increase in TVL for a given pitch aperture grille using his interpretation vs allowing only black and white lines for the measurement. I think thats a ridiculous interpretation of the definition of TVL, but thats what the man said.

Code: Select all

                                                                     RGB    RGB    RGB    RGB 
TVL using the widely accepted interpretation:                         1      2      3      4

TVL using BlackSheeps interpretation:                              123     456    789   10 11 12

Oh my, what did I do...
I definitely want to distance myself from that.
True, the definition only talks about "light" and "dark" lines. But I always had assumed that this meant shades of gray. An assumption on my part, I had not thought this through. Since it seems that the definition also does not establish any requirement about chrominance, by my logic, colored lines would be allowable.

However, everything is consistent: I don't think you would be able to achieve higher TVL with that, since the voltage (and luminance) difference will be lower. The highest voltage and luminance difference is between black and white. With black and white, you will be able to still count higher alternation frequency than with lines that contain chrominance, see my previous post. And also, I'm actually with you that you can't display more than one line per triad (I was only arguing against using the term "granularity"). So yeah, that would be complete BS.

[Derf]

TVL seems to allow for some wriggle room, and it would indeed be great if there could be some other metric for measuring and describing sharpness of CRT displays in an objective way. But luminance (which candela per square metre, aka "nit" is a unit of) is not it. It's very odd that you propose this. Luminance does not say a thing about sharpness. If you want to measure (improved) sharpness, you need a unit that measures that, not something that measures luminance. This should also be obvious.

He's hinting that "alternating light and dark vertical lines" is an arbitrary definition since it uses vague words of "light" and "dark", thus justifying a need to define light and dark with something like a difference in lumens or nits. Luke is saying that theoretically higher TVL counts could be pushed, but as a side effect it screws up the difference in lumenance of the light and dark lines - at what point is that difference too small to be considered a useful TVL count?

Saying a set is 1000 TVL doesn't mean much if the whole display displays black and whites as white and slightly darker white, even though that fits the TVL definition. For the purposes of the measurements here, it'd be great to have a requirement of "alternating light and dark vertical lines with a difference of 100 lumens between the two" or something to that effect (though with an agreed upon value, not a random value of 100).

[Blacksheep]

Luke is saying that theoretically higher TVL counts could be pushed, but as a side effect it screws up the difference in lumenance of the light and dark lines - at what point is that difference too small to be considered a useful TVL count?

The difference is too small when you can't visibly discern the lines anymore.

Here lies the mistake. "Theoretically" does not matter. What matters is what is seen on screen, and that is what TVL measures. It's practical.

Saying a set is 1000 TVL doesn't mean much if the whole display displays black and whites as white and slightly darker white, even though that fits the TVL definition. For the purposes of the measurements here, it'd be great to have a requirement of "alternating light and dark vertical lines with a difference of 100 lumens between the two" or something to that effect (though with an agreed upon value, not a random value of 100).

TVL measures sharpness. Saying one set is 1000 TVL means a lot. Again, luminance =/= sharpness. Max peak-to-peak amplitude may be lower, but one thousand lines are still resolved. A given display could resolve a high number of lines that have small luminance difference. Measuring luminance or luminance difference would be another metric, not a metric that can replace TVL.

That said I do agree it could make sense to support TVL by also providing a luminance difference for the lines that must not be underrun, especially for engineering.

[LukeEvansSimon]

Maybe I should just rename this thread to the true goal: mod a CRT to increase its OOF!

[Derf]

That said I do agree it could make sense to support TVL by also providing a luminance difference for the lines that must not be underrun, especially for engineering.

This is exactly what we're talking about... a need to have a different standard of measurement (or standard combination of metrics) to hold as the meaningful spec instead of TVL. Nobody is arguing that sharpness = luminance, just that TVL is not well enough defined for the purposes here and it'd be nice to have a more well defined [set of] metric(s).

Maybe I should just rename this thread to the true goal: mod a CRT to increase its OOF!

I support this. Perhaps the super neckboard could tip the scale from a medium to large OOF load.

[LukeEvansSimon]

I've been distracted by my XStation and MemCard Pro, and so I haven't touched this TVL mod hobby in weeks. However, yesterday I had time to simulate cathode amplifier designs in PSpice. Before I had been using the free LTSpice, and I had concerns around the accuracy of LTSpice's small signal AC sweep analysis. PSpice gave more accurate results, which show that a common emitter design using the modern OmSemi high voltage transistor will be enough to achieve around 650 TVL. A multi-stage amplifier design will be needed to amplify signals that are higher resolution than 650 TVL, which explains why PVMs use a cascode + push-pull design. Here is the PSpice schematic, where C1 and R5 are used to simulate the parasitic capacitance and the resistive load of the CRT's electron gun:



Here is the PSpice AC sweep analysis that shows the voltage gain versus RGB signal frequency:


Assuming I stay focused this upcoming weekend (the XStation is amazing btw), I will prototype this amplifier in a breadboard and use a signal generator to perform a real life frequency sweep to confirm the results that I am seeing in PSpice simulations, and if so, then the following weekend I will mod my consumer Panasonic CRT TV to do in situ testing.

[ElBartoME]

I'm interested to see some measurements, although a breadboard is not the best for high frequency designs. A perfboard would be better.

I'm also distracted because I will be moving soon. Then I finally have enough space to have all my TVs and monitors with me in the basement.

[LukeEvansSimon]

I'm interested to see some measurements, although a breadboard is not the best for high frequency designs. A perfboard would be better.

I'm also distracted because I will be moving soon. Then I finally have enough space to have all my TVs and monitors with me in the basement.

Breadboards have a higher parasitic capacitance than in situ, so the results will look worse in a breadboard than what can be accomplished in situ. I am incredibly lazy and so will try the breadboard first I have perf board and so if I don't like the results I get with the breadboard, then I will try perf board. In the simulation, if I remove the C1 capacitor, which simulates the parasitic capacitance in the CRT's electron gun, then the amplifier's bandwidth looks incredibly good. Parasitic capacitance is a real amplifier killer. I measured my CRT's electron gun's capacitance between cathode and G1. It measured at 8 pF, but all of the published literature on CRT amplifier design uses 10 pF. So maybe my measurements were a bit off.

[vol.2]

I think the confusion surrounding the description of the mod as making it "high TVL" might be alleviated a bit if it's phrased more specifically "decreasing spot size to increase TVL such that the pitch of the tube becomes the limiting factor".

I would be interested in replicating your results using some common arcade monitors. If the results are as sharp as you describe, then I expect they can not really be well conveyed in photographs, so seeing it in person would be the best way to judge it.

I'd describe it as increasing resolvable (as opposed to addressable) TVL by decreasing electron beam spot size from the original two phosphor triads width to be an electron beam spot size with a width of one phosphor triad.

This is very verbose though, and people need to get over the fact that two CRTs can have the same TVL, even though one CRT has a much higher mask dot pitch than the other CRT. PC CRTs can change spot size on the fly, as they switch between different resolutions, and clearly PC CRTs are not adjusting mask dot pitch on the fly. At 640x480 resolution, a PC CRT's resolution is 480 TVL, and when it switches to 1024x768 resolution, the resolution is increased to 768 TVL.

Using electronics to control for spot size in 15khz consumer CRT TVs is just taking a known technique from PC CRTs and porting it over to big chonker consumer CRTs. The research articles that I link in the OP are from the PC CRT industry 20 years ago.

I think it's a really interesting project. Thanks for putting so much work into it.

RE: the elephant in the room. I kind of agree with Matt in that you would be well served to adjust the title. The way you word it is (obviously) generating a lot a blow-back, and that's limiting the bandwidth for your positive discussion. Off the top of my head, a really good description for this is "Maximizing the Resolvable TVLs of a Consumer Set," or something along those lines. It appears that the scope of your project has gone beyond a "mod" and some kind of change like that would probably make sense anyway.

[maxtherabbit]

I think the confusion surrounding the description of the mod as making it "high TVL" might be alleviated a bit if it's phrased more specifically "decreasing spot size to increase TVL such that the pitch of the tube becomes the limiting factor".

I would be interested in replicating your results using some common arcade monitors. If the results are as sharp as you describe, then I expect they can not really be well conveyed in photographs, so seeing it in person would be the best way to judge it.

I'd describe it as increasing resolvable (as opposed to addressable) TVL by decreasing electron beam spot size from the original two phosphor triads width to be an electron beam spot size with a width of one phosphor triad.

This is very verbose though, and people need to get over the fact that two CRTs can have the same TVL, even though one CRT has a much higher mask dot pitch than the other CRT. PC CRTs can change spot size on the fly, as they switch between different resolutions, and clearly PC CRTs are not adjusting mask dot pitch on the fly. At 640x480 resolution, a PC CRT's resolution is 480 TVL, and when it switches to 1024x768 resolution, the resolution is increased to 768 TVL.

Using electronics to control for spot size in 15khz consumer CRT TVs is just taking a known technique from PC CRTs and porting it over to big chonker consumer CRTs. The research articles that I link in the OP are from the PC CRT industry 20 years ago.

I think it's a really interesting project. Thanks for putting so much work into it.

RE: the elephant in the room. I kind of agree with Matt in that you would be well served to adjust the title. The way you word it is (obviously) generating a lot a blow-back, and that's limiting the bandwidth for your positive discussion. Off the top of my head, a really good description for this is "Maximizing the Resolvable TVLs of a Consumer Set," or something along those lines. It appears that the scope of your project has gone beyond a "mod" and some kind of change like that would probably make sense anyway.

lol bandwidth modding the thread title of a bandwidth mod thread

[vol.2]

[LukeEvansSimon]

I found a way to add a potentiometer for adjusting the G1 anode voltage without having to add a new winding to the flyback. This greatly simplifies the mod and allows it to be an unpluggable board that is added alongside the flyback (which already has pots for G2 and G3 voltage). CRTs have a flyback pin that is used to create a voltage rail for the vertical deflection circuit. In my TV, it is flyback transformer pin 8, and it generates a 250 volts peak to peak waveform



The idea is to move D561 and C562 to a new circuit board with the following schematic:

Moving D561 and C562 (C9 in the new board schematic) makes room on the main circuit board to add quick disconnect terminals that allow for unpluggable wires between the main board and the new circuit board. The new circuit board's design is straightforward. Diode D561 and capacitor C9 are to ensure that the vertical deflection circuit's voltage rail is preserved. R4 limits the current flow during powering on. Capacitor C1 and diode D2 accomplish what the additional winding accomplishes in ElBartoMe’s design. C1 is a simple coupling capacitor that provides galvanic isolation, yet provides a "new" 250vpp power source. Diode D2 ensures the new power supply is a negative voltage. Diode D3 is a rectifier to ensure the new power supply is DC. Cap C8 is a switching noise dampener. C5, R1, L1, and C3 act as a CRLC filter so that the power supply's DC voltage has minimal ripple and minimal noise. D5 provides more ripple reduction, and potentiometer U1 allows the user to control the voltage of the G1 "grid" anode.


In simulation the ripple voltage is a single microvolt peak to peak, but in real life, the ripple is 6 millivolts, which is not bad.



My CRT had a jumper that connects the G1 anode to ground. I desoldered that jumper and installed two quick disconnect terminals so that I could plug cables in that run from the new circuit board's ground and vcc terminals respectively.


Left picture is the potentiometer set to 0 volts, and right picture is the potentiometer set to -229 volts. If the potentiometer dial is tuned to a setting in the middle, then the picture looks halfway between the scanline thickness of the left and of the right.

Next weekend, I will solder a permanent perfboard version and install it inside the TV chassis next to the flyback so that G1, G2, and G3 potentiometers are all easily accessible.

[SuperSpongo]

That's so cool! So with this pot you can very easily demonstrate the effect of the G1 voltage on spot size. I bet it's so fun to play around with that pot