The Nanao MS9-29 is a CRT monitor featured in many arcade machines of the 1990s and 2000s. It feature a 27/29" tube (depending on measurement methods). It has a reputation for high reliability, and excellent compatibility with different arcade video sources.
This monitor may be best known for its usage in the Sega New Astro City and Taito Egret II cabinets, but it can be found in others as well.
This page has been written to facilitate configuration, maintenance, and enhancement of this wonderful monitor.
It is not exhaustive, and is updated irregularly.
Disclaimer: This is a writeup of my experiences and observations. I take no responsibility if you plow forwards without double-checking your work, or make mistakes, or don't think check that I didn't make some critical error. If you break it, you did it! Not my problem!
As the successor to the similarly popular Nanao MS8-29 / MS8-25, the MS9 has some big shoes to fill. Fortunately, in virtually every category, the MS9 is a strict improvement over the MS8. The only advantage the MS8 holds over the MS9 will be addressed in this document as easy enhancements that can be made to the MS9.
This page was written under the presumption that the reader has some familiarity with CRT monitors, and arcade hardware. However, in order to make it more palatable to those from other backgrounds, some insider-common terms should be explained.
This list of specifications is not exhaustive - minutia about the monitor dimensions, compatibility, peak luminosity, etc. can be found elsewhere.
"MS9" is a catch-all term for a number of configurations that were sold to OEMs. Unfortunately, I can not find any sales information about the monitor from its day, so the only data we have is information from cabinet manuals, and observations.
Here are some variants that I've encountered, and associated notes. In short, they are all basically the same - at most, add some X-caps, change out some S-caps, and one becomes another.
This is the model generally found in the Sega New Astro City, and from what I've heard some other Sega cabs. You might find it in an Astro that was upgraded down the line.
It is paired with the A68KJU96X tube and associated yoke. It is mounted in an open-frame steel frame (the "number muncher"). Some later MS9-29A feature the automatic switching functionality. I have never seen a Versus City internally but that is where I would expect to see them.
This model I have always found in Taito Egret II cabinets. Contrary to popular belief, this chassis has no differences to accommodate the auto-degauss function of the cabinet; that is solely a matter of how the cabinet is wired. Any chassis can do that.
What does set this model apart - and I still don't understand why - is that the primary section does not feature the following capacitors: C994, C992, or C991. The latter two are X-capacitors. If you use this chassis in another cabinet, you are likely to experience interference. You can use this chassis in other machines with no trouble, but populate those caps if you do.
Also, the MS9-29T features a small laminated metal shield underneath the signal section. The reason is not clear.
It is also completely okay to do the reverse: use a -29A or other variant in an Egret II. Having those caps in there will not cause problems.
As the Egret II has its own chassis mount, this monitor does not have a frame like the MS9-29A does.
This model is frequently seen in American and European machines. It is paired with a Hitachi tube, rather than a Toshiba. The inductance measurements are more or less the same, but the S-capacitor values are different. It can be interchanged with other chassis, but the horizontal geometry may not be as good as intended (compression near the edges). Swapping the S-caps should make it a complete conversion.
From what I can tell, this appears to be comparable to the MS9-29A - it's "totally normal". I've never seen an auto-switching one of these.
These are some things that should be done when getting a monitor for the first time. Some time spent now can save you a lot of time, money, and hassle later. Not doing this would be comparable to buying a car with broken headlights, not replacing them, and then complaining that you can never see in the dark.
As mentioned earlier in this page, these monitors are 20+ years old. Components age over time, more so when exposed to long hours of usage and high temperature. Electrolytic capacitors - the little electronic components that often look like colorful soda cans - quite often drift in equivalant series resistance (ESR) and capacitance, making them less effective at whatever job they have been assigned.
Truly, it is not necessary to replace all the capacitors in a monitor, as many of them likely still work. Both ESR and capacitance should be measured out of circuit. As a rule of thumb, a cap is likely to have both specifications drift off in tandem, but not always.
Personally, given the age of these monitors and the interest in doing preventative maintenance to save headache later, I opt to just replace all of the electrolytic capacitors on the chassis. By the time you desolder them to gauge their condition, you've already reached the same level of effort required to put in new ones. Some capacitors can be absolutely fine now, but you might in five years find yourself pulling the chassis to replace the ones you skipped now.
Again - I can't stress this enough - these electronics are old workhorses that have seen some shit! Thermal and physical stress takes a toll on the solder joints that weld the electronic components to the PCB.
A cold solder joint is one that has began to crack or gone brittle, and no longer represents a reliable connection. These solder joint cracks encourage galvanic corrosion, which increases resistance, leading to signal instability/variance and increased heat - which promotes further corrosion and oxidation. This can make operation intermittent, or even outright cause the chassis to not work, if it gets bad enough.
Signs of a poor solder joint:
This category of problem is one that leads a lot of hobby newcomers to parrot instilled fears that monitors are some unreliable house of cards that fail frequently. Arm yourself with knowledge, and the fear of the unknown will dissolve.
Using a soldering iron, reflow solder points that may attach to a large component (and thus place mechanical strain on the point), things that operate at high voltages, or just look suspicious. Definitely reflow all of the connections to the potentiometers. A little flux goes a long way. Even if things look good, it's not a bad idea to just do a once-over on a chassis, and feel a lot more sure about it for a long time.
I am not going to describe the process of replacing capacitors - you desolder the old ones and solder in new ones - but here are some points that should be kept in mind.
For a monitor to look good, it's got to be able to display an image properly! It will simply look a lot better if the three color channels are well balanced, bright, and focused.
Our goal here is to make the three primary colors of the monitor (red, green, and blue) be represented on the screen with the same proportions as one another. Here are the steps I take to do so.
HOT TIP! Monitor manufacturers really goofed off with their terminology. Here's an addendum to the terminology defined above.
First, if it is practical, take the monitor out of the cabinet. Or, make sure at least that the pots on the chassis can be easily reached.
Turn down the contrast adjustment as much as possible (counter-clockwise), set the per-channel gains to the middle, and set the "brightness" just a smidge above center.
Turn down the R, G, and B cutoff pots on the chassis almost all the way (counter-clockwise). Leave a tiny bit for wiggle room.
Pull up a dark screen or RGB bars pattern. The Donpachi one pictured here is okay but the bright white grid might make black level adjustment challenging. I actually would prefer to use a mostly-black menu screen for this part. Yes, it's an Egret 29 cabinet, but I have an MS9-29 in it!
Turn down the "SCREEN" pot on the flyback transformer a ways (counter-clockwise); the image will become dark.
SLOWLY raise this level until you can just barely "see" the black part of the raster, as pictured.
The point here is that we want the SCREEN grid voltage to "take us most of the way" to the desired black level and only use the cutoff pots to balance the result. Keeping the signal bias low going to the amplifier on the monitor's neck board gives us a higher ceiling for the signal, which translates into allowing a brighter image.
Using the cutoff pots on the chassis, you want to "balance" the three channels so that the barely-visible black area of the screen becomes a dark grey color. The reason we left the "bright" pot on the remote board a tiny bit high is so that when we center it later, this nearly-black area can become true black.
When a good grey level has been set, you can use the remote board to bring the black level back down to a true black, as pictured.
Now, we have to set the white point. Using the R, G, and B gains on the remote board, balance the peak level of the three color channels so that the color white is a true white, and not tinted with another color. You can turn up the contrast pot for this part as well. The end result should be similar to the grid pattern shown above.
You can use a colorimiter here to target a specific white point and color temperature, if you like.
Here is an example of a poorly set white point. In this example, the green gain is far too low.
Here is a well set white point (admittedly difficult to photograph)
Finally, set the focus using the top pot on the flyback (FOCUS). Often this is a little bit of a tug of war between having good focal clarity near the edges of the image, versus the center. I suggest doing this in-game and not on a sterile grid pattern, lest you second guess yourself all the time.
The "B+" voltage is the main relatively high DC voltage used to drive the flyback transformer and horizontal circuit of the monitor. This voltage should be set correctly for the best operation.
B+ adjustment is done with these two pots. The 24Khz B+ voltage is higher than that used for 15Khz, so there is a separate adjustment.
The B+ voltage is measured at TP2, in the center of the chassis. You may use the heat sink as reference ground, as this is all secondary side (unlike the stupid MS8).
First, adjust for 15Khz. You want to get to 76V DC. Use the 15Khz pot to set the B+ voltage while the monitor is running.
Second, adjust for 24Khz, where you target 116V DC. It is crucial to do 15Khz adjustments first, as the 24KHz ones are offsets from those ones rather than being totally independent.
Now, B+ does not need to be precise. The MS9 runs fairly cool (for a monitor) and does not run right at the edge of what components can tolerate. If bringing B+ up five more volts makes your image a lot better, go for it. It's not going to blow.
Along the side of the MS9 chassis are numerous potentiometers related to geometry - that is, the size and shape of the monitor.
I feel these adjustments are very self-explanatory, and don't feel a need to go into great detail in this section (I expect somebody will write me a colorful message saying otherwise). SPC stands for Side Pincushion!
A note about resolutions: Like the B+ voltage, the 15k values for pincushion and hold frequency are used as a base for the 24k ones, so make sure those are correct first.
Modification: G1 Voltage for Improved Beam Shape
This is a spicy modification to the chassis that can provide a dramatic improvement in sharpness.
The one thing the MS8 has had over the MS9 is that it was always a little bit sharper. It was as if an additional dimension to the common FOCUS adjustment was simply set differently. A shmups post (a little strangely titled) brought up the idea of placing a negative voltage on the CRT's G1 pin (as done in many higher-end monitors) to improve the shape of the beam that strikes the phosphors.
Coincidentally, I had come across an MS8 monitor around the time of that post. I also observed that it was simply brighter and sharper than my beloved MS9, even with its design warts.
I looked at the back and observed that - sure enough - the neck board has a trace going to the G1 pin that is not simply GND (0V). I measured it with a multimeter, and it showed about -68V DC.
Here is a picture of the MS8-29's neck board. Take a look at pin 5, which is G1. On the MS9, this pin is linked to ground through a jumper.
Amusingly, the MS9's neck board actually has provisions for a non-ground G1 voltage, but is wired to ground from the factory.
...and, the chassis itself actually has an unused circuit to produce the voltage, just like on the MS8. It appears to be an unused feature.
So... let's see what kind of voltage that flyback pin puts out. This is something we can safely do with a scope on the MS9, but not the MS8 without an isolated probe.
That negative voltage is perfect for G1. If we populate the circuit, the diode will only pass a negative voltage, and will charge the capacitor to hold it.
This explains what this unused capacitor spot is for.
The neck board's jumper that grounds G1 must be removed. I populated the neck board with a resistor and a cap to be a goody two-shoes, but this really isn't necessary. A wire link for R332 would have sufficed.
It's convenient that the MS9 provides a wire for G1 already, but we must actually hook it up with the voltage.
On older revisions of the MS9, all the required traces are actually in place. Here, J404 needs to be populated, which will then link with the -68VDC trace from above.
On later revisions (more common) that trace has been removed, and the G1 pin is just grounded outright. The pin has to be isolated by cutting some copper, and then a wire has to be run manually from the negative leg of C521 that we installed earlier. You don't have to desolder it like I did, and in the end, it was resoldered anyway.
After all that slicing, all that must be done is a recalibration of the color. A negative voltage on G1 means less electrons can freely pass through the control grid - it's more picky now! But, that's what gives it the improved shape. To compensate, G2 (SCREEN) must be raised so that more electrons are passed through. This will give a very tiny reduction in total luminosity but if you've set G2 correctly to give yourself enough headroom for amplification, it will end up extremely bright and sharp.
Populate these parts:
Isolate the G1 pin on the chassis, if applicable (based on revision) and make sure it is hooked up to the negative leg of C521. Then, recalibrate color.
This mod is a must for the image to look any good if you do the other mod (31k resolution). However, the improvement is strong even for 15k content.
The MS9-29 supports a 24Khz horizontal scan rate, for "medium resolution" games. Think System 24, some older Konami rhythm games, Atari APB and Paperboy, and... the NEC PC-98 series.
24k is cool! ...but, there are very few arcade boards that actually support it. On the other hand, there are many 31k (VGA) resolution boards, maybe most famously Sega Naomi.
Later arcade cabinets like the Sega Blast City, New Net City, Taito Egret 3, and Konami Windy II are "Tri-sync" cabinets featuring later monitors that support 31k in addition to the other modes. So versatile, yet so fragile. These are all cabinets without great reliability track records when compared to the sturdy Sega Astro City.
Well, the goal here is to introduce that versatility to the good ol' MS9.
The thought process here was "31Khz isn't that much higher than 24KHz!" and a look at the datasheet for the PLL IC (LA7853) shows that it is designed to work across a broad range of scan rates, well above 31Khz even. Immediately the idea comes to mind that the 24Khz mode might just be tweaked a little to go up to 31Khz, and we pray that the deflection and requisite voltages can cope.
I will spare the details of the discovery process except where relevant. The short of it is that a resistor is changed to offset the range of the 24k hold pot. Caveats will be described in-line.
First things first, make sure you have done the G1 modification above, and have also set your B+ voltages correctly.
Now, the star of the show is this resistor, R563. It is a 22k ohm resistor.
The horizontal hold frequency is defined by a resistance to ground for one pin on the PLL IC. These pots set that resistance.
If you look at the circuit, you can see that one side of the HH24 potentiometer goes to a transistor. This transistor is only active when the hi-res mode is enabled, and it grounds that side of the pot. Otherwise, it leave it effectively as an open circuit. The other side is the 22k resistor to the signal we are interested in.
Thus, when in 15k mode, the HH15 pot does all the work setting the resistance. When in 24k mode, the HH24 pot kicks in, and provides an additional path to ground (minimum 22k). From this we can discern two points:
As an aside, when doing the first test, it was found that the monitor can actually be tuned to ~29Khz from the factory using these pots. So, 31k is really close!
So, R563 is swapped out with a value of lower resistance - 15k is a good value, but it's really quite flexible as the range of the HH24 pot is pretty good.
Simply changing that resistor is the bulk of the work. So long as your voltages were set up correctly, and your colors calibrated to get the greatest range... voila.
This mod works well also with chassis that automatically switch between low and high resolutions. Here is a video of it being done with Butasan on X68000 (which does 31k and 15k modes).
Now, for the Caveats and Mitigations:
This modification is not yet perfect. The following would be good to do.
Two capacitors form the S-correction caps for the E/W circuit on the MS9.
C519 (0.82uF) here is always connected - it is the base, minimum capacitance for S-correction. When the N/W jumper is set to "W" this is the only cap that is active.
This other cap, C531, (0.33uF) is only active in 15Khz mode, and when the N/W jumper is set to "N". Together, these caps form a total capacitance of 1.15uF.
To improve S-correction for 31Khz operation, but leave 15K unaffected, we have to adjust the proportion between these two capacitors. I pulled the caps:
..and got a small variety set of caps. A 0.68uF and 0.47uF capacitor allowed me to substitute two caps so that the base capacitance is lower, but the total (for 15k) remains the same.
All in all, I don't have a picture to share here because it was not a significant change. 15k was indeed unmodified. At 31k, things are a bit better, but at the very extreme edges the image curls, and the total width is reduced.
By changing these caps, we are only looking at a first-order improvement. Some slightly more complex modification will be needed to do higher order corrections to adjust for the edges in particular.
The reason the image is darker in this mode is because the supply for the amp on the neck board comes from the flyback.
Nominally, the neck board is supplied with 180V DC. In 31k, with the raised B+, it was something more like 140V. As a result, the signal "clips" somewhat early, making the image streak to the right when it is saturated.
We don't want to crank B+ any more than needed, so a better solution would be to make a little isolated power supply whose only job is to provide a dedicated 180V supply to the neck board. This can be a rather small PCB, and it will not need to supply a lot of current. The only thing stopping this is actually getting around to it; this is low-hanging fruit and will fix the main issue.
While the auto-switching chassis is convenient, many people do not have it. Fortunately, reproducing it is simple. It is not necessary to strap on a big relay to switch all four connections as seen on some aftermarket hacks. Here is what a resolution switch entails...
Now, there is an unused signal on the remote board - the "S.S." switch, which is a bandwidth selection holdover from the MS9. If that switch is populated, we have a perfectly usable 0 / 12V selector on that line. That takes care of the "D" signal that indicates the voltage. If Q951 is populated, then that handles B+ switching. All that is left then is to depopulate the selection jumpers, and install a relay in their place. Then, that gives you a remote-board-enable resolution switch. This is theorized, but untested.