Keohi HDTV



Powered by Keohi Web Design


WS-73905 Calibration

<< Back to Mitsubishi Tips

Tip from Mr Bob:  "This review started a long time ago. Rick and I got together to do his Mit awhile ago, and then a trip happened to Oregon, where I had NO extra time  because of helping both my brothers with time-critical stuff; then I was in a traffic accident that currently has me 8 hours more behind now than I was before the accident, and this coming week I fly both to San Diego and Texas for calibrations.

But this review is important to both of us, seeing as how this is the first calibration I've done for a moderator on one of these forums -- Rick moderates on the AVScience Forum -- so I'm going to give it a shot right now, before things get intense again.

The calibration took even longer than this review, to happen. Rick's TV was not showing a convergence grid, not for love or money. Not in the user menu, nor in the service menu.

The Mit warranty people took his convergence board down to another 73905 at their store -- their service dept. is part of their store -- and it worked perfectly. Even they were stumped, and wanted to take all his TV's innards out, work on them, and bring them back.

This made both Rick and me see red. Why not just replace the bottom section of the TV and be done with it???

Eventually that was done, but it was 4 months before Rick's TV was fully replaced -- FULLY replaced, with a brand new unit -- and we could finally get to his calibration.

When I walked in, Rick's picture was actually very crisp, for an out of the box unit. Except for the edges and a few trace convergence errors in the picture, I was quite impressed. The Wild, Wild West was playing, and I had not seen it before, except for snippets at Costco. I told Rick that it was a very good convergence I was seeing.

Then I said, "For an amateur."

He said, "For an amateur???!!!" We both laughed, but I immediately realized that even tho Rick swears off doing all the fine tuning and tweaking that I do in the field which makes me technically a professional and him technically an amatuer, both of us know that there's nothing amatuer about his systems, his attitude, nor his knowledge. To not be acknowledging that was a faux pax I will not be making again. He has some of the finest equipment available, and the Dwin system upstairs is an official knockout. Robert Busch, true to form, had done an extremely good job on its fine-tweaking setup. The look and feel of 35 mm film is all there, a job to be proud of, and a system to be proud of. Rick has definitely done his homework.

But back to downstairs...

The geometry was off very substantially, around the edges.

This is very typical of out of the box units, be they Mit or Toshiba, or whatever.

In Rick's case, the bottom edge was really bad.

On the AVIA Widescreen Enhanced circlehatch pattern, the bottom circles were both squinched up at the bottom, their linearity was as if you had dropped a peeled hard-boiled egg on a hard surface, and the resounding "thud" had put a permanent flat spot on the bottom of the egg, on both right and left bottom circles, and of course on the bottom edge of the master circle. This sat there and radiated the demand for concentrated attention.

The first thing that needed to happen was that the picture had to be raised at least an inch. But before I DO ANYTHING, I check and correct the focus and condition of the lenses, even on a brand new unit.

The first thing I noticed once I opened it up was that the NEW 73905 had a MYLAR mirror.

I have not seen a mylar mirror on a Mit, EVER. Usually I see them on old RCAs and Magnavoxes. But no, here on a $9,000 unit was a mylar mirror.

Clarence Hermary, owner of Hermary's in San Carlos, CA., uses mylar reflectors in his big, custom folded rear-projected projector installations, and he says that mylar is actually a better reflector than glass. Perhaps that's why Mit is now doing it, and perhaps that's the wave of the future.

Mylar has 2 major advantages I can think of immediately:

  1. You never have to clean the mirror. PERHAPS many years down the line, and if so VERY CAREFULLY, but not one-twentieth as often as glass needs to be cleaned, seeing as how mylar doesn't statically attract dust the way glass does. I have seen 10 year old mylar-equipped RPTVs with impeccably clean surfaces.
  2. Mylar by its very nature is a first surface mirror. No guessing anymore on that one.

Just don't get your hair near it, if you're a wethead, while you're cleaning the rest of the optics; getting the smudge mark off could cost you a few scratches...

The cleanliness factor on Rick's optics was great, since it was a brand new unit, just past the 100 hour burn-in time constant. There was a smudge on the red lens, but it was easily removed, and I then went to my photographer's brush, which I used -- only -- on all 3 lenses, rather than cleaning solution, since it was only very lightly dusty. In a year of course, due to the high voltage inherent in CRT use, the entire light path -- excluding the mylar mirror -- will require cleaning, including under the objective lenses themselves.

his has to be done every year, if you want your unit to stay new looking for 10-15 years. The calibration you only need every few years. The optics cleaning, every year.

I then checked the mechanical focus, and found it dead on, on all 3 lenses. This was impressive, and credit goes to Mit, whose RPTVs, to be frank, don't usually perform quite this well.

This was true on Rick's unit, but on most out of the box units I've worked with, regardless of brand, at least one and usually two of the lenses need redoing mechanically, if whatever focus error you have is to stay innocuous -- ie., balanced and symmetrical at all points on your screen.

The best way is to use the Cantilever approach, which you can find described by me on the Keohi website. This approach guarantees you that if there is any mechanical focus error due to whatever the manufacturer set in stone in production for that particular model of unit, in terms of correct angling of the lenses to accomodate such a slimline design as is used in RPTVs today -- just shooting straight at the screen is NOT all that has to be done, in designing a RPTV's optics -- that it will be averaged out over the entire picture. You can have two people doing a mechanical focus on a RPTV, and it ultimately may look good in the center, but no matter how good you get it in the center, one side may be off while the center and the other side are on, unless you absolutely center the focus point, in terms of throw distance. This means so that the correct-focus image hits dead center at the center of your screen, throw-wise, rather than half and inch before or beyond the screen's centerpoint, measured throw-wise from the mirror's centerpoint. The bottom line here, is PRECISION throw distance, and in a RPTV, this can't just be measured from a chart. It HAS to be done by eye.

If not perfectly centered -- such that the focus is still correct at diameter'd center whether the screen is half and inch out, or half an inch in, from throw center -- if not perfectly centered, the center of the screen will still be in focus on a crosshatch grid, but the outside edges will not be symmetrical with each other, focuswise. Again, this means that the center and the side closest to that lens will be in focus, but the other side will be out of focus. Or vice versa. This error, if there is any, must be averaged out between the 2 sides, the closer and the farther.

When I did the focussing, I made sure that all this was taken into account, and that his mechanical focussing could not possibly be more tight, nor more symmetrical, on all 3 images.

All of this, of course, has to be done BEFORE the convergence and geometry, or all that changes again afterwards, if you have to change the mechanical focus on any particular color.

As they say, timing is everything...

Now back to the positioning of the picture.

I tried to use the VPOS of the deflection jungle commands, but that doesn't work, either on the 903 nor the 905 series. I tried to find it elsewhere in the menus, but the secondary one that worked for the 903 did not do the same for the 905. I even tried calling Paul Carleton in Seattle, creator of, but couldn't find him. Guess why? He was right here in Redwood City that day, just a few miles away from us, we later learned. Thanks anyway, Paul -- CALL when you're around next time! We would always welcome a visit from you.

I approached Rick about actually resetting the centering magnets, and he approved, so I very carefully marked them on the green CRT neck, and reset them to raise the picture. It then occurred to me that there was another "VPOS", in the coarse convergence circuitry, and if it worked for all 3 colors, and had enough play, we could use it. It is called "VSTA" for "vertical stationary positioning".

Bingo. It was there all right, with green having just as much play in it as red and blue.

So I went back to the centering magnets and reset them to exactly where they had been before. As I have said in other posts, a Sharpie pen is great, cheap insurance, and will always take you back to where the factory had you, when you started.

I raised the image on all 3 colors, with each color's VSTA control. Luckily, on a Mit, the red and blue follow wherever the green is set, so all I had to really change to any great degree was the green, which took red and blue with it. Then finetrim the red and blue's VSTA settings, and my pic was exactly where it needed to be on the screen.

I then went to work on the anomalous bottom edge, which was badly contorted still.

It took awhile, but the edge finally cooperated with me, and straightened out the lower circles. It took linearity and a strong dose of 64 point redoing. In the process, I found that the edges on the Mit are not easily predictable, esp. when things are so far off.

The second-to-the-last set of vertical bars of the grid, side to side, help direct the outer edges in their angling, and for the furthest edges, you have to actually go OFF the picture itself to finally get them straight. The cursor is actually invisible at that point. You keep holding the remocon buttons down until you see that farthest edge come back to center, even tho you can't see the cursor at that point. It may take till plus or minus 511 before that happens, on the edge lines at the sides and/or top and bottom. But when the last inch or two of the edges are angled off so much each, right at the edge of the picture, that's what it takes to get it right. That's what I found on Rick's unit, along with the egg-shape anomaly. The grid lines at the edges veered off very sharply before they disappeared. The majority of the picture was OK, but if you are a discrimitating viewer, these edges were really distracting.

After that, you have to go completely to the other side edge of the screen to compensate for what you have now done to its side because of the other side -- or top and bottom -- because the cursor point that's offscreen affects both right and left. It's essentially in the midpoint between the two, because it comes around again, if you just push and hold a right or left button at the remocon's arrows. Just keeps coming around to the left, say, then reappears at the right, and keeps coming around again, etc.

On the top and bottom, a 65905 I just did out in Benicia had final top and bottom edges which required that you go BACKWARDS on the left and right, once cursor's off the screen, to get them right. This is the only time I have found this. All other 903 and 905 series units have been the correct direction at this point on the grid. Don't ask me why...

When I was done, his picture passed the AVIA test, the round circles being round in the Widescreen Enhanced pattern, plus in the Resolution pattern at the end of that chapter. When the circles are all correct - perfectly round or at least nearly so, and the bullets are all perfectly parallel and equidistant with each other from the screen frame at t/b, sides and corners, then and only then do I know I'm done with the geometry and convergence on a 16x9. The resolution pattern then provides a nicely, moderately stressed picture to do some ultra high precision convergence with.

Having to go off-screen is not how it works on the Toshiba's 64 point, BTW.

On the Toshiba, it's much more crucial that you observe how the second-to-the-end lines of the grid control the angling of the outer lines, because with a Toshiba, you can't go off-image to make your corrections at all, like you can with the Mit. As far as this second-to-the-end set of lines business works, itdoes it a little with the Mit, but it does it a lot with the Tosh.

At the end of all this, I checked the grayscale, and to my calibrator's eye, it looked dead on at D6500K. I did not happen to have my color comparator there with me that day, as this cal was actually a spur of the moment thing that day -- I had been looking for a spot on my calendar, and finally it came up, but abruptly, and as it was took well into the night, before I was through -- so I declared it good and not in need of any further resetting, and will eventually take the comparator over there and check it out further sometime, as Rick only lives a couple of miles away from me. But it was VERY well done at the factory, from my point of view, and the colors his TV generates are simply dazzling. They sparkle with realism and high-precision balance.

This is not the first time I have seen highly impressive grayscale out of Mits, nor out of Toshiba. Both have become very serious about achieving D6500K faithfully, right at the factory, over the last few years. This has carried into the field and into the customer's living room right out of the box, far more often that not these days, on these two brands.

I did not need to redo the precision blue defocussing on Rick's unit, either. I don't think I've had to redo any of the Mit 903 nor 905 series units, on this parameter.

Of course, I always recenter the blue electrostatic focus for the convergence section of the job, and re-defocus it back to where it was before, later. This assures the tightest possible convergence on blue, seeing as how typical 64 point service menu grid lines are WAY too thick for the kind of tightness I want to bring to a RPTV's image.

It makes the picture dingy for awhile, but the whites come back to full color linearity when re-defocussed later, back to where it was before.

I then recalibrated the color and tint to the blue filter test, found I didn't need to reset the Brightness/Black Level -- this is also good work on Mits's part -- and I was done.

All in all, I find the 73905 and the 65905 to be some of the hottest images available on the HD market today. The picture that leaps off those screens is simply dazzling, when finely -- and finally -- wrought by your friendly neighborhood calibrator. It's sizzling, velvety, crisp and at the same time full of the incredible feeling of 3 dimensional depth that we should all come to know and expect out of HD. The suspension of disbelief is real, palpable, and for most folks, of 35mm quality. I, unfortunately, can still see even line-doubled scan lines from 10' away -- it tells me when the fidelity of a HDTV is truly maxed out, and performing at its peak of potential. So I guess it's not quite 35 mm quality for me yet.

But it's the best we have at this time in the evolution, except for line quadrupling and a few other features that only the REAL expensive RPTVs would have -- like possibly the $35k Faroudja -- and more than does the job of achieving super-high degrees of suspension of disbelief, even for me. I LOVE to get lost in the images, to get spiritted away, to some far distant land--or galaxy...

This was fun. It turned out well.

And Rick is one of the finest, most hospitable hosts a guest in his home could ever hope for. I will be looking forward to returning, sometime in the future, hopefully, for when it's time to do this all over again."