Performance Testing of Computer and Video Monitors
Copyright (c) 1994, 1995, 1996, 1997, 1998
All Rights Reserved
Reproduction of this document in whole or in part is permitted if both of the following conditions are satisfied:
This document provides a guide to the testing of computer and video monitors for functional characteristics like color purity, convergence, geometry, focus, resolution, Moire, switching between scan rates (where applicable), and acoustic noise. WARNING: No monitor is perfect. Running these tests on your monitor or one you are considering may make you aware of deficiencies you never realized were even possible. You may never be happy with any monitor for the rest of your life! Note: the intent of these tests is **not** to evaluate or calibrate a monitor for photometric accuracy. Rather they are for functional testing of the monitor's performance. Obviously, the ideal situation is to be able to perform these sorts of tests before purchase. With a small customer oriented store, this may be possible. However, the best that can be done when ordering by mail is to examine a similar model in a store for gross characteristics and then do a thorough test when your monitor arrives.
(From: Will Nott (email@example.com)). As with every piece of machinery built in mass production, monitors have performance specifications which define tolerances on performance, as well as operational limits. This has always been the case for monitors, and, in fact, front of screen performance specifications have not changed appreciably over the last ten years. Other items, such as for video & scan rate capabilities have changed, but those items do not appreciably affect the front of screen issues which most users find disappointing. Most users never see the detailed specifications which the manufacturer works to, but may see a sub-set intended to "feature" the product's strong and/or competitive points. There are some variables which may be compensated for with more sophisticated circuits, but that usually results in higher product cost (*not* necessarily sales price; see below). There are other variables which are subject to normal state of the art production tolerances, and because of this, some units will actually come off the line looking better than others, even though they all meet spec. For an example, misconvergence is a variable subject to production tolerances, and practically all monitors have similar spec's, even though most people would not like to settle for a unit which is at the limit of the spec. The typical spec. is 0.3 mm in the area within a circle equal to the height of the image, and 0.4 mm outside that area. That is actually more than the theoretical pixel size on a 14" monitor, and hardly anyone will want to use a display with that performance. However, the spec. allows such variations to exist. An example of a feature which can be affected by additional circuit sophistication is Geometry, or line straightness - each manufacturer has specifications to meet, but here some units may allow as much as 6 mm of non-straightness, where other units may have the ability of being adjusted to a straightness of less than 1mm. So, what the user runs into are variations of two kinds: 1. Normal manufacturing tolerances explain differences that may be expected among various examples of the same brand & model monitor. 2. Differences in design (circuits) which provide the possibility of adjusting out other kinds of distortion explain differences between different brands and models. Of course, this probably has very little correlation with the price at which the product is being offered, because of other competing factors in the marketing game, so it's not really fair (although we all are likely to do it) to expect a more expensive product to perform better - maybe that company is doing better at "charging what the traffic will bear". That's why it's always advisable (within practical limits) to try and understand the spec's, and to try and examine the actual unit which you will be taking home, not a showroom example.
The following are among those characteristics of a monitor that should be evaluated: * Screen size and general appearance. * Brightness and uniformity, purity and color saturation. * Stability. * Convergence. * Edge geometry. * Linearity. * Tilt. * Size and position control range. * Ghosting or trailing streaks. * Sharpness. * Moire. * Scan rate switching. * Acoustic noise - buzzing and whining. For monochrome monitors, use the appropriate subset of these tests. The descriptions below usually assume an auto-scan SVGA type of monitor. Modify accordingly for fixed scan computer monitors and studio video monitors. Note: we use the term 'auto-scan' to describe a monitor which accepts a wide (and possibly continuous) range of scan rates. Usually, this refers mostly to the horizontal frequency as the vertical refresh rate is quite flexible on many monitors of all types. Fixed scan or fixed frequency monitors are designed to work with a single scan rate (though a 5% or so variation may actually be accepted). Multi-scan monitors sync at two or more distinct scan rates. While not very common anymore, multi-scan monitors may still be found in some specific applications. CAUTION: since there is no risk-free way of evaluating the actual scan rate limits of a monitor, this is not an objective of these tests. It is assumed that the specifications of both the video source/card and the monitor are known and that supported scan rates are not exceeded. Some monitors will operate perfectly happily at well beyond the specified range or will shut down without damage. Others will simply blow up instantly and require expensive repairs. Note: throughout this document, we use the term 'raster' to refer to the entire extent of the scanned portion of the screen and the terms 'picture', 'image'. or 'display', to refer to the actual presentation content.
Three kinds of test patterns will be needed: 1. Solid, saturated primary colors (R,G,B) and combinations of these (Y,C,M,W). These will be used for brightness and color purity tests. Total black will also be required to set background level and evaluate black level retention. 2. White cross hatch, fine dot, alternating vertical and horizontal line patterns and bounding box outline. These will be used for convergence, geometry, size, and position, focus, and Moire tests. 3. High quality graphic or photographic image for general color appearance and overall aesthetic tests. If no suitable material is available, a Windows desktop with a vibrant color scheme (computer monitors) or an outdoor scene on a sunny day (studio video monitors) may be used. For computer monitors, software programs are available to conveniently generate the required test patterns. However, an application like Windows Paint in conjunction with a picture with vibrant colors can be used to create anything that is needed. It just will not be quite as easy to switch between patterns. Video cards like those from ATI come with a software Install program which provides a bounding box pattern and at all (PC DOS/Windows) resolutions and scan rates. Test pattern generators for TV monitors are readily available and relatively inexpensive. However, a camcorder viewing appropriate printed material or a prerecorded tape can be used in a pinch.
Before starting the series of tests, allow the monitor to warm up for at least 1/2 hour and make sure it is completely degaussed (see thedocument: "Notes on the Troubleshooting and Repair of Computer and Video Monitors" for details on degaussing techniques. Power the monitor up in the physical orientation you will be using it (front-back facing North-South if you do not know how your setup will be arranged) and don't rotate it on its tilt swivel base after degaussing and for the duration of these tests. Make sure the monitor is not near any sources of electromagnetic interference (i.e., other powered monitors, loudspeakers, motors, transformers, etc.). Subdued lighting is best. Use only the original video cable that came with the monitor or high quality BNC cables as appropriate. If you have a choice, opt for a BNC cable - the quality at higher scan rates will be noticeably better. Do not use any extension cables, any video switch boxes, or excessively long BNC cables. Obviously, if you can do any of this in a store at all, then you may not quite have the luxury to fully control your environment! Note that for an auto-scan monitor, all tests should ideally be performed at several points including the extremes upper and lower limits of each scan range. Most aspects of a auto-scan monitor's performance are affected by scan rate. Unless otherwise noted, all controls are those available to the user. For each test, adjust the size and position controls so that the raster fills as much of the screen as possible or as appropriate for the particular test.
Screen dimensions are normally measured diagonally - corner to corner. So you thought you were buying a 17 inch monitor, right? Wrong. A monitor that is advertised as 17" (or 15" or 21") will rarely give you anything near that viewable size. The specification is of the CRT - including what is covered up by the front bezel and not considering the actual maximum size picture that may be possible in all scan modes. Easily, 1.5 to 2 inches of your valuable diagonal screen real estate may be lost to marketing hype. As a result, you might find as much as a 20% difference in usable screen area between monitors which claim to be the same size. This is like buying a 17" monitor and getting one that is only 15"! At certain scan rates, it may not be possible to get a picture corner to corner so even more usable area will be lost. Check out the section: "Size and position control range". Some advertisements now include actual viewable screen size. Hopefully, this will become a universal practice but I will not hold my breath. Other aspects of the monitor to check out: * Screen curvature - some are curved, others are nearly flat, Trinitrons are cylindrical. For the most part, this is strictly a matter of preference. However, if you intend to photograph (still or video) off the screen, a flatter screen (all other factors being equal like geometric accuracy) will result in less distortion. * Screen surface finish - this may be highly polished resulting in annoying reflections or dull (flat) which if excessively graining will decrease the crispness of the display. A highly polished surface with an anti- reflective coating (see below) is probably best. * Antireflective and/or black (smoked) glass - usually, a CRT with a dark appearance will produce a higher contrast picture but possibly at the expense of overall brightness. Antireflective coatings like on camera lenses help also but are easily smudged and may be damaged by excessive cleaning. * Convenience of the user controls. Digital controls are nice in that the settings for each resolution and scan rate are stored in memory. However, knobs are much easier to adjust in many cases. I prefer knobs for at least brightness and contrast. * The aesthetics of the case. While this is usually unimportant from a performance point of view, it could be of great significance to your happiness. However, I know of some incredibly ugly monitors with great performance!
Display a black - totally blank screen (but don't just pull the video cable as the monitor will be running with the normal voltages and signals) so that the raster is just barely visible. This adjustment should be possible but as noted below, not all monitors have totally independent brightness and contrast controls - you may need to bring up contrast a bit also. The raster should be of uniform intensity and neutral gray. If it is not of uniform intensity or has hum bars - full width areas of varying brightness - or full height vertical rippled areas, the monitor's power supplies may be defective or of poor design. The only way to be sure is to compare several samples of the same model. If the raster is not a neutral gray but has a tint, the color balance may need to be adjusted. See the section: "Brightness and screen uniformity, purity and color saturation". Make sure you can actually set the brightness for total black (turn out the lights and check). If not, you will never have truly dark shadow areas in your display. Blacks will never be black and the display will always look a washed out. This may be adjustable internally. However, inability or difficulty in adjusting for a totally or nearly black raster and vibrant bright display - or if the background brightness shifts as the picture content changes - may indicate a deficiency or defect in the black level retention circuitry. Since video signals are usually AC coupled, a clamping circuit is needed to set the black reference. If this does not work correctly or is poorly designed, the black level may shift depending on the picture content. Modern monitors should be nearly perfect in this regard. This problem is still seen in some cheap TVs, however. Now set brightness to make the raster just disappear. Set the contrast control all the way up. Display a pure red (R) full intensity raster. The entire screen should be a pure, fairly uniform fully saturated red. There should not be any areas that are not pure red. Repeat with the other two primaries - green (G) and blue (B). Follow this up with tests of pairs of primaries resulting in yellow (R+G), cyan (G+B). and magenta (R+B). Again, these should result in pure vibrant colors. Finally, display a pure white full intensity screen. The raster should be pure white. There should be no patches of discoloration. It may be a warm white (somewhat redish) or a cool white (somewhat bluish) but not objectionably so. Some monitors permit this color 'temperature' to be adjusted by the user (e.g., NEC AccuColor models like the 4FG or 5FG). If the screen is noticeably colored and there are no user adjustments, then the internal video gain controls need attention - see the section: "Brightness and screen uniformity, purity and color saturation". Note that the term 'vibrant' here is a subjective term but relates to the boldness, saturation, and, well, zappiness! However, this is affected both by the choice of spectral output of the primary color phosphors and to your perception. For a given CRT, the phosphors set the spectral characteristics of the display. Expensive studio monitors can be ordered with a specific CRT to suit the needs of the video standard. This is rarely an option with computer monitors. You probably do not have control over your visual perception (but do take off those autocompensating sun glasses!) In other words, your mileage may vary. If any of these tests detect a problem, the color purity may need to be adjusted. (By you or by a professional as appropriate). A brand new monitor should not have purity problems unless it is near a source of magnetic fields like a loudspeaker or MRI scanner. The pure white screen can also be used to evaluate the brightness uniformity of the CRT. Don't expect perfection. The actual quantitative measured brightness may vary quite a bit even with a high quality monitor. The corners and edges may be noticeably darker than the center even on the low deflection angle CRTs used in high performance monitors. With the high deflection angle CRTs in TVs and cheap monitors, this may be even worse. However, local significantly darker or brighter areas could indicate defects in the CRT dot/slot mask or aperture grille or phosphor screen which should not be tolerated. Should you see color or uniformity problems at full intensity, try turning down the contrast control. If the uniformity improves after a few seconds, the shadowmask in the CRT may be heating and deforming. This is not unusual with color monitors. One of the advantages of an InVar shadowmask is that it is less prone to these problems, however. Also note if there is a brightness limiter circuit that is kicking in with the full white screen. As you turn up the contrast, is there a point where further increase has no effect on the intensity of the display or where the intensity actually decreases? These brightness limiters are designed to maintain the beam current at supposedly safe levels to minimize X-ray generation and/or to mimize the shadowmask heating effects. Is the maximum brightness adequate for you needs and viewing enjoyment? Using the full white screen, adjust the contrast and brightness controls through their full range. The size of the raster should not change noticeably. A significant change in size - more than 1 or 2 percent - would indicate poor power supply regulation. Examine the entire screen closely for blemishes both in the surface finish and for dead spots. If you see any dead or darker spots, confirm that these are not due to the video source: if they are CRT defects, they will not move as you adjust the position controls. There is a specification for the number and size of acceptable CRT blemishes so you may have to whine a bit to convince the vendor to provide a replacement monitor under warranty. Before the purchase is the time to find these.
Display a picture having a complete range of colors and intensities. At refresh rates beyond 70-75 Hz, even a very bright display should appear rock solid. Turn off any fluorescent lights (whose possible flicker at twice the power line frequency can confuse the test) and examine the screen closely. There should be no shimmering, wiggling, jittering, or dark or light flashes. Any of these would indicate either (1) external electromagnetic interference or (2) a poorly designed or defective power supply in the monitor. It is also possible, though less likely, that the incoming AC power is noisy but modern monitors generally do a decent job of filtering the power lines to eliminate most of this.
Ideally, all three electron beams in a color CRT should be precisely coincident at every point on the screen. While this is never quite achieved, the degree of convergence even at the corners is generally quite impressive - less than .5 mm for many moderately priced monitors. Display a white crosshatch pattern with boxes about 1/2 inch square. IF you do not have this, use a white-on-black graphics or text screen with a lot of fine detail - small fonts or intricate patterns. The lines should appear white without excessive color fringing. The individual primary colors should not be visible at a normal viewing distance. First, examine the center. This should be nearly perfect. If your monitor has any user adjustable convergence controls, set these for best center convergence. Convergence will be worst in the corners but even there, it should not be objectionable. A serious convergence problem in the center of the screen is definitely an indication of a defective monitor or one that needs internal adjustments. Slightly poorer convergence at the corners may be within specifications. A new monitor with significant convergence problems should be rejected.
Display a bounding box image - one that extends to the very edge of the raster on all sides. Adjust any user pincushion controls (amplitude and phase) for minimum distortion along the vertical edges. Amplitude moves the sides in and out. Phase sets where, vertically, this effect takes place. If there are any other user controls that affect raster shape, optimize these for a perfectly rectangular display. Now, examine all edges for curves, wiggles, dips, keystoning, or trapezoidal deviations from a perfect rectangle. These are all considered defects in the geometry of the raster. These will likely be more pronounced at high scan rates - near the limits of the specifications for the monitor. In particular, you may see a wiggle or wave on the left and right edges near the top of the screen which will be come more pronounced as you approach the highest scan rate (this is a deflection problem, however, not strictly a CRT geometry problem). During manufacture, various magnets are strategically glued around the CRT or carefully positioned on rotating swivels on the deflection yoke frame or elsewhere. You need to decide if any remaining errors in geometry are acceptable or not because improving upon these settings is not something that is easy or fun to do - by you or a professional! As noted, if the geometry becomes noticeably inferior at high scan rates, this indicates a problem in the deflection circuitry - adjustments will probably not help. Consider another monitor if you intend to run at these rates.
Display a crosshatch pattern of roughly 1/2" spaced lines. Take a tape measure and compare the exact spacing of vertical lines in the left, middle, and right areas of the screen. Do the same for the horizontal lines in the top, middle, and bottom of the screen. Modern monitors should have very little variation - probably undetectable using a tape measure. Linearity may or may not be adjustable.
Inspect the bounding box for tilt - is it perfectly aligned with the cosmetic bezel of the CRT? If the monitor has a tilt control, see if it will compensate. Few do. The only way to correct tilt on monitors without a tilt adjustment is to rotate the deflection yoke or entire CRT - not recommended. If the degree of tilt bothers you in the slightest and you are a perfectionist, reject the monitor or insist that the tilt be corrected - and be present if possible to make sure that the adjustment is done to your satisfaction.
Determine if the size and position controls have enough of an range to fill the screen totally (for computer monitors) or (possibly in conjunction with an underscan switch) allow for a suitable reduced raster size (studio video monitors). For computer monitors, these tests may need to be done in conjunction with the video card you will be using and the software setup program for that video card. Ideally, all size and position adjustments can be done in software with the monitor's controls left at their center (default) setting. However, this is not always the case. Some people want their computer monitors to extend to or past the edges of the CRT. Many monitors may leave a large border around the picture particularly at higher scan rates. Test at the resolutions and scan rates you expect to use. Obviously, a monitor that will not fill the screen is shortchanging you in terms of how much screen size you purchased! Often, slightly reducing the scan rate at a given resolution will allow for a larger picture. This is one alternative if the flicker is not objectionable.
Display a picture with a large number of high contrast vertical edges - a Windows desktop with many open folders, for example. Vertical edges should be crisp and clear. Examine these for smearing, ghosting, or trailing darker or lighter lines. Any deficiencies will be most evident at high scan rates since these require the most bandwidth from the video card, cables, and monitor. Also, any ringing, undershoot, or overshoot, will extend for a longer space following the edge. Without substituting video cards, cables, and monitors, it is not usually possible to determine which is the limiting factor. The most common cause of these types of problems are inferior, defective, or excessively long cables; use of cable extensions or video switch boxes, or improper termination if there are termination options on the monitor. Full brightness vertical edges should not smear or bloom to the right - possibly with a color change. This is due to the internal video gain controls being set too high and may be correctable but possibly with a reduction in maximum brightness. There should also be no trailing lines to the right of long bright or dark horizontal areas. Similarly, the edges of the raster should not bulge out where the picture is very bright. These types of problems would indicate problems with the power supplies or just poor design.
Display a white screen at the highest resolution and scan rate your system is capable of (or the highest you ever anticipate using). See if you are able to make out the individual scan lines. Turn down the brightness - this will decrease the effective spot size and make the scan lines more visible. Display a fine dot pattern. The individual dots should be tiny, crisp, and fairly symmetric. If the spot size changes drastically with brightness, focus may need to be adjusted or the monitor's power supplies or CRT may be mediocre or defective. Note that it is not always best to have super sharp focus as long as the spot size is small enough. A slightly defocused spot will result in a smoother display and less likelihood of Moire effects.
Moire is causing by interference - beating - between the picture or raster and the phosphor dots or lines that make up the display. Technically, it is an aliasing artifact due to the relative sampling rates of these two structures. There are several causes of Moire. The following will address two of these: scan line Moire and pixel Moire. Try these tests with any 'Moire reducing modes' both off and on. However, the use of such 'features' may reduce the quality of the display in other ways like reducing sharpness or stability. Display a solid white screen at mid brightness. Look for patterns that look similar to contour lines on topographic maps. Adjust the vertical size and position controls to see if these move around or change their severity and spacing. Repeat with a display of alternating black and white horizontal lines. Now, display a pattern consisting of alternating black and white vertical lines at the maximum possibly frequency (alternating dark and light pixels for a computer display - make sure your software is not doing any dithering). Look for serious contour lines in this display. Adjust the horizontal size and position controls to see how these affect any Moire. Try these tests at multiple resolutions including the highest you will ever use. However, the highest may not necessarily be the worst with respect to Moire. Other than using any 'Moire reducing mode' provided by the monitor, there may not be anything you can do to reduce the severity of Moire other than running at resolutions which do not exhibit a serious problem. Nonetheless, I think it is ironic that some people will end up returning otherwise superb monitors because of moire - when in many cases this is an indication of most excellent focus - something many people strive for! You can always get rid of it - the converse is not necessarily true!
When running Windows or DOS with a auto-scan monitor, switching scan rates may be done quite frequently. Some monitors take longer than others to perform this switchover. There may be clicks (due to relays) and other sounds. Better monitors will blank the video until the new scan rate has stabilized. On cheaper monitors, you may see the image as it locks in. Some monitors are very quick. Others can take several seconds - an eternity if you are doing this frequently. Try switching between scan rates at the limits of each scan range as this will be the toughest situation. While it is hard to pin down what makes for a suitable outcome of this test (unless there is an actual failure to properly sync), a monitor that appears to be struggling or which doesn't always make it may be trying to tell you something. Note: there have been some monitors that blow up - fail completely and require expensive repairs - simply as a result of the video card initialization at boot time due to its power on self test when the video signal may be unspecificed and driving the monitor at an invalid scan rate. However, this is not likely to be a problem with any modern auto-scan monitor.
Ideally, a monitor is seen and not heard. However, there are a variety of components inside that can vibrate and this may be quite annoying in a quiet room or late at night. A buzz may originate from the switching power supply or vertical deflection components. A high pitched whine, squeal, or twittering may originate from the switching power supply, flyback (LOPT) transformer, or horizontal deflection components. Some people aren't bothered by these sounds at all or cannot hear them. Others will be driven stark raving bonkers. Listen carefully through the grille in the cover for any indication - even momentary - of annoying sounds. Try all scan rates - very often various resonances will only occur at particular horizontal or vertical scan frequencies. Even with high quality monitors, these problems sometimes occur erratically and no quick test will identify such faults. Obviously, doing this in a quiet location is best. Note that any decreasing hum or buzz that may be heard at power-on is due to the internal degaussing coil and is usually normal and unavoidable.
If after these tests, you determine that your monitor or the one you are considering is perfect - let me know as this is extremely unlikely. More likely is that you found a number of deficiencies. If this is a monitor you are considering purchasing, you need to decide if the benefits outweight the defects. For certain problems like color balance, the vendor may work with you to tweak the needed internal controls. Alternatively, using the information contained in the document: "Notes on the Troubleshooting and Repair of Computer and Video Monitors", you may decide that you will be able to take care of the problem yourself. However, for problems like severe misconvergence, uncorrectable pincushioning, serious Moire, or audible buzz, there may be no easy solution and searching for another monitor may be the only option. If you inherited the monitor or are getting a really good deal, then many of these problems can probably be dealt with but with some risk that significant improvement may not be possible.
So you have this great deal on a used TV or monitor. How can you tell if the picture tube is about to die on you? (From: Andy Cuffe (firstname.lastname@example.org)). The best way to tell is to look at the picture quality. There is no way to tell the exact number of hours. Also, the life of CRTs varies quite a bit. some will go down hill much faster than others. * It should be sharply focused over the entire screen and all 3 colors should be equally sharp. * Set the picture brightness and color to maximum. If you see any bleeding or smearing to the right of bright objects don't buy it. * When you first turn it on the picture should look normal in well under a minute. If it is dim, tinted, or blurry for more than a minute or two the CRT is getting weak. * A B/W picture should not be tinted. * The picture should have decent brightness with the picture at about mid range. Apart form that, if the overall picture is good the CRT is fine. CRTs usually fail very slowly. Even if it's starting to show it's age it probably has several years left. (Portiongs from: Jerry G. (email@example.com)). You cannot tell the hours used by just looking or even measuring a tube. A tube can go at any time. There are no hour counters! Turn on the unit and see if there is any unusual bleeding of the image in the picture at high contrast levels. When turning the brightness up and down, the color temperature should not change, only the brightness. When turning the contrast up and down, the focus at the center should also be very stable. It may change only a little bit. When turning on the set, the color temperature should be stable within about 3 to 5 minutes. Look at the colors in the corners to see if the purity is good. Bad purity can be attributed to a miss-adjusted yoke assembly, to a bad shadow mask. To know the manufacture date of the unit, it us usually on the back with the model and serial number. Most TV sets are on about 5 to 8 hours a day if it is a family TV. If it is a bedroom TV the hours may be 1/2 that amount. Monitors may be on 24 hours a day - or much less. A good way to know if the emission of the CRT is up to specs is to get a CRT analyzer and measure the gun emission. Some service centers own one.
There are a variety of PC compatible software programs for testing of SVGA computer monitors. These display various test patterns and color charts which are appropriate for the procedures discussed in this document.
Here are a few pointers:
"Screen Test" is a program for those who repair monitors. It has color bar, gray scale, purity, dot, crosshatch, and linearity patterns. It is available for downloading from Rosewood Software at:
(From: Mark E. Nikl (firstname.lastname@example.org).)
In the download section of the Web site, there is a file called monitors. It will give you all the test patterns and setups for gray scales, HV regulation, tell you about you video card and much more. I just ran across it the other day. You can even set up the pincushion and lots more.
A demo version with a few test patterns, more information on their products, and some video tech tips, and some test patters are available at:
A subset of these test patterns is available courtesy of PC Magazine at: http://www8.zdnet.com/pcmag/features/monitors/test/. I don't know how long this will be available but for now, it is really easy to use and doesn't require that you even exit Windows. You may use it on 640 x 480, 800 x 600, 1024 x 768 screen resolution or windows of these size. Test patterns are provided for color rendition, horizontal resolution, dot/cross hatch, and more. The images files may be downloaded and saved as well.