Measurement of modulation transfer function for cathode ray tubes

J .R. B A N B U R Y , F.B. W H I T F I E L D

The measurement of modulation transfer function for cathode ray tube displays is discussed. The problems are considered in relation to the non-linear characteristics of arts, and the consequent importance of making tests under conditions representative of actual applications. A versatile test method is proposed which can be applied to complete tv displays without modification, as well as to separate cathode ray tubes.

Avoidance of the fixed-pattern noise contribution introduced by phosphor grain structure allows measurements with low input modulation. An electronic window reduces the art duty cycle permitting low modulation tests to be carried out over the full dynamic range of the display, without affecting the supplies, or operating high performance arts beyond their m~an cathode current densi~ limits. Electronic control makes this method applicable to automatic testin 0 by the addition of a small digital computer.

The sensitivity and spatial resolution of the equipment are given. Results are presented to demonstrate the wide range of mtf curves obtained from different errs, and the need for normalization to a sufficiently low spatial frequency.

In recent years considerable effort has been devoted to the analysis of imaging systems, particularly for military applications. The impetus has come from progressive development of electro-optical sensors for use at visible and infra-red wavelengths. The continual search for improve- ments in performance has been accompanied by a growing interest in theoretical modelling of complete systems, so that the capabilities of the equipment can be predicted as accurately as possible, and optimized for a given set of" operational conditions. A realistic quality measure is required which can be traced through the whole system and related to practical performance. Modulation transfer function - mtf - has been increasingly used as a basis for this description, either directly or through derived para- meters such as m t f a , log m t f a 1'2 , which provide a measure of the maximum information content in terms of the area under the mtf curve.

Modulation transfer function can be defined as the output signal obtained for a constant amplitude sinusoidal input variation, plotted as a function of input frequency and normalized to zero frequency. Methods of measurement are well established for linear systems such as a group of cascaded optical lenses 3,4 , but there are additional difficulties when non-linear components are included.

The most common non-linear electronic display device is the cathode ray tube. The purpose of this paper is to discuss the problems associated with accurate measurement of crt performance in mtf terms.

C A T H O D E R A Y T U B E C H A R A C T E R I S T I C S

Before considering suitable measurement methods, it is appropriate to examine those aspects of crt character- istics which have a direct influence on the problems of

The authors are at the Royal Aircraft Establishment, Farnborough, Hampshire, UK

accurate mtf measurement. The most important points to be noted are as follows

1 Screen current is not a linear function of grid cathode input voltage. The result is distortion of the input sine wave.

2 The electron trajectories vary with gun drive, resulting in defocus of the spot and therefore an mtf dependent on luminance.

3 Some secondary electron emission from electrodes can reach the screen, causing stray light output which may change in amplitude and position, relative to the main focused beam, as the total current is varied.

4 Phosphor light output is a non,linear function of electron current density. The light output profile is therefore different from the spot current profile, especially at high luminance.

5 Internal reflections at the t'aceplate-air interface followed by diffuse reflection at the phosphor surface, and also lateral transmission of light within the phosphor layer, cause veiling glare which is a function of distance across the screen from the illuminated area. The mtf measured at one point will be a function of the illumination nearby, and to some extent also of the illumination over the whole screen.

6 The electron path length is greater when the beam is deflected, especially for a flat faced err. The deflection coils also introduce aberrations, resulting in a defocused spot off-axis which is predominantly astigmatic. The aberrations are a function of the electron beam diameter within the coils. Therefore the shape, as well as the size, of the off-axis beam will be dependent on focus and on luminance.

The significance of the above effects is determined by the crt design and the type of phosphor used for a particular application. The overriding problem is non-linearity,

DISPLAYS. JANUARY 1981 0141-9382/81/040189-10 $02.00 © 1981 IPC Business Press Limited 189

in the transfer characteristic, the spot current density profile and the phosphor efficiency. The basic principle must be to measure performance as far as possible under those conditions which will be used in practical applica- tion of the display. This is especially important for highly stressed avionics tubes operating up to the limits of current technology.

G E N E R A L A P P R O A C H E S TO MTF M E A S U R E M E N T

The modulation transfer function is analogous to t h e frequency response of an amplifier. It can be obtained either directly, by measuring the output resulting from sinusoidal inputs at different frequencies, or indirectly, by deriving a frequency characteristic from the response of the system to an impulse or step in'put.

Both methods are used for optical system evaluation. The otf (optical transfer function) consists of the mtf and the ptf (phase transfer function). For most purposes the mtf gives an adequate description of system performance. For mtf measurements on a linear system, the best method is a matter of convenience and of the accuracy required. Rosenbruch a gives a summary of the main methods employed, including several 'indirect' methods where a Fourier transform is derived from the line spread function.

An accurate direct approach was evolved from work at Imperial College, London and the Scientific Instrument Research Association (Sira), resulting in the EROS range of equipment produced by R. and J. Beck Limited. The method is outlined by Rosenbruch a and described in more detail by Baker 4 . In the usual form of operation, an optical Fourier analyzer is used as a test object generator, effectively producing a square-wave object of variable spatial frequency but constant contrast. The result is made equivalent to that which would be obtained using a sinusoidal test object, by rotating the grating about a displaced axis to produce a temporal variation in the photodetector output, and feeding the signal through a band-pass electrical tilter to reject the harmonics.

It is important to note that line spread function methods, and the ingenious direct metlaod outlined above, depend for their validity on the overriding principle of aqinear system. Thus it does not matter, in the latter case, that an effectively sinusoidal test object is achieved by filtering of harmonics at the detector rather than before imaging by the optical system being measured.

One type of electro-optical system on which mtf measurements have frequently been made is the image intensifier. This consists of an objective lens, one or more intensifier tubes, and an eyepiece presenting a collimated or partially collimated image to the observer. Because it is inconvenient to generate a sine-wave object, systems of this type are usually tested using square-wave object patterns at several discrete spatial frequencies. The equi- valent sine-wave response can be readily derived s , and the measurement procedure is well known ~ . This approach is quite adequate because space-charge effects are negligible, also the photocathode and phosphor have practically linear characteristics at the low current density

levels involved in image intensifier tubes. The system can therefore be regarded as linear, and derivation of the mtf for sinusoidal input is valid.

Since the crt is a non-linear device, measurement of mtf in this case should be approached with care, taking account of the characteristics discussed earlier. It is not possible to think in terms of a single mtf curve; a low contrast input is essential, and several sets of results are required to describe the performance over the full dynamic range and screen area. The line spread function method and two forms of direct measurement using sinusoidal inputs will be discussed briefly in the next section.

MTF M E A S U R E M E N T F O R C A T H O D E RAY TU BES

The line spread func t ion m e t h o d

This method (Fig. 1) may be applied to crts by displaying a single line on the screen, using an auxiliary lens to produce a magnified real image of the line. A narrow slit is traversed across the image plane and the light passing through the slit, collected by a photodetector, is used t o obtain the line intensity profile. A Fourier transform derived from this profile yields in mtf curve for the crt.

The narrow line displayed on the screen can be produced, for a tv display, either by a very fast input pulse applied through the video amplifier, or by bright-up of a single horizontal line from an otherwise blank raster. The first option includes the effects of the video amplifier for a complete display, and the second option tests the tube relatively independent of the video amplifier because of the much lower switching speed required. Whenever electron gun and phosphor non-linearities are significant, a horizontal line obtained by disabling the vertical scan will not yield a reliable curve, whatever the gun drive level chosen.

Fibre I r l Optic I

Lens slit ynC I [ er

Timeboses I L Display under test j -

Computation A(w) = f :e -i~t A(y)dy

Fig. 1 Line spread function method (horizontal line)

190 DISPLAYS. JANUARY 1981

The Fourier transform is usually derived by computation from the line profile but an interesting variation has been used by Bedell 7. In his equipment a direct mtf output from a displayed single line was achieved by using a grating to chop the detected light at constant frequency, and an extra slit rotated relative to the grating to select the component of the output corresponding to a particular spatial frequency. This method is closely related to the principle of the EROS off equipment 4 mentioned in the previous section.

There are several problems which can limit the accuracy of measurements when the line spread function method is applied to non-linear crt displays. First the amplitude of the input pulse must be restricted in order to test the crt over a sufficiently linear portion of the characteristic. It is not sufficient to use a large excursion and then to correct for non-linearity of the transfer characteristic, because the crt spot size changes with varying drive. It is advisable to make measurements with at least two amplitudes to ensure that an effectively linear range has been chosen. This is a necessary but not a sufficient condition for validity of the subsequent Fourier transform, since other forms of non-linearity may also be present in the system.

A second problem arises from the need for accurate normalization. The line spread function for some crts has a significant amplitude, over a distance approaching a hundred nominal spot diameters, caused by secondary emission from electrodes and light reflections within the phosphor and face-plate. As well as the need to scan well beyond the nominal line width, veiling glare in the auxiliary imaging lens can be another contributor to uncertainty over the effective zero position. An incor- rectly drawn zero has a considerable influence on the value of mtfa, the area under the mtf curve frequently used as a measure of display quality.

A third problem which is perhaps less well known concerns the basic assumption that the display behaviour will remain linear with application of a test signal having a quite different frequency distribution from that normally encountered in practical use of the display. Two examples will illustrate the potential difficulty. First, it is not possible to guarantee that every phosphor, when excited with short, high-intensity pulses of varying width and different slew rates, will give the expected light output under all test conditions. Some types of non-linear behaviour will cause assumptions inherent in the Fourier transform operation to break down. Second, for the case of vertical lines in particular (assuming horizontal scan), video amplifier ringing, which changes with

,, frequency, amplitude and slew-rate, can interfere with measurement accuracy. Some displays will give correct results because they meet the necessary criteria sufficiently closely and others will not. For this reason tests should be applied to ensure validity before the Fourier transform method is used.

The moving slit m e t h o d

This alternative approach (Fig. 2) is more frequently used for crts because it is relatively easy to produce the necessary sinusoidal input test pattern for an electronic display. A sine-wave bar pattern is displayed on the crt and a portion of the screen is imaged via a magnifying lens onto an analyzing slit. The light passing through

I multiplier I eht

~rJ:.~ > Fibre I

1 . Display under test _j A

generator | I Low I~.~

Fig. 2 Moving slit method

the slit is detected by a photomultiplier and the slit is driven mechanically, parallel to the screen, to record an output signal proportional to the bar intensity variations. The output signal changes are measured as a function of input frequency.

The slit method of analysis gives a direct result for mtf, but a number of precautions are required to reduce errors which readily appear unless considerable care is taken. Many of these precautions are also relevant to the line spread function method discussed in the preceding sub- section.

1 The slit width, referred back to the crt screen after allowing for lens magnification, should ideally be less than 5% of the bar pitch at the maximum spatial frequency to avoid correction factors.

2 The mtf of the lens, at the aperture used, must be high and preferably known for the maximum spatial frequency.

3 A glare shield is required to prevent reduction of the measured variations by light originating outside the area of the screen being analyzed.

4 Precise alignment is needed to avoid defocus on lateral transit of the slit mechanism. Depth of focus is important for measurements on crts with curved faceplates.

5 A low-pass fdter with sharp cut-off is desirable to minimize noise from the photomultiplier. This applies especially at low luminance levels or when using narrow slits for high spatial resolution.

6 The photomultiplier (and associated preamplifier if present) must become non-linear as a result of the peak light level emitted by the phosphor. The peak- to-mean ratio can be very high for some phosphors, and the pulses must be integrated without departure from linearity.

7 The length-to-width ratio of the slit is limited by practical rotational alignment problems, and to a lesser extent by curvature of field in the magnifying lens.

It will be evident that the larger the modulation of the bars, the greater will be the signal-to-noise ratio, but, because of tube non-linearity, the result will not be

DISPLAYS. JANUARY 1981 191

valid unless a relatively low modulation is employed, so that the output waveform remains approximately sinusoidal.

Important limitations of the method are the need for large mechanical transits of the slit for low frequencies on large crts, precise jitter-free movement for high frequencies on small crts, and the basic fixed pattern noise which limits measurement accuracy with low input modulation,

Schade's method

A method of mtf analysis using a fixed slit and moving sinusoidal bar pattern has been described by Schade s . As before, a magnifying lens is used to image a portion of the screen onto an analyzing slit but, by moving the bars rather than the slit, an important advantage is gained because fixed pattern noise due to phosphor granularity is removed. Schade used a horizontal slit, reducing the raster size until the scan lines were no longer resolved, and drifted the sinusoidal bars in a vertical direction. In his apparatus, demands on the performance of the electronic test equipment were moderate because of the relatively low signal frequencies involved. However, a non-interlaced scan was needed, and the electron-optical conditions were modified because the full raster scan amplitude could not be used with a tube of good resolution.

In the next section a method is proposed which uses the principle of a static slit and moving sinusoidal pattern, but which is suitable for standard tv displays without modification and which allows low modulation tests to be carried out over the full dynamic range of the crt.

M E T H O D General principle It has been stated in the section on crt characteristics that, because of the multiple effects of non-linearities within the crt, accurate mtf figures Can only be obtained when measurements are made under conditions close to those used in practical display operation. A new method will now be discussed in which the moving sinusoidal pattern and fixed analyzing slit used by Schade are applied to mtf measurements in the horizontal direction. This is achieved by the use of more modern control circuits, with an electronic window which allows tests to be carried out over the full range of crt operating conditions.

The experimental arrangement is shown in Fig. 3a. A sine-wave video signal is applied to the display, which is scanned in an interlaced tv raster using timebases controlled by standard synchronizing pulses. The method is suitable for any standard, although in practice it will almost certainly be the European 625/50 or one of the two US standards 525•60 and 875/60. The correct raster size must be set accurately on the crt screen. A small area of the phosphor is focused onto a f'ixed slit through a lens with high mtf over the range of spatial frequencies to be considered, and a glare shield is included to prevent unwanted light from entering the lens. The effect of this glare shield is demonstrated later.

r I /-Glare r~ieltl

Li.e.nc I I ~ I I

, t I ~ o r l ~ I ° '~ '~ " JSinusoiaal - J=,. . . . . .

Ja OsciljO10r 1~ J -

Line sync

Delay

Oscillator 9ate % )

Oscillator output

Oscillator

b

q3

I I

II II R h I I I I

I,e .~

~ Frequency ! Vf Amplitude V o Delay (t 4)

tg = flxecl,lxeset pulse wiclth d Vr l d t controls bar soon speed

Fig. 3 Fixed slit method with electronic scan; a - system block diagram; b - timing diagram.

Fig. 4 Photograph of the apparatus

The current output from the photomultiplier is integrated immediately and applied to a fourth-order Butterworth low-pass filter to give a flat response from dc to approximately 2 Hz (-3 dB at 3.3 Hz). Since most well-known phosphors (with the notable exception of P31) exhibit a virtually constant spectral emission over their usable excitation range, the photomultiplier is not

192 DISPLAYS. JANUARY 1981

Fig. 5 Signal generator outputs viewed on a tv monitor; a - fixed frequency bars scanned by variable delay (Fig. 3); b - as for a but with window restriction applied; c - electronic frequency sweep using a triangular wave at line frequency, d - frequency sweep at line frequency, and amplitude variation at frame frequency

normally fitted with a photometric filter. This would cause an unnecessary decrease in signal-to-noise ratio, particularly when measuring crts which give a predomin- antly blue or red light output.

The apparatus and the display under test are mounted on a small lathe bed to give good mechanidal stability. An additional xy table is fitted to provide convenient adjustment of alignment and focus. Figure 4 is a photo- graph of the practical arrangement.

Signal generator The equipment uses a special signal generator which provides sinusoidal bars of variable frequency and modula- tion. These can be moved at various scan rates across the screen of the display under test, and can be restricted in area by an electronic window. The signal generator was developed from a tv test pattern generator described by Thomas 9 , now~ffi~modular form with extended facilities. Most of the adjustments including frequency, modulation level and bar scan speed can be externally programmed and are therefore suitable for computer control where necessary.

As shown by the block diagram (Fig. 3a), the output of a slow ramp generator controls the pulse width of a mono- stable which is triggered at the start of a line. The mono-

stable output in turn provides a delayed enabling pulse to start the main oscillator. This produces a sinusoidal Output until the enabling pulse is removed after a time tg, just less than one line period. The oscillator is re-triggered at the same point on each line to build up a series of vertical bars. A timing diagram is shown in Fig. 3b. The slow ramp signal is applied to the graph plotter x-axis, and the low-pass f'flter output to they-axis. This method can provide very smooth bar scan across the crt screen (Fig. 5a). Unlike mechanical methods, it is equally convenient for all frequencies and crt sizes.

There are several options available for operation of the signal generator, because its functions are under electronic control. In Fig. 5c, for example, a triangular wave generator alters the bar frequency across the display screen producing a continuous frequency sweep. This pattern could be scanned across the crt instead of fixed

• frequency bars. Simultaneous control over amplitude and frequency is shown in Fig. 5d, in which amplitude is swept by a sawtooth waveform at field frequency, whilst bar frequency is swept by a triangular wave at line frequency.

At present the output from the low-pass f'dter is applied to the plotter-axis to give direct results, and the mtf curves are plotted separately. This is an adequate solution for laboratory work in which a relatively small number of

DISPLAYS. JANUARY 1981 193

displays of different types and sizes are evaluated. However it will be evident that the control method described above can be readily extended to suit repetitive work on a larger scale by connecting the ramp generator, frequency and modulation controls, and the low-pass filter output signal to a small digital computer with modest calculation and storage capabilities. Once the equipment has been set up and focused onto a selected area of the display screen, it would be feasible to carry out mtf measurements at various luminances, averaging the readings for several cycles at each frequency where necessary on noisy displays, and to plot the results automatically. One method of accomplishing this is shown diagrammatically in Fig. 6.

An electronic window is included to allow measurement of low-modulation mtf over the upper part of the crt dynamic range, without exceeding the display specifica-

i I 0 I

,

' | =* '= I I

T,i ~o G.oph ptoS~r

5eS

m

Fig. 6 Extension to automat ic plott ing o f m t f curves (at one specific point on the crt screen)

tions. Maximum large-area luminance is often specified on the assumption that the duty cycle will not exceed 50%. If the percentage of the screen area illuminated were not restricted, the mean cathode current density limit of the crt could be exceeded and the eht supply overloaded. This would damage the cathode, defocus the spot and give rise to errors in the measurements. The window used is a square with side dimension adjusted to 10% of the picture height. This has been chosen to be large enough to ensure that degradation of mtf due to internal reflections and other causes is generally similar to that obtained when the sine wave is displayed over the whole screen. At the same time, the duty cycle and pulsed load demands are small enough to avoid affecting the tube eht supplies. A window size of between 10% and 30% of picture height is normally satisfactory (Fig. 5b).

PERFORMANCE TESTS

Tests have been carried out to establish the limitations of the apparatus and to check repeatability of the measure- ments. The performance was tested directly in order to avoid assumptions about the adequacy of the glare shield and the lenses used between the display screen and the analyzing slit.

A high definition chrome-on-glass target (Fig. 7) was illuminated using diffuse light from a backing screen to fill the imaging lens aperture with an approximately cosinusoidal distribution. Failure to do this would have produced an over-optimistic result not representative of the resolution for diffuse emission from a granular crt phosphor screen. The target was traversed laterally, using a precision mechanism, to provide spatial resolution figures for the equipment. The influence of lens f-number is demonstrated in Fig. 8a, and varying sizes of glare shield aperture in Fig. 8b, for square-wave bars of 40,

I

40p.m bars ,o_ q

0

l

J

2C tLm bars 10p.m bars

Bar width 40/~m 2OF.m ~ Imm for fig 8

Bar width 160/~rn ~ 4Obtm 4ram for ficJ I0

Fig. 7 Photograph o f performance test object

1.0

0

b

Shiek;I aperture

.-.,~--~14 mm ~ 7 mm ' '¢q---35mm

Fujinon lens f2.8 40p, m bars

Fig. 8 Influence of; a - lens aperture (Fujinon lens with 3.5 mm glare shield aperture; b - glare shield aperture

194 DISPLAYS. JANUARY 1981

Table 1. Performance chart

Conditions

Lens Fujinon 25 mm Fujinon 25 mm Pentax 50 mm f-number 2.8 1.4 2.8 Magnification x 12 x 12 x5 Glare shield aperture (mm) 3.5 " 3.5 5 Analyzing slit (actual size,/am) 25 x 1000 25 x 1000 25 x 1000 Low pass filter (-3% response) ~ 2 Hz ~ 2 Hz ~ 2 Hz Photocathode (photomultiplier type 9558) $20 $20 $20

Sensitivity

(broad-band illumination 2854 K nom) At noise level of-+1% nom deviation 100 cdm -2 25 cdm -2 25 cdm -2 (approximately 0.3% rms)

Resolution

(square-bar test pattem) Traverse for 10% to 90% signal change 4 /an Peak to peak amplitude*

40/am bar width (80/an pitch) 97% 20/am bar width 89% 10/am bar width 70%

12/am 24/am

97% 86% 78% 30%

*The Fourier series for a square wave of unity peak amplitude is given by F(t) = -4 ~r

Thus the fundamental has an amplitude ~ = 0.786.

1.0

Q5

0 IO

Green illumination (llford 625 filter centred at 540 nm ) f / I 4

Q5

I I I

f (2854 K nominal)

o I I I I0 20 30

C(mm)

Fig. 9 Tangential mtf of Fujinon lens at x12 magnification

80 and 160/am width. These results are for a Fujinon lens of 25 mm focal length, a magnification of 12 and

40 an anlyzing slit 25/an x 1000/an. Some mtf curves were recorded for the lens at the same conjugates, and are given in Fig. 9.

I sin (n tot), for n odd n=l n

Glare shield / /~aper ture diameter

1.0 - ~ ,~nmi~ " ' - - ~ m m r ' - ' ~ I . ,

o J " ~ 160p.m bar 40pro bar

Fig. 10 Spatial resolution using Pentax lens (for [onger working distance) at f /2.8

Although the Fujinon lens is used wherever possible, it is sometimes necessary to employ a lens with a longer work- ing distance, eg in displays in which the crt has a separate contrast enhancement filter or is recessed behind the front panel. In this situation an Asahi Pentax lens of 50 mm focal length was used, with a magnification of five and the same analyzing slit 25/am x 1000 tma. It will be seen from Fig. 10 that this lens, stopped down to f / 2 . 8 and with a glare shield aperture of 5 mm diameter, is still satisfactory for measurements on crts with spot sizes down to approximately 50 lma. The resolution and sensitivity achieved using these two lenses are summarized in Table 1.

The noise reduction arising from the elimination of freed pattern phosphor noise for a monitor of intermediate size (23 cm tube diagonal) is demonstrated in Fig. 11. A relatively low output modulation signal obtained using a fixed slit and moving sinusoidal waveforrn is compared with the result obtained using the same slit moved laterally to scan a flied sinusoidal pattern. Alongside this result

D I S P L A Y S . J A N U A R Y 1981 1 9 5

a comparison is recorded between fixed slit and moving slit, with a blank raster on the screen. Although the level of fixed pattern noise in any given case is clearly a function of slit length and screen particle size, the example given is fairly representative of a wide range of crt screen sizes and manufacturers. Phosphor noise is rarely less than 5%, varying only slowly with slit length for large slit length-to-width ratios, and it is convenient to be able to use a single slit size with a wide range of display dimen- sions. A cautionary note here concerns displays with bad interlace jitter; they require a slit long enough to ensure that correct account is taken of the mtf degradation from this source•

Two sources of noise are removed by the use of a fixed slit. First there is a light output variation from point to point due to variations in the efficiency of the phosphor particles, and secondly the mtf itself changes slightly because of varying screen thickness.

It will be seen from Table 1 that the basic photomultiplier noise is not normally a practical limitation at the luminance levels of interest for crts. This fundamental limitation has, however, been analysed by Slaymaker 1° , in con- nection with mtf measurements on optical systems. Figure 12 shows the basic noise level from the photo- multiplier with the Fujinon lens at f[ 1.4 and/72.8, a filter bandwidth of 3.3 Hz (-3 dB) and test surface luminance 100 cdm -2 (2854 K illumination). The $20 photocathode will give better results for green phosphors, and rather worse performance for red, when compared with the sensitivities quoted in Table 1.

3OO

'N.~ 200

100

Sy~m oulput wh~ vi~,a~l-~e~l~l

[ E74clrcmic ~ of Medlanicol traverM of Mech~ieAI fro..v~ with Sl~ioe~y ~yz in9 liar paltlrn (400 tv Matioeary ~ r pot~r n zero l ip m~ulaflo~ mjstem ,~-o l ip I limm/l~nn I~1t~ (400 tv Itn~lplctm~ mo~l~ton Ilo'/. Vpmo~kmo.) ,eiSht I0% l iP

2s cm -I

Fig. 11 N o i s e r e d u c t i o n b y u s e of fixed slit and pattern shift shift

100

f ! 2.8

0

f /1.4

Is cm -I

Fig. 12 Photomultiplier noise level (2854 K il lumination) Fujinon lens

DISCUSSION The equipment described has been used to obtain results from a variety of crts, both standard laboratory monitor types and highly stressed avionics tubes•

A selection of curves is shown in Fig. 13 to illustrate variations in measured mtf characteristics, and to empha- size the need for normalization to a low frequency in order

1.0 i ,4, B lob monitors C high-qual i ty display

O.8 C

O6

0.4

0.2

I I I I I I I I 0 200 400 600 800

Iv lines/picture height

Fig. 13 Typical mt f curves

1 I.O

0.8

0.6

E

(34

0.2

0 200 400 600

tv lines/picture height

Fig. 14 MTF curves showing several undulations at low spatial frequencies

800

196 D ISPLAYS. J A N U A R Y 1981

E

1.0

0.8

06

0.4

0.2

Drive level

+ LOW

x High

20% input modulation

0 200 400 600

tv tines/picture height

Fig. 15 MTF characteristics as a func t ion o f luminance

to avoid the possibility of major errors. Curves A and a are typical of many displays, they were taken from two

good quality laboratory monitors of similar resolution, and it is evident that the results would have looked quite different with incorrect normalization. Curve C represents a special purpose display in which particular efforts were made to improve contrast rendition. Occasionally crts show significant rises above the normalization level, occasionally to over 1.1, before dropping away at higher spatial frequencies.

800

Many crts show an undulation in their mtf curves between 2~) and 150 tv lines per picture height. This probably results from a spot profile with small side 'shoulders' which represents a departure from the nominal Gaussian shape. Within the range of crts tested, some undulation of the mtf character|stic has been very common. Where it has occurred, the effect has always changed with luminance, showing that the beam profile as well as the spot diameter is dependent on the gun drive level. Figure 14 was measured from a display showing several undulations at low spatial frequencies. The mtf was recorded and checked at ten-line intervals up to 100 tv lines. The detail was only identified because of the low-noise results produced by this apparatus with stable display electronics.

Figure 15 has been-included to show the effect of drive ~defocusing on mtf. Three low-modulation curves were recorded from the centre of a crt for low, medium and high drive levels. Note the change in shape at low spatial frequencies, as [he drive increases. The electronic window was used as described earlier.

Figure 16 gives the light output recorded from a display of indifferent performance, where the accuracy of mtf measurement would be limited by display electronics

2.5 I0 3 5 5 0 I00 IKX) 300 4 0 0 5 0 0 600 700 800

hi lines/picture height

Fig. 16 Output trace from a noisy display

2.5 tv lktes/pictum height 500 Iv limm/pcture height Z~o I/P modulation 10% I/P modulotio~ 10% I/P n,,odulotio~

Fig. 17 Low noise trace from a high quality display

with poor noise rejection. Some crts give unstable light output because the cathode is in bad condition, and for airborne displays visible 400 Hz interference can be a problem. It is possible to compensate for some instability effects by differential input from a second photomulti- plier, viewing a separated window containing a plain raster. However, the extra complication involved has not been conslder6~l worthwhile because the effects are avoidable, and therefore small, for displays of reasonable quality. A master oscRlator, locked to 50 Hz or 400 Hz as appropriate, is sometimes a useful alternative to crystal control.

By way of contrast, Fig. 17 demonstrates how the equip- ment described in this paper can produce a clean signal output, given a good display. The result shows a noise level lower than 1% on the light output from the monitor. It is usually unnecessary to change the analyzing slit over a wide range of crt sizes. However, shadow-mask colour crts, particularly those with circular phosphor dots, require a larger slit and also need a small transverse shift to optimize output when changing between red, green and blue guns; It will be evident that electronic bar scan is necessary for measurements on all tubes with a dis- continuous phosphor structure.

CONCLUSIONS This paper has discussed a method for accurate measure. ment of crt modulation transfer functions, using a fixed analyzing slit together with electronic control based on a video oscillator gated at tv line frequency. This is a versatile approach which allows measurements with low input modulation, by eliminating the noise contribution introduced by phosphor grain structure. Use of an

DISPLAYS. JANUARY 1981 197

electronic window to control the crt duty cycle permits tests to be made at the top of the dynamic range without overloading the cathode or the eht supply.

The sensitivity and spatial resolution of the equipment have been given under stated test conditions, which are suitable for general purpose measurements on crts with spot sizes from several millimetres to below 20/am. The apparatus has interchangeable lenses to allow for measurement at longer working distance when the crt screen has a bonded filter or is recessed behind the front surface of special purpose display units.

Table 2. CRT characteristics affecting mtf measurements

1 Non-linear transfer characteristic 2 Spot size variation with tube drive 3 Secondary electron emission from electrodes 4 Non-linear phosphor conversion efficiency 5 Internal light reflections 6 Longer off-axis path lengths 7 Deflection coil aberrations m ,

Table 3. Features of proposed measurement method

1 Most accurate method for a non-linear display 2 Electronic control suited to automatic operation 3 Slit size can remain constant for different crt sizes 4 10% input modulation practicable mtf curves 5 Repeatability 0.5 to 1% of mean luminance

Fixed pattern noise removed

1 Luminance variations from granular screen 2 Screen thickness variation giving mtf noise

Noise sources remaining

1 Display electronics 2 Power supplies to display 3 Cathode emission variations 4 Photomultiplier basic noise

Test results are quoted which demonstrate the wide range of mtf curves obtained from practical errs and the impor- tance of normalization to a sufficiently low spatial frequency in order to avoid major errors in some eases.

The suggested method of mtf measurement has been compared with alternative approaches. Characteristics of crts and features of the proposed method are summarized in Tables 2 and 3. A discussion of err non-linearities has highlighted the importance of making tests which are representative of the actual display application. It is hoped that the greater accuracy possible with the method described will make a contribution to better understanding of crt characteristics, and permit more reliable mathe- matical modelling of systems using crt displays.

References

1 H.L. Snyder et al 'Visual search and image quality' AMRL-TR-73.114 (1974)

2 H.L. Task 'An evaluation and comparison of several measures of image quality for television displays' AMRL-TR-79.7 (1979)

3 KJ. Rosenbruch Proc SPIE 46, (1974) 19-26

4 L.R. Baker 'Automatic recording instrument for measuring optical transfer functions' Jap J A p p l P h y s 4, Suppl 1 (1965) 146-152

5 LW. Coltman 'The specification of imaging properties by response to a sine wave input' J Opt Soc Am 44 6 (1954) 468-471

6 J.A. Hall 'Photoelectroni c imaging devices, Volume 2' Edited by L.M. Biberman, S. Nudelman (Plenum Press, 1971) 53-76

7. RJ. Bedell 'Modulation transfer function of very high resolution miniature cathode ray tubes' IEEE Trans ED-22 (1975) 793-6

8 O.H. Schade 'A method of measuring the optical sine-wave spatial spectrum of television image display devices' J Soc Motion Pict Telev Eng 67, (1958) 561-566

9 W.V. Thomas 'A television test pattern generator for display assessment' RAE Tech Mere FS 144 (1977)

10 F.H. Slaymaker 'Noise in mtf measurements' Appl Opt 12 (1973) 2709-15

Copyright © Controller, Her Majesty's Stationery Office, London 1980

198 DISPLAYS. JANUARY 1981