| Next: Setup | Previous: Features | Contents |
Screen. I'm pleased with this screen. Colors are
quite good even without The only significant problem with the screen is that in unknown
circumstances involving backpacks and carrying bags, the screen touches the
palm rest, and a mark is made on the polarizer. A number of people with
various Inspiron models have mentioned this screen damage. It is caused by
textured edges on the plastic between the touchpad and the small mouse buttons
above it (which I never use). I affixed a piece of soft cloth over the buttons
using folded duct tape (so the adhesive won't get on the screen) and the mark
actually seems to be fading. When the lid is closed and the machine is sitting
on a table, there is enough clearance that a piece of paper can be slid through
that area, so the issue must be extra pressure in the backpack, plus jiggling.
As identified by the X-server, the screen
is a Samsung LTN150P1-L02, 15 inch diagonal. From the screen dimensions
it would appear to be 14.1 inch, corresponding to what the Dell ordering
information says. From the
Samsung datasheet for the products, it has these characteristics:
xcmsdb, and better when
properly calibrated; q.v. for details and DCC files. It is bright
enough (150 cd) to be seen in daylight if the sun does not shine directly on
it. ``Seen'' does not mean that the visual quality is equal to that achieved
in a shaded location. The viewing angle is ±45° in yaw with good color, and a
lot more with adequate but degraded color, so it's quite feasible for several
people to look at the screen. But the pitch (vertical) range is less, +10°-30°.
See below under ``color'' about consequences for the
graphics artist.
Response time is fast: when scrolling text quickly, there may or may not be
slight ghost images, but in several quick-moving Mesa demos, no ``comet
trails'' can be seen.
| Nomenclature | LTN141P1 |
| Nominal Size | 14.1 inch |
| Pixels | 1400x1050 (SXGA+) |
| Contrast (light:dark) | 250:1 |
| Brightness | 150 cd |
| Viewing Angle | 10/30; 45/45 degrees up, down, yaw |
| Interface | LVDS (2ch) |
| Supply Voltage | 3.3 volt |
| Number of Backlights | one |
| Mass | 540 grams |
| Rise-Fall Time | 20 msec, 30 msec (from Dell specs) |
| Power Consumption | 7.0 watt (backlight on, from Dell specs) |
The web site doesn't seem to tell the LCD technology used, but related Samsung products are Super Twisted Nematic, and the illustrations in the product technology guide represent STN. The screen is made of two sheets of glass (liquid crystal between them) with a plastic polarizing film on top. Exercise due care and follow the manufacturer's instructions when cleaning. Otherwise, do not touch the screen.
Dell has conveniently omitted from the printed manual or their web site
the instructions for cleaning the screen. Quoting from Samsung's web site
(here):
It is recommended that the product surface be cleaned if it is dirty
with isopropyl alcohol or hexane. Ketone type material (acetone),
ethyl or methyl chloride must not be used as they can cause damage to the
polarizer.
On the viewing angle, look for Samsung screens starting in 3rd quarter 2002
with an up/down range of 25/45 degrees, or one with 55/55 degrees up/down
and 70 degrees yaw, so they say.
Color. In a LCD display each pixel has three light
valves, and over each one is a red, green or blue filter, with white light
coming in from the back. While I have not done any photometry, the passband of each
filter on the Inspiron 4100's display seems fairly broad, more so than a CRT
would produce, and in particular the blue is equally as bright as red and
green, suggesting a generous passband. Using non-photometric (estimated) color
standards, I have produced ICCCM Device Color Characterization Files (daytime; night)
for the display. The intensity curves indicate a gamma of about 3.0,
compared to 2.3 for a CRT; thus the darker colors are stretched. This effect
is at least partially produced by a lookup table in the graphics subsystem,
because 4 units of RGB are needed to produce a visible brightness step for
the darkest tones, whereas in the mid and light ranges one RGB step produces
one visible step.
It is commonly stated that a ``serious graphic artist'' will be
``disappointed'' with the color rendition of a LCD panel, preferring a CRT. I
am not a professional graphic artist, but to my eyes the colors on this display
look pretty good. Certainly it is fine for photo editing. It helps if good DCC
data is loaded into the X-server, and if the user and the application program
agree on the color space (perceptual or linear light) in which the image is
interpreted.
For graphic arts work, the brightness changes significantly as you pitch
up and down, so when it's important to judge brightness, you should get in a
consistent position such as viewing exactly perpendicular to the screen,
and you should compare test images between your laptop, a CRT monitor, and
printed hardcopy, both for consistent brightness and for color fidelity.
Touchpad. I find the touchpad easy to use for
text work, though for graphics I prefer the force stick. But if your thumb
strays onto the touchpad it counts as a mouse click; you have to train yourself
not to do that. Most of the time I use the firmware's tap-for-mouse mode, for
button-1, for double click and for drag, but it's handy to have the alternative
of the actual buttons. Button-3 (middle) is simuated by the X-server if you
hit both buttons at the same time. I would prefer an actual third button, or a
substitute through the configuration utility, but the simulation works
reasonably well.
There is an adjustment utility program
for the touchpad. The latest
version as this is written is tpconfig-3.1.2, maintained by
Bruce Kall <kall@compass.com>. The original author was C. Scott Ananian,
whose version 0.1.1 is still available. All the fancy modes can be configured
on touchpads so equipped.
However, Synaptics touchpads have evolved in a direction which is
commendable technically but not favorable for the lazy Linux user.
Specifically, pads of hardware version 4.x and above (as on the Inspiron 4100)
do only relatively limited filtering and gesture processing in the firmware,
leaving more complex processing to the (Windows) driver. There is no special
Linux driver, and so Linux gets ``vanilla'' mouse behavior. tpconfig can turn
off the vanilla modes, but most people want them on.
| ``Vanilla'' Modes (on by default): | ||
| Tap to Click | A brief touch on the pad is interpreted as a click on Button-1. | |
| Pad Drag | Touch, up, touch and move is interpreted as holding down Button-1 and moving the mouse. | |
| Coast Mode | After motion, the pad will claim that your finger is still moving even after it stops. I haven't figured out the gesture that engages this mode. One would expect this to be an enhanced mode. | |
| Enhanced Modes (with driver, not for Linux): | ||
| Edge Mode | When your finger hits the edge of the pad, the mouse coasts, continuing the motion up to that point. | |
| Scroll Mode | Motions in the right margin of the pad appear to be wheel rotations of a Microsoft® IntelliMouse®. | |
| Corner Mode | Tap in the corners and get user-selected special effects such as emulation of Button-3 (middle). | |
| Locking Drag | Having done the ``tap and a half'' gesture to begin dragging, you need to tap again to turn it off, so you can lift your finger from the pad for a long motion. | |
| Confinement | The pointer is confined to the focus window unless you positively push it out, improving the accuracy of hitting corner buttons. | |
| Ignore Palm | If the width of the area touched is larger than usual, the touch is assumed to be accidental and is ignored. | |
References for Synaptics Touchpad protocols:
Developers' Support, and within that, particularly see the
"Synaptics TouchPad Interfacing Guide".
Hard Disc.
IBM's product
summary has some details about the hard disc:
| Nomenclature | IBM Travelstar 30 GN series | |
| Model | IC25N020ATDA04-0 | |
| Capacity | 20 Gb | |
| Rotational Speed | 4200 RPM | |
| Seek Time | 12.0 msec | (average) |
| Start-stop cycles | 3.0e5 | Tested to 1.0e6 cycles |
| Startup type | Head load-unload | Versus contact start-stop |
| Bearing type | Fluid dynamic (oil) | Versus noisy ball bearings |
| Head type | Giant magnetoresistance | |
| Magnetic surface | Pixie Dust | Antiferromagnetically coupled media |
| Platter type | Glass | More rigid than metal |
| Operating power | 2.0-2.5 w | 0.6 watts spinning with heads unloaded, 0.3 watts not spinning; autonomous adaptive power conservation. |
In the previous generation of disc drives, when the drive stopped the heads landed, physically touching the media. For various reasons this is unacceptable in today's drives, and on the 30GN and related drives the heads are pushed up a ramp, off the surface, when the drive stops. Thus the number of survivable start-stop cycles is much increased. 3.0e5 cycles would be used up if the disc stopped once every 5 minutes for three years.
``Pixie dust'' is a marketing term for the antiferromagnetically coupled recording medium. It has two thin magnetic layers separated by a layer of ruthenium, 3 atoms thick. As a result, each bit forms a closed vertically oriented domain with the magnetization in the two layers oppositely oriented, resisting spontaneous demagnetization at small sizes. Here is the URL of the press release on pixie dust.
The disc has an adaptive scheme for deciding when to enter various power-saving states such as unloading the heads or stopping the motor. IBM claims that a laptop's battery life can be up to 8% more if the disc is allowed to do that without interference, compared to a power-saving regime based on a fixed delay after the last disc access.
Reading the Dell documentation, it was not obvious (to me) how to remove the
hard disc. There is a screw on the bottom of the drive
cover; remove it. Now slide the cover downward firmly to unlatch
the drive, after which it can be slid out of the chassis. The Inspiron 3800
worked similarly.
Inserting the CDROM. On my Inspiron 3800 I got
burned on this one, and the Inspiron 4100 is similar. When you insert a drive,
specifically the CDROM, into the media bay, the data connector is stiff. You
will need to rock the drive left and right to make sure it's fully seated,
beyond the point where the safety latch clicks. Otherwise the drive may have
electrical power but it will not read things, appearing (falsely) to be
defective.
Also when removing the drive, rock it to get the connector to let it go.
Modem. The modem's PCI interface is an Intel
product, referred to by WinXP as a ``Sun MDC 56K D/F modem'', probably similar
but definitely not program compatible with the I810. The codec is a
PCTEL 2304 WT. http://www.medres.ch/~jstifter/linux/pctel.html
(Jan Stifter) has a driver for PCTEL products,
which I compiled --with-hal=i810intel. The pctel.o module loads, but the
ptserial.o module refuses to recognize the device, despite considerable effort
with PCI IDs and module parameters. When I essentially bypassed the device
recognition subroutine (setting the two iobases and the IRQ as reported by
lspci), the module loaded, but when I tried to open the device I got a kernel
OOPS.
So, unfortunately, this modem is a writeoff, unless someone else can
figure out how to make it work.
All the modems in the Linmodem howto:
Fan and Temperature Sensor.
There are two fans. The main fan, with two speeds, takes
air in from the bottom right and exhausts it to the right side through the
processor's heatsink. There are enough clearance holes that clothing does not
(usually) block the air intake. The smaller left fan, with one speed, takes
air in from the rear into the memory and PCMCIA area. The main fan comes on at
low speed during continuous computation (compiling a large package), but it is
quiet and not annoying. On battery power with SpeedStep engaged, the fan is
not needed even for continuous computation. High speed does not occur in an
office environment and program development activities, but when the processor
and graphics chip both get hot, or in hot weather, they take serious steps to
cool off, turning both fans on at high speed.
A typical CPU chip temperature on AC power under modest load (text work
under X) is 50°C to 55°C. During continuous integer operation (e.g. compiling)
it will reach 75°C and the fan will come on at low speed. Extreme loads
including heavy graphics can raise the temperature to 82°C, requiring high fan
speed. The maximum allowed temperature is 100°C.
On battery power the
SpeedStep technology keeps the the CPU well below 75°C with passive cooling
alone, on the same test task. In fact, I couldn't get the temperature over
68°C. The CPU runs at 3/4 speed, but the voltage is also reduced, so the
power is reduced by about half. If you are not crunching numbers, it makes
sense to engage SpeedStep even on AC power, to lessen the thermal stress
on the CPU.
Here are the temperatures (in degrees C) at which the BIOS turns the
fans on and off. Actions occur at the transition through the setpoint,
so if you use i8kfan to turn the fans off, they will stay off until the
next transition point. There is not an exact relation between CPU temperature
and fan behavior; likely the BIOS is using an external thermistor rather than
the CPU sensor diode (not a wise choice). Setpoints are the same whether
you are on AC or battery power. Setpoints are different in the two
listed BIOS versions.
| Action | A03 | A07 |
|---|---|---|
| Low speed (1) | 75 | 63 |
| Low speed (2) | -- | 65 |
| High speed (1) | 82 | 72 |
| High speed (2) | -- | 74 |
| Fan(s) off | 60 | 44 |
Massimo Dal Zotto When compiling, you may need to change the Makefile variables KERNEL_SOURCE
and MODVERSIONS to match the reality of your kernel. When executing i8kmon
under SuSE, you will need (as indicated in the readme) to use the command
line ``wish /path/to/i8kmon -- --noauto'' (or whatever options you choose),
or else edit the script to begin with ``#!/usr/bin/wish''.
You will need to get the module loaded before running i8kmon or i8kctl.
I put ``modprobe i8k'' in /etc/init.d/boot.local.
The author warns that the BIOS of the Inspiron 8000 has bugs in the fan
control logic, and
therefore he designed i8kmon to actually control the fans, overriding the BIOS.
This was not within my safety profile, since inadequate cooling can fry the
CPU, and I decided to run i8kmon with the --noauto switch. The BIOS controlled
fan setpoints seemed to me to be reasonable. At 75°C the right (processor) fan
comes on at low speed. At 82°C both fans come on at high speed. Decreasing to
60°C both fans shut off, whatever speed they were at. The left fan is capable
of operating independently, but I've never seen the BIOS turn only it on. Fan
actions occur at the transition through the setpoint, so suppose you exceed
75°C and the fan comes on, but you use i8kfan to turn it off; it will stay off
until 82°C when both fans come on. If you turn them off again they will stay
off until your CPU fries. Use i8kfan to turn the fans on again.
See also my own graphical monitor for
the temperature, fan status, battery charge, system load, and wireless signal.
In
ftp://download.intel.com/design/mobile/datashts/29834003.pdf "Mobile Intel®
Pentium® III Processor-M Datasheet", Page 70, ``Vcc Thermal
Specifications'', they say: ``. . . ensure the junction
temperature never exceeds 100°C. . .'' The ``THERMTRIP'' emergency
shutdown signal is asserted when the chip temperature reaches 135°C.
(Note when downloading: they sometimes change the rev number, so if the
listed file has vanished, browse in the containing directory for a new
version.)
On the Pentium 4 (not Pentium III), when a certain temperature is exceeded
(not stated in the documentation), and if enabled in a Model Specific Register,
the clock will be gated off for a MSR-configurable fraction (typically 50%) of
each 3 usec period, achieving a drastic reduction in power. This aggressive
power throttling is essential because the chip temperature can rise 50°C/sec,
and given typical analog averaging times in reading the thermal diode, the
value available to the BIOS for power and fan management can be as much as 6°C
below reality. The Pentium 4 also has a thermal sensor like the Pentium III.
IRQ Assignments.
These are the assignments of IRQs in Windows, and in Linux using default
selection during boot and during driver loading. The ``improved'' column
was achieved by explicitly setting IRQ options on the respective drivers
(indicated by **). Under WinXP, IRQs in parens are traditional assignments
which did not occur in my setup because the relevant devices were not present.
Under Linux, only assignments recognized in /proc/interrupts are shown.
lspci shows additional IRQ assignments, mostly to 11, some of which were
recognized at boot time.
Device WinXP Linux
Default Improved
Timer 0 0 0
Keyboard 1 1 1
Cascade 2 2 2
COM2,4 (3)
COM1,3 4
Dell 1150 Wireless 5 11 5**
Modem 5
Sound codec 5 11 6**
Floppy 6
Printer (7)
Realtime clock 8 8 8
ACPI 9
3c920 Ethernet 11 11 5**
Radeon Graphics 11
USB host controller 11 11 10**
PCMCIA slot 2 (empty) 11 11 11
PCMCIA slot 1 (empty) 11 11 11
PS/2 Mouse 12 12 12
Numeric Processor 13
IDE 0 (disc) 14 14 14
IDE 1 (cdrom) 15 15 15
Vacant 3,7,10 3-7,9,10,13 3,4,7,9,13
Next: Setup
Previous: Features
Contents