10W LCD screen part 1: Anatomy of an LCD
In this blog series on display power consumption:
Part 1 (You are here) - Part 2 (Ideal display)
My passion is electrical efficiency: getting high performance out of electrical equipment while using as little power as possible. To this end, I love knowing exactly what contributes to power consumption in IT equipment. When you know that parts A, B and C of some device use X, Y and Z watts of power, you will want to attack the biggest contributors first. I have done this before in my 8.5W Core i3 build and it worked out quite well. However, now that my computers are down to sub-10W power consumption, the display has become the biggest consumer. In the coming weeks, I am going to try to get a PC display to use under 10W as well. But to make this happen we first need to know what constitutes PC display power consumption.
This blog series accompanies the blog series about my newest computer build, an Ivy Bridge desktop quad-core all-in-one with IPS display consuming only 20W when idle. If you like this series or have found the information in my blogs useful, please consider donating to me (donation link at the bottom of each blog) to help make my newest build as awesome as possible.
We are going to look at display power consumption from four angles in roughly four blog posts. First off (in this part), we are going to look at the components that make up a display to get a feel for some of the terminology and to understand certain design decisions in modern displays. In part two we will do a theoretical study to find the ideal display from a power consumption point of view, taking into account some display technologies. Part three will be a literature study: how much power do modern displays consume and does this depend on screen size, brightness, panel type and other factors? Which screens are best, which are worst, how can you choose the most efficient screen? Lastly, in part four that is, we will be looking at the electrical side of things: how does a typical display distribute electrical power and how do I expect on improving on this? All of this will be treated from an (opto-electrical) efficiency point of view - I will leave things like image quality for another blog series.
0. Index for part 1
- 1. The anatomy of an LCD display
- 2. Sidestep: OLED and other active pixel techniques
- 3. Conclusion of part 1
1. The anatomy of an LCD displayAlmost all PC displays are very similar internally. There is some variation in component choice, but the basic theme has remained the same since the switch from CRT to TFT. OLED is the only exception here, being one of the first active pixel technologies (besides plasma which is rarely used for PC displays and not discussed in this blog post). The main idea is that the backlight produces a very evenly lit rectangular shape. Then in very small rectangular subsections of this area ('pixels') the light is dimmed and passed through color filters, creating a small area in a shade of a solid color ('subpixel'). Many of these subpixels exist next to each other in such a high density that our eyes cannot discern between the subpixels and so the colors flow together, creating many different colors. Also, by making the pixels small enough it is possible to create the illusion of smooth curves (which are actually made out of very small colored rectangles). That is the basic idea. The implementation of this idea requires four things:
1.1 The 'panel'The part that modulates (dims) the backlight in small rectangular areas is called the panel. The panel looks like.. well, a large greyish glass or polymer sheet. Let us drop the scientifically correct term and call those small areas pixels and subpixels. There are various ways to do this dimming. For instance, you could put very little blinds in there and turn them around with a piezoelectric element. This blocking of light by putting something in front of it is called opacity or aperture modulation. Even though this method has never been used (commercially), it is still relevant to the discussion as we will see later on. There are in reality many more methods, but for this discussion I would like to pretend that the only other method is what we call polarization modulation. This method is used in all liquid crystal panels and so in practically all PC displays.
Light has a property of polarization that can be visualized like this: let’s pretend light waves are in fact wave-shaped two-dimensional thingies that exist in a 3D world. Light travels in a certain direction, but along this direction it can be rotated. This rotation is called the polarization angle. Now, there is such a thing as a polarizing filter. A polarizing filter will only let pass light that is oriented (polarized) at a certain angle, but no other angles. If we now put two polarizing filters behind each other we can do something interesting: if we align the polarizing filters so that they both let light through at for instance 0 degrees polarization, light rays with the correct polarization will pass through both filters just fine. Now we rotate one of the filters in respect to the other, such that their polarization angle is different. A light ray passing through the first filter is now blocked by the second.
Polarizing filters don’t only allow one angle; they let a small range of angles through. So if you put two filters behind each other, both with a small angle-range, they will still let some light through. Not as much as when they’re perfectly aligned, but still some. See where we’re going here?
Well, that is not really how LCD displays work. They do not have rotating polarizing filters. Instead, they have two polarizing filters behind each other on separate glass sheets. In between the polarizing filters are liquid crystals: a solution that has a very interesting property: it can 'bend' the polarization of light. In this arrangement of filters and liquid crystals, light from the backlight first hits the back polarizing filter. Then it passes through the liquid crystals where the polarization is 'bent' a certain amount before it passes through the second polarizing filter. And yes, you guessed it, it is these liquid crystals that we can control with some voltage, thereby varying how much the light is 'bent' and as such stopped by the second polarizing filter.
Why did I go through the hassle of explaining all this when you can get a much better explanation on Wikipedia? Well, I have touched upon a few things that massively influence efficiency here. Let us start with the whole idea of polarization. It is safe to assume that the backlight in an LCD display produces light with all possible polarizations, however, even when displaying a purely white background (i.e. as much light as possible is transmitted through the liquid crystals) the light is still filtered by polarization. Some light will not pass because it will not even make it through the back polarizer. Exactly how much will be discussed in the next blog post.
Next up is the fact that these liquid crystals are not perfect light conductors. Great strides have been made since the beginning of the LCD era, but still a significant portion of light is lost in between the polarizing filters.
Individual subpixels obviously need their own voltage control to create individual colors. That means that a lot of wiring and quite a few transistors need to be placed on the screen, pretty much next to each pixel. If we zoom into a screen, we can see this:
All that black space around the pixels, that is wiring and transistors. Actually, just a small 'corner' cut out of the subpixel is a transistor area, the rest of the blackness is mostly wiring. Fortunately, large displays with large pixels (low pixel density, or pixels per inch [PPI]) do not suffer very much from this light blockage. The blocking area is usually only 10-20% of the active pixel area. However, on very small displays with high pixel density (e.g. high-resolution smartphone displays), wiring and transistors blocking light can take up more than half of the screen area. Screen technology makes a huge difference; older VA and IPS panel technologies necessitated larger transistors and/or other elements that are just not very good light transmitters. Also, in general TN displays are the most transparent while VA and IPS still are at a bit of a disadvantage there (although more recent implementations have improved a great deal in this respect).
Anyway, there is of course one more obvious cause of light blockage: the color filters. These filters usually block anything but red, green or blue. Again, this is not super-absolute: they do not just transmit one exact wavelength, they transmit a narrow bandwidth of wavelengths. Still, every filter only passes about 33% of the incident light.
All this blockage of light means you need a stronger backlight to get the same light output on the user side of the panel, increasing power consumption.
1.2 The backlightThe backlight is the source of light in a display. It needs to be a white light emitter because it delivers light to all subpixels (red, green and blue). Keep that sentence in the back of your mind for the upcoming theory blog post. This white light is usually produced by an array of either CCFLs or LEDs - CCFLs are fluorescent lighting while LEDs - I think - need no introduction. Unfortunately, both CCFLs and LEDs are either point- or line light sources, but the light needs to be distributed over a much larger surface (the backside of the LCD panel). To this end, diffusor and reflector layers are used.
The diffusor is not perfect. Unfortunately, I do not have efficiency data for diffusors, but a certain percentage of light is lost in the diffusor and adjoining layers.
The backlight needs to be as bright as the brightest pixel on the screen. Well, in a lot of displays it is actually at maximum brightness all the time, even if all of the screen is displaying black. There we have a major, major inefficiency inherent to LCD displays.
The light source itself is of course a very obvious source of power loss. The amount of light that a light source generates is measured in lumen (lm), and how much light is generated per watt (W) of electrical power is what we call lighting efficacy (lumen per watt, or lm/W). The average lighting efficacy of CCFLs is about 60-80 lm/W, while LEDs can go a bit higher (typically around 100lm/W but up to 160lm/W with commercially available devices). Note that this number has no direct relation to electrical efficiency (i.e. how much light energy and heat is generated per unit of electrical energy). This is better left for a separate discussion.
1.3 The power supplyAn LCD display requires a few different voltages which are not commonly available in your average household, so you will need a power converter (or actually, multiple power converters) inside an LCD display to work. It is not that easy to design power converters with better than 80% efficiency, so like with pretty much all electrical equipment, power conversion is one of the leading sources of losses. Power conversion and distribution in LCD displays is fairly 'standardized', that is to say, there is no universally agreed upon standard but very few displays stray far from the following build-up:
This picture is valid for both LED and CCFL displays. Mains voltage is converted to some intermediate DC voltage (often called 'bus' voltage) - usually something between 12 and 24V. CCFL displays tend to be on the higher end of this scale, LED displays are often 12 or 19V designs, because that enables the use of industry-standard external power adapters, enabling thinner display designs because effectively part of the display is dangling off a cord.
Power flow then separates: part of the bus power is converted to a voltage suitable for the backlight (36-80V for LED backlights, 300+ V for CCFL backlights), another part is converted to usually 5V and possibly 3.3V for the microelectronics. That 5V is really essential because most, if not all, LCD panels use this voltage for their electronics. It does not stop there though: the 5V fed to the panel electronics are again converted to - usually - 1.8 or 2.7V for the LVDS receiver and about 16V for the panel bias driver.
There are a few variations on this concept. Older designs often directly converted mains to bus voltage and to 5V, even designs that had external adapters (they would be 4-pin adapters like the ones you may still find on external hard drive enclosures). Why would you do that? Well, as you might know a conversion efficiency of 80% and above is considered pretty good, while anything under 65 to 70% is rubbish. If you put a few converters in series - i.e. first you convert mains to a bus voltage, then the bus voltage to 5V and then 5V to 16V - the losses multiply. If all of those converters were 80% efficient, i.e. decent converters, the overall efficiency from mains to 16V would be 0.8 x 0.8 x 0.8 = 51.2%. Awful! In power electronics, you will want to have as little series power conversion (called 'cascading') as possible. But as we will see in part 4, cascading of power converters is something that still happens, even in modern 'efficient' LED-displays.
But anyway, let's not dwell too much on my favourite blogging topic. We have got more to discuss!
1.4 The microelectronicsThe microelectronics (i.e. electronics involved in signal processing) inside LCD displays are usually split up into two functions: the video signal receiver and the panel driver. The video signal receiver in the olden days was something comprised of many different single-function chips: one for receiving digital video, one for receiving analog video, one digital signal processor (for instance for scaling, color correction, etc.) and one LVDS transmitter (used to transmit the final video data to the panel). Nowadays, you will probably find one large integrated chip behind the video ports and that is it.
The other part, as I have already touched upon, is the panel. Usually, integration is really the best way to go in electronics, but panel suppliers and display assemblers are still very different companies these days, so making a fully integrated panel+video input has not happened yet. On the panel driver you will usually find an LVDS receiver chip, a small dc-dc converter that delivers about 16V to the pixel transistors and a lot of so-called line drivers that send brightness signals to each of the subpixel transistors.
Now we get to the fun part: what uses power, and why? Two offenders here. The first one is actually more power conversion than anything else. The video receiver chip on the video input and the LVDS receiver on the panel itself are chips that really do not need much power, but they are often fed by linear power converters. Also, strangely enough they often have very poor power management by themselves: they run at full voltage and frequency even when displaying things that do not require such high operating frequencies and voltages. But on the grand scheme of things, they are really not that bad.
The second and biggest offender is the combination of dc-dc converter and panel drivers. You see, for most panel technologies, when you put 0V on a pixel it is white and when you put some high voltage on it (e.g. 16V) it is black. The most current and voltage is used when a pixel is black, so that is why most if not all LCD displays actually use more power when displaying a black image than when displaying an entirely white image. Interesting, huh?
The reason that this uses up energy is twofold: first, there is actually some current that flows inside the liquid crystal channels, so the pixels 'use' energy by themselves. But it is actually the transistors themselves and the line drivers that are throwing away energy by the bucketload (relatively speaking). The line drivers need to send information to lots and lots of pixels every second, so they run at a very high frequency internally. This is aggravated by a technique called response time compensation (which you should look up if you are not familiar with it yet) that increases the necessary working frequency for the line drivers by at least a factor of 3-4. Not only do they work at high operating frequencies, but also at a fairly high output voltage (roughly 16V) using very lossy transistors. All due to the fact that in for instance an HD display, almost 6 million subpixels need to be refreshed 120 times per second.
Pocket calculators only have a few pixels with very low refresh rates, and that is the reason why they are so damn thrifty - only in the order of tens of nanowatts are needed for such a display. A modern-day 24-inch HD display panel may use 5-10W just for refreshing the pixels and keeping them in the color they are at. Compared to the less-than-one watt for the video receiver, that is a big deal.
2. Sidestep: OLED and other active pixel techniquesPassing light through all kinds of barriers seems like a very wasteful way of making colors. There is of course a much more straightforward way: making pixels that emit light themselves. That is what we call active pixel techniques. Plasma screens were the first successful commercial active pixel technology, OLED (in particular Samsung's AMOLED) is a bit more interesting from a space- and energy-saving perspective.
Obviously, with pixels that emit light themselves, you do not need a backlight (with their respective losses and more importantly, bulk) and you do not need a diffuser. Just the 'panel' itself. Also, there is no polarization filtering, no color filtering and no transmission losses. This must be much more efficient than traditional LCD displays, right?
Wrong! I love these counterintuitive things. In the case of plasma screens, it is just a matter of using a fundamentally horribly inefficient way of generating light. Also, interestingly, plasma screens still have color filters. There is so much wrong, from a power and light generation point of view, with plasma screens that we will not be discussing this in the context of efficiency.
OLED screens really are a very nifty design. You really only need a panel. Each pixel by itself is a small LED that can be turned on in varying degrees to make a shade of red/green/blue and mixed, again, with nearby pixels to create the illusion of color. There is still wiring and transistors going on, but this time they are behind the light-producing element, so it is no problem. However, they are not efficient by any means. They are still fundamentally problematic.
This is because of two things: first up is the fact that the pixels are not white LED emitters. Colored LEDs in general, but really OLEDs in particular, have horrible lighting efficacy. This is partly because in a display, it is important to focus on color accuracy more than lighting efficacy. Second, when displaying less than 100% brightness the pixels in an OLED display are, at least in current technology, dimmed with a linear circuit rather than a switching converter or PWM methods. This means that, for instance, a pixel at half-brightness does not use half the power compared to the same pixel at full brightness, but more. Lastly, OLEDs as of writing still have a much lower lighting efficacy than silicon-based LEDs. All of this means that the effective lighting efficacy is pushed down to less than that of the already very light-inefficient LCD approach. Really a missed opportunity!
Theoretically, though, OLED can actually become more power-efficient than LCD. LCD is fundamentally limited by the opacity of polarizers and color filters. OLED can be improved on per-pixel efficacy, while newer display drivers can employ PWM for pixel dimming to negate its other major shortcoming. Combine this with the possibility of building razor-thin, cheap displays using OLED and you might be spelling the future of personal displays. I do not necessarily agree with that statement, but you will have to wait a couple of weeks for part 4 of this series to come out to find out why.
3. Conclusion of part 1In this part we have discussed the parts that make up LCD and OLED displays - from an efficiency point of view. In a way this part is a teaser: if you were able to follow what I wrote down you can probably read between the lines as to what I will be doing to improve on power consumption. Qualitatively, the answers are mostly there. However, this blog series is more than just a series of thoughts and qualitative explanations. We need hard numbers! Science!
That is exactly what the next part in the series will bring. In part 2, I will delve into the maths behind LCD power consumption and try to answer the question: is it even theoretically possible to make an LCD display (of a decent diagonal size, say 20+ inches) that consumes less than 10W? What about less than 1W? Seeya!
Thanks to my girlfriend, Devilly, Infant, pientertje, sebastius, Snowmiss and TheMOD for proof-reading
|No PayPal? Message me or e-mail me|