Librem 5 Backlight PWM frequency

Nothing changed on that front as far as I’m aware.


Ok, that’s great. Thanks for your help!

So this is just a suggestion for marketing, I think that mentioning that as a feature in the specs list and shop would be a good idea. I think there are people who would be interested in the phone because of that. I know other manufacturers market their screens as flicker-free or eye care or whatever. If I was looking at the specs for the first time, it certainly would have made me even more likely to purchase one.

1 Like

Well, I can still speculate and wait for final specification concerning the PWM controller/chip. Maybe Purism already decided to give a try to LP8555 High-Efficiency LED Backlight Driver anyway (around 1,25 € vs 0,45 € for 6000 pcs.). Furthermore, Purism self do not need to confirm PWM chip specification (before or after release) so I am not making here an issue out of this (as I am not manufacturer) other than hoping to be satisfied consumer. And even without willing to understand MacAdam Ellipse numbers this may be useful to read:

1 Like

Without looking too deeply into technical details, the LP8555 does seem a lot better than the LM36922 to me. In my understanding of what I have read in the specs, the LM36922 sounds like it does output PWM as @TungstenFilament suggested. The frequency is certainly an improvement over 200Hz as I mentioned before, so I am thankful for that.

I’m not sure if the LP8555 would be suitable for the phone design or not. Hopefully I am wrong but it’s possibly too late for Purism to consider something else now anyway unless it’s a drop in replacement. And honestly, I wouldn’t want them to change it if it would jeopardise the release of the phone because of another delay.

The “Hybrid PWM and Current Dimming” in the LP8555 sounds like it does the same as some of the displays I have seen where it goes down to 25% brightness before switching from Current to PWM dimming. I guess those displays probably use that device or something similar.

It would be really good, and I would certainly be very grateful, if Purism would consider the LP8555 or something similar for the Librem tablet. It can also be used for laptops according to the spec sheet. I would also very much appreciate if they would consider using it for their laptops, although my next purchase will probably be the tablet. Thanks @Quarnero for pointing it out.


Disclaimer: I am not an electronics engineer.

If you look at the block diagram (section 7.2/page 10 in the LM36922 datasheet), these frequencies appear to be for the boost converter which is used to generate a voltage higher than the input voltage in order to drive long series strings of LEDs.

The boost converter comprises some control circuitry, a FET shown in the block diagram and a 10µH to 22µH inductor shown in the simplified schematic on the first page of the datasheet. Like other kinds of switching power supply, as found inside phone chargers and PC power supplies, this will generate pulsed DC at the required voltage, which can then be smoothed into ‘normal’ DC. There is a 1µF capacitor shown on the simplified schematic, which may well smooth the pulsed DC to create non-pulsed DC. There will always be some ripple from this kind of power supply. In applications where low-ripple is important, voltage regulators can be used to chop the ripply bit off of the top of the waveform, after smoothing with a capacitor. The LM36922 probably doesn’t support adding a voltage regulator, but it’s less important for LEDs than it is for microprocessors and the like. If the waveform is smooth enough, then the LEDs won’t be leaping from fully on to fully off and back again. Instead, they’ll just be dimming by some amount at the frequency of the ripple.

The other end of each of the LED strings is connected to a current sink circuit, which is what actually controls the brightness. There isn’t much detail about how this part works, but in the absence of any information about its switching waveform or anything of that nature, I would assume that there is no switching and it just sinks the specified DC current with no measurable funny business.

The LM36922 does accept a PWM input, which can be used to specify the current with which to drive the LEDs. Alternatively the current can be specified using an I²C serial interface. Both methods result in the current being set in the same manner, they’re just different ways to input data into the chip.

So, I can’t see anything in the LM36922 datasheet that suggests that the LEDs are driven with PWM.

(Edit: After re-reading my own post, I decided I should be a bit more circumspect about the smoothness of the DC supply for the LEDs, so I’ve edited my wording. I still believe it’s likely to be pretty smooth, but i want to be careful not to make claims I’m not expert enough to support, and I don’t want to imply that the supply to the LEDs will be as smooth as the DC you might get from a PC power supply.)


Thanks for your post @patch. You probably saved wasting a dev’s time responding. :wink:

Excluding the necessary capacitor and inductor, the output would be essentially pwm though right? And how ripple free the output is will depend on the effectiveness of them. Or am I misunderstanding?

I do hope one of the reviewers will test the screen to see how it fares using an oscilloscope. However from a practical point of view, I’m really just hoping for a screen that will be as usable as possible for me. It seems it will likely be a lot better than some displays I have used, so I’m encouraged by that.


It definitely won’t be PWM. PWM is quite a specific thing. It might be a similar sort of waveform though. I don’t know enough about boost converters to say.

How ripple free the output is will depend on the overall circuit, including the capacitor, the LEDs, and the inductor. The components need to be chosen together.

You can’t actually take away the inductor or the capacitor. An inductor is fundamentally necessary for a boost converter, and this particular chip requires a capacitor for stability.

I think I will try doing that when I get my phone. I have a solar cell that I could use for that kind of test.

(Sorry to anyone observing my many edits of this post! I’ve not been thinking before I post!)


Ok, well thanks for your help. I’d like to do some reading on the subject when I get some time.

That would be good if you did test the screen. Will you share your measurements? What is the sampling frequency of the oscilloscope you’ll use?

I have an analogue oscilloscope with 20MHz bandwidth and I will share my results. (It might take me a little while to get around to taking the measurements. Remind me if I seem to forget.)

1 Like

Ok, that would be great. Will do.

I’ll buy it when librem phone comes with e-ink screen. This will be a perfect phone.

1 Like

I suffer from migraines and use that on debian. Specifically, I use

xrandr --output VGA1 --brightness .5

where the number (I’ve got it set to half hence the “.”) is the global brightness setting so just mess around with it till you find something you’re most comfortable with.


As well, Debian 10 offers new messing around within Settings/Devices/Displays and it is called: Night Light with Schedule and Color Temperature. Going from around 6000 Kelvin down to around 2000 Kelvin helps to some a lot, in short, without discussing the blue light dark side. I am very thankful to Debian on Wayland for this feature! But don’t expect that yours display Red color (No.9) quality will be more red as it depends on initial technology used (easily might be less than 50%). For those interested, here is link to the currently best Color Rendering Index (from LED technology for general use, not for backlights in LCDs) just for comparison (even though real display data are not commonly available especially if they don’t offer 10-Bit video output/true 30-Bit color output) and reference.

1 Like

will never happen. the software costs of developing that isn’t worth the 5 people who will buy it

1 Like

I’m nocturnal, but that sounds really cool for people with day jobs in respects to protecting their circadian rhythms :slight_smile:


don’t use that if you can afford not to … just use an external monitor with built in hardware OSD profiles for Reading and switch to sRGB when you need color fidelity.

or better yet (if you can afford it) buy a hardware color calibration device and use displayCAL under X (wayland needs more love with this to work without error)

a little old but will update when wayland is resolved …

1 Like

Yes, you can look at my thoughts just another way around: LCD screens between 6500–9300 K + minimal PWM frequencies response times of around 250 Hz gives us just right formula not to sleep at all for several hours after done with our homework late at night (but some young people still need to keep up with their circadian rhythm as scheduled, very next morning). Lower brightness might help of course, yet not with low PWM number, afaik. Anyway, I would rather, after sunset, look at something with no-PWM (or higher then 20000 Hz), even if this mean full brightness, and 2600 K instead. I’ll leave at this just not to go way out of topic, but interested in this conversion to lower Kelvin numbers may take a look at “The values as a handy python dictionary” from Andreas Sieß and adjustments made on his photo Grand Traversata Delle Alpi.

Yes, please! And this play around with Kelvin adjustment on wayland was actually for reading usage.

1 Like

or if you have 2700k classic light bulbs or LEDs with that color temperature … if you are keen to do ambient light white-point matching … contrast will suffer …

1 Like

Ah, this explains why the only viable bedtime reading is strictly traditional book for me and why on bad days the lighting in the new subways will trigger a migraine but never the older cars. Hey, to you, patch, and everyone else posting here…thanks because this is rather interesting. I’m only barely comprehending it (and only because I used to date a DOP years ago who would go on about colour temperature, etc. as it relates to film) but it is intriguing.