The Screen-tablet assembly seen as a capacitor.

[DaBotz]

These are my reflections, after having built two working X...Qs and having found, in both cases, that jitter could be lowered (virtually eliminated in one case, reduced down to 3-4 pixels first, and then 1-2 px in the second) by adding common mode chokes on the USB line

Along with EMF disturbance,  we should also consider that the interface between screen and the tablet circuitry act as a capacitor and is, thus, amenable to the passage of alternating currents between screen and tablet without any major magnetic influence.

A possible solution to this is to "ground together" the screen and the tablet base plate, and this has been observed to prove quite useful in plenty of occasion.

However, in some situations (if the tft circuitry inside the LCD is not really grounded with the ground presented by the TCon board), it is useless. Also, if the Wacom plate is actually "floating" and there was a decoupling capacitor between its signal and the actual ground of the usb cable, the result could also be to actually augment the capacitance of the assembly, and so lower the impedance seen by the noise.

As I said, these currents would flow through the interface between tablet and screen, and would be impervious to be fixed by classic EMF shielding, because they are a by-product of the very geometry of our adored X...iQs: two conductive, parallel plates (full of "holes" as it may be, the active semiconductors in the LCD matrix make still for a big plate of conductive material) with a dielectric in the middle.

Some rapid calculation (with a lot of assumptions in it, done using http://www.daycounter.com/Calculators/Plate-Capacitor-Calculator.phtml ) give a capacitance of anywhere between 300 and 1000 pF (depending on the nature of the diffusor, the latter value being for glass)for a 22" assembly which, at 666khz (a good enough frequency for Wacom tablets, per the researches of Aerendraca ) would mean that the impedance of the screen-tablet assembly is somewhere between 1600 and 500 Ohms. It is not very much.

Anyway, one good thing of this noise is that it can be hampered down (alas, not completely  eliminated) by adding a big "common mode choke" http://en.wikipedia.org/wiki/Choke_(electronics)#Common-mode_choke  on the USB cable (or a series of smaller chokes).

Adding chokes to the screen power and the video cable may help too, as in the end they all become impedances for the high frequency noise.

Also, a distinct reason to add at least a choke to a vga video cable is that the same kind of noise can also disturb analogic video signals.

I hope you found this interesting, and to let me know if you have any objection (I may have got wrong what was supposed t be ground together, for example).

Also, I would like to know if anybody else has find this trick useful.

_DB

[Aerendraca]

I'm not sure that I can agree that the screen is acting as a capacitor, at least not one of any significant value, as the total surface area of the wires within the screen would be tiny. There is also the issue that every pixel has a capacitor integrated into it further changing the dynamic of any capacitive behavior which may exist, and for these reasons I would suggest that you cannot treat this system as a parallel plate capacitor.

Now I'm not saying that the following calculation is true and correct, but I attempt to tackle the parallel plate problem in a slightly different manor by means of comparison.

Lets say that you have a fairly bog standard TFT - 17" 1920x1080 widescreen.
1. Let's assume that each tiny wire joining the pixels is something very small (but larger than an LCD could realistically have), let's say 0.001mm in diameter - [1x10^-6] metres in diameter.
2. Let's also assume that each column of pixels has 1 wire that joins them vertically as part of an array. We then have a total of 1920 vertical wires of length 0.23m (from the active area of a generic 17" 1920x1080 screen being 360mmx230mm).
3. We ignore the curvature of each wire since they are tiny and probably flat anyway, and calculate the total surface area of a 'plate'. Here this will be [1920x(1x10^-6)x0.23] = [0.4416x10^-3] m^2 or [0.4] mm^2 (with a little rounding).

Incidentally we could have equally chosen to use the horizontal lines instead; [1080x(1x10^-6)x0.36] = [0.4] mm^2 (again with a little rounding).

That's a pretty small area, but then that makes sense as you wouldn't want to see the connecting wires.

4. If we plug this in to the calculation for a parallel plate capacitor using a couple of further assumptions - 0.001mm thick dielectric medium with a relatively high value (by way of exaggerating a result), say perspex at around the 4 mark, we can generate a vastly boasted value of 15.6pF. Realistically it's unlikely to be anywhere near as high as this and it's because of this that I don't think this can be the reason. 

I do like that we're applying some thinking outside the box to try and work this problem out, and I really don't want to put you off. Perhaps someone disagrees with me, or maybe my calculation is wrong. Let's get a discussion going on this! 

[Aerendraca]

I've been doing some stuff at work with sensitive amplifier circuits recently, and this combined with my recent bumble of forgetting to buy a power supply has made me wonder something regarding jitter causes.

It's extensively advised everywhere that sensitive amplifier circuitry (including audiophile stuff) should be power with a regulated linear power supply to keep the amplification as pure as possible and reduce any noise which might originate at the power supply. Now this makes sense when you understand the difference between a linear power supply and a switch mode power supply.

A linear power supply takes the AC from the mains and steps down the voltage via a physical transformer coil. The signal is put through a bridge rectifier to turn it to DC, then smoothed with a capacitor or two and rectified using a semiconductive device such as a 7812 integrated circuit (for12V). The beauty of this system is that the noisiest part of this process is at the conversion to DC, but a good selection of capacitors after this stage and you can pretty much eliminate the noise altogether.

Switchmode technology is quite complicated so for the purposes of this I shall just say that this type of power supply essentially employs a transistor to switch the power from the mains full on and full off very rapidly cycling through a charge to discharge loop continuously. Because the power is being switched rapidly, very little strain is imposed on the circuitry reducing heat and vibration losses and increasing efficiency. However, one caveat of this process is that the switching on and off causes spikes in the power signal which under the wrong conditions can cause real problems. Thing is that a majority of devices do not require a silky smooth signal, devices such as mobile phone batteries will quite happily charge via a switch mode power supply; It is because of this, the demand for smaller electronics, the desire to keeps costs down, and the increase in efficiency (being kind to the planet goes a long way when marketing) that switch mode power supplies are extremely popular and much more easily obtainable.

Linear power supplies are poorly efficient in comparison, they're big, they cost more because they use a lot more metal, BUT......They produce a nice clean output. My thinking is that perhaps a part of the key to resolving this is to try using a linear power brick as a power supply and put the new school switching psu to use illuminating some fairy lights.

I'm going to buy a linear psu and see what happens.

[DaBotz]

I understand what you say about the switchmode power sources being some nasty clients. In fact, most of the issues of my CabinetiQ were created by a slowly failing power bric.

At the same time, I am not sure that we can discard the surface of the transistors in the tft matrix from the equation, or the electrodes that twist the nematic molecules, when looking at the "CintiQ as a capacitor".

If you consider them too, the area is probably half that of the screen.

At the same time, it is not that I came to think at this out of a superior moment of illumination.

When my last power bric burned down on the CabinetiQ, in a moment of desperation (the CQ had become nearly unusable, in those last days, probably because the failing power supply kept pushing out more and more noise into the lcd panel) I used its transformer nucleus to wind up a common mode choke on the power cable and the USB cable.

It was a "let's try this, before I give up and buy a YiYnova", that completely solved my jitter problems and the false second  buttons clicks. Never saw one of those ever since (before, I had to disable the second button one day a week or so).

If I take away one of the chokes, jitter is back.

The same happened with the UBiQ - testing after having striped the lcd, it was a mess, .

Shielding the tcon, moving the power source away from under the tablet helped, adding some chokes made it usable, adding some more smoothed things more.

So, there is a common mode noise - a current, not just "static" voltage spikes, or the choke wouldn't work - of some kind, that messes things up in my machines, and hampering it helped a lot.

The only model for this that I can think of is this: the assembly acts like a capacitor, and a signal pass trough it.

Then, the mechanism for this passage could be completely different (and, as always, I should not forget that my house grounding is shitty as hell... so, may have grounded the armour of my lcd panel the best possible way, and it would still be a damn "floating" electrode hovering 0,35 mm above the Wacom circuitry - crap! - directly soaking in noise from the power unit shield - bi-crap!)

A way to see what really happens, would be to find a "X...q" where adding chokes helps, wind an "inspecting" coil in the choke, and inspect the induced currents.

Probably, this would give us some ideas of what is really happening in these cases.

Unfortunately, I have access to no equipment at all, or I would have already tried to look into it.

[Aerendraca]

I'm not convinced about the transistors and electrodes taking up half the screen, if they did they would impact on how much light could pass through for sure. Whilst I appreciate that they may contribute I'm don't think they would make all that much difference, plus even 1000pF (1nF) is a tiny capacitance. Also, shouldn't the screen being switched off continue to produce jitter?

I know that the screen does affect the jitter on the tablet as I have seen this and recreated the jitter myself by broadcasting a frequency across a wire and forcing the cursor to jitter (think it's detailed in the original Serenitiq build log). This makes sense to me as, similar to the switch mode power supply, quick switching of pixel voltages causing radio noise seems logical and unavoidable. I think that the reason some panels affect the jitters more are due to the way in which the arrays are setup and potentially how much voltage is sent through to each pixel.

You've got me wondering about the USB power now though. I know that computer power supplies are pretty crappy in terms of noise produced which may be why you're seeing an improvement using a choke. I wonder if the power for the tablet was driven from the power for the screen (using a regulator and hacking the USB cable) if this would level the common ground between the devices. In principle I genuinely believe it would. I know that in the past people have tried using powered USB hubs, but this would just harp back to what I was suggesting about noise created by switch mode power supplies (which USB hubs always use) and inevitably no improvement is seen, however this method might work if the power is taken from the supply for the LCD. Infact I think that this is even the way that the Cintiq 13HD utilizes power; it uses a 3in1 cable that links USB, Power, and HDMI, all of which I bet share a common ground.

I will be trying this too then.