EEVblog #717 - How To Hack Vacuum Fluorescent Displays

[EEVblog]

Dave explains what Vacuum Fluorescent Displays (VFD's) are, how they work, and then hacks an interface and reverse engineers a surplus display from an industrial machine to make it work with an Arduino.

TI TL4810 VFD Driver chip: http://html.alldatasheet.com/html-pdf/156237/TI/TL4810B/84/3/TL4810B.html

[mikeselectricstuff]

Some nice info from Noritake Itron on VFD principles :
http://www.noritake-elec.com/vfd_operation.html

[dentaku]

That was quite interesting and educational.
It was more of a Mike's Electric Stuff style project/hack but closer to something a beginner could do.

[bktemp]

Great explanation of how VFDs work, without going to much into details!
There is also a reverse construction type with the anodes at the front glass and the grids and filament mounted behind.

Some notes for people who want do drive bare VFDs:
The anode/grid voltage depends on the multiplex ratio (the number of grids). For static drive displays (single grid) it can be as low as 10V or up to 70V for large displays with >50 grids.
Up to around 10 grids it is possible to use 4000 series logic (4094 as anode and grid driver) at up to 18V with decent brightness.
There is also an easy way for estimating the filament voltage:
Because the filament needs a certain temperature to emit electrons, all you need to do is measure its temperature and adjust the voltage. Since tungsten has a positive temperature coefficient it can be used as a temperature sensor: The hot resistance is typically in the range of 4x the cold resistance. This should give a dim glowing filament wire.

It is a good idea to start at a lower anode voltage and add some watchdog timer or other fail safe mechanism that shuts down the anode supply if the grid multiplexing fails. Otherwise the phosphor at the anodes burns away really fast at high voltages.

[Ketturi]

I love VFD's. Still best looking display technology excluding OLEDs. Dave got easy one to reverse engineer, those are kinda off the shelf modules that are best for own projects.

I have project of my own with much more complicated driver board, but I just hacksawed a part out where display and buffers & shift register were, and coded my own multiplexed display driver. Easier than reverse engineering old 8085 machine code from cpu that drives 3x 6 digit 7-segment and one 32 digit 14 segment alphanumeric VFD-displaydisplays. And seems to be not standard ASCII. They do not have their own power supply, so I have problem of finding 5VAC + 40V DC VFD psu module.

[firewalker]

Any chance for it to be "Hitachi" compatible?

Alexander.

[helius]

Nice video! I like these displays, too.
A note on the need to use special high-voltage driver circuits: there are some special VFDs referred to as "Chip-In-Glass" or CIG that have integrated drivers and a shift register. So when interfacing these you only need to supply the B+ anode voltage, and clock bits in at regular TTL levels.
http://www.noritake-itron.com/NewWeb/GVFD/AppNotes/CIG.asp

Any chance for it to be "Hitachi" compatible?

There are graphic VFD modules compatible with HD61202: http://www.noritake-elec.com/guu100.php
They are advertised as drop-in replacements, but the glass panel may not be exactly the same size.

[nitro2k01]

I don't know what you tried of course, but an unknown pin may be an address ("A0") pin for sending special commands. Another good bet is that it may be using ANSI escape sequences. Byte 27 (often written as oct 33 or hex 1B) followed by a command sequence.

Would have been fun to see a mini reverse engineering of the drive circuit.

[rollatorwieltje]

If you need a VFD, try to get a "customer display" for a cash register. These things seem to always use RS232 and listen to the standard ASCII commands + escape codes to define custom fonts etc. A typical controller for these things is a Samsung 20L203DA5.

[deephaven]

Interesting video.

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

[dentaku]

Interesting video.

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

It was from Mark Butterworth of Santa Clara California.

[Hydrawerk]

See also https://www.eevblog.com/forum/microcontrollers/itron-noritake-gu256x64-312-avr-library-needed/
http://kimboricha.blogspot.kr/2013/10/noritake-itron-gu256x64-901a-vfdvacuum.html

[Dinsdale]

Hay, what about "clear screen"?
What about "cursor position"?
Only 35 min.? You got a half hour to go!
And 25fps? I can't see anything!
 
But, really, I enjoyed it.

[deephaven]

Interesting video.

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

It was from Mark Butterworth of Santa Clara California.

Well, spotted, thanks dentaku and thank you Mark if you're watching 

[gslick]

Maybe the NEC microcontroller is an MCS-48 architecture uPD80C48 / uPD80C49?

http://www.cpu-world.com/info/Pinouts/8048.html

Do the data line inputs match the DB0-DB7 I/O pins 12-19?
Is the 12MHz crystal connected to XTAL1 and XTAL2 pins 2 and 3?

[nitro2k01]

Maybe the NEC microcontroller is an MCS-48 architecture uPD80C48 / uPD80C49?

http://www.cpu-world.com/info/Pinouts/8048.html

Do the data line inputs match the DB0-DB7 I/O pins 12-19?
Is the 12MHz crystal connected to XTAL1 and XTAL2 pins 2 and 3?

Yes and yes.

[TheEPROM9]

I have designed my own discreet logic driver for a VFD, the clock part has never been built but the display is now fully driven.

Here's a link to my website on the subject: http://www.eprom9.comeze.com/VFD%20clock.html





If you want more info on the project just message me. Voltage is 5V for the filaments and 12V for everything else.

[EEVblog]

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

I normally do that, but in this case must have forgotten the letter.

[deephaven]

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

I normally do that, but in this case must have forgotten the letter.

Thanks for responding, Dave, and for not flaming me. I still gave the video a 

[dentaku]

Bad form that Dave didn't acknowledge who sent it in (it wasn't me!). Whoever it was went to the trouble and expense of sending it in so at least Dave should have mentioned who it was. Being too busy isn't an excuse, it would take no time to stash these items on a shelf with the original sender's name on them.

I normally do that, but in this case must have forgotten the letter.

I don't think it's terribly important but if you wanted to, could you use the Youtube annotation tool to add "Mark Butterworth of Santa Clara California" at the point in the video where you say you can't remember where you got the VFD from?

[twice11]

Maybe the NEC microcontroller is an MCS-48 architecture uPD80C48 / uPD80C49?

http://www.cpu-world.com/info/Pinouts/8048.html

Do the data line inputs match the DB0-DB7 I/O pins 12-19?
Is the 12MHz crystal connected to XTAL1 and XTAL2 pins 2 and 3?

Yes and yes.

Good find, but likely, it's not the master processor, the 8048, but the slave processor, the 8041/8042. The pinout is quite similar - and we get the following connector pinout:

• Arduino 2 = /CS (Pull low to activate the interface)
• Arduino 3 = /RD (Pull low to read data from the 8042)
• Arduino 4 = Address (often 0 = 'data', 1 = 'command')
• Arduino 5 = /WR (Pull low to write data to the 8032)

So what Dave is doing while he writes the letters of "H" is toggling between the invalid pattern where /RD and /WR are both low (active) and the pattern where only /WR is active. This explains why in the end the rising edge of PIN 3 activates the read, while a write is typically initiated by activating (pulling low) the write pin: Deactivating (pulling high) pin 3 results in a valid write instruction.

[EEVblog]

Found the datasheet:
http://www.eevblog.com/files/Babcock-VF0240.pdf

[gslick]

Good find, but likely, it's not the master processor, the 8048, but the slave processor, the 8041/8042. The pinout is quite similar - and we get the following connector pinout:

• Arduino 2 = /CS (Pull low to activate the interface)
• Arduino 3 = /RD (Pull low to read data from the 8042)
• Arduino 4 = Address (often 0 = 'data', 1 = 'command')
• Arduino 5 = /WR (Pull low to write data to the 8032)

Yes, that makes more sense. Taking another look at the still image of the PCB card edge connector it looks like the control signals from the card edge connector go to pins 6, 8, 9, 10 of the 40-pin microcontroller.

With the standard 8049 pinouts that would be:
Pin 6: /INT (input)
Pin 8: /RD (output)
Pin 9: /PSEN (output)
Pin 10: /WR (output).
The direction of those signals wouldn't make sense for the usage scenario here.

With the standard 8042 pinouts that would be:
Pin 6: /CS (input)
Pin 8: /RD (input)
Pin 9: A0 (input)
Pin 10: /WR (input).
The direction of those signals does make sense here.

[AlphZeta]

I did a VFD clock project a while ago, and explained from how to properly drive a VFD filament to building the driver circuit.  You can find the series below:

Filament driver:
http://www.kerrywong.com/2013/06/05/vfd-filament-driver-using-555/

VFD driver:
http://www.kerrywong.com/2013/06/13/a-diy-vacuum-fluorescent-display-driver/

And the VFD Clock:
http://www.kerrywong.com/2013/07/01/vfd-clock-putting-everything-together/

[bktemp]

I did a VFD clock project a while ago, and explained from how to properly drive a VFD filament to building the driver circuit.  You can find the series below:

Filament driver:
http://www.kerrywong.com/2013/06/05/vfd-filament-driver-using-555/

VFD driver:
http://www.kerrywong.com/2013/06/13/a-diy-vacuum-fluorescent-display-driver/

And the VFD Clock:
http://www.kerrywong.com/2013/07/01/vfd-clock-putting-everything-together/

Looks a bit too complicated. All those level shifters around the ULN2003 can be easily avoided: Since you already have a floating filament supply, connect the center tap to a negative voltage instead of +5V. Now you can use simple PNP transistors directly driven by 5V oder 3.3V logic to drive the grids and anodes to the required voltage levels. To prevent ghosting you can add pulldown resistors from each anode/grid to the cathode potential. This is how it is done in many cheap VFD drivers using a single chip microcontroller/VFD driver.

[vlad777]

Why does the tungsten cathode have burn marks on one end? (and not the other end)
I am guessing it is metal atoms deposited on glass, but why?

[bundy]

Why does the tungsten cathode have burn marks on one end? (and not the other end)
I am guessing it is metal atoms deposited on glass, but why?

It is caused by the getter: http://en.wikipedia.org/wiki/Getter

[vlad777]

Why does the tungsten cathode have burn marks on one end? (and not the other end)
I am guessing it is metal atoms deposited on glass, but why?

It is caused by the getter: http://en.wikipedia.org/wiki/Getter

Thanks.

Damn, once I broke off a CRT electron gun and until now I wondered what this is:

[max_torque]

Found the datasheet:
http://www.eevblog.com/files/Babcock-VF0240.pdf

^^^ More of an "Infosheet" than a Datasheet really  ;-)

[vlad777]

Can someone help me out with this beauty?
I had it laying around for quite some time. But now I can use pic2550 with USB to control this.
Chip symbol is Rockwell International.
I don't know about that T_scale symbol, anybody?

Is LE latch enable?
What is VCT?

Is VFIL AC few volts RMS?

Many thanks.

(Made in USA, USA, USA, USA, ... )

PS:

P/N is product number. What is A/N?

[bktemp]

Is LE latch enable?
What is VCT?

Is VFIL AC few volts RMS?

VCT is probably center tap (center tap of the tranformer winding for the filament, if you do not have a suitable transformer, a simple voltage divider build from two 100 ohm resistors should do it for the first test). The filament voltage must not be connected to the -27V supply, because it is connected internally probably using a z-diode to float a few volts above the -27V.
Guessing from the size around 4Vrms. As I said on the first page, either measure the filament resistance and then adjust the voltage until the hot resistance is around 4x the cold resistance, or until the filament glows dim.
It could be either a complete controller like the VFD in Dave's video, or just a simple shift register.
LE (latch enable) sounds more like a simple shift register. RST maybe resets the grid counter to the first digit and each LE pulse stores the serial data and switches to the next digit.

[mikeselectricstuff]

Probably data,clock and latch, like a 74HC595. clock in all the data bits then toggle the latch pin.

[KD0CAC John]

I've got a unit that I took out of a commercial convection oven and trying to fix [ if not too much for friend ] or if a lot labor / parts , I found a replacement on Digikey site for $145.49 .
No engineering at this end just fix stuff if I can .
This does not light up at all , It had power to board close 5v , the SMD fuse was OK , the 2 sets of 3 pins at either end of board had just under 3 volts .
I tried to download PDF for some of the chips , but can not seem to get , some times just keep on getting linked to buying , even after clicking on PDF logo - just jumps to sales site ?
Any help ?
Thanks
John

[BoomBrush]

Hi Dave, I liked the video but it reminded me of the more conventional arduino displays, you know the 16 x 2 ones.

Anybody know how you would light up custom sections of a digit? Another way to word it is you dont want a character that exists in ascii, but maybe where you define the custom sections of it. Kinda like making your own font. Would that be possible with a display shown in the video? I would be interested to know.

eg if you wanted a digit where just only the outer sections of the digit were on.

Also Dave: good job explaining how these vacuum tubes work but why is a vacuum necessary? Would it be because it would be picking up the electrons from the air instead of the ones being emitted? So all the segments would always be on... is that correct? Thanks guys!

[bktemp]

Anybody know how you would light up custom sections of a digit? Another way to word it is you dont want a character that exists in ascii, but maybe where you define the custom sections of it. Kinda like making your own font. Would that be possible with a display shown in the video? I would be interested to know.

eg if you wanted a digit where just only the outer sections of the digit were on.

It depends on the instruction set the display supports.
If it only has a fixed character set, it is impossible.
Many VFDs use the non printable ASCII codes 0-31 as commands for changing brightness, setting cursor or changing font. Since most of those displays use a standard microcontroller it is easy to customize the instruction set and supported features by changing the firmware. Most of the VFDs I have seen do not support custom characters.

[JackOfVA]

Hi Dave, I liked the video but it reminded me of the more conventional arduino displays, you know the 16 x 2 ones.

Anybody know how you would light up custom sections of a digit? Another way to word it is you dont want a character that exists in ascii, but maybe where you define the custom sections of it. Kinda like making your own font. Would that be possible with a display shown in the video? I would be interested to know.

eg if you wanted a digit where just only the outer sections of the digit were on.

Also Dave: good job explaining how these vacuum tubes work but why is a vacuum necessary? Would it be because it would be picking up the electrons from the air instead of the ones being emitted? So all the segments would always be on... is that correct? Thanks guys!

If not evacuated, electrons emitted by the cathode would not be able to freely flow from cathode to plate (grid bias permitting) but rather would collide with gas atoms. Some would be ionized creating space charge, also not a good thing. 

Additionally, the hot cathode - in "modern" receiving tubes would likely be poisoned reducing emission. The heating element itself would oxidize and burn.

That being said, a few specialized tubes were designed to operate with specific gases inside the envelope. Voltage regulator tubes, such as OA2, might use a combination of neon and other gasses that would break down at a defined voltage. (The vacuum tube version of a Zener diode.) Some rectifiers were also made with a similar design, OZ4, for example.  And, high power / high voltage rectifiers uses some more exotic designs, such as mercury vapor (tube must heat up before HV applied so that the mercury vaporizes) or Xenon. 

For a high level view of vacuum tube design look at the first section of the RCA tube manual. www.tubebooks.org/tubedata/RC30.pdf  and the transmitting tube manual. w5jgv.com/downloads/RCA-TT-5.PDF

[Macbeth]

Maybe I've lost my puerile sense of humour...

But the cringe is strong with this one 

[vlad777]

@mikeselectricstuff

Hi Mike, I got it displaying some random stuff.
Didn't connect to PC yet, I am using debounced switches.

So here is what I got:
Reset is active low.
LE does  nothing.
Last bit on Data must be one , out of eight Clocks (low to high).

To get it started I have to click in eight ones.
00111111 is F
00110011 is 3
00001111 is clear
After inputting, next char is accessed.
Sometimes on random input random segments lit up within one char.

Does this remind anyone of some protocol ?
Many thanks.

[Len]

00111111 is F
00110011 is 3

Guessing from these two examples: If the first 4 bits are 0011, the next 4 bits are displayed as a hexadecimal digit.

[vlad777]

00111111 is F
00110011 is 3

Guessing from these two examples: If the first 4 bits are 0011, the next 4 bits are displayed as a hexadecimal digit.

Seem so but if last bit isn't 1 , nothing happens.

Edit:

Actually you are correct , must have been my clicking finger  .
0011 seems to be an opcode

[vlad777]

@bktemp

Filament cold resistance is 15 Ohm.
I am driving it with DC. Is that OK or will it damage something, in the long run?
At 5V DC resistance is about 50 Ohm.

[twice11]

@bktemp

Filament cold resistance is 15 Ohm.
I am driving it with DC. Is that OK or will it damage something, in the long run?
At 5V DC resistance is about 50 Ohm.

Typically, the filament is AC driven, so the average voltage between filament/cathode and the grid is the same over the whole length of the filament. With DC drive, you will get a brightness gradient. If the voltage at the negative end of the DC driven filament does not exceed the expected average voltage, you should be definitely save.

[nixfu]

Dave, two words:

Bus Pirate

[nitro2k01]

Dave, two words:

Bus Pirate

Bus Pirate is great for playing with serial protocols but not parallel ones. Adweeeno was a good choice for this quick hack.

[vlad777]

What kind of display would have five-two-seven groups of digits?
Any idea what was this displaying?

[blue-v]

Thank's for the video, Dave!

Speaking of reverse engineering these displays:
I'm struggeling for quite a while with an VFD in an HP Designjet 650C plotter.
First time the heater windings in the transformer were broken. I fixed that by adding a new winding to the transformer.
That worked a while.
Now the display went completely dark. One of the OKI C1162 had a short. Replaced that one but the display is still off.
Unfortunately you need the display for operating the plotter. And vice versa: The plotter needs the display too. Removing
it completely, the plotter hangs on startup.
I found out the display has a serial bus with two handshake pins. 16 bit are shifted in consisting of 8 data bits and 8 control bits.
Also status information is read out of the display. This explains, why the plotter hangs without the display.
Unfortunately I could not find any detailed information.

I really would like to replace that thing with a controller and HD44780 LCD. But I have not enough information about the display to build a replacement.
Does someone have the specs of that thing?
Or maybe an idea what might be wrong?
High voltage, heating and handshake to the interface are fine. But no data is shifted out by the controller to the tube.

Thank's!
Lothar

Here's some photos:

[bktemp]

Maybe the same happened as here:

The driver failed and shorted the high voltage and 5V together and this damaged the microcontroller.

[richfiles]

I picked up a pair of IEE Model S03601-40-040 VFDs from surplus. They are 2x20 character VFDs with HUGE 11mm characters! They were made by IEE, and feature an 8-bit parallel interface very similar to the VFD Dave has.

I contacted IEE, and unlike the data sheet Dave has, mine is VERY detailed... It's 17 pages long, with a complete listing of all the characters and sets, all the control codes, timing, modes, and everything.

[aargee]

I've got one of these that I'm thinking of putting to use some how. All the VFDs are run by OKI shift registers, as are all the push buttons.

A bit of work involved...

[blue-v]

bktemp,

yes that might be the case.
The OKI C1162 I replaced, shorted the high voltage to nearly zero.
But with the chip still in place I measured the other voltages and they were just fine.
The inverter got a little warm (not hot) and also the defective OKI had some temperature.

After replacing the OKI all the voltages are within expected ranges.
And the 8052 controller seems to work, because the plotter recognizes the display and starts up fine.
It's just, the controller doesen't output anything to the display side.
Of course I also would replace the 8052, but at the moment I have noreplacement handy.
I think I'll desolder the thing and try to read out the ROM contents. Hope that is still possible.
Atmel produced a AT89C52, which is compatible, but comes with a flash on chip. This will replace the 8052 just fine.

The thing is, I'm fed up with this non-reliable VFD board and would really like to replace it with a common LCD.
It's the second time now the thing breaks!
Also it was already very pale before the OKI chip died.

Is there a way to refresh faded VFDs, btw?

Lothar

D'oh!
I just read the 8052 has a security bit that prevents readout of the ROM if set, which it most likely is. 

[MLXXXp]

Dave's video prompted me to dig out an old VFD I've had for over 20 years, sealed and unopened in the original packaging. A date stamp on the back reads May 16 1989.

It's made by IEE, part number 03601-82. It's a single line of 20 characters and each character is a 5 x 7 dot matrix. I never had any documentation for it and a web search for that number didn't turn up anything useful.

However, I did find a document for the IEE 03601-86-080R, which is a 2 line by 40 character VFD. Both displays use the same single row 12 pin connector. By doing some signal tracing, similar to what Dave did, I determined that the interface was very likely the same. It requires a single 5V supply. The protocol is write only 8 bit parallel ASCII with a strobe, and there's a busy output.

I wired it to a breadboarded Arduino Pro Mini. Even though the Arduino has enough I/O pins to connect everything directly, I decided to also use a 74HC595 serial to parallel shift register, just for the heck of it. I didn't have the proper female connector so I hacked up a PC floppy cable and wired it to a header strip.

I wrote a quick sketch to send to the display anything received by the Arduino serial port. It worked a treat! 

[richfiles]

I've got one of these that I'm thinking of putting to use some how. All the VFDs are run by OKI shift registers, as are all the push buttons.

A bit of work involved...

I'm DROOLING!!! I know EXACTLY what I'd use that board for! KERBAL SPACE PROGRAM!!!

Orbital Altitude+Radar Altimeter
Apoapsis+Time to Ap, Periapsis+Time to Pe
Orbital Velocity+ Vertical Velocity, Inclination+Eccentricity
...and more!!! That only used the big top display and half the lower ones, if I multiclass! 

I could slap my semi major and minor axis up there, or my coordinates, or my vector heading, or air density+intake air for jet engines...

You're making me drool! Can I ask what the board was from, or if there is a manufacturer or model number on it... Something like that might be worth tracking down! That board's almost ideal for a Kerbal Command Module controller! VERY cool find!

[Gryphonmeh]

I was watching one of your old mailbags, #686, in which you received a Electric Imp Internet of Things. I was thinking what if it could be adapted to be used with the vacuum display. Possibly?

[aargee]

Richfiles,

The board is from an XRay machine control panel originating from a large dutch company  . Pretty hard to get one surplus I imagine. This one is faulty, but not sure what the fault is as I'll have to try and power it up to check it out. Absolutely no detailed info, no info on VFDs, no schematics, etc.

Good luck with your search.

[richfiles]

Richfiles,

The board is from an XRay machine control panel originating from a large dutch company  . Pretty hard to get one surplus I imagine. This one is faulty, but not sure what the fault is as I'll have to try and power it up to check it out. Absolutely no detailed info, no info on VFDs, no schematics, etc.

Good luck with your search.

Hehe! Thanks for getting back to me. Yeah... That's not likely to be easily found... NICE score though! 

I happen to have a pair of the large 2x20s (parallel) that I posted about above, and another pair of 1x16s that are much smaller (parallel with a serial board attached). I've got a pair of POS displays hiding out somewhere in my shop. I forget if they are 1 or 2 line, or if they are 16 or 20 character displays, but I do know those have a serial cable leading to them. The honest truth... I've considered actually using nixies for my altimeter, with a toggle to select orbital or radar altimeters... Just the altimeter though. I only got so many of them, and I despise those lazy Russian ones with the flipped "2" instead of a real "5"... I only get good quality legit "5" nixies that originated from western or asian makers when I build stuff!

I got a drawer full of nice Rodan GR-111a nixies that were left over from the donor machine from when I restored a Sony Sobax ICC-600W calculator. I got about a dozen more (part number not determined yet) from the worlds' most "fragmented" Sobax calculator... I have never seen a case cracked into so many pieces, while still being "fully assembled"!  As a collector, I will only tear it down if I find the need to do so. As... full of character as it is, I don't have another of that particular model in my collection, nor do I have another machine by which to restore using parts from it yet.

Regardless, VFD character display are not uncommon, and can be easily found online... It's just so RARE to see a single board with SO MANY!

Aaaaand it looks like it's a Philips.