Nixi Tube vs. LED latency

[noreply]

I recently have seen a video on the characterization of a nixi tube – where its ‘on’ and ‘off’ latency was measured with a high speed video camera (we are talking about camera capable of 1M frames / sec)

It was interesting to see this – because a visual source (the camera) was used instead of some electronic test equipment (ok – you could say the camera is a form of electronic test equipment)

When the nixi tube was ‘warmed-up’ – the ‘on’ latency was extremely low – you could switch the ‘on’ at 100KHz with no problem

The ‘off’ latency was much longer – however this was caused by the ‘decay’ of the emission of photons – once the charge (voltage) was removed from the digit in question. It still persisted – but the ‘on’ of another digit will drown out any persistency – so in effect the ‘nixi’ could be driven at the ‘on’ latency alone – at least 100KHz

The background for the above test with nixi tubes – was to develop a ‘digital’ clock – accurate to at least 1/100,000 of a second – to calibrate high speed cameras.

In essence – the camera ‘films’ the digits of the clock (the nixi tubes in this case) – which appear to be ‘on’ continuously (at least the digits representing < 1/10 sec) – but in reality are being switched at 100KHz – so we can resolve 1/100,000 of a second.

The project was very interesting and proved to succeed in what it set out to do.

In hindsight (other than the beautiful ascetics of the nixi tube) – purely from the test results – why nixi tubes and not , say 7 segment LED’s?

This made me think about the latency of a LED (typical) and it is much lower than the 100KHz of the nixi – so is there something , not so obvious, that prevented LED display from being used instead of the nixi??

Has anyone ‘clocked’ an LED 7 segment at > 100KHz – and checked if all segments are illuminated at the clock cycle (not sure how you would do this without a high speed camera – unless you use some photodiodes at ‘each’ of the 7 segments of the display)


So, some questions for discussion ...

How can we 'test' latency of 7 segment LED display using , say, oscilloscope and AWG ??

How could you test the 'on' and 'off' latency of a nixi tube - electronically (using oscilloscope and AWG)- rather than a high speed video camera??



Any feedback or comments would be appreciated 

[NiHaoMike]

A lot of LED displays are multiplexed. But I don't see why you couldn't use one that isn't.

[schmitt trigger]

It would be very interesting to build a fast photodiode preamp, where one could shine the light from an LED being clocked at successively higher rates.

I can tell you though, that many years ago I attempted to build a stroboscope with some Luxeon 3W white LEDs. The phosphor latency rendered them unusable for such task above 100 Hz approximately.

This is a photo of bouncing marble at a rate of 50 Hz, and one can already see a little motion blur.

[rdl]

NASA used LED displays in some of their cameras filming the Space Shuttle. You can see them in this video. At some point they explain what the numbers represent (I think, been a while since I watched it). The video has some great views of the shuttle during launch.

[noreply]

The phosphor latency rendered them unusable for such task above 100 Hz approximately.

Yeah ... precisely

I suspect that despite an LED having a very low latency (the diode 'on' time) - when it comes to the actual 'visible light' - there might be other issues especially when dealing with optoelectronic devices - like you mentioned 

The NASA video - and the 'LED' time stamping in the camera - falls short of the 100KHz - but perhaps they did not need this (100KHz) resolution??

There might be 'more than meets the eye' as to why nixi tubes were used for the clock - looking at it from a purely electronic point of view - not withstanding that the guy who made the clock also hand makes the nixi tubes as a business.

Below is the link to the video which inspired me to make this post, as it raised some interesting questions – which were not covered in the video ...

[StillTrying]

It would be very interesting to build a fast photodiode preamp, where one could shine the light from an LED being clocked at successively higher rates.

I've looked at the light shapes from some fast flashed LEDs here.
https://www.eevblog.com/forum/chat/20w-halogen-bulb-viewed-by-a-photodiode/msg2411274/#msg2411274

Photo diode's linear BW was around 22 MHz.

The delay and rise times for phosphor LEDs to be fully on or off varies from about 200ns to a few us.

[Siwastaja]

LEDs are typically specified between about 10..40 ns, measured between 10%<->90% intensity. This means, up to some 10MHz you get very sharp edges (relative to the period). I have modulated LEDs up to 20MHz in time-of-flight systems. This is close to the limit, for faster switching you need to consider laser diodes. This is for bare LEDs, i.e., no phosphor.

White LEDs have a phosphor which will be slower than this, but can still be surprisingly quick, 100kHz for example usually isn't impossible but it will have some off-time glow.

[ve2mrx]

If you want to use phosphor LEDs, build a bargraph and time on the brightest LED... The decay wouldn't be as important?

Of course, native colors will respond much faster!

If you want something funnier, use a synchronized color wheel ;-)

Martin

[james_s]

I don't really see this being an issue, most 7 segment displays do not use a phosphor. I'm not fond of the ones that do, one of the products a former employer made used white 7 segment displays in the front panel clock and within only about 6 months you could tell which segments were most used as they were starting to dim and turn bluish gray.

[jmelson]

We had a guy doing a project to measure some aspect of scintillating crystals, and he needed a fast, pulsed light source.
He fiddled around with UV LEDs and avalanche transistors, and eventually built an optical pulser that could give a pulse less than 1 ns wide, with 100-140 ps rise and fall times.  He was pumping several amp pulses into the LEDs that lasted only several hundred ps.

So, LEDs are actually quite fast.

I built a laser photoplotter, it uses a 5 mW 670 nm laser to record images on film at 1000 X 1000 DPI, in raster mode.  The pixel time of the laser is 5 us.  The turn-on and turn-off of the laser is so fast, there is no motion blur of the pixels.

Jon

[james_s]

I built a laser photoplotter, it uses a 5 mW 670 nm laser to record images on film at 1000 X 1000 DPI, in raster mode.  The pixel time of the laser is 5 us.  The turn-on and turn-off of the laser is so fast, there is no motion blur of the pixels.

That sounds quite interesting, is this project written up anywhere?

[jmelson]

I built a laser photoplotter, it uses a 5 mW 670 nm laser to record images on film at 1000 X 1000 DPI, in raster mode.  The pixel time of the laser is 5 us.  The turn-on and turn-off of the laser is so fast, there is no motion blur of the pixels.

That sounds quite interesting, is this project written up anywhere?

Not as a complete set of plans, but I could make a bit more info available to anyone who is interested.  One of the issues is it was completed in 1996, so used all parts in common use in 1995 and earlier.  Ie. "old school".  Now, it is just another one of the machines in my shop.  But, if I was going to design it from scratch now, I'd use a lot of more modern parts.

See :    http://pico-systems.com/photoplot.html     For some more info and a picture.

It does a VERY accurate 1000 x 1000 DPI, which is required if two films were to be used for a double-sided PC board, or to make a solder stencil for a commercially-made board, which is what I now use it for mostly.  I did make a few boards that were 7 x 11 inches, and the stencil lined up perfectly with the commercial PCB.

Most of the effort actually went into the program to convert Gerber files into raster format.

Jon

[james_s]

Ah, it's drum based, cool. I was picturing more of a classical XY plotter style thing.

[jmelson]

Ah, it's drum based, cool. I was picturing more of a classical XY plotter style thing.

In the interim, while I was fooling with all the complexities, I added a "light pen" to my big Calcomp plotter.  Itr DID work, but with a 0.013" writing beam, it took FOREVER to complete a single plot.  Somewhere in the 45 minutes to over an hour range for a larger board.  Then, for a PC board, you'd get black traces on a clear background, so you had to then contact print it onto another film to get clear traces on black background, for the photoresist.  So, that really wasn't very workable.

The drum plotter produces 0.6" of plot per minute, no matter the complexity.  So, a small board can be plotted in 5 minutes, a bigger one takes maybe 15 minutes.

Jon

[james_s]

I was contemplating ways to make a photoplotter at one point, I forget what I was even trying to do but it occurred to me that plotting on a sheet of xray film would be a convenient way of doing it. These days I wonder if the easiest route would be to just use a TFT panel pressed right up against the film and give the backlight a quick flash. Maybe removing the backlight assembly and using a point source would give better results. Some of the displays used in larger phones and tablets can do some impressively high resolution and in many cases they are crazy cheap.