High voltage LED display

[gui_tarzan]

I want to built a high voltage display using 7-segment LEDs. The problem is I haven't been able to find a module capable of 600v that's less than a couple hundred dollars. Most of the modules are 30v or less. Does anyone know of a module, reader, whatever, that will handle the high voltage and allow for LED readouts?

[Jana]

What is it that you want to do? If you want to read voltages, then just get a 0 to 5 volt display or whatever range is reasonable and scale the voltage to the input with a resistor divider.

[gui_tarzan]

I'm building a tube matcher and I want to display plate, grid and bias voltages. I didn't know if a divider would still give me accurate readings or not. I checked with Mouser, Digikey and Newark and none of the techs there knew of one that would read that high either.

[romberg]

I didn't know if a divider would still give me accurate readings or not.

Your DMM probably uses a voltage divider when measuring high voltages. I would image using precision resistors in a voltage divider would be good enough. Or you could incorporate a trim pot as part of the divider and then calibrate it using your DMM.

Mike

[gui_tarzan]

Yessir, you are correct. I will give that a try.

[R.G.]

Welcome to the can of worms labeled "Accuracy".

Accuracy is expensive. Your first task is to definitely decide what minimum accuracy you need. If you can make do with an *indication* instead of a measurement, you can get by much more cheaply. For instance, if you want to measure to 600V, but most of them will be in the range of 450-500V, and you don't mind if an indicated 500V is actually, say, 450 or 500V as long as it's consistent, that will be inexpensive. Wanting a reading of 500V to be uncertain to between 499 and 501 V is going to cost you.

And it's easy to confuse accuracy with resolution. Accuracy is how close to the real, true, known-only-to-God value the measurements are. Resolution is how finely you can measure, whether it's accurate or not. If the real, true value is 493.5742V, and you can measure it to 480, 490, or 500, then your resolution is 10V, but if you measure it at 490 every time, your accuracy is perfect. If you can measure it to between 485.1V and 501.3V, and you may get any value in that range, then your resolution is 0.1V, but your accuracy is - well, disappointing.

Worse yet, humans confuse the number of digits with both resolution and accuracy. Just because you can read 495.348V does not mean the real, true, known-only-to-God value is that. Your meter may be lying to you.

But you knew that, or you would not have asked the question.

Some fundamentals:
- digital displays cannot be trusted to be closer than +/- one least-significant digit - ever
- analog to digital converters have many kinds of errors; they may have nonlinearity errors, offset errors, and missing codes; this last means that the thing may **never** return certain values at all
- even if the digits are perfect, you are down to how accurately you can supply an analog reference to the converter, and how accurately you can scale the input voltage to be measured, which is what you were wondering about.

Sizing scaling resistors is tricky. Resistors come with a tolerance. That tolerance is a statistical indicator. What "10K +/-1%" means is "we will guarantee that if this resistor is actually outside the range of 9.99K and 10.1K, we will replace it as being defective". The actual resistor you get may be any value at all between those two values and it meets specs. There will be a few resistors in each batch that fall outside the range, as some defects slip through.

There are two competing effects going on. With only two resistors involved, you can compute the possible variation of results in a divider by taking each resistor to its worst case limit, then relying on not getting a defective one. The more resistors you add, the more they (a) spread the possible results because of their worst case tolerances and (b) narrow the spread by tending to lie close to the center of their tolerance by statistical averaging. Computing tolerances for complex divider networks gets complicated.

Many digital meters have a 2V or 200mV range. If you are scaling 600V to 2V, you have a problem with reading the meter, as you can only use one less digit than the meter already reads, as the biggest digit can only be a 1 or a 2, not a 6. So you likely have to buy an extra digit just to get the most significant digit to be a 6. Otherwise, you have to buy a meter with a full N digits, not N 1/2 digits, as the half digit is the one at the MSD side.

And you cannot read it to finer than +/- least significant digit.

So this gets complicated fast. And we haven't even talked about how close the reference voltage inside the meter is to the nominal value it was designed for. Most countries have something like a national bureau of accuracy and standards or some such. In the USA, it's NIST. When your meter says "calibrated and traceable to NIST" it's about as good as an individual can get, and probably more than an individual would pay. Back at TIC (Three Initial Corporation) they supported a many-person calibration lab just to calibrate all the measurement devices for development to some reasonable - and traceable - NIST standard. This kind of calibration is *required* for ISO9001 certification, by the way.

So the pool is deep. How much accuracy do you really need, and how much extra are you willing to pay for above that?

[Structo]

I bought a cheap AC voltage digital module from China a couple years ago.

I use it to monitor the wall voltage on my line conditioner.

Check eBay.

[Jana]

I stick my tongue on 9-volt batteries (I am calibrated to + or - 1 volt, although I have been drifting due to age).

[JMFahey]

http://www.adafruit.com/product/705

[img:970:728]http://www.adafruit.com/images/970x728/705-04.jpg[/img]

As is, measures 0/99.9V ; with a 10:1 resistive attenuator (2 resistors) measures 0/999V DC

Needs 4.5 to 28V DC supply.

Oh, I forgot:

$7.95
IN STOCK

[gui_tarzan]

R.G., you are absolutely right. Fortunately we're dealing with guitar amps that don't require precision measurements. If it's off 2, 3 or 10 volts either way (on the plates and grids) it won't make any difference in the end product, the most important thing is that I will be able to load up four tubes and check their readings with the flip of a switch.

When it comes down to it, how accurate are our meters? I'd like to think they're pretty darned close, and close is close enough in this situation.

[R.G.]

Good. Defining what accuracy, resolution, precision, and repeatability is indeed the first step. And for guitar amps, 2, 3 or 10V is not needed as absolute accuracy, and three digits is plenty of precision. Repeatability is important.

Freeing yourself to be "close enough" is an important bit of design work. It lets you know when you're done!