Need recommendation for EPROM eraser

[brouhaha]

(09-07-2023 07:38 PM)JoeG Wrote:  I did find this video on using a UV LED to erase a eprom:
https://youtu.be/m0y9IO95PRU?feature=shared

It looks interesting. I just don’t know the detailed specifics of what he ordered, and what power source he uses (milliwatts?). So I asked those questions via YouTube.

I can't speak for the specific LED he used, but let's take as an example the Luminus Devices XST-3535-UV-A60-CE270-00, which costs $35.41 in quantity one from Digikey Electronics. This is a surface mount component, 3.65mm on a side, so probably isn't well-suited for simple tabletop experimentation, bur can serve as an example for calculations.

The peak wavelength is 270nm, and the radiation pattern is a 60 degree wide cone. The maximum current is 800mA, and the forward voltage drop is 5.0V min, 6.45V typical, and 7.5V max. Note that a heat sink (large copper area on PCB, or an aluminum core PCB) will be required to operate at maximum current to keep the LED junction (inside the package) below the maximum rating of 100 degrees C. Also, radiant flux (intensity) decreases almost linearly from 25C to half at 100C, so keeping the LED temperature down yields best performance.

For a fixed voltage power supply of 12V at 1A max, we need to compute the necessary series resistance to ensure that the current does not exceed 800mA. For a regulated 12V power supply, which should have a current rating of at least 800mA, say 1A to provide margin, and assuming the typical voltage drop of 6.45V, the resistor will have to drop 5.55V. By Ohm's Law, the resistance necessary for that, R = E/I = 5.55/0.8 = 5.625 ohms. The resistor will have to dissipate P = R*I^2 = 5.625 * 0.64 = 4.44W, so that requires a power resistor. One could use a power rheostat to get close to that resistance, but in practice one uses the next higher standard value power resistor, which would be a 7.5 ohm +/-5% 5W resistor, such as the Ohmite 805F7R5E. The resistor will need to be mounted to a heat sink.

Now we need to do the calculations using worst-case values. Assume that the power supply regulation is 5%, so it could be as low as 11.4V or as high as 12.6V, the LED drop could be as low as 5.0V, and the 5% resistor could be as low as 7.125 ohms or as high as 7.875 ohms. If we run the calculations for those eight cases, we find that all cases with the LED drop at 5V result in over 800mA of current and 5W power. Further, even at the nominal 6.45V drop of the LED, if the power supply is 12.6V and the resistance is 7.125 ohms, there will be overcurrent and overpower.

What this means is that to satisfy the current limits, we have to go to a 10 ohm 7.5W or 10W resistor. The higher resistance means that we might get current as low as 371mA, running the LED at less than half brightness.

The best way to deal with this is to use a voltage regulator configured for current regulation rather than voltage regulation. This can be done easily with three-terminal linear regulators like the LM317. The load regulation will be within +/-1.5%, so we actually set the current limit to 788mA or less. The regulator will dissipate the same power the series resistor did in the simpler method above, so worst case it will dissipate up to 7.6V * 800mA = 6.08W. The TO220 packaged LM317 can do that if it has really good heatsinking and maybe forced air cooling.

The same could be done with a switching regulator, and will dissipate much less power, but the circuit design is much more complex. There are also dedicated LED driver chips that may be useful; some use an external pass transistor which would be necessary for the described case.

For running an LED at significantly lower power, e.g. 100mA, the power dissipation issues would be less extreme, but would still need to be analyzed.

Note that the LEDs are binned into multiple smaller forward-drop voltage ranges. When you buy in small quantity from a distributor, you don't get to pick the bin, but if you buy in large volume you may be able to.

A spreadsheet I've used for this example is attached. I did this in LibreOffice, but had to save to Microsoft .xlsx format in order to be able to attach it. The spreadsheet should work in Microsoft Excel, but I'm not sure whether the conditional formatting that makes overcurrent and overpower cells show as red background will work.