Pattern of HV laser power supplies burning out

I am at a makerspace where we currently have three CO2 lasers, a 150w BossHP3655, a generic 90-100w, and a generic 60w. They get a lot of community use.

Over the last two years, we have had four HV laser power supplies fry out on us, two of those in the last month. Three on the Boss, and one on the 90w. (Well we actually determined that the most recent power supply failure on the Boss was actually a failure of the insulation on the power cord attached to the supply, so we were able to cut the wire before the break and buy and install the fix. But it still seems related.) We are starting to suspect this might not be just crappy replacement parts, but perhaps our electrical system could be contributing to the problem. We are in a 100 year old building and run a lot of power hungry equipment in various departments, including kilns and a central dust collector.

A volunteer who knows somewhat more about electrical systems than I do did some basic testing with a multimeter, turning items on and off to see if we were getting any spikes in the voltage. We only got a variation of maybe a couple of volts when turning something on or off. The voltage was on the upper end of normal––around 121v––though the volunteer had tested the multimeter at his home workshop before helping me out, and it measured about the same there. Maybe the multimeter is not entirely accurate and runs a bit high?

Of course we could be getting some other types of infrequent power fluctuations perhaps caused by utility work. Or some spikes that are so quick that we can’t see them on the multimeter. We’ve also got some other testing we hope to do and are trying to get an actual licensed electrician in to look at our system.

Meanwhile, here are some specific questions that I would love to get some opinions on:

  • What is the typical upper safe voltage range for a 115v laser power supply? Would it be in danger of killing one at 125v? 130v? 150v? 220v?

  • Might a power conditioner have any downsides? Even if we’re barking up the wrong tree it might not be a bad idea to cover the possibility.

  • How fragile have others found these power supplies to be? Would we need to have something going seriously wrong to kill one?

  • I presume four dead power supplies on two machines in two years is well out of the ordinary.

  • Any chance a massive laser power supply failure could result in a broken tube? We have a seven month old tube that had a very nicely shaped beam (revealed by a sharp mode burn) and ran about 90w output. Just a couple weeks ago it dropped to about 65w and a still decent but not quite as sharp beam. By a week later, the mode burn test looked nothing like a gaussian distribution. Taking a smartphone video looking into the laser nozzle, I was able to see some sort of damage at the far end of the tube that does not seem to be visible from the outside. Could this be caused by its power supply frying?

Thanks for any help.

IMO, that kind of duty cycle burns through equipment lifetime in short order. A supply (optimistically) rated for 10000 hours will last twenty years (*) for my usage, but maybe four years with yours.

Worse, if the machines are of a similar age, then they’ll all start failing at about the same time.

The power company delivers 120 VAC with a ±10% tolerance at the service entry, so anything between 108 and 132 VAC is just fine by them. Typically, it’ll be a lot tighter than that, but small variations are no big deal.

So power variations aren’t likely to be much problem.

If memory serves me right, @jkwilborn had a tube fail about like that, going from “just fine” to “junk” in a matter of days. Tubes have a definite lifetime due to gas deterioration that happens as a result of elapsed time as well as duty cycle, so this may be just another end-of-life event.

With that in mind:

  • Typical upper safe voltage range: not to worry
  • power conditioner: ditto, particularly for the size you need
  • How fragile: not particularly, but you take your chances
  • well out of the ordinary: not really, as described
  • result in a broken tube: a different problem

Fix 'em as they break and move on …

(*) The electrolytic caps will die long before that and, alas, so will I.

The limiting factor in dc excited co2 is the response of the lps. A faster lps will allow faster speeds, at minimum.

When an lps is tested they pass/fail on the basis of does it reach 90% of the placarded voltage <= 1mS.

Here is a comparison of a pair of 100W lps from Cloud Ray. First link is the dy-13…

Here is the specifications for each… the left is the dy-13…

I think you can figure out which has a much faster response time…

Good luck…

:smile_cat:

You may want to investigate the internal components of the broken power supplies to determine what might have broken. This investigation should provide valuable insights into the possible cause of the issue.

1 Like

I know this thread is from several months ago, but I was revisiting it for other reasons and realized that I never asked my dumb question of you. You included a screenshot of two different power supply specs and commented: “I think you can figure out which has a much faster response time…” I don’t have a prior background in electrical systems, so I’m afraid you’re giving me too much credit. Could you please explain to me which is faster and why? Would it be the one with the lower max voltage, because it would take longer to reach 40kV than 28kV? Thanks.

Or is the other one faster because it can reach a higher voltage in the same timeframe?

If you don’t understand something, please ask… we don’t learn by osmosis…

LPS specification state that is must reach 90% placard voltage <=1mS.

If your tube lases at 25kV (90% value), it takes the other lps the same time to reach 36kV. So it’s close to twice as fast.

Make sense?

:smile_cat:

1 Like

Thanks so much for that! So does the DY-13 actually operate at 40kV when running the laser? If so, I’d think that would change the amperage needs for the tube relative to a ~25kV power supply. But as I said, the electrical systems are by no means my specialty.

These lps are for >100W glass tube or dc excited lasers… You need a more voltage to get them to lase because of the increased tube length.

Glass co2 tubes will draw as much current as the lps will allow, so the lps has to limit maximum current.

These are negative resistance devices … Ohms law doesn’t apply.

:smile_cat:

So, by this logic, would a vastly overpowered LPS for any given tube give a faster response time, as long as you limited the actual amperage output to levels safe for that tube?

Also, I wonder about the calculations, given the information input of “≤1ms” for both of the LPSs. How much less than 1ms? Is it possible that the lower voltage supply reaches 90% power in 0.4ms while the higher voltage power supply reaches 90% power in 0.8ms? The lack of precision in that number seems to obscure our knowledge of the actual response time to reach a given voltage.

Not trying to be contrary, just trying to understand some of these finer points. Any insight you could share would be valuable. Thanks!

I did some measurements on power supply bandwidth a while ago:

In round numbers, the PWM demodulation filter limits the bandwidth to about 200 Hz, with a 1.5 ms risetime. I very much doubt both of those numbers, but they’re in the right ballpark.

Because the filter cutoff probably doesn’t change with the laser output power (because that would affect the PWM demodulation), it suggests the power supply’s peak output voltage (probably) doesn’t have much to do with the final result. Yes, there’s more oomph available to fire the tube, but the signal after that point can’t change the results any faster.

The linkies in that post lead into a rabbit hole of Moah Datah …

1 Like

@ednisley As you pointed out, there isn’t much point in varying the pwm period as it goes through the low pass filter. The response time of the pwm isn’t much help, IMHO… Unless you are changing the power, like a grayscale, the Ruida produces the same pwm output for the entire time that layer is executing.

This wasn’t the response time I was referring.

The L input has a limited toggle rate based on the actual response time of that lps.


This is has never been mentioned in any of the documentation that I’ve seen… There are lots of questions in a lot of areas besides the lps that are not and probably won’t be answered.

I suspect if it makes 1mS specification, then it’s ok as far as they are concerned.


For example, if we take the basic lps response time of 1mS and use it for a basic evaluation of speed … or how easy it is to outrun an lps response time.

At 1000mm/s your response time limits the control you have to 25.4 dpi, since you are covering a mm in 1mS… that is the best resolution you can get…

At 500mm/s your best resolution is twice the previous or 50.8 dpi, you now have the time to turn it on and off twice within a mm.

250mm/s is 101.6 dpi… not gaining much are we?

The only way to really know is to know what your specific response time is…

There isn’t much we can do about it either. China will never give us much information about their products… It’s tough enough to get a legible manual from them, let alone connections and technical information.

I’ll get off my soapbox…

Have fun

:smile_cat:

1 Like

Aaaand the tube current doesn’t have a nice clean rising edge, so I think the “≤1 ms” spec applies to a resistive test load, not an actual CO₂ laser tube.

For example:

The traces:

  • Yellow = X axis DIR signal for scope sync
  • Magenta = L-ON = negative-active Enable
  • Cyan = IN = analog voltage 2 V/div
  • Green = tube current 10 ma/div

Over on the left side, the L-ON signal goes low with the IN signal high for the first bar of the test pattern. The tube current jumps from 0 mA to a huge off-screen-high value, then rings like a bell for maybe a millisecond before settling down.

Yeah, we all switch L-ON for a millisecond or two, but the tube current has nothing to do with any reasonable assumption based on the good old single-pole risetime we all think should describe it.

So the PWM (and analog) inputs have a low bandwidth and the Enable signal isn’t much, if any, faster.

We just assume everything works the way we think it should, but the truth is that we’re just putting up with whatever it does. :grin:

1 Like