The "Golden Rule" of MOPA settings

Dannym · May 13, 2025, 1:28am

TI worked on a “magic formula”. I found the biggest secret of making the MOPA work on MOST things is spacing your LI and pulse SPACING the same at about 0.7 of the spot size. Further apart and they have gaps. Closer and the overlapping pulses create a high localized surface temp and hot metal cuts WORSE. The galvo mopa “cuts” with explosive pulses that cause a particle of surface to fly off, often as a spark. If the metal gets near its melting temp, it becomes soft and elastic and pulses get absorbed as heat but no spark flies off. If you actually cause melting it will flow into the channel you’ve created and fill it in.

It’s not just metal. Ceramic chips fly off when cold. If it’s hot, little or nothing flies off, and it may melt into the cut.

Speed is always of value. Max you can get accurate performance at, always.
Heat is your enemy. Lower overlap minimizes heat burden per particle of material removed in a pulse.

First, make sure your timing is good. Doing a fill with no rotation and not using bidir is the least critical on timing, but ton/toff being off together in the same direction will offset the whole thing. If one one- ton or toff- is off, then the cut will be wider or narrower than commanded. I’d actually done the cal for vectors and thought I had it correct, then made up a new test for raster timing and found mine was totally wrong. I really need to investigate this, either I did that vector timing right but LB or this whole system may be doing something different when I changed from vector to raster, or maybe it’s different with different q-pulse types? I really have to investigate thoroughly soon.

So this is the “golden rule” of MOPA overall:
speed=max
LI is ~0.7 * spot size
Freq=speed/(LI), this makes the laser fire with the max spacing it can do without leaving gaps by using the same spacing as between the lines. e.g. your LI is 0.005mm, speed=7000mm/s, use 1400KHz. This spacing may leave little scallops on the edge. You can increase the freq further but material heat increases a lot here so you will have to hold back a lot to limit it and it all gets slow. LB can do an amazing fix here if the even passes offset by 1/2 pulse. It will make a smooth edge without actually making pulses overlap on one pass. And it will get even better if you can do it with 3 steps of 1/3rd of a pulse in groups of 3 passes. You would decrease the freq so each spot doesn’t even touch the adjacent ones within a pass thus no heat overlap

And in that same vein, LB needs to be able to do the same by line, “interlace” the passes. e.g. I have a 0.007 spot size and generally need to use a 0.005 LI. But I may want to get it smoother by reducing LI, but the overlapping heat makes that impossible. Interlacing would be to say, like, an LI of 0.002 but with 3x interlacing it will do one pass at 0.000, 0.006, 0.012, then 0.002, 0.008, 0.0014, then 0.004, 0.010, 0.016. So ultimately you do have lines 0.002 apart but on each pass they don’t even overlap so no overlapping heat patches.

Anyhow, nexy, go to your specific laser’s performance chart and pick the LONGEST q-pulse with a cutoff above that freq. This is your best starting point. Don’t have the performance chart? Oh… well, you totally MUST have that to make settings or you’re just shooting blind and will get nonsense results from testing. Get it from the mfgl

100% power always. If you need to reduce the output, I don’t think this is the way to do it. It will reduce amount of heat burden delivered to the stock, but also reduces the explosive shock that actually removes material more. The removal-to-heat-burden ratio is key, but reducing power makes that ratio worse.

I have a JPT 300W though. That’s a beast. If I do this, it will often melt anyways. Even if I increase LI and pulse freq spacing so much they have a gap between and thus unacceptable.

I found the next step to take is not all that intuitive- you need to keep the LI/pulse freq spacing meeting that criteria. We’re already rastering as fast as the head allows, no gain there. Reducing power will cost more in removal rate than it gains in heat load improvments.

Therefore, the next step is to reduce the Q-pulse to a shorter type. It works. To engrave a coin, usually the q-pulse has to be dropped by several types to get below the melting point. Sucks but any other measure just makes it worse. Either drastically slower cutting, or melts even easier.

One thing I did discover- I could melt a coin in just the wrong way and create a really shiny mirror in some spots. Then even if it cools down and reduce the settings, that spot does NOT want to cut at first. The surface condition matters. I’m seeing about an initial “roughing” pass to first, esp for stainless. Or maybe the other way, go slow, lowest q-pulse type that actually makes a mark, and with a freq signficantly higher than the spot size, to smooth it.

Dannym · May 20, 2025, 9:04pm

More view-shifting discovery.

The best removal-to-heating may actually be kind of the opposite. I’m not one to dogma on a single concept.

look up the LONGEST q-pulse you have. The most energy
Freq is going to control your net heat delivery.
The LI=~0.7 of spot size and “square spacing” (LI=pulse spacing) principle still holds. So Speed is not a variable, freq is.

I’m trying to cut 0.1mm wide, 0.2mm spaced teeth into a 0.5mm razor blade.
I want to keep the razor blade’s temper.
The teeth aren’t going to be very long, obviously, but this is a challenge. As the cut deepens, the teeth become more thermally isolated and can’t keep the temp down.
For my M7 300W, 110mm lens, 500us q-pulse, 40KHz, 0.02mm LI, then speed must be 40,000 * 0.02 =800mm/s

I think I need to be doing an offset cut FIRST to make a trench immediately around the tooth, like 10% of the thickness, then alternate by rastering out 5% and adding another 5% to the trench.

The weirdest, most frustrating thing is LIGHTBURN’S OFFSET FILL DOES NOT DO OFFSETS. NOR DOES FILL. Only “Line”, which is weird.

Correct me if I’m wrong. I probably am, TBH there’s precedent here, sure. Offset fill wants to loop from outside in, all the way to the center. Once it starts getting close to the center the heat and cut go kind of wild. I just want to do 3 fill spots in, 0.06mm, from outside-to-in to make a clean product edge and thermally isolate the cut zone, then raster 1 spot- 0.02mm- inward offset from the edge so it never actually brushes up against the product edge.

I can do this in LB by:

start with a closed shape of what needs to be removed on Layer 1. We’re not going to cut this one, it’s best to keep an untouched copy around in case we don’t like the offsets we created in step 2 or 3. You’d just wipe layers 2 and 3 and make fresh copies off layer 1.
Copy to Layer 2 and layer 3. PROTIP: ctrl-c then ALT-v then immediately click layer 2 without clicking anything else. Then hit ALT-v again and click Layer 3. ALT-v pastes in the exact same spot, deselects what you cut and your pasted copy will be selected. Clicking Layer 2 moves that selected copy to layer 2.
Offset inward layer 2 by 0.01, delete original.
Offset inward layer 2 by 0.06, don’t delete original.
Offsets inward layer 3 by 0.04, delete original
OFFSET FILL layer 2. Step 3 creates an inner boundary so it effectively does (almost) what I want- an offset fill but only for a small distance from the product edge. The pulse spacing should be 0.5 the spot size or less to keep it smooth. This may necessitate reducing the q-pulse size and taking multiple passes, or going slower, because we’re overlapping the spots and combining heat. The goal is to get a smooth surface WITHOUT heat-affected zone effects. That heat can undo my carbon steel’s temper. For more general use, the heat of overlapping pulses can quickly lead to melting the substrate stainless steel
FILL layer 3. It’s been offset from the product edge so you throw you can put a of of energy into the cuts
Offset the raster so it won’t actually touch it the work edge. We can optimize the raster by using an LI With a of 0.7 to 1. Low overlap, heat doesn’t overlap, you can go boom-boom with the pulses. But you get faint lines and the edge of the raster can have scallops- but we’ve already done a finished cut away from the product edge and aren’t going to touch it. Go wild.

I’m using
Offset pass- 0 to 0.06 offset, 800mm/s, 80KH, Q100ns, 0.01mm LI, 4 passes
Fill: 0.02 offset in, 800mm/s, 40KHz, 250us, 0.02 LI, 4 passes 45 deg increments bidir (you must have your TON/TOFF timings exact, otherwise when it changes dir or angle they won’t line up

Then Repeat as much as needed

It occurred to me perhaps I could replace the Offset Fill with a series of lines. It saves adding a new offset profile copy in CAD every time I add new content to cut. But that seems even more net difficulty overall, since every time I change the cut parameters on only line, I need to change them all. If I were to do this I guess I’d make a cut setting library entry and “link” all the layers to use the current lib entry, so a change in the lib should update all those layers for you.

The thing is, it would be very useful if this were possible as Layer settings. This is pretty much becoming my go-to method, I started it while making crazy fine-trace PCBs.

We shouldn’t need to go into the design and make these copies and redo for EVERY cut we add.

Both Offset Fill and Fill need a STARTING offset for beam width (kerf). Let’s say beam spot size is 0.02, Offset Fill needs a starting offset of 0.01 then make 6 steps inward to 0.06 with 0.01 steps. A starting offset, ending offset, and stepover.

Well, for the bulk raster Fill, there’d only be one offset which includes beam kerf and enough to put the raster’s stop/start edge inside the Offset Fill area.

I did find Lightburn’s machine config limited the MOPA’s PWM to 20KHz by default. This might have sourced from the machine’s mark7.cfg, but might just be a default based on nothing but an assumption that a min like this is a good idea. I found that for blasting out 250ns pulses on the 300W had to be SLOWER than that to avoid overheating the stainless past what I needed.

A lot of this could be solved if LB can follow up on my Interlace feature request. Pulses could overlap but not at the same time. Fire all the odd pulses, then come back and fire the even pulses. With a minimum pass time (would be a great additional element if this were a layer setting) the surface heat from the first pass will dissipate before the second pass fires the even overlapping pulses.