JLCPCB Aluminum PCB manufacturing notes + USB connectors

[thmjpr]

Its hard to find some information on JLC aluminum process. For most cases the specs are the same as their FR4, but, I wasn't sure how it would actually turn out. Here is what I found.



Rules:
The two rules they specifically call out for Alu, in place of the regular rules are
- minimum 1mm drill
- minimum 1.6mm slot

Routing:
- Curved route path for an internal slot I put was almost perfect
- The outer route of the entire PCB was also very clean, just slight bump in where they started or ended the cut.
- Some of the holes had more burrs than others, none are bad but its still a sharp edge. No annoying FR4 fibers that I noticed.

Clearances:
- Appears to be the same as their regular FR4 spec.
- Plated holes, the copper comes right up to the edge of the aluminum, maybe within 1mil of the edge of a hole. There is no electrical connection of course. True plated holes that connect with the aluminum are available from other manufacturers at much higher cost.
- non-plated holes, there is an added gap to the copper
- Board edge to pad: seems not quite as close as a plated hole, but still very usable.



USB connector options:
- Through hole is possible (type B), what I did was heatshrink the "+" lead as thats the only one I care about shorting. Then solder the tops as usual to the PCB pads. Or glue could also be used to fill in the area around the pins with a small needle. Glue can be applied around the USB housing for support as well.





- SMD, multiple options are possible. The one I went with is mid-plane type-C 6 pin. You can solder the sides of the connector to large pads on the PCB, providing good strength. Pads at the edge of the PCB work OK.





I don't know if these are options for serious production, as the gap between the solder connecting Vbus and the aluminum is incredibly small.

Soldering:
- Wasn't quite as bad as expected, any pads I had could be soldered with a T12 iron. The large ones did require a long dwell time.
- Preheater will help a lot if you have one.
- Hot air just didn't work at all, could work with the preheater though

Board spec:
- 1W/mk
- Thickness info: https://www.eevblog.com/forum/manufacture/jlcpcb-now-do-aluminium-pcbs!/

[thmjpr]

Aluminum PCB Bending Notes:
- I used a bending tab width of 6mm, spacing of 2mm (the minimum routing out spec for JLC), and PCB thickness of 1.6mm (1.0, 1.2, and 1.6mm are available).
- Distance from the center of the bend to the nearest component was 2mm. This was a bit close, the part is a 1206 resistor so its robust and had no issue, but smaller components or capacitors should be further away as they could crack.
- An angle of 18 degrees was tested without issue, no cracking or delaminating of the copper traces were visible. In the future if I get some spare boards I'll test one to destruction.

The bend was fairly easy, if you needed a stronger PCB for mechanical reasons you could go thicker. Or if you wanted an easier bend for a longer tab size, could go thinner. To me 1.6mm seemed like a good trade off. Although the actual board received measures closer to 1.45mm.

Bending jig:
- Bending jig can be 3D printed in PLA to get consistent bend angle
- Springback can be calculated online, but for aluminum it is very low. I only ended up needing to add 0.5 degrees or so on the over-bend angle.

Initially I was considering how to make a bending jig out of aluminum. But its just not necessary, the softness of the al-core is easy to work with. The jig would also end up fairly complex to machine. 3D printed PLA seems fine for low volume tests.

Cutaway of the jig:


Parts:


Fit:

[Infraviolet]

Are aluminium PCBs mainly just about better thermal conducitivity to take heat away from high components (motor drivers, high power LEDs...), or are there other common uses too? I gues with the right tolerances one could use the aluminium PCB service as a way of getting precicion cut 2D aluminium parts for mechanical use?

[thmjpr]

Are aluminium PCBs mainly just about better thermal conducitivity to take heat away from high components (motor drivers, high power LEDs...), or are there other common uses too? I gues with the right tolerances one could use the aluminium PCB service as a way of getting precicion cut 2D aluminium parts for mechanical use?

That is the main common use, yes. The board I am working on can dissipate 3-5W or so, trying to do the same with FR4 would not be realistic.

Your mechanical use idea is good, if you need something that is roughly 1 to 1.6mm thick, in low volumes. I have seen a few postings of people using them for front panels, guitar pedals, etc.
But that is what I am trying to get at here, there are many more possibilities. With bending, you could form the board into a box shape, triangle pyramid, etc. FR4 will not bend and stay in place, you need flex PCB and some structure to hold it.

[JohanH]

Are aluminium PCBs mainly just about better thermal conducitivity to take heat away from high components (motor drivers, high power LEDs...), or are there other common uses too? I gues with the right tolerances one could use the aluminium PCB service as a way of getting precicion cut 2D aluminium parts for mechanical use?

I've created front panels for some project cases. White soldermask and black text works nicely. No circuits, but I guess you could theoretically have a PCB integrated in the front panel.

[tooki]

Are aluminium PCBs mainly just about better thermal conducitivity to take heat away from high components (motor drivers, high power LEDs...), or are there other common uses too?

I don’t even think they existed until high power LEDs came along, since they’re really the only power semiconductors where the heat must be dissipated through the back. For everything else you’d just put a heatsink onto the top of the package, but with LEDs that’s obviously impossible. I don’t think I’d use aluminum for serious non-LED applications anyway, since the thin FR-4 layer is still going to add a lot of thermal resistance; a heatsink is going to be better anyway.

I gues with the right tolerances one could use the aluminium PCB service as a way of getting precicion cut 2D aluminium parts for mechanical use?

I’ve thought about using them for this, too!!

[jirij]

Sorry to revive this, but I would like to warn people that things might have changed since 2022.

This thread gave me some confidence to submit an alu PCB design with PTHs (thinking the isolation layer would extend throughout the hole, without a pad on the other side), and I can see them on the PTH gerber layer, with PTH drill map correctly identified, just as it would for a 2-layer FR4 board.

But the boards I received don't have plated through-holes, they're literally just top-layer pads with NPTHs drilled through them:


In many cases, these are shorted together by the tearout reaching into the hole, past the isolation layer, and contacting the aluminium.
In case you were wondering - these were 1mm holes, 2.54mm apart, not some tiny vias.

The thing is - this sucks for mounting holes (like for USB connectors) as well, as the solder will likely only bind to the tiny pad on the top, not the aluminium (at least not without some aggressive chemicals as flux).

So unless I did something wrong, ... don't use PTHs on alu PCBs with JLCPCB .. ?

[thmjpr]

But the boards I received don't have plated through-holes, they're literally just top-layer pads with NPTHs drilled through them:

Yes, that is how it works and what I used in the original post. Unless you pay a lot more you are not getting plated through in aluminum.

In many cases, these are shorted together by the tearout reaching into the hole, past the isolation layer, and contacting the aluminium.

That is a major problem, either caused by dull bits or offset drills. In your photo the holes don't look well centered.
You can try letting JLC know and see what their response is, sounds like a process failure or spec issue to me.
Another option would be manually increasing the copper pullback in the hole center.

The thing is - this sucks for mounting holes (like for USB connectors) as well, as the solder will likely only bind to the tiny pad on the top, not the aluminium (at least not without some aggressive chemicals as flux).

The USB connectors shown above solder to large planes, you need to do that or enlarge the pad size. They are very strong.
There is no need to solder to the aluminum, I guess you could nickel plate it if you really wanted to though.

[JohanH]

Sorry to revive this, but I would like to warn people that things might have changed since 2022.

This thread gave me some confidence to submit an alu PCB design with PTHs (thinking the isolation layer would extend throughout the hole, without a pad on the other side), and I can see them on the PTH gerber layer, with PTH drill map correctly identified, just as it would for a 2-layer FR4 board.

But the boards I received don't have plated through-holes, they're literally just top-layer pads with NPTHs drilled through them:

Unless I'm mistaken, they have never advertised this capability.

But you have some important information that IF one decides to use PTHs in the gerber, there is a risk for shorting between the pad and the hole. It sounds safer to use custom footprints with NPTH instead, because they will have a clearance.

[jirij]

Yes, that is how it works and what I used in the original post. Unless you pay a lot more you are not getting plated through in aluminum.

I guess I saw the texture (in your photos) inside the holes match the texture of the pads (+ the rounded corner between the pad and the hole) and assumed the hole is plated using the same technique (at the same time) as the pad. In my photos, there's a pretty distinct break between the milled/drilled aluminium of the hole and the HASL-tinned pad.

The USB connectors shown above solder to large planes, you need to do that or enlarge the pad size. They are very strong.

Sure, I assumed as much when retrospectively looking at your photos, but - at that point - I might as well use a SMD version of USB-C (or bend the legs of a mid-plane one ).

But you have some important information that IF one decides to use PTHs in the gerber, there is a risk for shorting between the pad and the hole. It sounds safer to use custom footprints with NPTH instead, because they will have a clearance.

Yeah, knowing what I know now, that's what I'd do - use a footprint with a pad shape + NPTH in the center, ~0.2mm between the two.

But, honestly, I don't see the use here -- if a conductive connection is wanted, anything going through the hole can easily short to the substrate, so a THT doesn't make sense. ... For a purely reinforcing use case (ie. a connector shell), the tear-out is a non-issue.

[thmjpr]

Unless I'm mistaken, they have never advertised this capability.

But you have some important information that IF one decides to use PTHs in the gerber, there is a risk for shorting between the pad and the hole. It sounds safer to use custom footprints with NPTH instead, because they will have a clearance.

You can see through holes in their example photo here: https://jlcpcb.com/blog/unlocking-the-power-of-aluminum-pcb
I've had a few sets of boards made and they were all perfect with no shorts. If I had shorts I'd send JLC some photos and see what they think about the quality.

I guess I saw the texture (in your photos) inside the holes match the texture of the pads (+ the rounded corner between the pad and the hole) and assumed the hole is plated using the same technique (at the same time) as the pad. In my photos, there's a pretty distinct break between the milled/drilled aluminium of the hole and the HASL-tinned pad.

I'll quote from my post: "Plated holes, the copper comes right up to the edge of the aluminum, maybe within 1mil of the edge of a hole. There is no electrical connection of course."

Sure, I assumed as much when retrospectively looking at your photos, but - at that point - I might as well use a SMD version of USB-C (or bend the legs of a mid-plane one ).

Yes, regular SMD or mid plane SMD USB C is the way to go, no bending needed.

But, honestly, I don't see the use here -- if a conductive connection is wanted, anything going through the hole can easily short to the substrate, so a THT doesn't make sense. ... For a purely reinforcing use case (ie. a connector shell), the tear-out is a non-issue.

I showed its possible but could have been more clear that its a bit of a pain to do. I only did it because I needed a connector on the opposite side of the PCB, so SMD would not have worked.

There are some other through hole mounts found on LED lamps: https://electronics.stackexchange.com/questions/613578/what-is-this-through-hole-component-holder-used-on-aluminum-pcbs

[fourfathom]

Does anyone have a guess about the impedance characteristics of the trace / insulator / aluminum substrate?  This could be good for surface-mount RF dummy load or attenuator board heat dissipation.  JLC says the insulator layer is 0.1mm, and if I assume it's got a dielectric constant of 4 (not even a guess, just a number pulled out of the air, close to FR4) that gives me a 50 Ohm trace thickness of about 0.2mm.  My frequency range is generally 100 MHz or below, and the traces are going to be quite short, so it probably doesn't matter very much.  Still, it would be nice to know the εr and loss characteristics..

[tszaboo]

So unless I did something wrong, ... don't use PTHs on alu PCBs with JLCPCB .. ?

Or ever. It's not for that. The metal substrate PCBs are not for through hole components. There is no guaranteed isolation from the substrate. And there is also no plating to it, so you cannot connect to it with vias and whatnot.

Does anyone have a guess about the impedance characteristics of the trace / insulator / aluminum substrate?  This could be good for surface-mount RF dummy load or attenuator board heat dissipation.  JLC says the insulator layer is 0.1mm, and if I assume it's got a dielectric constant of 4 (not even a guess, just a number pulled out of the air, close to FR4) that gives me a 50 Ohm trace thickness of about 0.2mm.  My frequency range is generally 100 MHz or below, and the traces are going to be quite short, so it probably doesn't matter very much.  Still, it would be nice to know the εr and loss characteristics..

It would definitely be a weird stackup for RF. The substrate is not really GND, because you cannot just place dozens of vias to connect the GND to the substrate. So then it's a big (in RF terms) floating metal 0.1mm from your trace. I guess it would work up to a few MHz. It's also very easy to make a connection to the substrate accidentally, but difficult to do it intentionally, because it's aluminium, which has a non-conductive outer layer, that is also non-wetting for solder.
But you should test it. It's very cheap to order a PCB with a few traces, place some SMA on it and measure it with a VNA.

[Zoli]

Does anyone have a guess about the impedance characteristics of the trace / insulator / aluminum substrate?  This could be good for surface-mount RF dummy load or attenuator board heat dissipation.  JLC says the insulator layer is 0.1mm, and if I assume it's got a dielectric constant of 4 (not even a guess, just a number pulled out of the air, close to FR4) that gives me a 50 Ohm trace thickness of about 0.2mm.  My frequency range is generally 100 MHz or below, and the traces are going to be quite short, so it probably doesn't matter very much.  Still, it would be nice to know the εr and loss characteristics..

On my samples(from JLC) the specific capacity is 0.56pF/mm², with a Q of ~50, measured with DE-5000@100kHz.

[fourfathom]

It would definitely be a weird stackup for RF. The substrate is not really GND, because you cannot just place dozens of vias to connect the GND to the substrate. So then it's a big (in RF terms) floating metal 0.1mm from your trace.

You're right, I didn't consider that the aluminum substrate would be effectively floating.  But with top and bottom-side copper groundplane over aluminum, with the 0.56pF/mm² capacitance mentioned by Zoli, a 6cm x 6cm  board area would have about 2nF per side to the aluminum.  SMA edge-mount connectors would tie both copper sides together.  I don't have the sim tools to work it out, but it's cheap enough that I might just try it and see...

[Zoli]

It would definitely be a weird stackup for RF. The substrate is not really GND, because you cannot just place dozens of vias to connect the GND to the substrate. So then it's a big (in RF terms) floating metal 0.1mm from your trace.

You're right, I didn't consider that the aluminum substrate would be effectively floating.  But with top and bottom-side copper groundplane over aluminum, with the 0.56pF/mm² capacitance mentioned by Zoli, a 6cm x 6cm  board area would have about 2nF per side to the aluminum.  SMA edge-mount connectors would tie both copper sides together.  I don't have the sim tools to work it out, but it's cheap enough that I might just try it and see...

Another option to replace via stitching is substrate screwing(bad pun, I know): instead of vias, just place round pads with hole in the centre, tap the holes, install screws in, and you have connection to the substrate. Personally(in your situation) I would use 1/16" long screws(0-80, 1-72, 2-56), brass or stainless steel.

[T3sl4co1l]

Top picture appears to be unrelated. It really does look like PTH.

Perhaps you should take another picture that better shows the finish/texture/color inside the hole to clarify to readers how it's been done, and edit it in?

Tim

[shabaz]

I tried sharp bends on JLC PCB alu PCB (1.6mm) with an experimental amount/spacing of slots to make the bend easer. Perhaps to be expected, but it didn't work very well! I don't know what alloy is used, plus there's anodization for some insulation presumably, and it all behaves unlike more general-purpose typical alu sheets.

The bends were possible, but were a one-time thing, because it cracked the surface of the alu deeply, so I wouldn't be able to bend it back (e.g. if the initial bend had overshot the amount of bend required).

The alu PCBs make nice-looking front panels though. I've used it for a couple of DIY front panels so far, to replace the plain alu sheet on small instrument cases. That saves a lot of time, not needing to drill shaped BNC holes etc, letting the PCB mnfr do all that milling.

[T3sl4co1l]

Definitely too tight bend radius -- any cracking means it's severely weakened.  (A wrinkled or lumpy surface may be okay: this can happen due to how the crystals flow around each other, but it may also hide cracks.)  I would redesign such a part.

If I had to guess, the material is something T6, i.e. about as hard as can be; permissible bending is very little before it cracks.  A lower temper, or softer alloy, would be desirable (but you'd have to ask them if they have any options, or if they're at all interested in offering different in the future, or how much a custom run might cost).

(Considering JLCCNC exists, and PCBWay does kind of everything as well, among others, I wouldn't be surprised if they have other materials on hand that could be fed into the process just as well.  Getting a combined proto panel together might not be feasible though.  Alternately, I wonder if they accept Gerbers for laser-cut or CNC metal parts.)

You might get by with less angle per row of perforations, and use multiple rows to ease the overall bend.  This isn't as good as bending to a radius (each individual bend is very tight), but limiting the maximum bend angle per corner at least prevents cracking.

Radius bending isn't easy to do without tooling, I think; while that would be ideal, I understand the desire for low/no tool bending.

You might also do it by making two rows of slots, overlapping in such a way that a stiff end connection is made to a thin web of material inbetween. Thus the web is bent in torsion lengthwise, distributing potentially quite a large angle over a gentler radius.  The resulting joint will be weaker overall (there's less material in it), but the springiness may well be an asset in some builds (flextural joint).

Tim

[tszaboo]

It would definitely be a weird stackup for RF. The substrate is not really GND, because you cannot just place dozens of vias to connect the GND to the substrate. So then it's a big (in RF terms) floating metal 0.1mm from your trace.

You're right, I didn't consider that the aluminum substrate would be effectively floating.  But with top and bottom-side copper groundplane over aluminum, with the 0.56pF/mm² capacitance mentioned by Zoli, a 6cm x 6cm  board area would have about 2nF per side to the aluminum.  SMA edge-mount connectors would tie both copper sides together.  I don't have the sim tools to work it out, but it's cheap enough that I might just try it and see...

Another option to replace via stitching is substrate screwing(bad pun, I know): instead of vias, just place round pads with hole in the centre, tap the holes, install screws in, and you have connection to the substrate. Personally(in your situation) I would use 1/16" long screws(0-80, 1-72, 2-56), brass or stainless steel.

That's true. Placing it on a heatsink and placing a bolt every so often would bring the grounds together.
But for a RF dummy load I would just use one of those 5 EUR 250W resistors. Supposedly they are good till 2GHz, it's much less effort to use them. RFR-50-250

[tszaboo]

For the RF PCB, something worth considering is the Copper core PCB. There the substrate can be soldered and all the traces can be built as a stripline.

[fourfathom]

Yeah, the RF on Aluminum PCB was just a thought.  Seeing how inexpensive it is, and how loose my requirements are, I was wondering if the thermal conductivity would let me make a useful 10W dummy load and/or attenuator using cheap SMT resistors (I need a bunch of these for test and burn-in of a product I am making).  It's not a big deal, I can (and have) purchased proper dummy loads.  The most interesting part of all this was the "cheap".

[Zoli]

Yeah, the RF on Aluminum PCB was just a thought.  Seeing how inexpensive it is, and how loose my requirements are, I was wondering if the thermal conductivity would let me make a useful 10W dummy load and/or attenuator using cheap SMT resistors (I need a bunch of these for test and burn-in of a product I am making).  It's not a big deal, I can (and have) purchased proper dummy loads.  The most interesting part of all this was the "cheap".

If you are still looking for "cheap", LCSC has some non-inductive thick film resistors; one is rated for 7W in free air(IIRC), so two in series would do the job for you.

[3roomlab]

the below are some of my bend experiments last year
the best sample is the pic xxx610/xxx015

slot 6/64 bridge 8/64 (1.6mm ALU JLCPCB) slot 0.8mm
(measurement is to edge of cut, not radius of slot)
next to it drill 0.8mm slot 6/64

my experiment is to try out a quick way to bend 90 degree w/o tool
i also tried some fancy slots, but they dont seem to work well
i suppose based on the 8/64 slot, it would be more sturdy with 10/64 or 12/64 slotting  bridge

the idea is to hand bend a chasis box no tools and nice 90degree corner, it seem to turn out ok
so wrt shabaz case. i think it just need a wider metal bridge so it wont tear up itself

[shabaz]

Hi,

Great to see the results of your experiments!

Just to be clear, the values you used in the '610 screenshot are as shown in the diagram? And you're proposing that for it to be stronger, that the bridge value might be better at 10/64 or 12/64, right?

Just for info, the ones in my photo earlier had a width (drill/mill) value of 1.6mm, and slot of 3mm, and bridge of 2mm.

I took a look at the JLC website, since they do offer CNC capabilities as @T3sl4co1l mentions. The standard alu PCB service does have the great advantage of very low cost, but sadly I think some damage may just to be accepted on very sharp bends due to that type of material, so some techniques (like multiple rows with less tight bend per row) might need consideration as mentioned.

For the board in my photo earlier, I used EAGLE (since KiCad has poor shape handling). But if I were to try it now, I'd use FreeCAD since it has sheet metal capabilities (I've never tried it) and then convert to a Gerber file to try on the alu PCB service again for low cost.

It would be pretty neat to have a JLCPCB-like ultra-low-cost sheet metal service though, with normal steel or alu!

I suppose another option could be to try and use the steel stencil service for such purposes (but it's not as low-cost as the PCB service). And, due to the thinness, other techniques would be required for rigidity, e.g. more curves etc., which could get complicated..

EDIT: One other idea (not a very good one) was to maybe try filling in the bends with an epoxy for metal, in the hope of gaining back some strength. But it's an ugly solution, and is no longer as simple as making the bend and using it.