0.1mm/0.075mm (4/3 mil) trace width fab options?

[Jay1011]

I'm trying to design with a 9-ball BGA part that has 0.5mm spacing between the balls (a TI TPA2028D1 audio amp if it matters).

Per the datasheet, the recommended way to escape the inner ball is with a 75um (3mil) trace.  TI recommends 100um (4mil) traces for connecting to the other 8 balls on the perimeter:

12.1.2 Trace Width

Recommended trace width at the solder balls is 75 μm to 100 μm to prevent solder wicking onto wider PCB traces. For high current pins (PVDD (L, R), PGND, and audio output pins) of the TPA2028D1, use 100-μm trace widths at the solder balls and at least 500-μm PCB traces to ensure proper performance and output power for the device. For the remaining signals of the TPA2028D1, use 75-μm to 100-μm trace widths at the solder balls.

I am wondering what PCB fabs to try for these small 3mil/4mil traces.  Note that I'm using such small traces on just the one part; the remainder of the board will use the usual 6mil/6mil or better trace width/spacing.

This is just for a personal project, and I'm willing to accept the chance that none of these small traces will be fabricated successfully and the boards will be non-functional.  Basically, I just want to give it a try, mainly so I can see if I can hand-solder this tiny BGA.

I want ENIG for best solderability, so in terms of fab I'm thinking:

• OSHPark 2-layer, cross fingers because they have 6/6 width/spacing requirements, though my impression is that their fab can do better than this
• Aisler Beautiful Boards HD, which claims 100um/100um width/spacing is supported, so I'd be violating this only on the one 75um trace to the middle ball

Suggestions?

[Yansi]

The question is then why to drive cost exponentially up just because of one part in the design?

I think most of the local PCB houses would be willing to attempt that size of traces. Just ask them and accept the price.

[asmi]

WellPCB says they can do 3/3 mil traces. Haven't tried it myself yet - though I've done many boards with them with 0.1/0.1 mm traces and they always came out perfect, so I'm reasonably sure they can do 3 mil.

[Jay1011]

Yes, there are other parts I could use, but I'm really just trying an experiment here for a personal hobby, not claiming that this is the best part to use.

In terms of cost, honestly, I'm hoping to keep the same cost by just knowingly violating trace width/spacing DRCs and yet "get away with it" because of some combination of getting lucky and the fab being able to manufacture better than the fab service is willing to guarantee.

[Jay1011]

Yes, I saw WellPCB, but as soon as you move from HASL to ENIG the cost becomes more than using OSHPark or Aisler.  They do support 3mil trace/spacing, though, this is a good point.

[asmi]

Yes, I saw WellPCB, but as soon as you move from HASL to ENIG the cost becomes more than using OSHPark or Aisler.  They do support 3mil trace/spacing, though, this is a good point.

The quality of their boards is FAR above and beyond what you get just about anywhere else. Here is example of my recent small board (this is just a 2 layer board). They even sent a core sample along with this board!

[Drew Fustini]

I want ENIG for best solderability, so in terms of fab I'm thinking:

• OSHPark 2-layer, cross fingers because they have 6/6 width/spacing requirements, though my impression is that their fab can do better than this

We only guarantee our minimum specifications, but our minimum trace width and space specifications are not enforced.

You can go smaller than the minimum and often will get favorable results such as 5/5 on 2 layer.  The minimum drill is a hard limit as the drill machine has discrete drill bits.  Where as, the same LDI machine is used for all boards, so the minimum specifications are really a matter of guaranteed yield.

Please feel free to email suppport@oshpark.com - our support team is always happy to help.

Thanks,
Drew Fustini
OSH Park
<drew@oshpark.com>

[soFPG]

They even sent a core sample along with this board!

I always thought about ordering at WellPCB because of their 3/3 mil trace width and very small via hole size (I don't remember it at the moment) but unfortunately they don't offer a cheap (meaning as cheap as JLCPCB) shipping option - only DHL Express and TNT which are ~35$.
After contacting them about this they also said they are not going to offer another way of shipping.

Very interesting to see how their boards come out - thank you!

[KL27x]

I believe anything down to 6-7 mils is done with ferric or cupric. Below that, ammonium/basic etchants practically "have" to be used. So it's a completely different process to have practical yields. Probably why Drew says 5/5 is already pushing it.

It will have a much better chance of success if you spec half oz copper.

[JLCPCB Official]

Hello ;

Talking about supplying you with some good PCB quality, here is the link for JLCPCB production capabilities where you can find details about the BGA capabilities and you can try the service once you find that it suits your needs.
https://jlcpcb.com/capabilities/Capabilities

[daqq]

From our local ones, pragoboard ( https://www.pragoboard.cz/en/node/27 ) can do 75um/75um, and upon request 50um/50um

[soFPG]

From our local ones

Do you have details for how much that would cost?

[daqq]

From our local ones

Do you have details for how much that would cost?

Nope, sorry. Contact them, they are friendly enough, no minimal order size and can do really funky stuff.

[asmi]

Very interesting to see how their boards come out - thank you!

The quality of their boards is a whole another level compared to the likes of JLCPCB and other cheap places. They don't race to the bottom, instead focus on quality while keeping the prices reasonable. For example, the board from my previous post was $37 for 10 boards with ENIG and matte black soldermask with yellow silkscreen (I think this combination looks amazing). Even my clients regularly comment on just how great their boards look. Also their soldermask is very durable and withstands many rework cycles with no problems, while JLCPCB boards tend to not fare well on that front. This makes your prototypes look more professional in the eyes of your clients.
I think that is why they are not interested to provide cheap-ass shipping options, because they tend to provide crappy customer experience with slow and unreliable delivery. I usually order several different designs at once to spread shipping costs around a bit, but ultimately most of the time I want my boards ASAP and don't mind paying few extra bucks for that. Also most of my boards are 4 and 6 layer, so shipping cost is insignificant compared to their price.

[soFPG]

From our local ones

Do you have details for how much that would cost?

Nope, sorry. Contact them, they are friendly enough, no minimal order size and can do really funky stuff.

They just send me back an offer about 5 PCBs, 50x50mm, 4 layers, 3mil trace width, 0.2mm via hole size:

131.5€ (PCBs) + 120€ ("Tooling") + 25€ (Transport) = 276.50€

Yeah..no thank you

[lawrence11]

I want ENIG for best solderability, so in terms of fab I'm thinking:

• OSHPark 2-layer, cross fingers because they have 6/6 width/spacing requirements, though my impression is that their fab can do better than this

We only guarantee our minimum specifications, but our minimum trace width and space specifications are not enforced.

You can go smaller than the minimum and often will get favorable results such as 5/5 on 2 layer.  The minimum drill is a hard limit as the drill machine has discrete drill bits.  Where as, the same LDI machine is used for all boards, so the minimum specifications are really a matter of guaranteed yield.

Please feel free to email suppport@oshpark.com - our support team is always happy to help.

Thanks,
Drew Fustini
OSH Park
<drew@oshpark.com>

I heard rumors of 6 layer capabilities coming, for many years.

Will this ever materialize? There needs to be a 2-6 layer option right here on the continent.

[Jay1011]

OP here.  Thought I would followup now that I have the OSHPark boards in hand.

First, due to distributor stock issues, I ended up using a slightly different BGA part, one with 12 balls in a 4x3 array at 0.5mm pitch, rather than the original plan of a 9-ball BGA with 3x3 array at 0.5mm-pitch.  The 12-ball BGA requires the two inner balls each to be escaped with a 75um/3mil trace that routes between two 0.5mm-spaced balls on the perimeter of the BGA.

In terms of the OSHPark boards, there's good news and bad news.

The good news is that both 3mil traces seem to have been successfully fabbed on all 3 copies of the board.

The bad news is that each 3mil trace is shorted to one or both of the two perimeter BGA balls as the trace passes between the perimeter balls.  The short is present for all 6 traces (2 traces on each board, 3 boards).

Unfortunately I don't have a camera on my stereo microscope, so I'm not able to post a picture of the board.  The shorts are reasonably apparent visually, and I confirmed them with a multimeter using testpoints on the board.

Oh well, it was worth a try.  Out of sheer stubbornness/stupidity, I may try giving Aisler a try with the same board design just to see what happens.

[mbless]

Unfortunately I don't have a camera on my stereo microscope, so I'm not able to post a picture of the board.

I just use my phone, although, it takes a minute to get the angle right. I'd be interested in seeing the boards if you can manage.

[PeteH]

You can also distort the pad shapes slightly to get a better spacing between the pads you want to exit between. Oblong the BGA lands slightly and thin them out to get a trace by with better spacing.

[asmi]

Oh well, it was worth a try.

No it wasn't. Going beyond what's guaranteed by the manufacturer almost always end in tears.
Take a look at JLC - they guarantee down to 3.5 mil traces on their 4 and 6 layer boards. They are cheap, and the quality is acceptable for the price (don't even hope the solder mask will survive any kind of rework).

[Jay1011]

Oh well, it was worth a try.

No it wasn't. Going beyond what's guaranteed by the manufacturer almost always end in tears.
Take a look at JLC - they guarantee down to 3.5 mil traces on their 4 and 6 layer boards. They are cheap, and the quality is acceptable for the price (don't even hope the solder mask will survive any kind of rework).

It was worth it to me:  $6 for the OSHPark board, with ENIG, and in my hands in 11 calendar days.  JLCPCB would be $11 for HASL with arrival in "30-35 working days" or like 6 weeks.

Also -- the JLCPCB "Capabilities" page shows 5mil/5mil trace width/spacing for 1, 2, 4, and 6-layer boards.  Not sure where you're seeing anything about a 3.5 mil guarantee.  https://jlcpcb.com//capabilities/Capabilities

[asmi]

It was worth it to me:  $6 for the OSHPark board, with ENIG, and in my hands in 11 calendar days.  JLCPCB would be $11 for HASL with arrival in "30-35 working days" or like 6 weeks.

As long as the board is not much larger than a post mark. What about 10x10 cm one?

Also -- the JLCPCB "Capabilities" page shows 5mil/5mil trace width/spacing for 1, 2, 4, and 6-layer boards.  Not sure where you're seeing anything about a 3.5 mil guarantee.  https://jlcpcb.com//capabilities/Capabilities

Apparently they changed it recently. Here they still show old values (I've made a screenshot in attachment in case they mess with it too). If that change is their new standard, I won't be ever ordering with them because I need 0.1/0.1 mm trace capability for all but the simplest designs.

My apologies.

[asmi]

They still state 3.5mil on their Chinese site, and I just ordered a stupidly dense 0.1mm board, it came out good. I even violated their 5mil via to trace rule a lot..

I think it's just a change to ensure higher yield and a more favorable (for them) dispute resolution, rather than a hard limit.

Their solder mask quality kinda blows, so for production I prefer using WellPCB even if they cost extra (client pays for it anyway ). On JLCPCB boards if I touch a trace covered by the solder mask with an iron, the mask burns off and I end up tinning the trace. This never happens to the mask used by WellPCB. It is important for me that my prototypes look professional in the eyes of my clients (even after some rework which is typical for prototypes), to me the extra expense is totally worth it. And like I said above, I don't like playing chicken with my boards, because they contain thousands of traces and I will not be checking them all on each and every board manually to see if I won this time around or not.

[mbless]

JLCPCB account says it's still 3.5mil https://www.eevblog.com/forum/manufacture/bga-on-jlc-4l/msg3091703/#msg3091703. From that thread it looks like they got several numbers wrong when switching to the new capabilities format.

Their solder mask quality kinda blows, so for production I prefer using WellPCB even if they cost extra (client pays for it anyway ). On JLCPCB boards if I touch a trace covered by the solder mask with an iron, the mask burns off and I end up tinning the trace.

I've noticed this problem with more recent batches. I don't remember having this problem before.

[asmi]

I've noticed this problem with more recent batches. I don't remember having this problem before.

Yea I don't know either if it became worse or always been that way, because usually I don't have that much of a rework. However my most recent order with them is a 4 layer RF board, and wherever there is RF, there is always a ton of rework to tune things. I ended up having several lifted pads, even though I was desoldering-resoldereing using hot air gun specifically to protect traces from damage.