PCB Prototype manufacturer ROGERS in China?

[Yansi]

Yansi...  i am not surprised You were told off... It takes time and effort and additional process capabilities to handle Rogers and squeeze every little dB out of the materials in order to meet design objectives.

Just keep in the back of your mind... quality is like oats... once they have been through a horse... they cost a little bit less.

Why should I? I know that as a hobbyist every single company thinks I should f.ck off when requesting special technology. Prices of these are set only for corporates, that just don't care and pay whatever number is present on the bill.

As I like to play with microwave tech, you know, some circuits do not work well with FR4, hence the requirement for better substrate.

But at least am trying to find the cheapest manufacturer. So far PCBCart looks promissing.  A few days ago, I got a quote  from RayPCB.  They ask $230 for a 3pcs 100x100mm (2ly 20 or 30mil 4350B). That is just no!  I can get this about same same price from an (otherwise very expensive) local company too.

What the hell I am telling myself. I think the cost of the substrate is about 4-6€ per square decimeter.  Why the PCBs are so f*ing expensive?

[IconicPCB]

Yansi,

At the risk of keeping a civil tongue...
Just ask Yourself how much the local Mechanic charges to do routine oil change and maintenance on a car.

Then think about how much it costs to live in the economy You are immersed in and how much  an hour of Your labor will earn You.

Then think about how much it actually costs the company which provides You with the hour of labor oportunity.

And dont forget the company is there to turn a buck for the share holders.

And now settle on a ratio between the company hourly charge out rate and Your income hourly rate.

I am sure You do not whinge paying for Your beer. May be You should go and brew Your own?

OH and do use quality oats.   

[Yansi]

While keeping the civil language: Sorry, I do not drink beer. And do not steal. Though the simple prototype will cost me a bit more than half my monthly earning. Do you roger that or do you volunteer to sponsor me? 

Maybe $200 is not a lot for a westerner or Aussie like you, but for me, it is hell of a lot.

[xaxaxa]

in that case I think your best two options are:
1. figure out a way to make do with FR4; don't believe anyone who tells you FR4 is no good for RF (<5GHz). I have built everything from 2.4GHz low noise amplifiers + filters to 5.8GHz power amplifiers on FR4 (and the losses were negligible if you design it properly). My company's first commercial product which is a 3GHz vector network analyzer was done entirely on FR4.
2. etch the PCBs yourself

If your filters performed badly on FR4 it has a high chance of performing badly on rogers material too. If you look at the datasheets the difference in loss between the materials isn't THAT big, and plus a big part of your loss is the copper losses which will be the same for any material (but will be different depending on the trace width for a given impedance).

[pzw]

I know it is an old thread, but maybe you are still looking...
I just came across PCBWay, and they do offer Rogers 4003C and 4350B laminates for prototyping. For 10x 100x100mm PCB they charge $97 without shipping... I think that is a very reasonable price.  Lead time is 7 - 8 days, I guess business days :-) ...

Please let us know what the experience is if you decide to use them..!

Thanks
Paul

[piranha32]

I know it is an old thread, but maybe you are still looking...
I just came across PCBWay, and they do offer Rogers 4003C and 4350B laminates for prototyping. For 10x 100x100mm PCB they charge $97 without shipping... I think that is a very reasonable price.

I found this company through my twitter feed: https://micron20.com/
According to their calculator one 100x100mm board on 0.8mm 4003C costs 39.39E (13.00E setup + 26.39E for 1 board) plus shipping.

[dcarr]

Two ideas:

1. OSHPark 4 layer process uses Isola FR408 - $10/sq in, empirically about 2x better than FR4 by my measurements at 10GHz, etching tolerance is "adequate", does have ENIG finish

2. PCBPool in Europe offered (may still) a RF PCB service that I remember being reasonably priced

Would be nice to find a good cheap option for this.

[TheUnnamedNewbie]

1. OSHPark 4 layer process uses Isola FR408 - $10/sq in, empirically about 2x better than FR4 by my measurements at 10GHz, etching tolerance is "adequate", does have ENIG finish

After all, FR408 is based on glass fibers, which have orientation, hence impedance discontinuity.
Rogers RO boards are harder to handle due to the material nature. This very same nature (ceramic filled hard polymer) gives it unified, isotropic behavior, which guarantees RF performance.
Fiber boards can be improved by weaving the fibers denser, or to use interleaved weaving patterns, but after all, it's not isotropic on a microscopic level.

You would be surprised as to how anisotropic even rogers material still is. This is also why modern rogers datasheets list 2 permittivities: one measured by some IPC standard, and one to use in a design, because the IPC standard does not match practical circumstances. To give you an idea about the error: iic, Ultralam 3850 has a IPC DK of 2.9, and a actual DK for PCB structures of 3.15.

[Fabian]

Two things that might be helpfull for someone:
1) Here are a few prices I collected over the past three years or so. Each for signle quantity of a 12"x18" sheet:
RO4350B, 6.6 mil: 24€
RO4350B: 60 mil: 130€
RO4003C, 8 mil: 50€
RO4003C, 12 mil: 53€
RO3003: 60 mil: 225€
RT5880, 10 mil: 220€
RT5880, 62 mil: 430€

2) There was the discussion why this stuff is "so expensive" even for a small board. You may also take into account that this is quite special stuff and the manufacturer will probably have to do a special run just for you. So he will use a full panel, no matter how large your PCB is. You have to pay for that

Has anyone had experiences with https://micron20.com/ ? They are very cheap (at least I think so... I just got an offer from a different manufacturer for a four layer PCB, 150x150mm, RO4350 on the outside, 1.3mm FR4 in the center: 900€)

[Georg - PY5ZSE]

I am bumping this old thread because it's a topic I have been thinking about for quite some time...

To understand why Rogers PCB (or Taconic, or whatever RF materials) are so expensive compared with today's chinese FR4 boards one should see why the latter have become so dead cheap:

   - Pooling. With the hundreds of Gerbers coming in each day, they can fill up their production panels. They might achieve an efficiency (surface turned into Boards / panel area) of 80% or more.
   - big panels (24 * 36" or more), state of the art equipment, standardized, high yield process.
   - very high degree of automatization.
   - running several identical panels that are drilled together (drilling is the most time consuming production step)
   - material itself is cheap.

In comparison, "special" boards, which includes RF Materials:

   - no pooling. They run a small panel with just one customer's design. Efficiency may be 30%, i.e. 70% of the expensive material ends up in the trash, and production costs have to be paid by the remaining 30%.
   - small panels (12 * 18"), 1/4 of the "big" FR4 panels. They are either run through a special, smaller prototyping line, or with an adaptor frame through the big one, where they occupy the same space/time slot as a large FR4 panel.
   - one panel instead of 3 in the drilling machine. even more expensive machine time per board.
   - the material itself is more expensive, but that, IMHO, explains only a small part of the higher price.

The key factor for cheap prototype PCB's is pooling. That only works, if the manufacturer collects enough orders to fill up a panel during his guaranteed turnaround time (- the one day really needed for production). So, quick and cheap are mutual exclusive. On the other side, this opens a possible way to reasonably priced RF boards:

   - EITHER convince one of the chinese manufacturers that there is a market for an RF PCB pool, offering only selected materials / thicknesses, and accepting significantly longer turnaround times than with standard FR4
   - OR convince somebody like Aisler, who's business model is pooling and having the panels manufactured by third parties, of the same thing
   - OR somebody sets up such a pooling service

The main question is, which turnaround time will be accepted. A hobbyist will probably have no problem waiting 4 weeks if he gets his little Rogers board for 20 $ instead of 200. For a business, that may be 3 1/2 weeks too long.

[OwO]

Another point is that business customers will usually want 4+ layer boards. Rogers material for RF is really overrated. I've done 5.8GHz power amplifiers on FR4 without a noticeable efficiency drop. One thing I'm *not* willing to do is go back to 2 layer boards. The applications in which using Rogers material is necessary are all very niche (no, you do not *need* interdigital filters at frequencies < 6GHz). Unless it becomes very cheap (no more than 50% more expensive than FR4) for 4 layers I can't see myself using the service. I would rather design my way around it.

[Yansi]

You do not need custom microstrip filters, sure, if you use just the damn standard ISM bands like 2.4G or 5.8G, where filters are available readily made as SMD parts (with mostly poor performance and extreme pricing otherwise).

If I want to do a custom whatever outside of the ISM bands, then neither do I have any monolithic filters available, nor any single-chip fully integrated solutions for anything.

Microstrip filters on FR4 almost can not be done (silly me have tried), unless running through handful of prototypes and even then it will have less than poor repeatability.  Better defined FR4 substrates such as Isola FR408 are not widely available as one would like and expect, so no go that way either.

Also, impedance matching of custom discrete LNAs (not integrated "LNA") is almost impossible. You just make oscillators, not amplifiers. They can be tuned, sure, but also you will run through lots of test boards yet the performance will not be really repeatable.

And don't forget, that some of us also want some power amplifiers (not just "power amplifiers"), I mean 50W on 3.4GHz. Lossy FR4 would almost melt.

[Georg - PY5ZSE]

@Yansi:
I basically agree with you. The other point is, that for hobby (ham radio) applications, I don't want to "design around". That's why I am looking for reasonably priced Rogers boards.

Where OwO certainly has a good point is the multilayer issue. a commercial product will in most cases have a "small" RF / microwave part and a lot of digital / general electronics around it. In a hobby project I just split that into a RF module and a digital one, and don't care about spending 10$ for cables and connectors.

Besides hobbyists and industries that either produce dead-cheap consumer stuff or, like aerospace and mil, don't care about money, there are others: university and research labs, companies that produce very specialized stuff in small quantities. The question is, is there a market for such a project or not. It's certtainly not the manufactureres of garage door remote controls. or 24 GHz doppler motion sensors (where I have seen RF boards for the patch antennas, and they are a consumer product)

[TheUnnamedNewbie]

In comparison, "special" boards, which includes RF Materials:

   - no pooling. They run a small panel with just one customer's design. Efficiency may be 30%, i.e. 70% of the expensive material ends up in the trash, and production costs have to be paid by the remaining 30%.
   - small panels (12 * 18"), 1/4 of the "big" FR4 panels. They are either run through a special, smaller prototyping line, or with an adaptor frame through the big one, where they occupy the same space/time slot as a large FR4 panel.
   - one panel instead of 3 in the drilling machine. even more expensive machine time per board.
   - the material itself is more expensive, but that, IMHO, explains only a small part of the higher price.

I think you forgot a very important reason: Some substrate materials such as the PTFE-based stuff from rogers, to my knowledge, requires a fundamentally different process. You can't just swap out the FR4 panel with a rogers panel (be it with or without a frame to compensate for different panel size) and go with it. This means that whoever does this cannot use their automated line unless they have something set up specifically for a certain rogers family.

Besides hobbyists and industries that either produce dead-cheap consumer stuff or, like aerospace and mil, don't care about money, there are others: university and research labs, companies that produce very specialized stuff in small quantities. The question is, is there a market for such a project or not.

I work with this kind of stuff in a university/research context, and honestly, there is little demand (at least from my research group) for lower cost. Main push is higher quality and repeatability, since our experience with many manufacturers with RF proto techs is that the repeatability is low: One week you get slightly over-etched, next week you get slightly under-etched, ...

And we want better features - when you are trying to demonstrate a fundamental concept, you don't want to be performance limited because your vias had to be 200 um apart etc...

[IconicPCB]

I have manufactured some 3D cavity structures in rogers     material.
Pressed the multi layer layup, activated and plated the structure only to find the layers did not bond well enough despite following instructions.
When asking the client if it were OK to roughen the copper surfaces i was told it is a no no.

Apparently there is more to achieving the requisite bond quality to eliminate possibility of plating electrolyte ingress in between the layers.

Let me just finish with... if you cant afford it..you don't need it.

[Georg - PY5ZSE]

I think you forgot a very important reason: Some substrate materials such as the PTFE-based stuff from rogers, to my knowledge, requires a fundamentally different process.

The problem with teflon based materials is that nothing sticks to it - neither steaks nor copper. For that reason the holes need an "activation step" for which to my knowledge there are 2 possible procedures, plasma etching and chemical with some solution containing metallic sodium. This is absolutely non - standard, completely different chemistry than the usual one in PCB manufacturing.

The "industrial grade" Rogers materials (series 3000 and 4000), however are designed to be treated in the same process as FR4.

Let me just finish with... if you cant afford it..you don't need it.

Nobody "needs" to do electronics as a hobby. But if you do anyway, aint it legitime to try to source your materials at affordable costs?

[IconicPCB]

Ghieber,

Ofcourse it is valid to go shopping for an affordable solution and yet in the final analysis if you cant find an affordable solution... thats too bad.
You cant go about demanding it.

Just so that you understand the other side of this comment...

Sell on price
Live on rice

Does Your supplier not deserve a decent living also... after all your "need" is only a hobby not a daily existence.

[Yansi]

They get likely beyond decent living, from the pricing that is often pushed for those "industrial grade Rogers substrates", that likely? can be treated by standard FR4 process, as even the manufacturer of the substrates is boasting in their papers with.

That is my point. I am not demanding any special substrates for 100GHz, I am not even demanding any special technology (blind or burried vias in weird custom stackups), I am not demanding any special manufacturing tolerances for etching or stuff.

So why the hell do I have to pay even up to $500 premium for a simple, small, 2 layer rogers job for a two transistor LNA and filter for a couple GHz?

If anyone can suggest an alternative substrate for RF work, I will happily ditch Rogers and will go with something else.

But never mind, I have already given up demanding reasonable. Easiest and pretty affordable solution is to just buy the raw substrate that ain't any expensive and etch it at home, then find an acceptable way for making the vias, such as soldering a piece of wire in here there - such as was done by microwave oriented radio-amateurs 30 freaking years ago, when the RT duroid and similar materials became available behind the iron curtain. So pretty disappointed no progress in this area since back then.

[IconicPCB]

if you are going to etch it yourself and activate the holes...look into soviet era hole wall activation process based on copper phosphite pyrolysis chemistry.

It should offer you a good outcome.

[IconicPCB]

Surface roughness ( pumice slurry ) treatment of FR4  is the classic method of improving multilayer adhesion in a multilayer process.
This option is not available in a GHz filter design.

Additional chemistry must be deployed to improve on adhesion without deforming surface texture.


To quote my old lecturer:  everything must be just so,,, anything more or anything less is from the devil...

[intmpe]

OK.. Ive been doing sub 5GHz for years using Chinese FR4 PCB vendors - low cost ones.

Sorry there is a lot of inexperience here - if you haven't been doing this for 10 years you are still learning - sorry to be frank that's the way it is plus you need some decent qualifications.

Firstly, you get 5G NR cell phones out completely on FR4 - and multilayer at that for a few primary reasons:

a). No distributed solutions are allowed. If you want distributed circuits its a teflon based material. You can probably get 4350 in China but traditional you could not. Arlon TC was the alternative because its made there. There were export restrictions on the better rogers materials. This has changed but the way China works you would not know if they substitute a local 4350 equivalent anyway. ALWAYS ALWAYS do a test board breakoff on any rogers or fr4 prototype so you can test dielectric constant and loss.

b). Because no distributed solutions are allowed you need to make sure your lumped circuits are on point - no weird impedance transformations. With the Q's available on small parts you just cant do it and get decent results. Don't go too small - you wind down your vendor choice and your Q's will suck and so will your design. Do tricks like edge mounting of caps if necessary.

c). You can't define a prepeg - they wont use it on FR4. They will use whatever the hell glue they use - it doesn't matter if you define the thickness - be careful with that if you are doing a high density multilayer board make sure your RF layer is thick enough so that the glue layer is not going to kill you as it varies. Thickness is usually not a big player unless you have chosen a very thin rf layer - I've seen that happen due to inexperience and its a show stopper. Think about the thicknesses for the RF and make sure the unknown glue layer is small. Forget about impedance controlled boards. They say they can do it but its hit and miss and not tested at RF.

d). Dielectric constants are not constant over frequency - they are just not - I don't care what the substrate suppliers say -  Ive measured them (remember - 10 years min). As a result you need to take that into account - remember the breakoffs.

e) The best way to take these last few points into account is via a pure lumped solution using high Q components packed as tight as you can. Forget about transmission lines - they don't work - NEVER lay out 50 ohm lines on FR4 - you are better to give up some mismatch than to put up with dielectric loss (remember - 10 years min)

f) Use their units. They don't give a shit if you are American and too fucking stupid to have converted to metric. The world has spun and what the fuck have you been doing for the last 10 years. Their education is typically better than yours. If you don't believe me sit in a Chinese University and sit in a US university and compare. I have and I know what the situation is. I also go to their conferences and they are superior to US conferences. Sorry. Forget the hype - US education has declined into garbage - remember they can do specialist electronic engineering degrees. In the US you can do a basic electrical engineering degree with none to few advanced options let alone product design. A US Masters may be equivalent to a Chinese Bachelors if you are lucky. I did a US masters and it was garbage. Too many adjuncts and not enough experience in the world. I haven't heard of imperial being the choice units in China for more than 10 years. Again - they don't give a shit what you want - they have already pivoted to supporting their local economy. Basically they are now driving the Electronics industry and are probably 5 years ahead. They have a process and you are using it because they are cheap. You don't like it? go to an American vendor - that won't last long because US shit is even worse - no discipline, management thinking sucks ass and education is poor. You get what you pay for and in China if you don't go so cheap you will get something much better. The US equivalent for that something better will cost you an arm and a leg - in the US board makers are still shipping non masked boards. They can't compete so everything is just worse. Electronics is almost dead in the US so just remember that - they are pumping and the US is dying - finding real R&D hardcore electronics porn work is getting hard in the US - at best its iterative improvements in the US 95% of the time. I go to US conferences and its hype with dickheads with big mouths and good hair - just dilbert stuff. I usually walk away with not a single usable idea.

[TheUnnamedNewbie]

OK.. Ive been doing sub 5GHz for years using Chinese FR4 PCB vendors - low cost ones.

Sorry there is a lot of inexperience here - if you haven't been doing this for 10 years you are still learning - sorry to be frank that's the way it is plus you need some decent qualifications.

Firstly, you get 5G NR cell phones out completely on FR4 - and multilayer at that for a few primary reasons:

a). No distributed solutions are allowed. If you want distributed circuits its a teflon based material. You can probably get 4350 in China but traditional you could not. Arlon TC was the alternative because its made there. There were export restrictions on the better rogers materials. This has changed but the way China works you would not know if they substitute a local 4350 equivalent anyway. ALWAYS ALWAYS do a test board breakoff on any rogers or fr4 prototype so you can test dielectric constant and loss.

b). Because no distributed solutions are allowed you need to make sure your lumped circuits are on point - no weird impedance transformations. With the Q's available on small parts you just cant do it and get decent results. Don't go too small - you wind down your vendor choice and your Q's will suck and so will your design. Do tricks like edge mounting of caps if necessary.

c). You can't define a prepeg - they wont use it on FR4. They will use whatever the hell glue they use - it doesn't matter if you define the thickness - be careful with that if you are doing a high density multilayer board make sure your RF layer is thick enough so that the glue layer is not going to kill you as it varies. Thickness is usually not a big player unless you have chosen a very thin rf layer - I've seen that happen due to inexperience and its a show stopper. Think about the thicknesses for the RF and make sure the unknown glue layer is small. Forget about impedance controlled boards. They say they can do it but its hit and miss and not tested at RF.

d). Dielectric constants are not constant over frequency - they are just not - I don't care what the substrate suppliers say -  Ive measured them (remember - 10 years min). As a result you need to take that into account - remember the breakoffs.

e) The best way to take these last few points into account is via a pure lumped solution using high Q components packed as tight as you can. Forget about transmission lines - they don't work - NEVER lay out 50 ohm lines on FR4 - you are better to give up some mismatch than to put up with dielectric loss (remember - 10 years min)

f) Use their units. They don't give a shit if you are American and too fucking stupid to have converted to metric.

That is odd. In my dealings with one the biggest telecomunications device manufacturer from China, they:

- Used transmission lines everywhere
- Used mils all the time, to my great frustration, because I prefer microns (incidentally, all design rules I've gotten from their manufacturers also are written in mils, not mm/um)
- Used lots of substrates, including Isola, Megtron, next to FR4
- Defined exact stackups that were to be followed to the letter
- Define impedance controlled boards and enforce that

So perhaps, in your great experience and wisdom (10 year min.), just failed to find the good manufacturers. Could be they were just to lowly for someone with your minimum 10 years of experience.

[TheUnnamedNewbie]

Perhaps the two of you are just in two different industries.

I imagine it is likely, but that is beside the point - I just wanted to pose a counterexample to the sweeping generalizations and 'you are doing it wrong' phrasing of intmpe's reply.

[OwO]

Transmission lines on FR4 are fine - on non impedance controlled boards I haven't seen more than 10% variation, and impedance controlled boards are usually better than 5%. Of course you have to vet your boardhouse well to be sure they are actually doing impedance control. I usually use a mix of discrete and distributed components. For example capacitors < 0.5pF or inductors < 1nH I will prefer to put a stub instead. For higher power/lower loss you can cut away the in1 layer and drop down to in2 as ground reference, to get ~1.2mm dielectric height. I've done this for a wilkinson power combiner on a 5.8GHz power amplifier.

[david_geng]

I think FR4 is ok for 6GHz even up to 8GHz. Here is a test with VNA on a 4ly TG155 board. The test trace is 1inch long with SMP connectors.