PCB Prototype manufacturer ROGERS in China?

[Yansi]

Hello!

the more I am interested in the microwave bands, the more I find need for something better behaved than FR4, especially when attempting to construct microstrip filters (I find it almost impossible to do on FR4).

Do you have any experience with prototyping PCB manufacturing in China on Rogers substrates? A quick google search shows that such PCB prototype houses in China exist, but usually they offer this on special request and I am quite afraid of the pricing there.

But still not as afraid of the pricing, as what the local manufacturers in Europe want for that. I would describe that as utter nonsense and ripoff, with regards what the substrate really costs and that very likely they do not need to change almost anything in the technology setup I guess. *

Do you have any recommendations for any specific manufacturers? What to look for, what to be aware of?

I will not need to manufacture any special PCBs, likely simple 2 layer stuff below 5 GHz. What substrate is the cheapest choice for this?  I guess the most common two are RO4003C and RO4350B, while the 4003C being the widely used cheaper option. (is it correct?)

Thank you for suggestions,
Y.


//EDIT: * yes. Straight from the RO4000 brochure : RO4000 series laminates can easily be fabricated into printed circuit boards using standard FR-4 circuit board processing techniques. Unlike PTFE based high performance materials, RO4000 series laminates do not require specialized via preparation processes such as sodium etch. This material is a rigid, thermoset laminate that is capable of being processed by automated handling systems and scrubbing equipment used for copper surface preparation.

[WPXS472]

I don't have any experience with Chinese suppliers and Rogers substrates. Sorry. But I do have experience with Chinese PCB suppliers and FR4. First, FR4, though good for a lot of things, is awful for RF. It doesn't even have to be microwave either. At my for employer, we were dealing with the Chinese branch of a major American company trying to get some products on their approved vendor list. They didn't like out test fixture at all and provided their own. Though I had some problems with how it was constructed, it did have excellent return loss. They provided the Gerbers for it and we had some made by our Chinese supplier. Even though they physically looked identical, the return loss was simply terrible. This was only up to 1.25 GHz too. We used a Chinese supplier almost exclusively, much to my dismay. They were difficult to deal with from an engineering perspective. Whenever we would send a set of Gerbers and the specs for the materials and panels, they would always send a big list of questions that had to be answered. Sometimes, it was next to impossible to even understand what they were asking. I don't know if they were using translation software, or they just didn't do well with English. One habit they had that I found particularly troublesome, though I admit others might not find it a problem, was that I always sent all dimensions in Imperial measurement format. I think in inches, not centimeters. They would always convert every measurement to metric form when replying. Just out of stubbornness, I would always convert everything back to imperial when I made my replies. If I could have had my way, I would have told them that as the customer, I demanded that they use my measurement system if they wanted to continue to do business. Even though wanted every aspect and detail documented, they had a habit of just changing things without telling us. My advice when dealing with a Chinese supplier is to be up front about what is important to you, provide every detail imaginable, and just let them know that if they want your business, they have to meet your expectations. If you specify something like Rogers material, expect to pay a heavy premium for it because it likely won't be locally produced. The idea that everything that comes from China is junk, is no longer valid. A lot of Chinese companies can do excellent work.

[rx8pilot]

Subscribing to the thread.

Short and misplld from my mobile......

[yl3akb]

Interesting topic, I also would like to learn more about low cost options.
https://www.pcbcart.com/ recently quoted me total of 154$ for 32 (minumum order, I guess) 35x55mm boards with RO4350, but I did not try it (yet).

[xaxaxa]

The problems with return loss aren't because of FR4, it's because you kept the geometry the same and switched to a material with a different dielectric constant. All microstrip widths must be recalculated for the new substrate material. Also most major pcb fabs in china support impedance control nowadays (often at no extra cost) so you should be able to get return loss better than 20dB up to many GHz.

FR4 is perfectly usable for RF up to 6GHz if you design with the losses in mind and use impedance control. There is very little reason to use a more exotic material other than for e.g. microstrip filters; power amplifiers (>20W @ >1GHz).

[Yansi]

Well.. as the original question didn't get much attention... I have another question:

What is the most common rogers material used for RF designs? As I mentioned above, I think that RO4003C and 4530B are the two most common.  Not sure which one is the cheaper or more common one.

But I have a different interesting question: What is the common substrate thickness used for the designs? I think that for most applications, where microstrip filters are typically required, you can not get away with 1.5mm substrate easily (everything 50ohm will become just insanely! wide - 3.6mm to obtain 50ohm at 3GHz on RO4003C)  and thinner must be used, like 0.8 or 0.6mm substrate.

Do you have any idea?

//One of the local EU manufacturers just replied and in short told me to fuck off, that they won't order any 0.8 rogers for me. Is this normal? I am pissed...

[ThomasDK]

These guys do 0.5mm RO4003 in the EU at pretty reasonable prices:

https://www.micron20.com/orders/calculator-production?lang=en

I haven't tried them myself yet.

[Yansi]

You’ve reached your daily limit for using the calculator!
For more more calculations and additional discounts, login first!
Don’t have a registration? It is fast and easy here.

Fuk em in the ass!!!   Had to launch a TOR browser.

Was about €200 for 5 pieces of 120x65mm  0.5mm thick RO4003C. Which is quite meh...  40€ for such small board, stupid simple 2 layer stuff, I don't know. Maybe my expectations are a bit different than reality.    But it is in EU, which is good to start with. One do not want to pay stupid import duties and the pelf for the shipping company.

But thanks for mentioning them.  Mayyyyybe I will try, if I do not find a different cheaper vendor.

//I have just  converted €200 to our currency. I knew 200€ is a lot, but did not expect that especially that much.   

[KJDS]

It's been a while since I ordered any, but I think that 4350 is about £100 for a single 18" x 12" sheet of plain material. 4003C is a lot more, or it might be the other way round.

I've generally gone down the route of using expensive but trustworthy board houses, as delaying a project by a month due to a crap board spin costs way way more than even a very expensive board.

[G0HZU]

What is the most common rogers material used for RF designs? As I mentioned above, I think that RO4003C and 4530B are the two most common.  Not sure which one is the cheaper or more common one.

But I have a different interesting question: What is the common substrate thickness used for the designs? I think that for most applications, where microstrip filters are typically required, you can not get away with 1.5mm substrate easily (everything 50ohm will become just insanely! wide - 3.6mm to obtain 50ohm at 3GHz on RO4003C)  and thinner must be used, like 0.8 or 0.6mm substrate.

Do you have any idea?

I've been using Rogers materials for over 25 years at work for RF design so I can offer a few clues.

The most common thickness at work is 0.020" and this is for designs up to maybe 7GHz or so. This gives a (50R) microstrip width that mates fairly well with SMD parts and it means that vias have low inductance.

But the price for this is higher dB/inch loss and also this skinny material is very brittle. Very easy to snap a 0.020" thick 4003C PCB like a dry biscuit! Also, the material is quite bendy so it is very easy to crack MLCC caps on a PCB like this unless you stiffen it up with some backing material before you solder parts to it. A typical design at work would have an 8-12 layer FR4 core with 0.020" Rogers layers as the top and bottom layers. This complex multilayer stack helps with rigidity and makes for a very versatile PCB with loads of options for signal/control and power planes etc. But it is very expensive!

We are slowly phasing out the use of 4003C in favour of 4350B and I think this is because of reasons to do with ROHS. But for general purpose RF design/dev work you would be best to choose whichever you can find cheapest as the RF performance is very similar.

At work we have always used a PCB mill to make rapid PCB prototypes with Rogers materials of various thickness. The most common are 0.060", 0.032" 0.030" or 0.020" but as I said earlier, we tend to go for 0.020" nearly all the time now. At some point it may be that we use 0.010" more often.

I do quite a bit of work here at home with Rogers material and I also have a PCB mill (T-Tech 7000S series) and I bought this machine as an old/used trade in about 16-17 years ago for about £800.

I think part of the reason it costs a lot to have a Rogers PCB made is that it is assumed that the PCB etching tolerance and the other PCB manufacturing tolerances need to be very tight. For this reason, we tend to use the top PCB companies like Graphic, Exception or Labtech here in the UK to make our 'real' PCBs that use Rogers material and the costs are very high even for prototypes because of the need for repeatable etching accuracy. Otherwise, what's the point in trying to make a high performance RF filter or coupler with expensive materials if the etching tolerance will spoil the results?

Have you considered buying bare material and etching it yourself? Maybe practise on FR4 until you can get good results with accurate etching?

[Yansi]

It's been a while since I ordered any, but I think that 4350 is about £100 for a single 18" x 12" sheet of plain material. 4003C is a lot more, or it might be the other way round.

I've generally gone down the route of using expensive but trustworthy board houses, as delaying a project by a month due to a crap board spin costs way way more than even a very expensive board.

Your price are a bit high. RO4003C is about $100 (not GBP100) for that kind of size, RO4530B is about half that.  I have only briefly looked through RFMW stock. I do not know where else I could buy these.

I am not afraid of the Chinese guys screwing up, they actually never did any of my PCBs. But the expensive local mfgs, yes they screwed a lot. Hair, someone's oily finger in a soldermask,  inconsistent solder mask color,  ENIG board that had a leakage due to gold depositing in a microfilm all over the board, not abiding to the extra cost express manufacturing deadlines, and such. The chinese guys can screw you in many different ways, but from my experience the failure rate there is much lower actually. But I am only mainly a hobbyist, so I do not spin off that much PCBs.


G0ZHU:  20mil substrate seems too thin, I understand it may be very fragile. I'd rather use something in the range of 0.5 to 0.8mm, for example the 30mil or 20mil RO4530B substrate, as it gives reasonable 50ohm trace width I am used to.

I do not have PCB mill, my friend does, but I would not trust that for precision. And it is tiringly slow. But I have no issues whatsovever etching my our PCBs using the photographic method, with as good precision as it can get at home. (600DPI laser printer) Can do both positive and negative solder resist film.
 
Problem is, where to obtain (in EU) the substrate. I only mainly find RT/duroid material, which I know nothing about. Only that HAMs used that a lot in the past, as there wasn't anything much different available, apart from plain very thin PTFE laminates, which are pain in the butt to work with. Do you have any ideas, where to buy some in the EU, especially RO4530B substrate in about a the 0.5 to 0.8mm thickness range, without the cost being stupidly excessive due to "exclusivity" of the goods?

However doing such RF PCBs at home has major drawbacks and that is no plated vias.  I am used to work with small SMT stuff, including fine pitch  QFN (which most RF/uW chips are these days in anyway), where it is mandatory to use small vias and vias under the devices itself (exposed pad GNDs and such). Hence looking for a cheap manufacturer, that can do the 4530B or 4003C. (Guessing, the 4530B should really be the cheaper and suitable option for a home-gamer. Still huge leap better than FR4!)

I do not think the price is high due to increased precision expectancy, for example in the Bulgarian manufacturer, one can clearly specify the precision, yet the price is a nonsense.  The high price is I think mainly due to exclusivity. The vendors think, they can just simply rip off people, as corporates almost don't give a sh!t about the price (they just want the work done) and home-gamers are only a small if not almost zero fraction of the customer base for such special materials.

//EDIT: Even one of the chinese guys I have asked for 4003C gave estimate $300 for a small, 45x30mm pcb x 5pcs, with almost no drilling, no solder mask and silk, ENIG.  I was like WTF?!

[G0HZU]

However doing such RF PCBs at home has major drawbacks and that is no plated vias.

It's possible to buy rivet kits (eg from LPKF) for the PCB vias and this is what we do at work for our milled boards. When fitted correctly they do look like plated holes. The rivets come in various sizes. However, it really is a labour of love to fit them to a board and it requires some skill and experience to do it quickly and reliably. Otherwise it can become a frustratingly long and clumsy experience. Even then they need to be soldered to make sure they are reliable.

One thing to watch out for is that the ENIG finish can wreck the insertion loss of coupled line filters up in the GHz region, especially combline or interdigital designs. I've been there and made this mistake many years ago.

The nice thing about the Rogers 4xxx material is that it is repeatable and the dielectric constant is well documented for the various PCB thicknesses and metal weight/roughness. It's also quite good in terms of how little the dielectric constant changes in the xyz axis. So it usually gives rewarding results if you first simulate with something like Sonnet (Lite) and then make a PCB.

However, if you just want to mess about with a few designs to gain experience then maybe try a cheaper substrate. Even FR4 can provide a useful learning experience even though the dielectric constant will be poorly controlled. Many years ago GIL GML1000 (0.030") was a really good choice as it was cheap and reliable. But it went obsolete. However, it might be worth asking if you can buy old stock of it from a PCB house. Some of them still have it. I still have loads of sheets of it here and it is really easy to mill and doesn't wear the tools out as quickly as FR4 or (even worse) Rogers 4003 or 4350. The Rogers material wears the milling tools even more than FR4!

[radar_macgyver]

One thing to watch out for is that the ENIG finish can wreck the insertion loss of coupled line filters up in the GHz region, especially combline or interdigital designs. I've been there and made this mistake many years ago.

What's the reason for this? I've seen recommendations to not use ENIG for RF boards on some app notes, but without an explanation.

The OP might be able to buy substrates at this Italian site: https://www.rf-microwave.com/en/rf-microwave-laminates/371/
You may also be able to find cheaper quotes from board houses if you specify non-Rogers RF materials.

[G0HZU]

I don't know a definite answer but I think it's because of the nickel content. Nickel has a relatively high permeability compared to copper and the conductivity is less. How much it affects the insertion loss probably depends a lot on the substrate thickness and therefore how significant metal losses are for a given microstrip line and how the microstrip is employed (eg in a coupled line filter).

I got stung by this when I swapped across to ENIG for an RF preselector board many years ago. This typically had a suite of printed combline filters across 500MHz to a couple of GHz or so. Each section had a five finger combline followed by an LNA followed by another combline filter. The overall loss was something like 2dB worse with ENIG but I can't remember for sure. This was way back in the 1990s. But all the boards failed the test spec.

It was possible to slowly rub away the ENIG coating with a fibre pen and watch the insertion loss slowly recover to what it should have been. I've tried to avoid ENIG on printed filters ever since. Maybe the process is better now but I think this is a fairly well known issue with ENIG even today.

[Yansi]

I have found this interesting to read, regarding the ENIG insertion loss problem. However they stay there is negligible effect on a filter, compared to for example a solder mask covering. (which I think is sort of expected)?  However I am not sure what surface to use then.  Bare copper? That sucks (oxidizes). HASL? I don't think the rough surface finish will be good for it. So I might just stick with the ENIG. The insertion loss is not that much different, according to the application note. Certainly good enough for me.

https://www.rogerscorp.com/documents/2341/acm/articles/Ambiguous-Influences-Affecting-Insertion-Loss-of-Microwave-Printed-Circuit-Boards.pdf

To sum up this thread: There are not much options for Rogers from china, if any at all.  There is also not many options, even sourcing rogers PCBs within Europe. Third of the companies don't give a fuck about your single PCB, third quote rip-off prices (because they outsource it, seems the same for the chinese) and third is about usable. But still the price is driven insanely high because of fees like "technology preparation", which are some type of fixed fee (usually charged even for FR4 PCBs). Arrrgh!

I was quoted this (so far the best quote, apart from the "technology preparation")  from a Czech company Pragoboard.cz (they seem to have many kinds of Rogers substrates in stock ready) for board like this:

100x100mm, 2L PTH Rogers RO4350B 0.762mm, 18um Cu, >0.3mm drill,  >150um trace/space, 2x green soldermask, 1x white silk, HASL or ENIG (+$1/dm^2):
1pcs $79/pc, 3pcs 28$/pc, 5pcs  $19/pc, 8pcs $16/pc.
Technology preparation: $72
Lead time: 10 working days.
(Pricing without VAT).

Not cheap, I can tell you that, but at least they have a stock of the material. What SUCKS is that they want a minimum 6dm^2 minimum order. Bastards!!
That is fucking to much for just a filter to be tested here and there. 


I will probably create a different thread, about the microstrip filter design. I have experimented a bit with it, never got any useful results out of it.  Don't know how to do it correctly. 

[WPXS472]

Quote: The problems with return loss aren't because of FR4, it's because you kept the geometry the same and switched to a material with a different dielectric constant. All microstrip widths must be recalculated for the new substrate material. Also most major pcb fabs in china support impedance control nowadays (often at no extra cost) so you should be able to get return loss better than 20dB up to many GHz.
Our problems with FR4 weren't because we changed substrate materials, it was because the FR4 from different board houses was sourced from different manufacturers and had different characteristics, even though you couldn't tell the difference by looking at them. We also made some parts using printed, inductors and they varied from time to time. The particular I mentioned seemed to have 50 dB return loss at 1.25 GHz. I say seemed to because I didn't trust either the calibration nor the VNA we were using to be that accurate. Our boards were only about 22 dB at 1.25 GHz. I have personally never used FR4 above 1.25 GHz because we didn't make parts that went any higher. I wasn't too crazy about the performance of FR4 even at 1.25 GHz. The fellow who replaced me says he is investigating alternate board materials. I wish him luck in getting away from FR4. Those particular parts are very cost sensitive.

[Yansi]

There is a quite common material "FR408" which is supposed to be better behaved at the higher end.

//Well I have just looked, the DK of the FR408 seems to be pretty well controlled, about 3.66  - same  as the RO4350B, huh?
Y.

[eb4fbz]

PCBCart has RO4350B, i have used them a couple of times.

[Yansi]

PCBCart has RO4350B, i have used them a couple of times.

Thank you for suggesting them. Could you provide more info about the size of boards, substrate thickness and the resulting price please?

The question of where to manufacture any kind of Rogers PCB prototypes is still open. It seem that most (if not all) manufacturers have the pricing set ill and only for corporates.  Which is not unexpected, but makes microwave/RF hobby a bit tough, when one wants to do something more serious, than a HF transceiver from a kit. 

Y.

[eb4fbz]

30x25mm individual boards:
30mil RO4350 (0.762mm)
0.5oz (18um) copper
No soldermask, no silkscreen

PCB Cart:
Tooling cost: 40€
Unit price: 0.7€/pcs for 100 units

[Yansi]

Thanks for sharing. However I am a bit afraid that my opposite requirements (larger board, less pieces) will invoke some nasty price.
However I find those 40€ still way better than anything I have found so far. As soon as I will have something finished, I will try asking for a price.

Y.

[rx8pilot]

30x25mm individual boards:
30mil RO4350 (0.762mm)
0.5oz (18um) copper
No soldermask, no silkscreen

PCB Cart:
Tooling cost: 40€
Unit price: 0.7€/pcs for 100 units

I just submitted a quote to PCB Cart for a Rogers design. 3 days to see what they say.

[Alex Eisenhut]

https://www.matrixelectronics.com/products/panasonic/megtron-7/

[vk6zgo]

I don't have any experience with Chinese suppliers and Rogers substrates. Sorry. But I do have experience with Chinese PCB suppliers and FR4. First, FR4, though good for a lot of things, is awful for RF.

Gee! I'll have to tell NEC, Harris, Thomson, etc, that they have been doing it wrong for the past forty plus years!

It doesn't even have to be microwave either. At my for employer, we were dealing with the Chinese branch of a major American company trying to get some products on their approved vendor list. They didn't like out test fixture at all and provided their own. Though I had some problems with how it was constructed, it did have excellent return loss. They provided the Gerbers for it and we had some made by our Chinese supplier. Even though they physically looked identical, the return loss was simply terrible. This was only up to 1.25 GHz too. We used a Chinese supplier almost exclusively, much to my dismay. They were difficult to deal with from an engineering perspective. Whenever we would send a set of Gerbers and the specs for the materials and panels, they would always send a big list of questions that had to be answered. Sometimes, it was next to impossible to even understand what they were asking. I don't know if they were using translation software, or they just didn't do well with English. One habit they had that I found particularly troublesome, though I admit others might not find it a problem, was that I always sent all dimensions in Imperial measurement format. I think in inches, not centimeters. They would always convert every measurement to metric form when replying. Just out of stubbornness, I would always convert everything back to imperial when I made my replies. If I could have had my way, I would have told them that as the customer, I demanded that they use my measurement system if they wanted to continue to do business. Even though wanted every aspect and detail documented, they had a habit of just changing things without telling us. My advice when dealing with a Chinese supplier is to be up front about what is important to you, provide every detail imaginable, and just let them know that if they want your business, they have to meet your expectations. If you specify something like Rogers material, expect to pay a heavy premium for it because it likely won't be locally produced. The idea that everything that comes from China is junk, is no longer valid. A lot of Chinese companies can do excellent work.

We had a bunch of amplifier modules which were supposedly built on Rogers material.
One, maybe the prototype, was probably the real thing, but all the others were on some stuff which looked the same, but after a few times when the LDMOS devices "cooked up", (another story*), started to come apart.
FR4 would have been a lot better!

*These amps were 300watt units, combined to make 600watts, then further combined to ultimately make 1.2 kW.
The problem was that the Chinese manufacturer didn't individually tune the amp modules, so they would cook the unbalance loads, in turn, unbalance load would cook on the  next combination down, & so on.
These were UHF amplifiers.

[IconicPCB]

I have had some experience with rogers 4xxx materials fabricating cavity structures.
not  purchasing finished boards, actually fabricating them myself.
Structuring copper and laminating /pressing multi layer structures in my shop.

There are a couple of things to observe when working in the 5 Ghz region.
Surf ace structure will affect the return loss. It will send the group delay crazy if surface is not maintained in untouched un scoured condition.
It is a difficult material to laminate despite claims that it can be processed with fr4 like process parameters.

Having addressed technical issues let me address commercial realities by saying:
"...  Sell on price, live on rice..."

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.