Suggestions for beginner - thin PCB (for >GHz RF)

[cdev]

Hi,

I an a beginner and despite my relative inexperience I enjoy building GHz range RF gadgets.

I've recently found the performance at higher freqs. is markedly better with really thin PCB. I also am on a budget and need every purchase to be a consistent performer, no drama. Most importantly I want to know what the dielectric constant of the PC is and I don't want it or the thickness to change from batch to batch. I want it to be predictable.

I cant afford the expensive (Rogers, etc) materials- Devices I make are all really small, Typical 5 x 5 cm. Just basic building blocks I put together for my own self-learning, I'm an experimenter. My main interest is antennas, for GHz range reception. I would like a PCB that I could both work with myself at home and also expect a PCB house to have if I wanted them to produce a bunch of them.

What PCB material would you use? Ive had strikingly better performance in my application with very thin PCBs. I don't know where the material came from, though. It was a scrap.

I'm embarrassed to say that I don't even know the standard thin sizes. The back of the SMA connectors I have from ebay have to be padded out with copper so they work on this thin PCB. There must be cheap SMAs available for thin PCB. If not, I guess I would rather hand pad them than pay $2-3 each. 

So, that gives you an idea of my budget. But the only way to learn is to do stuff.

[T3sl4co1l]

How thin is "very"?

1mm, or 0.8mm (1/32"), is the commonest "thin" material.  Thinner can be made, usually at greater expense (custom build).

Tim

[cdev]

How thin is "very"?

1mm, or 0.8mm (1/32"), is the commonest "thin" material.  Thinner can be made, usually at greater expense (custom build).

Tim

According to my old vernier caliper, the board I am having good luck with is 1/32 so its 0.8 mm,

Are any manufacturers whose boards are available inexpensively (a board I can buy a small quantity of and know its the same as the one a PCB house uses on my tiny production run if I make more)  and also at PCB houses particularly good with dielectric consistency from batch to batch?

[Howardlong]

I use 0.4mm for some prototyping when space is critical in the RF area, but I fab them myself, not sure if there are board houses offering this. CIF ABC16 which is also presensitised is what I use. In fact I only use CIF nowadays, too many inconsistencies with other boards to bother using them.

[tggzzz]

For commercial PCBs, you might like to consider looking at the stackup of 4 and 6 layer boards. There the electrical thickness can be a little as the prepreg thickness.

Different manufacturers have different products, and not all of them provide decent specifications.

[cdev]

I dont know anything about the photoresist PCB method at all, ive just been using toner transfer and hydrogen peroxide/vinegar/salt.  Which is nice because there is no toxic stuff to dispose of issue.

I'll have to investigate that, though. As i said I am still a beginner.. I like the way i do it now because its not a production at all, I can make a single PCB and etch it in a little tupperware container while I do other things.

The only thing that is still not really there yet with my setup is drilling precisely and my vias. Still trying to figure out the best way to do that.

[zapta]

IIRC elecrow has good deal on 5x5cm including 0.8mm, you can send them Gerber's for solid boards with no stop layer is you like.

[hagster]

By going thinner you are reducing dielectric losses. The downside is that you increase skin effect losses(as you have less copper width), but there is a balance to be struck. With a thinner substrate you need to be more accurate with track widths to get near your desired impedence. I would also expect thinner boards to be more effected by changes in permitivity, but you can do some tests with an impedance calc tool to confirm this for yourself.

I would guess that 0.5 or 0.8mm would be about right for most purposes.

Eurocircuits do an RF pool which is worth considering if you can afford it.

There are a few surface mount SMA options that sit in a slot in your board. These are largely unaffected by thickness. Generally I would advise that you don't buy the cheapest Chinese knock off for this. If it's not right you will probably never be able to tell without access to some serious rd test kit.

[tggzzz]

The only thing that is still not really there yet with my setup is drilling precisely and my vias. Still trying to figure out the best way to do that.

AFAIK all chemical ways of depositing vias imply usage of unpleasant chemicals. Mechanical ways are to use the leads on through hole components, wire stakes, or copper rivets pressed and then soldered.

If you are working at GHz frequencies, you may find that the woven structure of FR4 substrate affects your circuit's operation. It might be helpful to fo more research.

For aerials, have you considered sticking self-adhesive copper foil tape onto a thin piece of material of your choosing, and then cutting it to shape?

[ve7xen]

DirtyPCBs will do down to 0.6mm for their standard (reasonably low) price; standard FR4 of course. Their tolerances (and lead time, IME) are quite good compared to most budget PCB houses. They specify impedance consistency of 15%.

iTeadStudio I have experience with, they will do down to 0.8mm for a small surcharge, they're slightly cheaper for the boards, but with shipping cost more for small orders. I don't use them much anymore, but they're probably still more economical if you're ordering a bunch of designs at once and/or need express shipping.

[nctnico]

I use 0.4mm for some prototyping when space is critical in the RF area, but I fab them myself, not sure if there are board houses offering this. CIF ABC16 which is also presensitised is what I use. In fact I only use CIF nowadays, too many inconsistencies with other boards to bother using them.

You are using CIF boards? Those are the worse crap I ever got. I binned the whole lot of it for the sake of my sanity. Some have a very thick layer of resist and others have an extremely thin layer. I got mine from Farnell BTW.
Anyway, I wouldn't use too thin substrates because the thinner the trace the more effect you'll have from etching tolerances. For self etching I'd certainly use 0.5mm as a minimum width if a trace needs to have a certain impedance.

[nctnico]

I dont know anything about the photoresist PCB method at all, ive just been using toner transfer and hydrogen peroxide/vinegar/salt.  Which is nice because there is no toxic stuff to dispose of issue.

Copper is a heavy metal which should not end up in the sink. Where do you leave your etchant??

[Howardlong]

I use 0.4mm for some prototyping when space is critical in the RF area, but I fab them myself, not sure if there are board houses offering this. CIF ABC16 which is also presensitised is what I use. In fact I only use CIF nowadays, too many inconsistencies with other boards to bother using them.

You are using CIF boards? Those are the worse crap I ever got. I binned the whole lot of it for the sake of my sanity. Some have a very thick layer of resist and others have an extremely thin layer. I got mine from Farnell BTW.
Anyway, I wouldn't use too thin substrates because the thinner the trace the more effect you'll have from etching tolerances. For self etching I'd certainly use 0.5mm as a minimum width if a trace needs to have a certain impedance.

Oh dear! I spent quite a long time battling with presensitised boards from Mega and others. I would often get regular streaking across the copper, almost as if it wasn't a uniform thickness either on the copper or on the photo stuff. I suspected my printer at one time. Almost as bad was that the way they were supplied, a dozen or so eurocard sized boards taped together, there were often scratches despite the plastic protection. The same seemed to apply to a few other boards.

Then by luck I used some CIF stuff because they had the thinner substrate for an RF project. I've never looked back, I always have stock of the 0.4, 0.8 and 1.6mm stuff (part numbers ABC16, ABB16 and AB16 respectively for double sided presensitised. It's the most consistent I've ever been able to do my own boards by a very long way. I can't remember the last time I had a board fail on me and I has to re do it (famous last words!) These boards are the sort of things I tack onto an order to make up the minimum order value.

I guess it's down to that magic sauce that you get when you finally have the process working. CIF was a part of it for me, although looking at others' experiences they seem to manage well with other methods. The toner transfer method for me was a failure, although I didn't try too hard. Truth is, I can get the chemicals out on the counter and have a board out and all packed away well within half an hour using the traditional photo method, down to 6mil track and gap if I push it, but usually I stick with the thickest I can get away with.

The boards I do are generally small, maximum maybe 5x5cm, for unit testing prototypes.

With the thinner boards, you do need to be very careful about flexing, they are not at all robust. This can lead to ceramic caps cracking almost invisibly, as well as making it a crap shoot with BGA type hidden pads if the board lacks coplanarity.

For vias, I drill 0.7mm holes and insert 0.4mm I/D copper rivets. For RF, I've got into the habit of putting them very frequently to stitch the upper and lower groundplanes. These are quite big vias, no good for production really, but for prototypes it's fine. Trying to drill much below 0.7mm at home is going to break a lot of drill bits even with a decent drill press.

Just me findings, YMMV of course.

[nctnico]

I had the same order numbers as you but the thin substrates where impossible to get exposed right. Even with exposure times of 30s (normal PCBs require about 4 minutes on my UV exposure unit) the result was crap. Maybe I got a bad batch. Would you care to post a picture on how the resist layer looks on your CIF boards? And where did you buy them?

[kingofkya]

I have had good luck with theses guys in addition to elecrow (i use both interchangeably)with .8mm pcbs although i don't do rf so, they go down to .4mm but cost goes up at that size.

http://www.pcbway.com/

[G0HZU]

I've recently found the performance at higher freqs. is markedly better with really thin PCB. I also am on a budget and need every purchase to be a consistent performer, no drama. Most importantly I want to know what the dielectric constant of the PC is and I don't want it or the thickness to change from batch to batch. I want it to be predictable.

I cant afford the expensive (Rogers, etc) materials- Devices I make are all really small, Typical 5 x 5 cm. Just basic building blocks I put together for my own self-learning, I'm an experimenter. My main interest is antennas, for GHz range reception. I would like a PCB that I could both work with myself at home and also expect a PCB house to have if I wanted them to produce a bunch of them.

What PCB material would you use?

You could try getting friendly with the nearest PCB company to you. They may sell you scrap offcuts cheaply. I'm lucky in that I have two very close to where I live and they both deal with exotic PCB materials. But I'm also lucky to have access to offcuts at my place of work.

Or you could try asking ones that are further away if they sell offcut material. Sometimes this stuff appears on ebay and you can buy little offcuts of Rogers 4003 or 4350 quite cheaply.

However, if you work with 0.02" thick 4003/4350 then it is very brittle and very easily can snap in half like a biscuit.

Years ago I used to do a lot of design using GIL GML material as it was a very good performer and was cheap. But they went out of business and the material is very scarce now. But you might find a PCB house that will sell you old redundant stock of this? eg GIL
GML 1000-030 is (was) lovely stuff to work with at home because it is easy to mill and quite cheap to buy. Much cheaper than Rogers.

http://paginas.fe.up.pt/~hmiranda/etele/g1000_30.pdf

You could also try and find a reputable seller of thin FR4 material eg 0.032" or less and they may be able to tell you the dielectric constant at various frequencies if they always use the same FR4 from the same manufacturer.

[Dago]

Basically the available budget solutions are 0.8mm two-layer board (available from almost all vendors), there are some vendors that can do thinner but not terribly common and then there are 4-layer boards. 4-layer boards usually have a 180 or 200 um pre-preg which is pretty nice for RF. 4-layers boards are also way under 50$ nowadays from China.

If you want actual RF-boards with an RF-substrate then I know at least pcbcart does them, but you need to sen in your Gerbers to get a quote.

Inside EU I once checked that RF PCBs from Eurocircuits weren't as expensive as I thought, but still something like 300€ for a few small boards IIRC.

[Howardlong]

I had the same order numbers as you but the thin substrates where impossible to get exposed right. Even with exposure times of 30s (normal PCBs require about 4 minutes on my UV exposure unit) the result was crap. Maybe I got a bad batch. Would you care to post a picture on how the resist layer looks on your CIF boards? And where did you buy them?

The boards are from either rswww.com or Farnell, depends on who I am buying from that day, and I need to meet the minimum order value. They are supplied in red plastic sealed envelope as well as the peel-able plastic cover on both sides, the two layers definitely minimises the chance of scratching.

The resist looks a dullish purple.

2.5 minutes UV exposure each side (ingrained in my head!) using a Megauk LV202-E exposure unit.

When exposed there is very little contrast between track and gap. Once it's been in the developer, it comes up a very, very clear black with the etching areas clean copper. Maybe  it's 15s or so in the developer, but I do it by eye.

Two pics, I tried to get the best representation I could, but I'm no photographer.

[cdev]

I dont know anything about the photoresist PCB method at all, ive just been using toner transfer and hydrogen peroxide/vinegar/salt.  Which is nice because there is no toxic stuff to dispose of issue.

Copper is a heavy metal which should not end up in the sink. Where do you leave your etchant??

I have been disposing of it in the sink but the volume of spent etchant which is just the copper, hydogen peroxide and cider vinegar is very small as my boards are never more than a few inches in size. Think Raspberry Pi size, or a bit bigger.

Sure, copper is toxic but its so very diluted by the much higher volume of other stuff in there. For example, water. We really are talking about tiny PCBs.

I like making individual devices that can be connected in different combinations as a learning platform.

At some point I may want to do big boards and at that point I will start sending them out so somebody else will be disposing of the etchant as their local laws require. Probably a domestic PCB house.

[nctnico]

I had the same order numbers as you but the thin substrates where impossible to get exposed right. Even with exposure times of 30s (normal PCBs require about 4 minutes on my UV exposure unit) the result was crap. Maybe I got a bad batch. Would you care to post a picture on how the resist layer looks on your CIF boards? And where did you buy them?

The boards are from either rswww.com or Farnell, depends on who I am buying from that day, and I need to meet the minimum order value. They are supplied in red plastic sealed envelope as well as the peel-able plastic cover on both sides, the two layers definitely minimises the chance of scratching.

The resist looks a dullish purple.

That looks way better than the material I got! On the boards I binned the resist was much lighter (thinner) and had a marble effect. I guess you got lucky and got the good stuff.I think the conclusion here is that the quality from CIF can vary.

@cdev: a lot of little bits make a large chunk! I use ferric chloride which I reactivate with hydrochloric acid (10%). That way my etchant can be used for decades and many boards. And if I need to get rid of it I take it to a chemical waste disposal station.

[Howardlong]

....
The resist looks a dullish purple.

That looks way better than the material I got! On the boards I binned the resist was much lighter (thinner) and had a marble effect. I guess you got lucky and got the good stuff.I think the conclusion here is that the quality from CIF can vary.

Just to add, I use the LaserStar film from MegaUK, this is a translucent film, not transparent, it's like heavy tracing paper. 2.5 minutes is actually a bit longer than I used with the other boards I had, I used to do 1.5 to 2 minutes a side with those.