Designing my own Power Splitter: Connector

[LoveLaika]

Ahoy.

As my first foray into RF circuits, I'm looking to design my own power splitter to take in one voltage signal and output it to two coaxial outputs. I'm trying to replicate a design I saw (mildly surprising to see this here of all places) that used a resistive-Delta configuration with my frequency range being roughly DC to ~500 MHz, nothing in the GHz range. Looking at such a design, I have a question about the connector.

You can use any type of connector (provided it fits your needs), but is there any advantage to a board edge one vs a typical right-angle/through-hole one? I see a lot of them have edge-mounts, so I'm curious if there was any advantage to it. Following that, I'm a bit concerned about the spacing. Looking at the connectors below, they're edge mount, but would they fit a typical PCB board thickness of ~1.7 mm? Is this a general board standard?

• BNC
• SMB

[fourfathom]

If you have a choice, I suggest you use an edge-mounted SMA connector (like this: https://www.digikey.com/en/products/detail/molex/0732512120/1465158).  These fit perfectly on a standard 1.6mm PCB, and make it easy to provide a good constant-impedance layout.  And SMA is much more popular than SMB.  I also like BNC, but prefer SMA.  That Molex connecter from Digikey is slightly expensive, but you can get Chinese knock-offs on Amazon (https://www.amazon.com/gp/product/B009EOORXO/) for less, and eBay even more cheaply, and they are adequate for most uses.

I suppose, given an appropriate hole pattern, you could also mount these upright on a board, but there are better footprints for that.

When you lay out your board, you might want to place the SMA connectors far enough apart that you can screw on SMA/BNC adaptors (and still leave room for the BNC connectors).

[LoveLaika]

Thanks for replying, but i have a reason for my choice of connectors. So, the reason for edge mount is mainly for smaller footprint size? That's what it seems like from your reply, and thinking about it now, it would make things easier in some ways.

Now that I think about it, the distance is copper to copper (no mask or silkscreen), so maybe my board thickness is smaller than what I thought

[fourfathom]

So, the reason for edge mount is mainly for smaller footprint size?

  I use edge-mount SMAs because they work well with my boards and enclosures, and much of my test equipment uses SMA.  The larger test gear uses N (not going to be able to fit that in my designs, and don't need the performance), or BNC.  There's nothing wrong with SMBs -- I've used those in some commercial telecoms designs -- but they are less common.

I believe the edge-mount connectors may have better performance in the GHz region, at least with simple 2-layer board layout, but I could be wrong about that.

[LoveLaika]

Thanks. I guess that's what I'll go for. For the record, I use BNC and SMB is because that's what I'm using in my equipment. If I may ask, what do you think about the BNC connector I linked to? Looking at the gap of 1.65 mm, that seems a little tight, but is that okay for a standard PCB board thickness?

Also, more of a design question, but how do you deal with controlled impedances via a trace width? Given a target frequency (say for example 2 GHz), I can use Saturn PCB to estimate the max conductor length. Using that frequency to calculate the impedance of a conductor, the width has to be very large in order to get an impedance of 50 ohms (Microstrip, copper and plate thickness of 1 oz or 35 um). Is that normally how it goes with design?

[fourfathom]

Thanks. I guess that's what I'll go for. For the record, I use BNC and SMB is because that's what I'm using in my equipment. If I may ask, what do you think about the BNC connector I linked to? Looking at the gap of 1.65 mm, that seems a little tight, but is that okay for a standard PCB board thickness?

Also, more of a design question, but how do you deal with controlled impedances via a trace width? Given a target frequency (say for example 2 GHz), I can use Saturn PCB to estimate the max conductor length. Using that frequency to calculate the impedance of a conductor, the width has to be very large in order to get an impedance of 50 ohms (Microstrip, copper and plate thickness of 1 oz or 35 um). Is that normally how it goes with design?

I'm not a mech engineer, but with that edge-mount BNC connector I would be concerned about leverage forces peeling the traces off the board.  Perhaps it's not an issue, and of course it's going to depend on what's plugged into it.  As for the dimensions, I've got SMA edge-mount connectors that show 1.73mm and  1.78mm ground-pin clearance, but also one shows the center pin to underside ground pin clearance of only 1.47mm.  The gtound pins on my SMAs and your BNC all show a slight chamfer, and a snug fit is better than a loose fit, so I would guess that your BNC is going to fit.

I'm not an expert in impedance control, having only worked with FR4 boards, but for short traces I just make them as fat as practical, knowing that double-layer 50 Ohm microstrip over FR4 needs a trace width of about 2.7mm (in a 1.6mm thickness board, 1 oz tinned copper).  I can't usually get them that fat, but for electrically short traces it isn't so critical.  You can use a fat trace and neck it down where necessary, but I don't usually bother.  If I were designing GHz test fixtures I would be more worried about impedance control.  Generic FR4 doesn't have a tight dielectric constant or loss characteristics so I don't see much point in too much finesse.

I look forward to seeing some more-informed opinions here!

[LoveLaika]

Thanks for your reply. I realized suddenly that I was looking at it all wrong. I was envisioning my board traces as a microstrip when I think it should be a coplanar wave. See what I mean here? I have both copper pours on my 2-layer board tied to ground, whereas a microstrip only has one plane on the bottom side. Changing that in my calculation removed the frequency dependency that I saw in microstrips, allowing me to use a much smaller conductor width. Previously, with a microstrip, I had a trace width of over 2 mm (which seemed rather impossible for such a small distance), but treating it as a coplanar wave, the width was reduced to something much smaller.

Boy, that does make me think about all of my previous projects.

[fourfathom]

Yes, the coplanar layout definitely helps with the trace width, and apparently with the high-frequency performance.  Using this calculator (https://chemandy.com/calculators/coplanar-waveguide-with-ground-calculator.htm) for FR4, 1.5mm thickness, 4.5 E_R, 1.3mm trace width, 0.25mm gap width we get about 52 Ohms Zo.  But you need lots of feedthroughs to nail down that top ground layer.

[mc172]

What are your EMC requirements, or do you not care?

[LoveLaika]

At this point, with my application, EMC is not something to worry about.

[mc172]

OK, no worries then.