Fixture for testing tiny MMIC at 8GHz

[tkamiya]

I had a good fortune to buy some HP equipment dirt cheap for parts.  Inside it, I found quite a large number of MMIC.  These are really tiny ones where diameter of molded part is 2 to 3mm.  4  leads are sticking out radially, and the are 3 mm at most.  I'd like to test these to make sure I didn't damage them in removal process.  These amplifiers are good for 8GHz but power range is something like 15dbm.

I have tracking generator and spec analyzer.  What I can't figure out is how to rig up the fixture.  Something small, something that does not require solder, and something good enough to make sure each of them work.  Obviously, at this frequency, banana clip is of no use.  Does anyone have any idea how I can accomplish this? 

[Wolfgang]

Hi,

Plan A: Make a ROGERs PCB with gold plated tracks and a hole for the MMIC.
Use edge-mount SMA connectors.
Use a spring-loaded PTFE cylinder to press down you MMIC to the PCB.

Plan B: Solder them in, with only little solder. They normally survive this.

[tkamiya]

Hello.

What is ROGERs PCB?  That's a term I've never heard of.

[DaJMasta]

Rogers is a controlled dielectric PCB material designed for high speed work.  Odds are good your PCBs in question are made of something similar.


If I'm understanding the description correctly, these are fairly standard packages for MMICs, a small round package with flat leads in a plus shape coming out from them, and I've worked with a couple similar frequency devices repairing some RF gear, but making a fixture would be tough.  I agree that making a small PCB is the best route if you need to test them out of circuit, these are going to need to be biased to operate properly, so you'd likely need several passives in addition to the MMIC, and to try and ensure good 50 ohm matching, you'd need calculated stripline on a known dielectric, and at those frequencies FR4 is quite lossy.  Certainly possible, and maybe something like this already exists, but a costly way to test a few chips.

I think the much simpler way is going to be to test them in circuit.  The bias supply input should be obvious from the layout, as generally are the input and outputs, and with removing or rotating a few caps or resistors, you should be able to disconnect it from the circuit (if needed) to connect up a cable to monitor the signal or feed something into it.  If the board can power up, you could try using a piece of wire or an EMC probe to couple in a signal to your SA to get a relative idea of the input and output signal levels, though it can be quite variable.  I've found that monitoring the current passing through the amp and monitoring the temperature of the package relative to others in the chain with a thermal camera can be very good for diagnosing issues as well.

While sockets to remove it and test it without soldering in place may exist, they would be expensive specialty items.  It would probably be cheaper to replace dozens or hundreds of the amps in question without testing them than it would be to get your hands on such a fixture, and it may just have to be custom designed.


EDIT: If it helps, this is a pic of an RF FET in a similar package (at least, so I understand).  This one also needs a negative gate bias, but you can see the gate bias injected on one side, the DC power injected on the other, the filtering to keep the HF out of each line, and the RF path with DC blocking caps on either side.  Disconnecting a passive on the end before the voltage comes up through the board or probing the end of a DC blocking cap should give you an out of circuit in circuit testing platform that's at least well designed and on the right substrate - close to the best conditions you could expect a test jig PCB to offer.

[LukeW]

You can buy or make a general-purpose MMIC gain block board that gives you the grounding, coupling capacitor, connectors and bias tee.

https://www.w6pql.com/LNAs%20(preamps)%20and%20MMICs.htm

Here's an example.

Useful for lots of things.

Do you really need Rogers? What's the frequency?

Test it with a good MMIC first to give you a baseline.
Maybe you could clamp it down carefully, as a solderless test fixture onto the PCB.

[hendorog]

You will be fine with FR4, just keep it small with very short tracks between the MMIC and the in and out connectors.

The loss/inch/GHz is higher for FR4 than Rogers - so just use less inches to reduce the total loss.

You can see this effect here in the graph:
https://www.embedded.com/print/4398951

Keeping tracks short relative to the wavelength will also reduce the effect of transmission line impedance. The Er of FR4 is more variable than Rogers and so you are unlikely to end up with a 50ohm trace on FR4. So make the traces short and limit the damage.

[David Hess]

The extra loss of FR4 can be an advantage if it prevents spurious oscillation.

[radioactive]

Another inexpensive option for good RF performance is OSHPark 4-layer service if you have a small board size like this.  They use FR408 material (lower loss than FR4 and much more consistent dielectric from run to run).  I frequently utilize CPWG (co-planar waveguide) over ground for this type of thing.  I prefer to use layer 3 for ground and top layer for the feed.  Use lots of vias for these high frequencies. (should look something like the image).

[Wolfgang]

... in the first post the OM mentioned 8GHz as the frequency limit.
FR4 for 8GHz, and for *measurement* purposes ? Naah, never, sorry.
Not even *one* reputable manufacturer of test fixtures does that.
FR4 is lossy, its dielectric constant has large tolerances and is frequency dependent. Nobody would want such a material for measurements.
Whats wrong with a ROGERS or similar substrate ? Its very small, so cost will be manageable.

[radioactive]

@Wolfgang, 

OP also mentioned that he just wanted to test the devices and make sure he didn't damage them in the removal process.  A one inch line with coplanar waveguide using FR4 could  be less than 0.7dB loss at 7GHz and less with FR408.  I agree that this wouldn't be the way to go for device characterization, but more than adequate to test functionality.  Many real-world products utilize the lower cost materials at higher frequencies.  The cost savings is pretty significant.

This is a good write-up on the subject with some actual measurements:  https://www.edn.com/design/analog/4398951/1/What-PCB-material-do-I-need-to-use-for-RF-

[Wolfgang]

Consumer products and test fixtures are a different matter, I agree. When you look at ebay, a PCB with a microwave material for a MMIC costs about 12$.
I cannot see why this should be so critical, and results will be better and more reproducible.

[tkamiya]

I had an idea. 


I still have the original PCB these devices came from.  Associated (short) microstrip line and associated bias circuits are still in tact.  However, they were surface mounted like usual.  I suppose I could cut this board and separate the relevant section.  Then, punch a hole to receive the MMIC.  Then, to test, simply drop in the chip, align it with the trace, and use something to apply pressure.   It may just be enough to test....  I'd rather not solder because there isn't much lead left. 


Microwave region is a whole new thing for me.  I am forced to think and rethink my methods.

[xaxaxa]

Problem with rogers PCB isn't just cost, it's finding a PCB fab that will even do it; all of the prototype PCB services rely on combining many customer's designs onto the same panel and as soon as you deviate from the standard options either the costs skyrocket or they refuse to do it. It's not a 50% increase in cost or even a 2x increase in cost, more like 20x, and unfortunately the board being small does not even matter.

[TheUnnamedNewbie]

A lot of people talking about Rogers like it is one single material. Rogers is a brand, just like amphenol or molex. They have a lot of different materials, and not all are for the same purposes (and not all are even good for RF applications). And one of the reasons to get Rogers is often not just the lower loss, but the stability of things like the relative permittivity of the material v.s. temperature and frequency, and how well controlled it is over a number of different sheets of material.

Regarding making Rogers boards: This is one of those cases where it could be beneficial to make your own PCBs. Rogers material can be had for acceptable prices (few tens of USD/EUR), but nobody wants to process it for prices you can afford because it requires them to set up a process just for your one board. I've heard quotes from people working with multi-layer rogers/alumina boards, and they were talking about 10k EUR for a run of a dozen or so PCBs.

However: sometimes you can look at the local university EE lab. I know at my university, we do still have some in-house PCB production for exactly this reason: simple, small test-fixture boards that are one or two layer. You can always talk to them and see if they are willing to do a run for a hobbyist.

I agree with the 'pressure' method. This is in essence how we test on-chip - using probes that apply pressure. It could be interesting to make your own probe like this, but I don't know how hard it would be and what mechanical tools you have at your disposal.

A test PCB with gold traces would be best, since gold does not form an oxidized layer. I would also suggest going over the leads with something to get rid of the oxidation just prior to mounting the MMIC.

[Wolfgang]

Hi,

of course Rogers makes a lot of high-quality substrates. In the case here it is not so decisive which one you use, as long the 8GHz properties are good enough.
Whats wrong with using a prefabricated PCB for MMICs from ebay ? The MMICs the OM is asking for here is the standard propeller or SMD 4-terminal button style cases, so that should not be a problem. Why is it necessary to roll your own PCBs ?

[TheUnnamedNewbie]

Hi,

of course Rogers makes a lot of high-quality substrates. In the case here it is not so decisive which one you use, as long the 8GHz properties are good enough.
Whats wrong with using a prefabricated PCB for MMICs from ebay ? The MMICs the OM is asking for here is the standard propeller or SMD 4-terminal button style cases, so that should not be a problem. Why is it necessary to roll your own PCBs ?

I wasn't referring to you specifically, Wolfgang. And there are prefabricated PCBs for MMICs but I do not know what type of MMIC the OP has, so I suggested a general options. It is not necessary, but I just wanted to inform them about the option

[Wolfgang]

dont worry, I just wanted to make his life easier. 

[tkamiya]

Now that I know what "Rogers" refers to, I don't think I want to go that route for a temporary use jig. 

I don't have much of mechanical tools.  Besides common hand tools, I have a small lathe and a drill press.  (oh yeah, a weed wacker, too)  So nothing complicated is possible.  My current plan is to use a spring loaded mechanism and use 32 gauge 4mm long diabetic needles for probes.  It comes with a plastic piece that can be hot glued.  I'm not really doing performance tests.  I'm more of doing works/broken test. 


I'll keep you updated.  I'll post a pic when done, so we can all have a good laugh.

[Wolfgang]

Sorry for misunderstanding what you want. For a works/broken test FR4 ist fair enough, of course.