Geoffrey Hazelett – Impedance and Polar Instruments Tools
aDuring PCB West, we met with Geoffrey Hazelett to discuss one of the most important PCB topics: impedance. Vice President Sales at Polar Instruments, he also explained the importance of understanding electromagnetic models, solutions of Maxwell’s equations in a two or three-dimensional environment, etc., and shared some insight on PI’s tools.
0:10 When talking about a 50-ohm impedance line, do designers really understand what impedance means here?
Designers generally do. A lot of them come from the realm of, these days, electrical engineering backgrounds, RF (radio-frequency) backgrounds. Even in the design side they’ll have their layout people, will have a lot of physics backgrounds. It’s more an issue in understanding in some of the cases of communicating their interests and needs to fabricators. And fabricators communicating back what the tolerances are, what are the things that are going to change in the fabrication process. Because it’s actually kind of the reverse in a way of fabricators are making changes to a board that affect the designers. What the designers haven’t communicated are important things to the fabricator.
We have some field-solving tools that designers have been primarily using for a long time. They are a few steps above what the fabricators have been using. We have seen a significant growth in that, for fabricators stepping in to looking at things like copper roughness, changing dielectric values and observing how that’s affecting the insertion loss that as the designers are stepping into the realm, the fabricators are stepping into the realm of low loss laminates. Now things like slight variations in the width of the trace due to the etch factor and copper roughness have significant impacts now on what the designers are interested in. And that’s beyond impedance. You’re still 50 ohms. You still get that target impedance but changes elsewhere might make it so the signal is not getting through anymore.
1:59 Do designers need to understand electromagnetic models, solutions of Maxwell’s equations in a two or three-dimensional environment, etc.?
That’s interesting. Looking and saying how much do the designers need to know about the physics of what’s going on is a loaded question. The answer is yes. They need to know what’s going on so that they can make intelligent design choices about trace widths. It’s more than just fitting all of the traces into the board to connect the chips. They can do really wide traces or really narrow traces and still achieve all of the connections that they’re looking for.
But if their impedance isn’t matching, or the impedances are matching but… Referencing to what I said before about insertion loss due to copper roughness, you might not have a signal coming out the other end. But you are able to lay it all out. The more a designer understands and, likewise the more the fabricator understands, the physics of what’s going on enables them to make intelligent design choices for what’s going to actually be manufactured.
When it comes down to things like various copper – I’m going to keep saying copper roughness because that’s really a key thing these days that we here from designers in being able to get the right copper roughness out of their factory, the fabricator that they’re using fundamentally comes down to communicating that because if that’s not understood by both parties of why it’s significant, things can go awry – and I’ve heard that from a number of designers in the Bay area recently about that.
3:30 When does copper roughness become really relevant to the designer?
It depends. Eric Bogdan is famous for saying “it depends”.
It depends on the application, the materials that are going into that stack-up. If the designer has set out their design to have a sufficient loss budget that they can use RT foil, then that might work for them.
In other cases, they might need VLP or HVLP, or you named the copper roughness value out of the different laminate suppliers and copper foil suppliers. Those have different roughness values and it’s kind of a growing issue of trying to understand and characterize that. And that’s why tools like our SI-9000 show where the insertion loss is starting to drop off at different frequency ranges as you go up in frequency.
And that’s really the key thing as a lot of designers are stepping into PAM-4 and they’re shooting for targets of 40 and 50GHz for their designs – copper roughness is now the issue that the industry is looking at, partially because most factories that they’re working within that GHz range already have good, tight controlled impedance traces. They’re able to manage the pressing, the etching factors, and so now it’s stepping into what are the next things that the factory can do to improve quality and get the customer what they’re looking to build.
5:09 What can Polar tools do regarding losses at various frequencies?
In both in the modeling and the measurement sides, Polar Instruments has tools, like the SI-9000 which is our field-solver, and then when that’s paired with Speed stack – our stack-up tool – we can then send those structures that the designer puts into their stack-up, the factory can get that, use our CGen tool for coupon-generating – to actually create insertion loss coupons as well as impedance coupons. We also actually, we call it “completing the loop”, of then going on into our hardware test tools that – you’ve heard of the Sits – we also have an atlas line which does insertion loss testing.
So, we have a number of factories in the US and in Asia that have this set of equipment so that they can model the structures so that they know what’s going to happen ahead of time, build it, test it, and then take that data and information and put it back into the tools for post-fabrication analysis.
So after you do a cross-section of that coupon, put those values in, model it out. Then, take the measurement value, put that into the tool and compare on the same screen the two – the model versus the measured – to see, you know, may this was an issue of over-etching, under-etching. We have a different copper roughness value going on here. Or we have the wrong values for something. Because that’s the key thing, of taking modeling and measured values and getting them to connect, in understanding what is going on. But we do have a complete suite of tools for impedance and insertion loss.
6:49 What is the accuracy of the models for insertion loss measurements, especially for copper roughness?
We don’t like at Polar to limit what designers and fabricators are doing. We’re trying to equip you so that you can have the most flexible tool-set. And, to answer that of how do we accommodate those different roughness characteristics in putting them into the tool. We’re not making up our own in these, we’re implementing accepted industry methods. We started with Hammerstad, we had that in our tool-set for a long time, and we added Groisse and Huray.
Huray is the one that a lot of people had a focus on and interest. Unlike the other two, it didn’t have a saturation value where there was a limit within that formula. And Huray doesn’t have that. When he came up with this method, was looking at saying ‘Let’s actually do the physics modeling of this instead of just coming back from empirical measurements. Let’s do the physics modeling, let’s do the empirical measurements, let’s bring those together to come up with a more complete method.’
The Cannonball Stack Huray method
The slight problem with that is it’s fairly complex to get the values in, to put that into a tool-set. We worked with some other industry experts to simplify that in what’s called the Cannonball Stack Huray method. It’s just doing some simple math to converge the values that the fabricators already have of what the roughness values are. And put it into that model. Now, it’s not perfect, but it’s a method that does it quickly and gets a really, really good answer. I’ve talked to designers when I pulled this information from a factory. And they said, “This is what the rough values are.”
I gave that to the designer, aside on seeing not knowing anything else that he was doing, he plugged it in and he was shocked because he called me back and said, “Do you have, like, a spy in my office?”, because the value I gave him back, when he put it into the field software 13GHz, let’s say it was 0.01dB per inch different from what he measured.
So, the modeling and the measuring are really coming together these days and that’s beneficial for designers and fabricators because now we’re able to push those frequency ranges and get the products through without a bunch of scrap, without re-designs and that’s reducing cost and increasing throughput for everybody. So, we consider that a win all of the way around.
9:30 Are you able to separate out the tree components of the losses in the measurement?
No. Separating the losses from the measurement tends to actually be a fairly complicated analysis that needs to be done because, in order to do that, you need to fully understand the models. But you also need to fully understand what’s going on with the materials because in the measurement, by nature itself, all of those losses are lumped together. And so, in order to pull out one from it, it’s kind of like making Kool-Aid, or you know Gatorade or something: you pour in the mix, you stir it up, and now you’re saying, “I would like to separate the water from that.”
Do some measurements
Well, it will take some time. You need to know what those chemicals are that are in there right now. And you can go through that process. And the process would be doing a cross-section, etc., so you know what the dimensions are… Solidly on that, doing some other measurements of how the dielectric materials are behaving, their losses, the losses of the conductors. If you’re able to identify all of those pieces out, you can then take apply it to the next case over.
But I’m thinking, I just spoke with an industry expert a couple of weeks ago. He was saying the copper roughness wouldn’t account for this additional loss that he was experiencing. And it had to do with water coming into the board and causing some additional issues there.
It’s hard for fabricators and designers to recognize that it, there’s a lot of art in this science that we do.
11:17 How coupons for insertion loss are different from impedance coupons?
That’s a really large question. Insertion loss coupons really fundamentally aren’t that different from impedance coupons.
11:29 Do they use some different connectors?
Sometimes they can. Not always. You asked a really big question there. Most of the time they do use different landings and different connectors. But that has to do with the frequency range of those connectors and the landing. The connection onto the board that you are looking for. You don’t want to be measuring the additional losses due to the probe. You want to be measuring what’s on the board.
There are industry methods from IPC, like SPP, Set to Dill, Delta L which is being re-branded as the Eigenvalue method… Those actually all come from – believe it or not – the design side. It’s the designers who then stepped in. They requested the information from the fabricator. It was the designers, the Intels, the IBMs, pushing downstream to the factories. They wanted the fabricators to use these methods because they needed this information to make informed decisions.
And also quality control, to know what’s coming in and out. Now, all of those methods use different landing patterns and styles to connect to the board. In some cases, they don’t even spec specifically how it should be connected.
The designer should talk to the fabricator
So, when it comes to insertion loss coupons, it’s one of those ‘the designer should talk to their fabricator’. Or ‘fabricator talk to your designer’ of what they are specifically looking for in regards to what type of insertion loss measurement. And what the expectations are. Then also talking to the tools supplier, so they could give me a call. Within CGen, we have those patterns, we have the coupon designs. So it’s just a few clicks to bringing that in.
13:24 Do you have any instruments for insertion loss measurement?
We have a suite called Alice. It’s a suite, so it’s hardware and software paired together for doing insertion loss measurements. We’ve had that for a number of years now going back to 2012 and maybe even earlier.
13:41 What is the highest frequency that it can go?
The tool-set of the frequency it covers is based off of the hardware package that it is connected into. Most of the time, we do a range of 0-20GHz. But I also know that there are customers out there doing 0-50GHz. We don’t really go much further past that with our tool suite in there. We do direct people to other tool suppliers when it’s appropriate.
14:05 What is next for Polar?
I’m confident in saying that we build relationships. We build partnerships with other tool suppliers, with our customers. We really have a heart for helping our customers solve their problems. Our company is primarily made up of engineers. We have that hard-focus of we’re here to help you solve problems. And we don’t want to be a problem for you in that regard. So as much as we can come alongside our customers, both on the design side and the fabrication side. Practically where we see that going is a lot more into the things like copper roughness, insertion loss.
We were there supporting the fabricators when impedance started becoming an issue. And as much as Ken was around to help designer or fabricators, now we’re embracing that. We help designers and fabricators stepping into the realm of increasing frequencies and insertion loss.
Tags: controlled impedance, impedance, Polar Instruments