Previous Page  36 / 52 Next Page
Information
Show Menu
Previous Page 36 / 52 Next Page
Page Background

34

Chip Scale Review March • April • 2019

[ChipScaleReview.com]

Testing of high-frequency 5G applications and why

simulations are critical to success

By Jeff Sherry

[Johnstech International]

he pervasive drive towards

ubiqu it ou s con ne c t iv it y,

high content/fast data links,

and Internet of Things (IoT)

hyper-growth prompts the 5G air interface

standard to quickly supersede 4G. The

rapid adoption of the 5G standard poses

a dilemma for the supporting ecosystem,

particularly in providing the required

infrastructure and allied devices to launch

the overall functioning system. Faster

time to market means accurate and timely

device test models and simulations to avoid

costly iterations upon product release.

The 5G frequency bands per 3GPP Rel 15

consists of two spectral clusters namely FR1

(Frequency Range 1) from 450-6000MHz,

and FR2 (Frequency Range 2) ranging from

24-53GHz. More than 10X the incumbent

4G/LTE data rates are achieved on account

of much wider bandwidths (100MHz

for FR1, 400MHz for FR2) than 4G’s

bandwidth of 20MHz.

This paper demonstrates examples of

electrical modeling to support customers

working in 5G frequency ranges and shares

why the data is so important to make sure

the full test system implementation is

successful. Data that compares simulations

that only include measured dielectric

constant and loss tangent or dissipation

factor for one frequency, to results that

use measured material performance over

the full bandwidth will demonstrate how

having the correct material properties

can lead to more accurate simulations.

This leads to expected measured results

causing fewer redesigns, and therefore a

shorter time to market. In some examples,

simulated results will be compared to third-

party measurement data.

Simulations are critical to development

success in high-frequency 5G applications.

Carefully executed, modeled simulations

will closely match actual results, which

avoids the added costs and guess work in

high-frequency product development. This

paper will do the following:

• Produce examples of measured versus

modeled data in those introduced 5G

frequency bands; and

• Outline the effects of metal and non-

metal material properties on testing

results.

• Outli ne t he modeled ef fects of

S-parameter convergence to get

simulations to closely represent

measured results.

Introduction to 5G simulations

5G applications will move to much higher

frequencies to support the immense amount

of data. In the semiconductor industry,

higher frequencies translate to smaller

wavelengths, which means that smaller

contacts and shorter path lengths will be

standard to support 5G applications.

In any discussion of testing, it is

important to recognize that both the

assumptions and inputs into the model need

to be correct. The more accurate the inputs

to any simulation, the more accurate the

results of the simulation. However, if the

inputs are too simplified, there is danger

that the answer will be inaccurate and not

predictive of the real-world application.

Material properties are a big factor in

the system performance and many non-

metal materials sometimes

vary greatly over frequency.

In addition, the mater ial’s

properties may change over

temperature, humidity, or other

test conditions making it more

difficult to predict accurately

measured real-world results.

To further complicate things,

many material suppliers test

their materials at really low

frequencies that are well outside

the 5G frequency bands. Most

material specification sheets

have the dielectric constant

and dissipation factor or loss

tangent specified at 10MHz,

which meets the requirements of the

method used to call out the testing of

the material. Many materials properties

are less reliable the higher the frequency

range, and for 5G applications could yield

results that deviate significantly from what

is modeled resulting in costly redesigns

and the building of more prototypes to

test, causing delays for the system to get

to market.

F i gure 1

shows t he t h i r d - pa r t y

measured versus modeled results of

Johnstech’s 0.5mm pitch ROL

®

100A.

In the f igure, one can see measured

return loss as compared to an equivalent

circuit of contact and modeled data of

the contactor, which includes a 50 ohm

board and device pad. Notice that the

HFSS modeled data with board and

device matches the 50 ohm measured

results closely at 20GHz. In this model,

housing materials were tested to 20GHz.

In testing alternative materials, we found

that material properties can sometimes

cause results to vary by more than 30%

at elevated frequencies. At very high

frequencies, the best data model to use is

the transmission line model vs. lumped

elements that were used to develop the

equivalent circuit to match measured

results. As viewed in

Figure 1

, the

T

Figure 1:

Third-party measured vs. modeled results of Johnstech’s

0.5mm pitch ROL

®

100A measured return loss compared to an

equivalent circuit of contact.