West Point Precision's Guo Rongzhe: Technological evolution trend of in-vehicle high-speed interconnection

This is a comparison of SI performance between the two forms.

The left side shows the SI performance of the current automotive Ethernet. Due to some structural limitations, we had to work very hard to barely meet the 25Gbps rate. The right side shows the high-density concept we have just proposed. Since the link is shorter and many structures and processes are optimized, it is easy to achieve the 112Gbps performance. We can see that the insertion loss is very smooth all the way to 40GHz. The crosstalk is basically below -45Db at the operating frequency of 28GHz.

However, what features have we made special designs for? The following section will give you a brief introduction!

When designing the signal integrity of high-speed connectors, we pay great attention to a feature called Stub (Chinese name: residual pile). This thing is very unfriendly to the smooth transmission of high-speed signals, so we always hope to make it small, or even eliminate it. On this page, we list the commonly used board-side connector termination methods. We analyze the size, reliability, and especially the friendliness of high-speed support (mainly considering the size of the Stub). From this table, we can easily understand why, in the face of high-speed scenarios, our current automotive Ethernet connector termination method needs to be changed.

The first is through-hole reflow or wave soldering. The current 10G automotive Ethernet connector uses this method. We can see that this method is one of the following methods. The Stub and the largest size, the red line is the schematic Stub length. In addition, an additional via is required to introduce the signal into the inner layer of the PCB. Therefore, this structure does not have an advantage at high speeds. We believe that this method can cope with the rate of 25G and below, but when the rate is high and the size is required to be small, the weaknesses of this method will be magnified a lot.

The second is surface mounting. You can see that the stub is relatively small. This method is very friendly to high-speed signal integrity. We estimate that this method can also support the 224G stage, and the size can be achieved with a pitch of 0.25mm. However, it is subject to some restrictions for some applications. For example, it may be easy to drop parts during double-sided welding, and it is not easy to check the welding quality of the inner shot when arranging in a matrix.

Then there is crimping. I personally prefer this method. We also use this method for our high-density concept. The processing method is cold processing and does not require welding. In addition, the Stub length can be controlled after PCB back drilling. The rate can support 112G very well. Of course, 224G is also possible. The size is also moderate, which is more suitable for matrix arrangement. It can meet our high-density application needs and the cost is relatively cheap.

The last is the BGA method, which is commonly used in our current chips. It is very friendly to signal integrity, but its disadvantages are that the welding strength and reliability are slightly worse, and the cost of the device itself will be more expensive. There are also some risks of double-sided welding like SMT.

We have also made bold changes to bare wires. The picture on the left is the current commonly used form of 10G automotive Ethernet. The internal core wire is multi-core, twisted together after adding an insulator, and then wrapped with aluminum foil. Some are also wrapped with an intermediate insulation layer (that is, the gray layer of material), and then wrapped with a metal braided mesh, and finally wrapped with an outer sheath. The process is still quite complicated.

The one in the middle is a method we recommend. The middle core wire is a single-strand, parallel arrangement structure. It only needs to be wrapped with aluminum foil and no braided mesh is required. The outer sheath is made into an oval shape. The one on the right is a measured performance comparison. The red wire is the old structure with a core size of 26AWG. The blue wire is the new structure with a core size of 27AWG and a length of 5 meters. Judging from this performance, I think we can reduce the wire diameter by 2 wire gauges. For example, under the same loss requirements in the future, we can use 28AWG wire instead of the current 26AWG wire.

The table below makes a rough comparison. Under the premise of completing the same function, the cost is reduced by 45%, the weight is reduced by 60%, the size is reduced by 50%, and the performance is also greatly improved.

This is a size comparison of the core contacts of the two connectors. We can see that the length, width, and height are much smaller.

This report is a bit technical and may be a bit boring. Thank you again for your time. If you have any questions, please feel free to communicate with us!