Chinese Microwave PCB Substrates. Can They Be Used? My Experience
Hi! My name is Olesya – leka_engineer, I am an engineer, and in my free time I blog about development, my lab and a little about technology. My main job is designing microwave printed circuit boards, so I write a lot about it (check out the publications section!).
It's been almost two years since Resonit officially* stopped fulfilling orders for Rogers and any other customer-supplied material. Now boards can only be made there using the Chinese analogue of Rogers – Wangling. I have been researching the parameters of this material for some time now and writing about it in my blogs. In this article, I want to collect all the information, so there will be brief repetitions from previous articles and posts (and many links). And at the end – additions that are relevant today.
1
In this short article I described my first experience of ordering boards from Chinese WL (Wangling). Rogers analogues are offered on our market for quite a long time – Relong, Wangling, FSD. But Rogers is still sold (where you can buy, I wrote here). In the article I told about our new development at that time – the filter, that it was made on Rogers and on the WL analog. The filter works in the subgigahertz range. The result of the frequency response measurement is shown in Figure 1.
![1 Blue graph - Rogers, Red - WL 1 Blue graph - Rogers, Red - WL](https://habrastorage.org/getpro/habr/upload_files/a3e/d05/322/a3ed05322b06fece9c29aa29b4fc7703.png)
From the graph, we can conclude that the filter's operating band has shifted slightly to the left, which corresponds to the increased WL substrate epsilon. In numerical terms, the difference is approximately 0.1-0.15 units.
It is important to understand that 1- there is no perfect way to measure the permittivity of a material. John Coonrad of Rogers writes a lot about this in his blogs. You can read my translation (pay attention to the links) here. There are many ways, and if you look at any Rogers datasheet, you will see that there are two values and measurement methods. 2-epsilon decreases slightly with increasing frequency. This article was written by Leka_engineer
Therefore, the epsilon value we obtained is applicable only near the filter's operating range (and most likely only for resonant structures).
2
In this article I described how I now order microwave printed circuit boards. By the time this article was written, the entire batch of filters on WL had been measured. And all filters had a downward frequency shift, but when the frequency response of the filters on WL was superimposed, the graphs did not match, and the frequency response of the filters on Rogers were identical (The sample was about several dozen pieces in both cases). Based on this information, I then concluded that WL has a negative feature, which is that the loss tangent is not constant in the x and y coordinates.
3
Next, the filter topology was adjusted taking into account the epsilon value found in Figure 1. I purchased the material from Electreid and sent it to Electroconnect. The resulting filter boards were measured and the frequency response was where it was needed coincided with the frequency response of the filters on Rogers (and, accordingly, with the one required by the technical specifications). Link to the post.
4
In that post I showed boards on WL material of another series – boards without resonant structures or distributed elements. So there were no problems with them. When calculating, I took the epsilon design from the Rogers datasheet.
New information
After manufacturing a batch of filters at the customer's WL in Elektrokonnekt, we ordered a batch at “their” WL in Resonit. The frequency response matched the required one. Since then, quite a few have been manufactured filters on WL (with corrected topology). Each filter is measured before being sent to the customer.
![2 loss statistics based on a batch of filters on WL 2 loss statistics based on a batch of filters on WL](https://habrastorage.org/getpro/habr/upload_files/468/d45/ecd/468d45ecd18b8c06cf1fa66c350b3683.png)
Based on the measurements, we made the following conclusions:
– yes, in the sub-GHz range, epsilon WL differs by 1-2 tenths
– No, the judgment that WL has “floating” parameters across the sheet is erroneous. As a result of measurements of a large sample of filters, it became clear that filters on Rogers and WL have approximately the same statistical distribution of absolute losses.
-yes, the filter losses on WL are slightly higher
– yes, Chinese materials can be used, but with caution. Considering that almost always after the first sample is made, it is not so bad. Simple structures (just boards like “strips connecting chips”) can be safely made on WL (I think, on other Chinese materials as well).
Thank you for your attention,
Olesya (Leka_engineer)
Useful links:
1.Protocol measurements (my comments to this text can be read here) Resonite
2.Article with independent measurements Stepan Revutsky
(*I suspect that it is not for everyone)