Multilayer boards for RF?

An update to the question above:

Regarding the t/h<0.005 assumption we checked the referenced source “I. J. Bahl and D. K. Trivedi,  A Designer's Guide to Microstrip Line”. The electronic edition can be found at http://www.ece.ualberta.ca/~rambabu/EE470/class26s.pdf. From page 2 of the paper one can see that the basic assumption is W/h≤1, and assumed zero thickness; however in reality the thickness cannot be zero, and a further qualification on page 2’s left side states that when t/h≤0.005, 2<ϵr<10, and 0.1<=W/h≤5 the agreement between experimental result and calculation based on t/h=0 assumption is excellent.

1) Multilayer board: Yes, you can copy the design onto a 8 layer board. The most important thing here is the distance from the top layer to the first ground layer. If this distance is the same in your board as in the reference design you don't have to change the balun or the tracewidths. Note that it is only for traces that are longer than 0.1 times the wavelength you need to consider 50ohm traces.

Thanks for your very helpful answer.

TER said:

The most important thing here is the distance from the top layer to the first ground layer. If this distance is the same in your board as in the reference design you don't have to change the balun or the tracewidths.

1. If the distance between the top to the first GND is the same as a reference board, does it mean I can ignore all the other GND planes? There might be still one or two GND planes beneath the first GND layer. In addition, do I have to keep thickness and width of traces the same as the reference board, or can adjust thickness-width for a particular Zo?

2. In the case that the distance between the top and the first GND plane is different from the reference board, I know how to turn track width/height to match the impdeance; but for balun, I completely have no idea. AN068 talked about balun, but uses Smith Chart for evaluation/validation. Could you tell me In practice how is balun really adjusted? Do I have to change the balun topology / layout? Does traces in balun need to have their width adjusted?

3. Does the wavelength here mean wavelength in air, or in the dielectric? For the latter there is an sqrt(epsilon) reduction from the vacuum wavelength.

Matt

1) The EM field from the layer 1 will terminate in the first ground layer. The RF tracks on layer 1 will not "see" the layers under the first ground layer. If you adjust the copper thickness on your board compared to the reference design you should check if the trace width needs adjustment. If possible use the same copper thickness as the reference design, then you don't need to do adjustments.

2) If you adjust h some you don't have to adjust the component values. But we have seen that if h is down to ~250um the parasitic caps between layer 1 and 2 starts to influence the result. The same if h is too much larger then the via inductances will be larger and start to influence. If h is changed more than 50-100um compared to the ref design you should simulate in ADS or similar and adjust the component values to get the same S11, S12 as the ref design. 

3) I have used wavelength in air. This is probably not correct seen from a physics standpoint but this is a rule of thumb and the air number will give worst case.

TER,

The FR4 used in each layer on our board is thin. I am trying to combine several FR4 layers together to make the sum of their H close to 0.8mm of CC2500EM-31mil design.

Our tentative solution is to leave the copper area for several consecutive copper layers beneath the top unfilled. As shown in the figure, in this way the dielectric layer height accumulates across several layers so that several thin FR4 would combine to a height close to 0.8mm, and then we add a GND copper area right beneath the top layer region wheere the RF part is placed. We have explicitly marked the RF region copper in yellow to distinguish it from other parts in the figure.

Would this be a good solution?

First, we don’t know if FR4 would expand to fill the area vacated by copper. For example, if the original FR4 thickness is 1.2mm, then if it somehow expands thicker to fill the empty area, then its density decreases. I wonder if this would change the dielectric constant?

Second, we don’t know if copper thickness should be included in calculate the accumulated H in this manner. As shown in the figure, should be take H merely as 0.12+0.3+0.3=0.72mm, or add the thickness of the copper multiplied by two, even if the copper area would be filled by expanded (or not?) FR4?  The lower part of the image shows the view of a typical “stripline”, and the thickness is considered in its definition of height H. So should I include intermediate empty copper layers’ thicknesses in H calculation in our tentative layer stackup?

Matt