Decreased current consumption after change of PCB-supplier. 1

[3dB]

Any ideas what can cause a decrease in supply current of ~10mA (from 54 to 45mA @5VDC) after changing supplier for the PCB? The board is of 2 layer FR4, used in a RF unit module of size 10x10cm. All components are from the same batches as before, the only difference is as mentioned a new (chinese) supplier for the PCB, ordered with same specs.

 

[Comcokid]

The dielectric constant of FR4 material can vary, especially from vendor to vendor. You state that the unit is an RF module. Perhaps what has happened is that the impedance matching of your circuits has changed as a result of a dielectric change in the PCB. Example, if one RF stage is driving a RF gain stage, the first stage may be putting out less power into the second resulting in lower current consumption (and lower power output) of the second stage.

 

[buzzp]

Even specifying the copper can have an affect of the impedance of the traces. You would think if you told them 1 oz copper (thickness) that this is what you would get. This is not always true and will vary from vendor to vendor, to some degree. I found one vendor only purchased 2 oz plating to get more discounts. So if you asked for 1oz you still ogt 2oz copper. Even if you specified 2oz copper from two different vendors, your likely not to get the same thing. This could have some affect as far as what Comcokid suggested. I doubt the impedance of the traces alone, would cause a 10mA change in current. However, the impedance matching you designed is likely causing the problems, as a result of the impedance changes of the traces and/or dielectric of the board material.

 

[melone]

To expand on BUZZP comments on copper thickness. 2oz copper can mean different things to different PCB manufacturers. One manufacturer can give 2 oz of base copper, while another gives 1oz base copper plated up to 2oz. The problem with the second method is that the plated copper will average out to 1 oz, but the amount of variation in the plating process will create hills and valleys of copper.

 

[MacGyverS2000]

To expand even FURTHER on buzz's and mel's answers... While the copper is often plated, many times they will start with something like 0.75 oz, but plate another 0.5 to 0.75 oz on top, calling it 1 oz in the process.

To ensure correct impedance, you need to be as specific as possible in specing the board...finalized copper weight, percentage of trace over/undercut after etching, thickness of each laminate layer (particularly the die. constant), thickness of soldermask, etc.

 

[3dB]

Thanks all for your replies!

I measured the thickness of the two types of PCB and they differ with ~2,5mil. The one with lower I_supply is thinner (33.7mil). So I guess the next step is trying to put a test strip on the PCB for impedance measurements and write a bit stricter specs.

 

[felixc]

Depending on what electronics are on board, can it be possible that since your PCB comes from China, odds are that some of the ICs can be equivalents coming from other suppliers? Or can there be a "new and improved" version of a chip? Often these improvements are geometry scale downs that reduce power consumption.
20% of change in power consumption deserves a serious look. Don't let these boards go out before you understand what happens, or you may have bad surprises in the field. Aside from the power consumption do you see any difference in performance?
If your board hasn't too many parts on it, you may consider moving parts from one pcb to the other, and record any differences in power consumption.

 

[MacGyverS2000]

In your BOM, specify part manufacturers and the exact part number (include items like "Industrial Temp range" explicitly). I also place notes on almost every part (resistors can be replaced by another without much consequence) specifying "No replacements may be made without permission by ..."

 

[logbook]

Well there are some nice esoteric explanations given above, but perhaps that is being a bit over the top.

You have
1) 54mA x 5V = 270mW
2) 45mA x 5V = 225mW

So you have a "massive" 45mW difference between the two systems. It really depends what is dissipating the power in the system. If most of the power is being dissipated in one chip then that much change of dissipation is quite possible.

There is too little power dissipation for the copper thickness to make any difference at all.

If the whole board is thrashing around at high frequency then the dissipation factor of the pcb might be that different from board to board.

I liked felixc’s idea of swapping parts from one board to the other, depending on how good your re-work skills are.

 

[buzzp]

But the impedance of the traces may affect and impedance matching circuits he has set-up which in turn could realistically affect the current draw that much.
The impedance of the traces alone is likely not the cause (unless we are talking about a sizable board or trace length/width). However, the impedance change may affect any other circuits designed for maximum power transfer, which in turn would affect the current.

 

[3dB]

I'm quite confident concerning the components since the BOM is written as MacGyver suggests. There are a couple of "improved versions" that draws less current, but the total of that is less than 1mA.

The board is by the way a direct convertion thing workning @2.45GHz. One thing I forgot to mention is the fact that this new board has a lower P_out than the old one when using the same resistor values (as in the old) in a T-network used for trimming of P_out. But trimming P_out up doesn't affect the current...

This complies with the idea Comcokid had about the RFICs.

I will try the part swapping after I have finished all the %€#¤!! paperwork on my desk.

 

[logbook]

Ok, with FR4 at 2.45GHz, reduced output power, and chips with a low tolerance on current consumption, I am now strongly in favour of the mismatch theory.

More data gives a more accurate/probable theory