The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Bond wire coupling
https://designers-guide.org/forum/YaBB.pl?num=1390868489

Message started by RobG on Jan 27^th, 2014, 4:21pm

Title: Bond wire coupling
Post by RobG on Jan 27^th, 2014, 4:21pm

Question: If two bond wires are each 3 mm long and separated by 400 um, does anyone know how much current does one induce in the other? The current would be from a digital output. The current would be a 5-10 mA pulse about 600 ps wide.

Title: Re: Bond wire coupling
Post by loose-electron on Jan 27^th, 2014, 5:10pm

you can do a magnetic model of this to get something quantitative.

It is going to depend on the circuitry connected to the receiving bond wire for sure, so a simple number answe can not be given.

One comment - it is significant and is frequently an issue in a noise coupling scenario.

Title: Re: Bond wire coupling
Post by RobG on Jan 28^th, 2014, 6:31am

Yeah, in playing with the Ldi1/dt-Mdi2/dt formula I can see that loading is more important than I originally assumed. I guess I need a swag at the "k" in the "mind" mutual inductance model.

Title: Re: Bond wire coupling
Post by raja.cedt on Jan 28^th, 2014, 10:44am

Dear RobG,
are two bond wires are differential? or not. In general Mutual coupling decreases to less value if the horizontal distance between two wires> 4*wire diameter, in your case distance is around 400um which is Quite high compared to bond wire size(in my case the are around 60um).

Of course this approximation depends on what is the hight from the ground plane and shape of the wire ( the above approximation is for the rectangular waveguide, it may be slightly differ for circular shape)

Thanks,
Raj.

Title: Re: Bond wire coupling
Post by RobG on Jan 28^th, 2014, 10:58am

Raj, one wire is a sample clock input (sensitive to noise since it jitters the sample time) and the other is a digital output. I wasn't too happy with the bondout but was overruled cuz I'm just a hick from MT ::)

The 18 output drivers also share a common ground which I think is contributing to the problem, if not the major source of error.

Title: Re: Bond wire coupling
Post by RobG on Jan 29^th, 2014, 2:32pm

I found page 38 of "Digital System Engineering" by Dally and Poulton, which suggest k=~0.3, which gives a result similar to what I am seeing using the mind device to couple two inductors.

A link is here, although I don't know how stable it will be:
http://books.google.com/books?id=oDWRAxCU-g8C&pg=PA38&lpg=PA38&dq=coupling+coefficient+between+two+bond+wires&source=bl&ots=8_NBsrreZf&sig=aiMySGHOBQYR8XDbp7zSUSYdOPc&hl=en&sa=X&ei=w3_pUo_GKc6GogTgiIHQAQ&ved=0CCUQ6AEwAA#v=onepage&q=coupling%20coefficient%20between%20two%20bond%20wires&f=false

Title: Re: Bond wire coupling
Post by carlgrace on Jan 30^th, 2014, 10:05am

Rob,

We measured mutual bondwire coupling in 65 nm (80 um pad pitch) and we found a k of about 0.4. So there is one data point for you.

Can you put a ground pad (and bondwire) in between aggressor and victim? That makes a really big difference. I gather that the chip is already in process or back from fab so my guess is the answer is no.

Carl

Title: Re: Bond wire coupling
Post by RobG on Jan 30^th, 2014, 10:43am

carlgrace wrote on Jan 30^th, 2014, 10:05am:

Thanks - those data points help everyone. A few things conspired to cause the problem, all related to egos and an overly aggressive schedule of course. It is comforting to be able to reproduce the negative results in simulation with this sort of information.

As an aside and lesson learned, although I didn't like it from the start, I didn't go bananas because the pads were at the corners, thus perpendicular. They also had ground pads between them. Well, the ground pads were downbonded to the paddle and offered no shielding. Furthermore, even though the two lines in question were perpendicular on the chip, in the package the bond wire went to the package corners so they were the longest wires and almost perfectly parallel. You couldn't have picked a worse configuration!

Correct me if I'm wrong, but I don't think adding a ground wire between the two leads would help since it is magnetic coupling. We used LVDS this time so the fields should cancel, but most importantly we treated the clock like the analog input that it is and kept it far from the digital.

Title: Re: Bond wire coupling
Post by carlgrace on Feb 3^rd, 2014, 6:53am

The idea is just to increase the distance between the offending bondwires and therefore reduce the coupling. Connecting it to ground won't add any shielding but it will give induced a current a non-troublesome place to flow.

Title: Re: Bond wire coupling
Post by aaron_do on Feb 3^rd, 2014, 5:23pm

Quote:

Correct me if I'm wrong, but I don't think adding a ground wire between the two leads would help since it is magnetic coupling.

Hi RobG,

my previous group did quite a bit of study on this a while back so I can share a little of my experience. The coupling coefficient of 0.4 is quite reasonable. Also, regarding the perpendicular bondwires, sometimes it looks like the bondwires are perpendicular, but from a 3D perspective it turns out that they're not that perpendicular. This is more true when the chip is quite high so although the bondwires appear to be going in different directions, it turns out that they're both just going downwards.

As for putting ground downbonds in between, it can help, but as I remember you need to be careful. Basically inductance is always calculated around the whole loop, otherwise it is just partial inductance. Therefore you need to consider the return path. If you use a ground bondwire close to your signal bondwire, and they are tightly coupled, then the loop inductance will be reduced. The lower inductance will result in less coupling to other bondwires.

However, if you share the ground bondwire with other circuits, then you might end up just coupling noise through the shared ground. Also, the ground bondwire must be a current return path. If you just leave one end dangling it may not have any effect.

Aaron

Title: Re: Bond wire coupling
Post by RobG on Feb 4^th, 2014, 7:41am

Thanks Aaron - I hadn't thought of them being parallel in the vertical direction.

rg

Title: Re: Bond wire coupling
Post by loose-electron on Feb 18^th, 2014, 7:01pm

rule of thumb - iif there is any signal on the leadframe, it will couple into the front end of an RF LNA or into the externally placed VCO filter of a PLL.

Anywhere on the leadframe, not just beside each other.

No formulas, just seen a lot of learning experiences.