At what current will a 25um wire melt?

Curious · Guest

Recently during FA of customer returned samples, we found that one sample passed ATE and x-ray inspection; the wires appeared good. Upon curve tracing, the loop showed some abnormalityat higher current (slightly higher voltage drop) compared with a KGU. Upon acid decapsulation, we found that one wire (emitter) was broken with some fused mold compound surrounding (just a thin layer, not a whole lot). The die (this is bipolar transistor) was still good.
My question is at what current does the wire bond (25um) normally melts? I am puzzled at this case for 3 reasons:
1. Normally the junction gets shorted before wire melt occurs.
2. There was very little scarring on the wire that we could not detect it during x-ray. Normally for fused wire, we can see massive damage surrounding the wire.
3) ATE showed the unit passed all parametric tests.
My 2nd question is: is it possible to distinguish a failure caused by voltage, current, or power. If so, how do you do that.
Would appreciate your input.
Thanks.

spongklong · Guest

curious,

here's my 2 cents on this. i assume the wire is not depressed because xray image was good - so no short with the die. if the die is good and the damage on the wire is only in the small area where its broken, then my guess is the wire got a defect. if your wire was 'nicked' during assembly, the nick will have high current density during powerup, and the melt may occur there even if the current is not high enough to damage the die.

questions:
1) is this gold wire?
2) is your die attach conductive? is collector grounded to the die pad?
3) were you able to establish the path of current that damged the wire?
4) what's the highest current at emitter in customer's application?
5) just one unit returned from the customer?

maybe i can give more inputs if you can give answers to these.

spongklong

FARel Engr · Joined: 20 Jun 2026 Posts: 40

Hi Curious,

Hmmm...interesting case. The fact that the wire melted before the die simply indicates that in this sample, the silicon was able to dissipate the heat better than the wire. In what package is this transistor assembled? When you say that 'normally' the junction shorts first before the wire melts, are you referring to exactly the same device/package combination in the past? In short, for this FA, we need to know where the current passed and check why in the current path the wire is less able to dissipate the heat than the metal lines of the die.

Regarding your 2nd question, when the failure involves something that melts, then it is due to power dissipation that raised the temp of the affected area. Thus, both I and V may be involved. Usually the role of V in EOS is to 'turn on' junctions that are otherwise in cut-off or avalanche a reverse-biased junction. When these happen, 'abnormal' current paths are created and large surges of current takes over, resulting in the high power that melts the affected site.

If no meltdowns are observed, such as in simple dielectric shorts, then chances are that high voltages rather than large currents are involved.

Pls keep us posted on this case.

Thanks!
FARel Engr

Curious · Guest

Hello FArel engr and Spongklong,

Thank you all for your replies. I would like to provide more details on this case.
The transistor was assembled in a SC-51 package (slightly bigger package than TO-92). The wire bond is Au wire. The die attach is Au-Si eutectic. This means the collector is attached to the paddle and serving as the heat dissipation path; this is a 900mW device. When I referred to normal, I am referring to similar device(wattage) and same package. In most cases of wire melt that we encountered, we normally observed that the junction is shorted; there is massive burnt marks on the die surface. In this phenomenon, we assume that a high voltage has first caused the junction to be broken. Then, as a consequence of this short, a high current is drawn by the device which in turn causes the wire bond to be melted. However, in this case, the die is still good. So, I am puzzled as to how to explain the wire melt phenomenon.
(3 samples were returned by the customer; 2 pieces we could find the junctions broken and massive wire melt on the emitter wires--through x-ray inspection. Because wire melt occurs on the emitter wires, can I assume that the current is originating from this end? Only in 1 piece, the die was intact and the damage on the wire was minimal.)

Thank you and regards.

boston gardener · Guest

curious,

if the die is still good but the wire is melted, then large current was shunted from emitter to collector. only thing I can think of for this to occur is if emitter was shorted to collector. which means the emitter wire got connected to die substrate or the paddle; even just a temporary connection. you might no longer find this short because it probably got burned off already. is this transistor pnp? is emitter at higher voltage than collector, like +15V at emitter and collector is at 0V? how about the base voltage? if base was at 0V when the short happened, the emitter wire can also melt if it touched the base wire.

boston gardener

EE · Posted: Tue Jan 27, 2026 9:58 am Post subject:

Hi Curious,

One guideline you can use for the maximum current through your wire is
MIL-M-38510, which states that I = k*d^(3/2) where:

- for gold or copper, k = 30000 with a bond-to-bond length of <= 0.040 in. (0.1 cm) or k = 20500 for a bond-to-bond length >0.040 in. (0.1cm).
- d = wire diameter in inches.
- I = DC or rms current.

Thus, for a 1-mil gold wire < 40 mils long, your max current should be 0.95 ampere. If your wire is > 40 mils, then I < 0.65 amp.

If we are to follow this, your 1-mil wire may be a bit marginal for a 1-watt part, in the sense that a defect or damage in the wire that reduces its diameter significantly can really be deadly for the wire even if the current through it is within the normal operating range.

Wire damage is not uncommon - even debris accumulated at the tip of a bonding tool can reduce the wire diameter significantly. I'm not saying that the melted wire of yours was damaged to begin with, but it may be one possible explanation for your sample's failure. Of course, shorting of the emitter wire to the die substrate is another possibility, as the others have said.

By the way, at which part of the wire was the meltdown seen? Is it in the middle, near the bonding pad, or near the bonding post/finger? Also, do you see a ball at the end of the melted wire? At which side is this located - on the die side or the lead side? These may provide clues.

Thanks and keep on posting!

Best regards,
EE

Curious · Guest

Hello EE,

Thank you very much for your detailed explanation. I find it very informative.

To answer some of your questions. The wire bonding appeared good; there is no possibility of shorting to the substrate or paddle. The wire meltdown occurred at the middle span. No, there was no ball formation at the tip of the melted wires--they appeared as if they have just been broken off; I checked under SEM. In fact, I also checked the portions where the wire was still good. I couldn't find any sign of cracks. (I notice that sometimes when high current has passed through the wire but it has not melted, the wire will be degraded and if you inspect it under SEM, you can see on the wire surface, structures that appeared like micro-cracks. Do you have this experience? Can a wire become degraded (by high current/temperature?) but not melted and later on just become disconnected under normal DC application. Do you have such experience?)

Thank you and regards.

EE · Posted: Wed Jan 28, 2026 12:55 am Post subject:

Yup. The crack-like patterns that you see on the wire surface after it has been subjected to tremendous heat and then cooled down again are actually grain boundaries. Wire recrystallizes with larger grains after experiencing excessive thermal stress. These larger grains make the wire more vulnerable to shearing stresses.

Goodluck on your FA! I'll let you know if I think of something else.

Thanks and keep on posting!

EE

	SemiconFarEast.com Forum Index -> QA/FA/Rel Forum	All times are GMT + 8 Hours
Page 1 of 1