(MFC AXIOMS)

TOLERATE NO LEAKS

Leak check, leak test, examine for leaks, and verify that the system is leak tight

Small leaks in the gas lines are the slow, silent, invisible killers of MFC's. Large leaks are in some ways more welcome than small ones because at least a large leak will promote a more rapid failure, rather than having a process drift slowly out of specification, thereby affecting yield. With small leaks, on the other hand, often the calibration accuracy is affected first, by degrees, due to residual contamination in the sensor tube and bypass assembly. Gradually the performance of the MFC is degraded, control valves become sticky, sluggish or unresponsive. If fortunate (because hard failures are often easier to diagnose than suspected ones), the valve orifice or sensor capillary tube will clog completely, incapacitating the MFC, thus calling immediate attention to the problem. Also, reaction due to leaks of atmosphere or moisture in the gas lines sends useless contamination and circuit destroying particles downstream, like so many microscopic meteors downstream, bombarding and depositing themselves onto what was to become pure high-yielding profitable wafers (i.e., our collective paychecks).

DO NOT TWEAK THE MFC

NEVER tweak any pot except the zero pot

All pots, except the zero pot, have a direct effect on either the calibration or response of the MFC. Even if you are able to get your process "dialed in" by tampering with the other pots, you will not know precisely why, and the same "hit or miss" tweaking process will need to be employed each time the MFC is replaced, each time with unpredictable results. Time spent by the process engineer formulating recipes will be invalid, because you will have made your MFC, and therefore your system, an unreliable liar. 100 sccm reported, for example, will not equal 100 sccm actual, and all future changes in process will be founded on faulty data which nobody will be able to verify, or for that matter, reliably duplicate. Future recipes will have to be founded on similar lies. Even as with a computer, the principle for process engineering is the same garbage in-garbage out. Don't mess with the other pots. If there's a problem with growth or etch rates, MUM tests, etc, resolve it another way. Pass the buck.

A ZERO LEVEL OF GREATER THAN 2% OF FULL SCALE IS UNACCEPTABLE

If an MFC is mounted in a system, warmed up for at least thirty minutes, and the output (READ signal) is greater, plus or minus, than 100 millivolts (2 percent of full scale)

BEFORE ANY ADJUSTMENTS, with NO GAS APPLIED, ZERO THE MFC

Within the first one hundred millivolts of the MFC, the calibration curve is relatively unaffected. If zeroing becomes necessary beyond this range, however, the accuracy and linearity curve may be affected, sometimes substantially. Check the mounting attitude of the MFC in the system, and verify that it matches its calibrated mounting attitude Check power and signal grounds, as well as +1 5VDC and -1 5VDC levels AT THE MFC, verifying their stability. Also, note any extremes in either the gas or the ambient temperature which might cause the MFC to register abnormal zero levels with NO GAS APPLIED.

DON'T THROW TOO MANY GOOD MFC'S AFTER BAD ONES

If more than one MFC is replaced in the same location, and the system exhibits the same failure symptoms, STOP. Gather more data, and attack from another angle. Diagnose

Quite often MFC's are replaced in rapid succession, one after the other, each exhibiting similar, if not identical failure modes. If contamination is present in the upstream gas lines, causing an MFC to fail, and four or five or more MFC's later, the problem is resolved, it is more than likely safe to assume that the MFC's that failed before were used as nothing more than expensive filters, until contamination levels depleted to a level that would not immediately harm the MFC that finally worked. But this is only one of many possible explanations, which leads to ...

BE THOROUGH - SYSTEMATICALLY VERIFY ALL DATA

It is not at all uncommon for systems to be taken out of production for problems that do not exist. For example:

A correction factor is put in where one is not called for;

Data is misinterpreted during a routine procedure;

New or unreasonably stringent tolerances are incorporated into a process, causing "previously acceptable" devices to fall outside the new window of acceptable limits;

Ask questions and verify everything, gather and analyze all available data before proceeding with a cure.

DON'T STRESS AN MFC CASE DURING INSTALLATION

Use two wrenches to install MFC - Do not use one wrench while holding onto the case -

A Mass Flow Controller is a delicate calibrated precision instrument. A 316 stainless steel block is a sign of corrosion resistance, not indestructibility. When the case housing, sensor assembly and electronics is bent or stressed during installation, all too often the printed circuit boards develop fissures, or cracks along the traces, which leads to the worst kind of possible failures; intermittent, or thermal intermittent. Also, many cases are made of metal, or have some sort of metal shielding which can sometimes short out the PCB when stressed.

WHENEVER POSSIBLE, DON'T PURGE BETWEEN PROCESS RUNS

Purge gases are rated for different levels of purity, or ppm of moisture or oxygen content. Verify the purity of the purge gases used, taking special note of those being used in critical processes. The moisture contained in purge gases cause reactive, corrosive and pyrophouric gas contaminants to accumulate in the gas lines during long purge cycles. These contaminants adhere to the wetted surfaces within the gas lines and the MFC. This in turn can easily affect the calibration of the unit, and even a very thin film of contamination within a valve assembly (the only moving part within), can drastically impair the control performance of an MFC. Therefore, a system that has been verified leak tight should only be purged or cycle purged when absolutely necessary, or when there is to be a breach in the gas lines. Also, we recommend, wherever possible in a leak-fight situation, that pump down time be given priority over purge time. With pumpdown time, outgassing occurs in the lines, removing unwanted contamination.

WHEREVER POSSIBLE, PERFORM INCOMING QA

MFC's undergo rigorous QA before leaving Laminar's Facility. Presently however, there is no hard fast procedure for determining whether something has happened to an MFC from the time it leaves our facility, to the time it is pulled out of stock for installation. Many things can happen to an MFC between these two times; in spite of all care taken, it can be dropped, vibrated or shaken, possibly in transit, or in the stock drawers that contain them (opening and closing until it is finally taken out). But most importantly, so as to keep the black magic and diagnostic guess work to a minimum, the maintenance person contending with a suspected MFC failure must be able to see with his or her own eyes, whether or not the device to be replaced really does function as it should - before placing it into a system. Likewise, MFC's pulled should have post QA performed on them when necessary, to understand what happened to either the system or process.

SYSTEMATICALLY RECORD ALL DATA AND FINDINGS - FILL OUT ALL LABELS AND TAGS THOROUGHLY

The more we know and learn, the more problems we can both prevent from recurring. There is no statistical process control without accurate and complete statistics. Servicing a system is only half a maintenance person's function. Record the data, and the job will be far easier in the long term. Let us know what's going on, so that we can play a more vital role in your success. Don't suffer, and more importantly, don't allow your process to suffer in silence.

DOES SET (SET- POINT OR COMMAND VOLTAGE IN) EQUAL READ (OUTPUT VOLTAGE)?

This should be the first question asked during any suspected MFC malfunction or failure. This will not tell you anything about the calibration of the device under test, but it will tell you whether or not it is functioning properly. If gas is applied at the correct inlet pressure, and the set-point voltage in equals the read voltage out, then the first deduction that may be safely made is that the feedback circuitry of the MFC is functioning properly.

[_private/bottom.htm]