Where the line actually splits: closed-loop control vs end-of-line inspection

A GC run on a new lot of inlet liners failed at injection 412. The liner had passed final QC for deactivation, but the autosampler needle dragged a small glass burr into the septum, sheared a fragment, and the next three runs showed ghost peaks. The burr was 0.14 mm high, inside a 0.8 mm ID liner bore, and no one on the line had measured it. That is the difference between what a lab chromatography factory claims to control and what it actually controls on the line. Most spec sheets list finished-product dimensions. They do not tell you which parameters are closed-loop controlled during the run, which are checked only at the end, and which are never checked at all.

When you walk a chromatography consumables line, the first thing you notice is that the word “control” means two different things depending on the station. A closed-loop parameter is one where a sensor feeds a controller that adjusts the process while parts are still being made. An end-of-line parameter is one where the part is measured after the process is done, and if it fails, the part is scrapped or reworked. The factory does not control the second type. It sorts it.

This distinction matters because most published specifications for lab chromatography consumables are a mix of both, and the datasheet rarely tells you which is which. A liner may have a stated ID tolerance of ±0.05 mm, but if that ID is only verified on a 5-piece sample per 500-unit lot, the tolerance is an inspection limit, not a process control limit. I have seen lots where the sample passed and the remaining 495 liners had IDs drifting by 0.08 mm because the glass drawing die wore during the shift.

Closed-loop parameters on a well-run line typically include oven zone temperatures during column coating, extrusion melt pressure for tubing, injection molding cavity pressure for vial caps, and web tension during membrane casting. These are adjusted in real time. End-of-line parameters typically include septum puncture counts, liner deactivation quality, frit porosity uniformity, and final assembled dimensions. The factory controls the first group. It verifies the second group. You should know which group each of your critical specifications falls into.

Station-by-station map: what gets adjusted live and what gets measured after

Glass tubing draw and liner forming

Inlet liners start as borosilicate or quartz tubing drawn from a preform. The draw tower has three closed-loop controls: furnace zone temperature, draw speed, and laser micrometer feedback on outer diameter. The laser measures OD at 200 Hz and adjusts draw speed to hold OD within ±0.02 mm. That is real closed-loop control. But liner ID is not directly measured during drawing. It is inferred from the preform ID and the draw ratio. If the preform ID varies by 0.03 mm, the liner ID varies by the same ratio, and no sensor catches it until a post-draw air gauge check. That check happens on a sample, not 100%.

After drawing, liners are cut to length, fire-polished, and deactivated. Deactivation is a silanization or silicon deposition step inside a vacuum oven. The oven temperature and gas flow are closed-loop controlled. The deactivation quality itself is not. It is checked after the batch with a test injection of a polar probe compound. A liner that tails active sites on the probe is scrapped, but the entire batch may have already been processed. The factory controls the oven. It sorts the deactivation result.

Frit pressing and porosity

Frits for HPLC columns and SPE cartridges are pressed from sintered stainless steel or polyethylene powder. The press force, dwell time, and powder fill weight are closed-loop controlled on modern presses. Fill weight is checked by a load cell in the die, and the press rejects under-filled cavities before sintering. That is a genuine in-process control. The resulting frit porosity, however, is a bulk property measured after sintering on a bubble point tester or porometer. A 2 µm frit may be specified at 2.0 µm nominal with a range of 1.6 to 2.4 µm, but that range is verified on one frit per tray, not every frit. I have seen a tray where the edge frits sintered hotter than the center frits and the porosity shifted by 0.6 µm. The sample frit from the center passed. The edge frits caused backpressure spikes in the customer’s column.

Column packing and bed stability

HPLC column packing is the station where closed-loop control is most mature and least transparent. A slurry packer controls slurry concentration, packing pressure ramp rate, and final packing pressure. The pressure transducer feeds back to the pump to hold the target packing pressure within ±50 psi. That is closed-loop. What is not controlled live is the bed density distribution. The factory measures column efficiency after packing by injecting a test probe and calculating plates per meter. That is an end-of-line test. A column that passes the plate count may still have a void at the outlet frit if the bed settled after packing. The void does not show up in the initial plate count. It shows up after 200 injections when the customer sees peak splitting.

For GC capillary columns, the coating station controls oven temperature, carrier gas flow, and coating solution feed rate. The film thickness is inferred from the feed rate and the column ID, not measured directly during coating. Film thickness is verified after coating by cutting a section and measuring it under a scanning electron microscope or by a retention factor test with a standard probe. That verification is destructive for the section and is done on a sample from each coating run. The factory controls the coating speed. It verifies the film thickness after the fact.

Septum molding and puncture resistance

Autosampler septa are molded from silicone or PTFE-faced silicone. The injection molding machine controls barrel temperature, injection pressure, and hold time in closed loops. Cavity pressure sensors can detect short shots and reject them in real time. That is good in-process control. But the property that matters to the user, puncture resistance over repeated injections, is not measured during molding. It is measured on a puncture tester after the batch. A typical spec is 100 punctures with no coring for a 11 mm GC septum at a needle gauge of 23. The tester counts punctures until the septum cores or leaks. That is a destructive end-of-line test on a sample of molded septa.

The failure mode I have seen most often is not coring. It is septum particles dropping into the inlet when the needle withdraws. That happens when the silicone durometer is too low or the PTFE facing delaminates. Durometer is checked on every batch, but the PTFE-to-silicone bond is checked on a peel test sample only. A batch that passes durometer and fails the peel test after three weeks on the shelf is a latent defect that no in-process sensor catches. The factory controlled the molding parameters. It did not control the adhesive cure time between the PTFE layer and the silicone body, because that cure happened in a separate lamination step with only a time and temperature log, not a bond strength sensor.

Vial forming and dimensional stack-up

Autosampler vials are formed from glass tubing or molded from Type I borosilicate. The critical dimensions are the outer diameter at the crimp neck, the height, and the inner diameter of the opening. OD and height are measured by laser gauges at the forming station, and the forming machine adjusts the mandrel position in a closed loop. That is real-time control. The neck ID is more difficult. It is checked with a go/no-go pin gauge after annealing, on a sample basis. A vial that passes OD and height can still have a neck ID that is 0.05 mm too small, which causes the crimp cap to seat incorrectly and the septum to sit at an angle. The autosampler needle then hits the septum off-center and bends. The factory controlled the outside of the vial. It sampled the inside.

For COD vials used in digestion methods, the wall thickness and thermal shock resistance are the parameters that matter. Wall thickness is controlled during forming by the glass feed rate and the mold temperature. Thermal shock resistance is checked after annealing by quenching a sample from 150°C into 20°C water and inspecting for cracks. That is a destructive batch test. A vial that passes the quench test can still fail after repeated autoclave cycles if the annealing schedule drifted slightly. The factory controls the forming. It samples the thermal shock resistance.

Membrane casting for filtration

PTFE and nylon membranes for HPLC filters are cast on a moving web. The casting line controls web speed, dope viscosity, air knife pressure, and oven temperature in closed loops. The membrane thickness is measured by a beta gauge or laser profilometer across the web, and the line adjusts the casting knife gap in real time. That is the most sophisticated closed-loop control on any consumables line. The pore size, however, is not measured during casting. It is verified after the membrane is dried, using a bubble point test or a porometer. A 0.45 µm membrane may pass the bubble point test on a sample cut from the center of the web, while the edges of the web have pores that are 0.6 µm because the air knife pressure was lower at the edges. The factory controlled the thickness. It sampled the pore size.

For PTFE membrane filters used with aggressive solvents, the hydrophobicity is a critical parameter that is not controlled live. It is checked after casting by a water contact angle measurement on a sample. A batch that passes the contact angle test can still have uncoated spots where the membrane wets prematurely and the sample does not pass through. The factory controls the casting speed and the oven temperature. It verifies the hydrophobicity on a sample.

What the factory cannot control, even with a good line

Some parameters are outside the reach of any in-process sensor. The interaction between the needle and the septum is one. The factory can control septum durometer, needle gauge, and inlet temperature separately. It cannot control the combined wear behavior of a specific autosampler needle against a specific septum lot, because that interaction depends on the needle tip geometry, the injection speed, and the number of injections. The factory tests this with a standard needle at a standard speed. Your autosampler has its own needle wear pattern and its own injection speed. The factory’s puncture test is a proxy, not a guarantee.

The same applies to liner deactivation under a specific method. A liner that passes the factory’s standard deactivation test with a polar probe may still show activity with your specific analyte, because your analyte has a different polarity, a different concentration, and a different matrix. The factory controls the deactivation oven. It cannot control the interaction between the liner surface and every possible analyte. That is why method-specific liner selection exists.

Column selectivity is another parameter that is only partially controlled. The factory controls the stationary phase chemistry and the coating thickness within specified ranges. It cannot control the exact selectivity factor for every analyte pair, because that depends on the phase batch, the column age, and the mobile phase. The factory provides a batch-specific test chromatogram. That is a verification, not a control. If your method is sensitive to a selectivity shift of 0.02 in alpha, you need to buy columns from a single coating batch or verify each new column against a reference standard.

Decision checklist: what to ask the factory before you qualify a lot

When I qualify a new consumable lot, I do not ask for the spec sheet first. I ask for the process control plan. The spec sheet tells me the inspection limits. The control plan tells me which parameters are adjusted live and which are only sampled. Here is the checklist I use, with the numbers that matter.

Two field cases show why this checklist matters.

First case: a pharmaceutical QC lab switched to a lower-cost inlet liner from a new supplier. The spec sheet matched the previous supplier exactly: 78.5 mm length, 6.3 mm OD, 4.0 mm ID, deactivated borosilicate. Within three weeks, the lab saw increasing RSD on peak areas for a residual solvents method. The RSD went from 1.8% to 4.6%. The liner dimensions were fine. The deactivation probe test was fine. The problem was the taper angle at the bottom of the liner. The new supplier’s liner had a taper angle that was 2 degrees steeper than the previous supplier’s, which changed the flow path and caused the sample to hit the column inlet at a different angle. No spec sheet listed the taper angle, and no in-process sensor measured it. The fix was to return to the previous supplier’s liner or specify the taper angle in the purchase order. That parameter was never on the datasheet.

Second case: an environmental lab running EPA SW-846 methods for semivolatiles saw a sudden increase in column backpressure after switching to a new lot of 0.45 µm syringe filters. The filters passed the bubble point test. The problem was the filter housing, not the membrane. The new lot had a housing outlet ID that was 0.1 mm smaller than the previous lot, and the smaller outlet created a flow restriction that increased backpressure by 15%. The housing ID was not on the spec sheet and was not controlled live. It was a molded dimension checked on a sample basis. The fix was to specify the housing outlet ID in the purchase order and to request a lot-specific dimensional report. The lesson is the same: the factory controlled the membrane. It did not control the housing.

What a well-run factory does differently

A well-run consumables factory does not just have more sensors. It has a control plan that says which parameters are adjusted live, which are sampled, and what the sample size is. It also has a traceability system that links each finished part to the process parameters used to make it. For a column, that means the packing pressure trace, the slurry lot, and the frit lot. For a liner, that means the draw tower log, the deactivation oven log, and the probe test result. For a vial, that means the forming machine log and the annealing oven log.

The difference shows up when a customer reports a failure. A factory with traceability can pull the process data for the specific serial number or lot and find the drift. A factory without traceability can only say the lot passed final inspection. I have worked with both. The first type fixes the root cause. The second type sends you a replacement lot and hopes the problem goes away.

The same applies to the consumables you buy from major suppliers. Agilent and Thermo Fisher publish extensive technical documentation for their chromatography consumables, but the process control plan is not in the catalog. You have to ask for it. Restek publishes application notes and technical guides that are useful for understanding failure modes, but the lot-specific process data still requires a direct request. The factories that are willing to share the control plan are the ones you want to qualify.

One more parameter that separates good lines from average lines: the frequency of tooling changes. A liner drawing die wears over time. A good factory tracks the die life in number of parts drawn and replaces the die before the ID drifts out of tolerance. An average factory runs the die until the final inspection catches the drift. The first factory prevents the drift. The second factory sorts it. The difference is invisible on the spec sheet, but it shows up in your RSD after 500 injections.

The decision rule is simple. For any critical consumable, ask the factory which parameters are closed-loop controlled and which are only sampled at the end. If the parameter that matters most to your method is only sampled, ask for the sample size, the sampling location, and the lot-specific data. If the factory cannot or will not provide that, find another factory. The spec sheet is not the process. The process is what happens on the line, and the line is where your RSD is actually made.

References & further reading

Related resources from our lab consumables network:

• Lab Chromatography at labchromatography.com
• Hplc Filter at hplcfilters.com
• Ptfe Membrane Filter at ptfemembranefilter.com
• Cod Vial at codvial.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography