When a process changes, an analyzer does not report the change instantly. There is always a delay, made up of the time it takes the sample to travel to the analyzer and the time the detector takes to respond once the sample arrives. Understanding that total delay is essential, because it is what limits how well an analyzer reading can be used to control a process in real time. This guide breaks down the two parts of the delay, explains the T90 response measure, shows why the total dead time constrains closed-loop control, and describes how engineers work around it.
Analyzer Response Time & Lag in one line: Analyzer response time and lag is the total delay between a change in the process and the moment the analyzer reports it. It has two parts: sample transport lag, the time the sample takes to travel from the tap to the analyzer, and detector response time, often stated as T90, the time the analyzer needs to reach ninety percent of a step change once the sample arrives. Together they form a dead time that limits how effectively an analyzer can be used for closed-loop control.
The first part of the delay is sample transport lag. Most analyzers do not sit in the process line; they draw a sample through tubing from a tap to a sample-conditioning system and then to the analyzer. The gas or liquid takes real time to travel that path, so a change that happens at the tap is not even present at the analyzer until that transport time has passed. The longer the sample line and the slower the sample flow, the larger this lag, and on a long run it can dominate the total delay.
The second part is the detector's own response time once the sample reaches it. No detector jumps instantly to a new value; it approaches the new reading over some seconds. This is commonly characterized as T90, the time the analyzer takes to reach ninety percent of the way to a new value after a step change at its inlet. A fast detector has a short T90 and settles quickly; a slower one takes longer to reflect the true value. Some analyzers, like a chromatograph that runs on a cycle, add their own kind of delay because a fresh result only appears once per analysis cycle.
The total response time an operator experiences is these parts combined: the sample must travel to the analyzer, and then the detector must respond. Because the transport lag is largely pure delay, during which nothing changes at the analyzer at all, it behaves as dead time, while the detector response is more of a gradual settling. Knowing the size of each part matters, because they behave differently and are reduced by different means.
Dead time is the enemy of tight control. In a control loop, the system makes a correction, then waits to see its effect and correct again. If the measurement that tells it the effect arrives late, every correction is based on stale information, and the loop can overcorrect, hunt, or become unstable. The transport lag portion of analyzer delay is exactly this kind of pure dead time, and a large one makes an analyzer a poor sensor to close a fast control loop around.
This is why analyzers, despite measuring exactly the quality variable operators care about, are often not used as the direct feedback in a fast loop. By the time an analyzer reports that a stream has drifted off target, the process has already been off target for the length of the delay, and a correction made now will itself only be seen after another delay. For a slow-moving property this can still be workable, but for anything that changes quickly, the analyzer's lag makes direct closed-loop control on its reading difficult to do well.
The detector response portion is generally less troublesome than the transport lag, because settling toward a value is easier for a controller to cope with than pure delay, but a long T90 still slows how fast the loop can see and react. Both parts together set a floor on how responsive any control built on the analyzer can be, which is why quantifying and minimizing the total is a real part of analyzer engineering.
Engineers deal with analyzer lag first by shrinking it where they can. Shortening and heat-tracing sample lines, increasing sample flow within limits, placing the analyzer closer to the tap, and choosing a faster detector all cut the delay. When a fast primary measurement exists that correlates with the analyzer's variable, control is often closed on that faster signal, with the slower analyzer used to trim or bias the setpoint rather than to drive the loop moment to moment, so the loop stays responsive while the analyzer keeps it honest over time.
The lag also has to be respected in alarming and interpretation. Because an analyzer reports the past state of the process delayed by its response time, operators and logic must read its value as reflecting conditions from moments ago, not right now, and time comparisons or diagnostics that pair an analyzer reading with a live measurement need to account for the offset. Ignoring the lag can make an analyzer look wrong when it is simply late.
A cloud SCADA platform such as Merobix historizes the analyzer's readings alongside the faster process measurements, so the delay between a process change and the analyzer catching up is visible in the trends rather than hidden. Seeing both signals together helps operators and engineers understand the effective lag, tune how the analyzer is used in control, and set alarm timing that reflects the real response rather than assuming an instant reading. On remote sites where the sample runs can be long, having the analyzer's response captured and reviewable in the same place as the rest of the data makes the lag a known, managed quantity instead of a hidden source of confusion.
Transport lag is the time the sample takes to travel from the process tap through the tubing and conditioning system to the analyzer, during which nothing has changed at the analyzer yet, so it behaves as pure dead time. Detector response time is how long the detector takes to settle to a new value once the sample reaches it, often stated as T90. The total delay an operator sees is these two combined.
T90 is the time an analyzer takes to reach ninety percent of the way to a new reading after a step change appears at its inlet. It characterizes how quickly the detector settles toward the true value, so a short T90 means a fast-responding analyzer. T90 covers only the detector's settling and does not include the sample transport lag, which is a separate part of the total response time.
Because their total delay, especially the sample transport lag, acts as dead time in a control loop. When feedback arrives late, corrections are based on stale information and the loop can overcorrect or become unstable. For fast-changing variables that makes direct control on the analyzer difficult, so engineers often close the loop on a faster primary measurement and use the analyzer to trim the setpoint instead.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.