Technical Introduction · For instrumentation and process engineers · Published: October 2, 2026
Table of Contents
Key Takeaways
- T90 is the time a sensor’s output takes to reach 90% of its final stable value after a step change in the measured concentration. It is the standard way water quality instrument makers express dynamic tracking capability [1].
- T63.2, the time to reach 63.2% of the change, corresponds to one time constant of a first-order system. T90 and T95 are quoted alongside it on the same ladder [3].
- Technology sets the baseline for dissolved oxygen: polarographic cells typically reach t90 in roughly 15-30 s, while galvanic and optical sensors run roughly 30-60 s depending on membrane or cap design and temperature [2].
- In pH electrodes, membrane thickness is a deliberate trade-off. Thin membranes of about 0.1-0.15 mm reach 95% of the stable reading in 10-30 s, while rugged thick membranes of 0.3-0.4 mm take 60-90 s [4].
- Flow matters as much as chemistry. Steady-state electrochemical DO sensors consume oxygen during measurement and need sample movement across the membrane for accurate readings [5].
- Match the number to the loop: for fast control such as aeration feedback, 60 s or better is recommended, while 90-120 s standard models are adequate for trend monitoring [6].
T63, T90, T99: One Step Change, Three Numbers
Every water quality sensor answers the question “what is the value right now?” with a delay. Response time is the formal answer to how large that delay is, and T90 is the number that carries meaning in real specifications: the time for the instrument output to rise from the moment it contacts a new concentration to 90% of the final stable value [1].
The family of numbers around it comes from the mathematics of first-order systems, which most membrane and electrochemical sensors approximate. A step change produces an exponential settling curve, so the time to reach 63.2% of the change — T63.2 — is the system’s time constant, and T90 and T95 sit further up the same ladder [3]. T99 marks the point most engineers would call “settled” in daily practice.
Why quote T90 rather than the time constant? Because T90 matches what an operator cares about: how long after a real event — a storm inflow, a chemical dump, a pump trip — the reading in front of you will be essentially true. A specification that omits T90 is asking you to guess at the one number your control loop cannot tolerate.
What Actually Sets Response Speed
Membrane and glass thickness
The sensing barrier is the dominant physical delay. In pH electrodes, glass membrane thickness is a trade-off between speed and survival. Thin membrane designs of roughly 0.1-0.15 mm let hydrogen ions diffuse to the inner surface quickly, reaching 95% of the stable reading within 10-30 s of contacting a new solution. Rugged thick membranes of 0.3-0.4 mm extend that to 60-90 s in exchange for mechanical strength suited to gritty, industrial streams [4]. The same logic applies to oxygen-permeable membranes and to the cap designs that hold them.
Sample flow at the membrane
Membrane-covered electrochemical DO sensors consume oxygen as they measure, so they depend on fresh sample being delivered across the membrane surface. The effect is commonly called flow dependence, and without adequate movement the readings drift artificially low [5]. Optical (luminescence-based) designs do not consume the analyte the same way and are therefore less sensitive to flow, which is part of why they are favored for profiles and low-flow applications [5]. The practical consequence for response time: an under-flowed sensor does not just read low, it reads late, because the stagnant boundary layer thickens the diffusion path the signal has to cross.
Temperature
Diffusion slows as water gets colder, and manufacturers normally quote response behavior together with temperature conditions [2] [4]. In seasonal plants this is worth remembering: a loop tuned in summer behaves differently in winter not because the controller changed, but because the sensor’s intrinsic lag did.
Technology choice
The measurement principle sets the floor. For dissolved oxygen, polarographic cells typically reach t90 in roughly 15-30 s, while galvanic and optical sensors run roughly 30-60 s depending on membrane or cap design and temperature [2]. Faster specialized designs exist — high-precision fluorescent DO instruments with T90 controlled within about 40 s are on the market for capturing rapid oxygen dynamics such as aeration cycling and storm responses [1].
The Hidden Delay: Sample Transport
Sensor lag is only part of the story. In online installations the sample usually travels: tubing runs, filters, and flow cells add their own transport delay, and the longer and wider the line, the larger the delay. Adsorption on line materials and clogged filters slow the response further [3]. The loop designer’s honest dead time is therefore:
total lag = sample transport delay + sensor T90 + transmitter/PLC scan and damping time.
A sensor with a superb 20-second T90 installed behind twenty meters of narrow tubing can still feed the controller news that is minutes old. When commissioning a dosing loop, measure the whole chain with a step test at the process, not just the sensor on the bench.
Why T90 Matters in Control Loops
Dosing feedback lag. In a neutralization or disinfection loop, every second of measurement lag is dead time the controller cannot see past. The dosing pump acts on a value that was true T90-plus-transport seconds ago. The classic symptoms are overshoot, oscillation, and chemical consumption above theoretical demand. Shortening the sensor’s response time — or moving the sensor closer to the mixing point — directly reduces reagent waste and effluent excursion risk.
Aeration control. Dissolved oxygen in biological treatment evolves over minutes, and a lagging sensor makes blowers chase stale data: over-aerating after a load drop, under-aerating after a spike. Selection guidance for DO electrodes used in fast process control such as aeration regulation points to response of 60 s or better, while slower 90-120 s standard models remain perfectly serviceable for trend monitoring [6].
Event capture. Some of the most valuable data in water treatment is transient: the oxygen sag after a storm inflow, the chlorine dip during a backwash breakthrough, the conductivity spike when regeneration waste reaches the sampling point. A slow sensor does not merely delay these events — it can miss the turning point entirely, which is exactly why fast-response optical DO designs emphasize T90 within tens of seconds for dynamic tracking [1].
Matching Response Speed to the Application
| Application | Response expectation | Why |
|---|---|---|
| Aeration / blower control in biological treatment | T90 of 60 s or better recommended [6] | DO shifts within minutes; a lagging sensor makes blowers hunt |
| Trend and compliance monitoring | 90-120 s standard models acceptable [6] | Dynamics are slow; the faster sensor buys little |
| Neutralization and disinfection dosing feedback | Fastest practical T90; thin-membrane pH designs reach 95% in 10-30 s [4] | Every second of dead time becomes chemical overshoot |
| Intake and outfall early warning | Sub-minute T90 preferred — fast optical DO designs achieve about 40 s [1] | Storm, shutdown, and breakthrough transients must be captured, not smoothed over |
| Verification sampling and lab comparison | Not critical | The sample is taken after stabilization; T90 only sets waiting time |
Specifying and Verifying Response Time
Specify it. Put T90 in the purchase specification, together with the conditions it was measured at — temperature, sample flow across the membrane, and the magnitude of the step [2] [4]. A transmitter datasheet should be read alongside the electrode’s response data, since output filtering and scan settings add their own lag. Shanghai ChiMay’s industrial pH/ORP transmitter product pages, for example, list the measurement, compensation, and output specifications a loop designer needs alongside the electrode data (chimaytech.com).
Verify it. After installation, run a step test at the process: move the sensor between two known conditions and time the output’s rise to 90% of its final value — the T90 you specified [1]. Do this once at commissioning and again each season, and treat a response that has grown noticeably slower as a maintenance signal rather than a curiosity, because aging membranes and fouled surfaces slow sensors down before they fail outright. That is the practical meaning of “Buy the sensor, own the outcome.”
Design for it. Where multiple parameters feed one control decision — a dosing loop that keys off pH with conductivity and temperature as context, say — a synchronized multi-parameter probe keeps every signal referenced to the same sample moment and the same flow field. Four parameters, one probe, zero alignment error is also a response-time statement: the parameters cannot disagree about when the event happened.
References
- Easysensor — “T90 ≤ 40 s response speed: high-precision fluorescent dissolved oxygen analyzers capture DO changes in real time” — manufacturer technical note defining T90 (time for the output to reach 90% of the final stable value after contacting a new concentration) and reporting T90 within 40 s for optical DO instruments used for dynamic tracking. http://www.easysensor.com.cn/Article-4405175.html
- SpecForge (SourceBySpec) — “Dissolved Oxygen Meter” encyclopedia entry — cites t90/t95 quoting conventions and typical values by technology: polarographic cells roughly 15-30 s; galvanic and optical sensors roughly 30-60 s depending on membrane or cap design and temperature; references ISO 17289. https://www.sourcebyspec.com/encyclopedia/dissolved-oxygen-meter.html
- GPL Instruments — “Response speed of oxygen analyzers” — technical note giving the T63.2 / T90 / T95 ladder (T63.2 as the first-order time constant) and explaining how sampling-line length and volume, adsorption, and filters add transport delay. https://gpl-instruments.com/news-46%20Response%20speed%20of%20the%20oxygen%20meter.html
- Jixie100 — “Maintenance of measuring pH electrodes” — explains the glass membrane thickness trade-off: thin membranes (~0.1-0.15 mm) reach 95% of the stable reading in 10-30 s; rugged thick membranes (0.3-0.4 mm) take 60-90 s. http://www.jixie100.net/wiki/mrf2nz3u2o-55500329.html
- YSI — “What is affecting your dissolved oxygen measurements, Part 1” — explains flow dependence: steady-state electrochemical (polarographic and galvanic) DO sensors consume oxygen during measurement and require sample movement across the membrane, while optical sensors are far less flow-sensitive. https://www.ysi.com/ysi-blog/water-blogged-blog/2013/05/what-is-affecting-your-dissolved-oxygen-measurements-part-1-of-4
- Chem17 — “Selection guide for fluorescent dissolved oxygen electrodes: membrane caps and response time” — selection practice noting T90 differences by design and recommending response of 60 s or better for fast process control such as aeration regulation, with 90-120 s standard models suited to trend monitoring. https://m.chem17.com/st693015/article_4505518.html
About the author: Written by the Shanghai ChiMay Technical Editorial Team, which engineers online water quality analyzers and multi-parameter sensors for drinking water, wastewater, and industrial process applications, with hands-on experience in sensor selection, installation design, and control-loop commissioning.
