Conductivity to TDS: Why the Conversion Breaks Your Rejection Math

Your panel meter reads microsiemens. The data sheet, the lab report, the compliance binder and the customer all talk in ppm. So somebody converts, and the ppm value that comes out depends on a setting in the meter's menu nobody has looked at since the system was commissioned. For logging a trend, that's fine. It stops being fine the second you divide one converted number by another, because that's where a menu setting turns into a decision about whether to pull elements.

I'd bet most "my membranes are finished" calls are this and nothing else. The feed number came off one scale, the permeate number came off another, and the arithmetic in between produced a rejection figure low enough to look like a failed element. Healthy water, healthy membrane. Bad bookkeeping.

Convert when a report needs ppm. Don't convert inside a calculation.

One reading, two scales, two different ppm numbers

A TDS meter is a conductivity meter with a multiplier bolted on. Which multiplier you get depends on which standard solution the scale was built around, and the common ones differ enough to matter.

Myron L's 442 standard is 40% sodium sulfate, 40% sodium bicarbonate and 20% sodium chloride in deionized water, built to behave like natural fresh water. The NaCl scale is for water that's predominantly sodium chloride, and Myron L's own bulletin puts 1,000 ppm of NaCl at 2,000 micromhos, so that scale is built around a 0.50 multiplier at that one point. KCl is the international calibration reference for conductivity instruments rather than a model of anyone's feed water.

Read the cross-reference table in that bulletin and the tidy "0.5 and 0.7" factors everyone quotes turn out to be averages. On the 442 scale the implied multiplier climbs from 0.655 at 229 µS/cm to 0.769 at 3,900 µS/cm. On the NaCl scale it runs 0.475 to 0.520 across the same span. Myron L says it plainly: "The 442 to Conductivity correlation is non-linear and no single multiplier can be used to determine the relationship."

Which means one panel reading of 1,417 µS/cm is 1,000 ppm on 442 and 704 ppm on NaCl. Both are correct. Both came out of the same instrument. If you want the longer argument about which factor suits your water, that's in our piece on what a TDS number tells you about your RO system.

For picking a scale and leaving it alone: a brackish well or a seawater blend, where the feed is predominantly sodium chloride, belongs on the NaCl scale. General municipal fresh water is closer to what 442 was built to imitate. Then put a strip of tape on the meter with the scale written on it, because the next tech who picks it up has no way of knowing which one you chose, and that's how a feed and a permeate end up on different settings in the first place.

USGS puts the honest range at 0.5 to 0.9

True TDS is a mass measurement. Filter the sample, evaporate it, dry the residue at 180°C, weigh what's left. That's Standard Methods 2540 C, it needs a bench and a balance, and nobody runs it between service calls.

USGS Techniques and Methods 9-A6.3 expects the ratio of dissolved solids in mg/L to specific conductance to land between 0.5 and 0.9. Run that 1,417 µS/cm feed through both ends and you get 709 mg/L at one and 1,275 mg/L at the other. The top answer is 80% larger than the bottom one, and everything in between is defensible depending on what's dissolved in the water.

That's the whole reason a lab report and a handheld disagree by more than anybody expects. Neither instrument is broken. One weighed a residue and the other measured how well dissolved ions carry a current, then guessed at the mass.

Temperature, and why it only sometimes matters

Conductivity climbs with temperature. USGS gives the increase as 0.5 to 3 percent per degree Celsius, varying with the ionic species present, and notes that meters correct for it with either a linear coefficient of 0.019 to 0.020 per °C or a nonlinear one. Specific conductance is referenced to 25°C by convention.

Now the part that's less alarming than it looks. The error largely cancels in a ratio. A feed and a permeate both at 15°C, both read on uncompensated meters, both come in roughly 20% low, and the ratio between them barely moves. Rejection lands within a rounding error of where it belongs, which is lucky, because the coefficient isn't identical for two streams of different composition anyway.

The trouble starts when the two streams are at different temperatures, or when one meter compensates and the other doesn't. Say the feed is tapped off a cold incoming line at 15°C and reads 1,134 µS/cm raw against a true specific conductance of 1,417. Permeate gets sampled downstream of a warm storage tank at 25°C and reads 28. Rejection falls out at 97.53% instead of 98.02%. Half a point, courtesy of a thermometer.

Compute rejection on one basis, or don't compute it

Salt rejection is one minus permeate over feed, times 100. Trivial arithmetic, and entirely dependent on both numbers meaning the same thing. Take that same 1,417 µS/cm feed and a 28 µS/cm permeate, and read them three ways. Every ppm figure below comes off the cross-reference table at the conductivity it belongs to, interpolating between rows where it has to, so the low permeate reading gets its own multiplier instead of the feed's.

Same water, same element, and in the bottom two rows the same pair of meters. The spread is 1.26 percentage points. On a brackish rack that's the gap between an element you leave alone and a purchase order. Log conductivity in µS/cm with the sample temperature beside it, take both streams on one meter at one scale with one compensation setting, and the problem disappears. Our walkthrough on reading a normalized RO performance report covers where that number goes next.

The changeout the meter ordered

This is the failure that costs real money, and it almost never gets diagnosed as a measurement problem. An operator computes 97.44% on mismatched scales, compares it against the 99.65% minimum rejection on a NanoH2O (formerly LG) BW 400 R G2 data sheet, and concludes the element is passing seven times the salt it should. Elements get ordered. The replacements read the same, because the meter settings didn't change.

The data sheet number wasn't directly comparable to begin with. NanoH2O tests the BW 400 R G2 at 2,000 ppm NaCl, 225 psi, 25°C, pH 7 and 15% recovery. DuPont tests the FilmTec BW30 PRO-400 at 2,000 ppm NaCl, 225 psi, 25°C and 15% recovery as well, but at pH 8. Close, not identical, and neither one resembles the water in your feed tank. Comparing published percentages across makers without reading the test block underneath them is the same error one level up, and we went through it in detail on low pressure RO membranes and their test conditions.

Before anybody raises a requisition: pull a feed and a permeate sample within a minute of each other, run both through the same meter on the same scale, note the temperature of each, and compute rejection off the raw µS/cm. If that number clears the element's minimum rejection and the logged one didn't, the logbook was the problem and the elements are fine.

Which number belongs in the logbook

Conductivity in µS/cm with a temperature next to it, every time, and a ppm conversion only where a report demands one. Rejection gets computed on conductivity or it doesn't get computed. And when a clean paired sample does come in under the element's minimum rejection, then you've got a membrane to replace.

On an 8-inch brackish rack, the element worth specifying there is the NanoH2O BW 400 R G2. Read the test blocks before the percentages, same as always. DuPont and NanoH2O both rate these at 2,000 ppm NaCl, 225 psi, 25°C and 15% recovery on 400 ft² of active area, and the one condition that differs is test pH, 8 against 7. That's about as close as two makers' sheets get, and it still isn't a controlled head to head. On those sheets the BW 400 R G2 publishes 99.8% stabilized rejection against the FilmTec BW30 PRO-400's 99.6%, and 11,500 gpd against 11,000.

Enviro Membranes stocks genuine NanoH2O (formerly LG) RO elements in McAllen, Texas, with same-day US shipping. Published prices drop 5% at 5 to 9 units and 10% at 10+ per model; see current models at enviromembranes.com/shop. Questions about a rejection number before you order elements? Email us the paired readings and the meter model at enviromembranes.com.