Open almost any CW laser datasheet and you will find the phrase “< 1% RMS.” Open ours and you may find it twice — once against power stability, once against amplitude noise. Same number, same units, two rows apart.
They are not the same measurement. They are not even measuring the same physical behaviour. One describes how far your laser wanders over an afternoon; the other describes how much it flickers in a microsecond. Confusing them is the single most common mistake in laser specification, and it leads to laboratories buying stability they do not need while missing the noise that is actually widening their data.
This guide unpacks both numbers, explains why a noise figure without a bandwidth is meaningless, and gives you a method for working out which one your experiment cares about.
What RMS Actually Means
RMS stands for root mean square. Take your laser's output power, sample it many times, and you get a set of readings scattered around a mean. Subtract the mean from each reading, square the differences, average them, take the square root. That is the RMS deviation — the standard deviation of the power, in watts.
Divide by the mean power and you get a percentage. A 500 mW laser specified at < 1% RMS is fluctuating by less than 5 mW RMS about its average.
So far, so unambiguous. The ambiguity arrives with two questions the percentage alone does not answer:
- Over what period were the samples taken? One second, or eight hours?
- Over what frequency range? What was filtered out before the numbers were recorded?
Those two questions separate the two specifications completely.
Number One: Power Stability — The Slow Measurement
Power stability answers a question about time: if I turn this laser on and leave it running, how far will its output wander?
The measurement is made over hours, typically four or eight, after the warm-up period has elapsed. It captures slow drift — the gradual shift in output caused by the laser head reaching thermal equilibrium, by the ambient temperature of the room changing as the building's HVAC cycles, by pump diode ageing across the working day, by cavity alignment shifting a micron as aluminium expands.
Nothing in this measurement happens quickly. Every fluctuation it records unfolds over seconds to hours. Two other specifications on the same datasheet are really part of the same story:
- Warm-up time — the period before the stability spec applies at all. Ours is < 5 minutes on most lines, < 10 minutes on a few. Measure during warm-up and you will see drift far exceeding 1%, entirely by design.
- Operating temperature range — the ambient window within which the stability figure holds. Most of our lines specify 10 – 35 °C. Run a laser outside its stated range and the stability number no longer describes your system.
Who cares about power stability? Anyone whose measurement takes a long time, or is compared against something recorded earlier. Raman spectroscopy with minute-long integrations. Quantitative fluorescence where today's intensity must be comparable to last week's. Any calibrated instrument where absolute signal level carries meaning. Long-exposure imaging. Photolithography and materials processing, where dose is the product of power and time.
Number Two: Amplitude Noise — The Fast Measurement
Amplitude noise answers a question about frequency: within a defined band, how much does the output flicker?
This is why our low-noise specification is written out in full rather than left as a bare percentage:
Amplitude noise (RMS, 20 Hz – 20 MHz) < 1%
The bracketed range is not decoration. It is the part of the specification that makes the percentage mean anything at all. The measurement passes the detector signal through a filter that rejects everything outside 20 Hz to 20 MHz, then computes RMS on what remains.
Change the bandwidth and you change the number — and you can change it enormously. A vendor quoting “< 0.1% RMS” measured from 10 Hz to 1 kHz may well have a noisier laser at your working frequency than one honestly quoting < 1% across 20 Hz – 20 MHz. The narrower window simply excludes most of the noise before the calculation begins. A noise figure without a stated bandwidth is not a specification. It is a marketing number.
Why the window sits at 20 Hz – 20 MHz
Both limits are deliberate.
The 20 Hz floor separates noise from drift. Below roughly 20 Hz you are no longer measuring flicker — you are measuring the slow wander that power stability already covers. Setting the high-pass there stops the two specifications from double-counting the same physical behaviour.
The 20 MHz ceiling comfortably exceeds the response speed of essentially every detection system these lasers feed. A PMT-based flow cytometer, a confocal photodiode, a spectrometer CCD — none of them resolve events at 20 MHz. Noise above that ceiling is averaged away by your own instrument before it can affect a measurement.
Between those two limits sits everything that can genuinely corrupt your data.
Where the Noise Comes From
Noise is not a manufacturing defect. It is a physical consequence of how each laser architecture generates light — which is why the noise character of a DPSS laser differs fundamentally from that of a diode.
DPSS: relaxation oscillations
In a diode-pumped solid-state laser, the energy stored in the gain crystal and the photons circulating in the cavity form a coupled system that can ring, much as a struck bell does. Any disturbance — a flicker in pump current, a thermal transient — excites a damped oscillation between inversion and photon number. The resonance typically lands in the tens to hundreds of kilohertz.
That is squarely inside the 20 Hz – 20 MHz window, and it is the dominant noise feature of most DPSS lasers. It is also why the low-noise variant of a DPSS line is doing real engineering work: quieter pump drive electronics and tighter thermal control damp the resonance rather than merely filtering the output.
Frequency-doubled DPSS lasers carry a second, subtler mechanism. When several longitudinal modes oscillate together inside a doubling crystal, sum-frequency mixing couples them, and the resulting competition can produce large chaotic amplitude fluctuations — the effect long known in the literature as the “green problem.” It is managed through cavity design, and it is one reason DPSS lines such as 532 nm and 561 nm demand more engineering than their simple block diagrams suggest.
Diodes: current ripple and mode hopping
A direct diode has no separate gain medium and no doubling stage, so it has no relaxation oscillation in this band at all — its equivalent resonance sits up in the gigahertz, far above the measurement window. In exchange, the diode couples drive current straight into optical output. Whatever ripple the power supply produces appears in the beam, essentially instantly. Diode noise performance is therefore largely a question of driver quality.
Diodes add one behaviour DPSS lasers do not: mode hopping. As temperature or current shifts, the diode can jump between longitudinal modes, producing a small step in output power and a small shift in wavelength. Good thermal control suppresses it, and it is a large part of what tighter temperature regulation buys you in a low-noise diode variant.
The broader trade-offs between these two architectures are covered in diode laser vs. DPSS laser.
The Useful Trick: Match the Band to Your Measurement
Here is the practical method, and it is more informative than any rule of thumb about acceptable percentages.
Your detection system is a low-pass filter. It integrates light over some characteristic time, and noise faster than that time averages out before it reaches your data. Find that time, invert it, and you have the frequency where laser noise starts to matter to you.
| Measurement | Typical integration time | Noise band that matters |
|---|---|---|
| Flow cytometry — single cell transit | ~2–10 µs | ~100 kHz – 500 kHz |
| Confocal microscopy — pixel dwell | ~1–10 µs | ~100 kHz – 1 MHz |
| Widefield camera exposure | ~10–100 ms | ~10 Hz – 100 Hz |
| Raman spectroscopy — CCD integration | 1–60 s | Below 1 Hz — this is drift |
| Optical trapping / interferometry | Continuous | Full band, often to MHz |
Read the first two rows against the previous section and the picture sharpens considerably. Flow cytometry and confocal imaging both sample in the region of 100 kHz to 1 MHz — which is precisely where DPSS relaxation oscillations live. That overlap is the single strongest argument for specifying a low-noise variant on those instruments, and it explains why the tier exists at all.
Read the fourth row and the opposite conclusion follows. A Raman spectrometer integrating for thirty seconds averages away everything in the amplitude-noise band. What it cannot average away is drift between the sample scan and the reference scan. For that instrument, power stability is the specification that matters and amplitude noise is nearly irrelevant.
What Noise Does to Your Data
In fluorescence measurements the mechanism is direct: emitted signal is proportional to excitation power, so a 1% fluctuation in the laser produces a 1% fluctuation in the signal — indistinguishable from a real difference between samples.
In flow cytometry this shows up as broadened coefficients of variation. Worth knowing, though, is that the contributions add in quadrature rather than arithmetically:
CVtotal = √(CVbiological² + CVlaser² + CVoptical² + CVelectronic²)
That square-root relationship cuts both ways, and it is worth understanding before you spend money. If your other error sources already total 5%, reducing laser noise from 1% to 0.3% moves total CV from 5.10% to 5.01% — undetectable. But in a well-built instrument where every other term has been driven down to around 1%, laser noise becomes a co-dominant contributor, and improving it produces a visible tightening of your populations.
The honest conclusion is that low noise is worth paying for when the rest of your measurement chain is already good, and largely wasted when it is not. Our flow cytometry laser selection guide places this alongside the wavelength and power decisions.
Where neither number matters much
Plenty of legitimate applications are indifferent to both specifications, and there is no virtue in over-specifying them: alignment and targeting, optical pumping, machine vision illumination, laser display, most materials processing, and general laboratory illumination. If your application is on that list, buy the standard variant and put the difference toward power or beam quality.
Standard, Low Noise, and Pro
Our three-tier structure maps onto this distinction directly, and it should now read clearly.
| Tier | Grade code | Power stability | Amplitude noise |
|---|---|---|---|
| Standard | E | < 1% RMS | Not specified |
| Low Noise | Z | < 1% RMS | < 1% RMS, 20 Hz – 20 MHz |
| Pro | P | < 1% RMS | < 1% RMS, 20 Hz – 20 MHz |
Two points follow from that table, and both matter when you are comparing quotes.
First: every variant we ship holds < 1% RMS power stability. Drift performance is not what you buy when you step up a tier — every tier is actively power stabilised.
Second: what the Low Noise tier adds is a guaranteed, bandwidth-defined noise figure. A standard variant is not necessarily noisy; it is simply not tested and warranted against a noise specification. If your measurement lives in the 100 kHz – 1 MHz region, that guarantee is the whole point of the tier. If it does not, you are paying for a number your instrument cannot see.
The Pro tier carries the same noise guarantee and adds beam quality — TEM₀₀ at M² < 1.1 or 1.2, tighter polarisation, a smaller beam. It is a beam-quality tier that happens to include the noise specification, not a quieter one.
What to Ask Any Laser Vendor
- Over what bandwidth is the noise figure measured? If there is no answer, the number is not comparable to anyone else's. This one question separates real specifications from marketing copy.
- Over what duration is power stability measured, and after how long a warm-up? Eight hours after full warm-up is a meaningfully different claim from thirty minutes.
- Is the figure typical or guaranteed? “Typical” describes a good unit on a good day. Ask what is warranted on the unit you will receive.
- Over what ambient temperature range does it hold? A stability figure valid only at 22 ± 1 °C describes a different product in a room that swings ten degrees overnight.
- Is a noise spectrum available, not just a single number? A plot of relative intensity noise against frequency tells you where the peaks sit — and whether they overlap your measurement band.
Aimpico Noise and Stability Specifications
Every CW laser in our catalogue — from 261 nm through 1550 nm, DPSS and diode alike — specifies power stability of < 1% RMS, with active power stabilisation as standard. Warm-up is under five minutes on most lines, with operating ranges typically spanning 10 – 35 °C.
Low Noise (Z) and Pro variants add a guaranteed amplitude noise figure of < 1% RMS measured across 20 Hz – 20 MHz — a fully stated bandwidth, so you can compare it against any other vendor's number on equal terms. Both figures appear on the specification table of every product page, including 488 nm and 532 nm.
If you can tell us your detector, your integration or dwell time, and what the rest of your error budget looks like, we can tell you honestly whether the low-noise variant will change your data — including the cases where it will not.
Not sure whether your application needs a low-noise laser?