RoR Smoothing

RoR smoothing changes how stable and responsive the rate-of-rise curve feels. Use this page when the RoR line is too noisy, reacts too slowly, or you want to understand the smoothing controls in Settings.

For most users, the default setting is the right starting point: LWMA / Standard, RoR period 60s, and BT pre-smoothing off.

What RoR Smoothing Does

BT changes are affected by airflow, bean movement, probe placement, and sampling jitter. RoR magnifies those small changes because it measures the speed of temperature change.

Smoothing reduces short-term noise so the trend is easier to read. The trade-off is delay: stronger smoothing gives a calmer curve, but the curve reacts later when the roast changes direction.

Processing Order

HiBean calculates RoR in this order:

Stage	Default	Purpose
BT pre-smoothing	Off	Filters the temperature signal before RoR is calculated.
RoR algorithm	On	Converts BT samples into RoR. HiBean supports LWMA and Stable Regression.
RoR post-smoothing	On	Filters the RoR output to reduce remaining jitter.

Start with the RoR algorithm and post-smoothing preset. Only enable BT pre-smoothing when the probe signal itself is unstable.

Choose a Preset

Goal	Recommended setting
Fast response for manual adjustments	Low / Less smoothing
Daily roasting	Standard smoothing
Very noisy probe or demo-friendly chart	High smoothing

If you are unsure, use Standard. Change one setting at a time, then compare the curve in a roast log or practice roast.

LWMA vs Stable Regression

Algorithm	What it feels like	Good for
LWMA	Calm and easy to tune. Higher weights use more history, so the line is steadier but slower.	Daily roasting, noisy probes, users who want fewer knobs.
Stable Regression	Tracks a fitted temperature trend over recent samples. Lower levels can react earlier, while higher levels are smoother.	Users who want more detailed control over RoR response.

HiBean stores both preset and custom parameters. Switching algorithm changes which custom controls are shown.

Preset Reference

Algorithm	Level	Main parameters	Post filter
LWMA	Less	temp decay = 6, RoR decay = 6	Moving Average window = 1
LWMA	Standard	temp decay = 10, RoR decay = 10	Butterworth window = 3
LWMA	High	temp decay = 16, RoR decay = 16	Deadband EMA window = 4
Stable Regression	Low	sample count = 12, pre-smoothing tau = 2.25s	Built into the regression preset
Stable Regression	Standard	sample count = 16, pre-smoothing tau = 3.25s	Built into the regression preset
Stable Regression	High	sample count = 20, pre-smoothing tau = 5.0s	Built into the regression preset

Custom Controls

BT Pre-Smoothing

BT pre-smoothing runs before RoR is calculated. It can help with unstable sensors, but it adds delay that later stages cannot remove.

Use it only when:

• the BT line itself is visibly noisy;
• the sensor is loose or the environment is causing obvious sampling noise;
• you are testing hardware behavior and want to compare raw vs filtered input.

For normal roasting, keep it off.

LWMA Custom Settings

• Temperature decay weight: higher values use more history and produce a steadier curve.
• RoR decay weight: higher values smooth the RoR output more strongly.
• Post filter: Moving Average, Butterworth, or Deadband EMA.
• Post window: window 1 is effectively off for Moving Average; larger values add more smoothing and more delay.

Stable Regression Custom Settings

• Sample count: more samples make the curve steadier but slower.
• Pre-smoothing tau: higher values make the input trend calmer and less reactive.

Practical Tuning

1. If RoR jumps around too much, move from Less / Low to Standard before touching custom values.
2. If RoR reacts too late after a control change, move down one smoothing level.
3. If only BT is noisy, check probe placement first. Then consider BT pre-smoothing.
4. If the chart is readable during the roast, avoid tuning further during that roast.

RoR is a decision aid, not a control target by itself. Use it together with BT, ET, roast events, smell, color, and sound.