addressable.wiki

WS2815

1-wire RGB · WorldSemi · ws281x

Channels
RGB
Bits / channel
8
Supply ≠
12 V
PWM ≠
2 kHz
Backup data
yes
Packages
7

Datasheets disagree

The sources we hold state different values for the facts below. Rather than pick one quietly, both sides are listed with where each was read. Check the datasheet for the exact part number and revision you are buying. How this is shown.

Reset period

The original sheet asks for 300 µs; the WS2815B-V1 sheet for more than 280 µs.

What each source says about reset period
SourceSaysPage
ws2815 RES 300 µs or more 5
ws2815b-v1-c5446699 RES > 280 µs 3
FastLED led_timing.h (driver-tuned) no reset recorded —

What we show. Both datasheet sets are listed; 300 µs satisfies both.

Bit timing

The bit windows differ widely between the original and V1 sheets, and FastLED’s values sit outside the original sheet.

What each source says about bit timing
SourceSaysPage
ws2815 T1H 750 ns–2 µs, T0L 750 ns–2 µs 4
ws2815b-v1-c5446699 T1H 580–840 ns, T0L 900–5000 ns 3
FastLED led_timing.h (driver-tuned) T1H 1340 ns, T0L 1640 ns, period 1890 ns —

What we show. The datasheet sets are listed in the timing table. For the very wide windows the stored midpoint is not a usable timing; read the windows.

PWM frequency

WorldSemi sheets give 2 kHz and 4 kHz, and one sheet gives both.

What each source says about pwm frequency
SourceSaysPage
ws2815 refresh frequency 2 kHz 2
ws2815b-v1-c5446699 "scan frequency is of 2KHz" and "Refresh Frequency updates to 4KHz" 1
ws2815b-4020-c42417605 port scan frequency 4 kHz 1

What we show. 2 kHz is shown (the original sheet). Later B sheets say 4 kHz.

Supply voltage

The sheets give a window around 12 V that moves between generations.

What each source says about supply voltage
SourceSaysPage
ws2815 VDD +9 to +13 V 4
ws2815b-v1-c5446699 VDD +9.5 to +13.5 V 2

What we show. The nominal 12 V rail is shown, not either window, so the part is found by the “12 V” filter. Stay inside the window on the sheet for your revision.

The 12 V member of the family. Running at 12 V means far less voltage drop along a long run, which is why it dominates architectural installations, at the cost of coarser cutting granularity and more heat per LED.

Timing

Sources disagree. This is normal and worth understanding: FastLED's numbers are tuned to drive real parts reliably, including clones, while the datasheet gives the manufacturer's nominal figures. Both are recorded here rather than picking one and hiding the conflict.

WS2815 bit timingWS2815 bit timing: a zero bit is 350 nanoseconds high then 1375 low; a one bit is 1375 high then 350 low; tolerance plus or minus 300 nanoseconds; reset 300 microseconds low1 · Bit encoding, to scaleDINhighlowbit 0 period 1725 nsbit 1 period 1725 nsT0H 350 ±300 nsT1H 1375 ±300 nsT0L 1375 ±300 nsT1L 350 ±300 ns"0" bit"1" bit050010001500200025003000time, ns± tolerance on each falling edge2 · One frame: 24 bits, most-significant bit firstDIN170110010007100110101700101100bitG · 8 bitsR · 8 bitsB · 8 bitsWire order GRB · frame = 24 × ~1725 ns ≈ 41.4 µs · example data, first bits 1 0 1 …3 · RESET / latchlast bits of frame 1first bits of frame 2LEDs latch the data they receivedRESET: DIN low ≥ 300 µs (not to scale)4 · Cascade: each DOUT feeds the next DINMCULED 1DINDOUTLED 2DINDOUTLED NDINDOUTLED 1 keeps the first 24 bits, re-shapes the signal and sends the rest out on DOUT;LED 2 keeps the next 24, and so on. The last LED's DOUT is left unconnected.BIN (backup): tied to the previous pixel's DIN
T0H 350 / T0L 1375 · T1H 1375 / T1L 350 ns · ±300 ns · period 1725 ns · reset 300 µs source: datasheet
SourceT0HT0LT1HT1LPeriodRateReset±
fastled ↗ 250 ns 1640 ns 1340 ns 550 ns 1890 ns 529 kHz 280 µs —
FastLED lists WS2815 among the 400-600 kHz chipsets; the derived bit period is 1890 ns, i.e. about 529 kHz - noticeably slower than the 800 kHz WS2812 family.
datasheet ↗ 350 ns 1375 ns 1375 ns 350 ns 1100 ns 909 kHz 300 µs 300 ns
Datasheet ranges, midpoints recorded here: T0H 220-480 ns, T1H 750 ns-2 us, T0L 750 ns-2 us, T1L 220-480 ns. The stated bit period is 1.1 us +/-300 ns, i.e. 800-1400 ns - but FastLED’s T1/T2/T3 derive a 1890 ns period, well outside that window, and a 550 ns T1L against a 220-480 ns spec. FastLED’s figures are the ones that reliably drive real parts; the datasheet’s are what WorldSemi publishes. Also note FastLED records a 0 us reset for this part while the datasheet requires 300 us or more.
datasheet ↗ 300 ns 2950 ns 710 ns 2800 ns 3250 ns 308 kHz 280 µs —
A LATER, DIFFERENT table from the mirrored one above: T0H 220-380 ns, T1H 580-840 ns, T0L 900 ns-5 us, T1L 600 ns-5 us, reset above 280 us (the older sheet: T0H 220-480 ns, T1H 750 ns-2 us, T0L 750 ns-2 us, T1L 220-480 ns, reset 300 us). Midpoints are recorded, and because T0L and T1L are WIDE windows (up to 5 us) their midpoints are not a usable timing; read the windows. Whether the silicon changed or only the documentation did is unknown. Supply 9.5-13.5 V (older sheet: 9-13 V).

Clones & compatibility

“Compatible” is never stated bare here — each relationship carries how strong the claim is, and what bites if you assume too much.

  • near clone WS2813

    Same backup-data-line concept, but 12 V and a different bit period. Do not assume WS2813 timings will drive it.

  • electrically similar of this GS8208

    Both are 12 V single-wire parts with a backup data line, and both are driven with WS2812-family timings, but they are different silicon with different specifications.

  • near clone of this WS2818

    Both are 12 V parts with a backup line; the bit periods differ (1200 ns vs 1890 ns), so they are not interchangeable without changing timings.

Library support

FastLED
WS2815 ↗
Adafruit_NeoPixel
NEO_GRB + NEO_KHZ800
WLED
WS281x

Parts using this protocol

Dating evidence

Evidence, not a release date. A library commit proves the part existed by that date and says nothing about how long before.

  • 2020-01-13 first library support WLED — first commit introducing WS2815 ↗ day
  • 2024-09-12 first library support FastLED — first commit introducing WS2815 ↗ day