If you check an outbound SMS with message.length <= 160 before sending it through a carrier API, three things will eventually surprise your invoice:
- One non-GSM-7 character flips the whole message to UCS-2, dropping the single-segment limit from 160 septets to 70 UTF-16 code units (and 67 per segment once split).
-
Ten ASCII-looking characters in the GSM-7 extension table (
^,{,},\,[,~,],|,€, and form feed\f) stay in GSM-7, but each one costs 2 septets instead of 1. -
Most emoji sit above
U+FFFF, so a single emoji takes 2 UCS-2 code units (a UTF-16 surrogate pair) while switching the entire message to UCS-2.
Here is how the segment math works in 3GPP TS 23.038, along with a small JavaScript function that counts segments the way carrier gateways do.
The capacity rules
An SMS frame carries 140 bytes (1,120 bits) of payload:
-
GSM-7 (7 bits per septet):
1120 / 7 = 160septets in a single SMS. When a message exceeds 160 septets, a 6-byte User Data Header (UDH) is prepended to each part so the handset can reassemble them, leaving160 - ceil(48 / 7) = 153septets per segment. -
UCS-2 (16 bits per code unit):
1120 / 16 = 70code units in a single SMS. With a 6-byte (48-bit) UDH, each concatenated segment holds(1120 - 48) / 16 = 67code units.
The function
const GSM7_BASIC = new Set(
"@£$¥èéùìòÇ\nØø\rÅåΔ_ΦΓΛΩΠΨΣΘΞÆæßÉ !\"#¤%&'()*+,-./0123456789:;<=>?¡" +
"ABCDEFGHIJKLMNOPQRSTUVWXYZÄÖÑܧ¿abcdefghijklmnopqrstuvwxyzäöñüà"
);
const GSM7_EXT = new Set("\f^{}\\[~]|€");
function countSmsSegments(text) {
if (!text) return { encoding: "GSM-7", units: 0, segments: 0, perSegment: 160 };
let isGsm7 = true;
let extCount = 0;
for (const ch of text) {
if (GSM7_EXT.has(ch)) extCount++;
else if (!GSM7_BASIC.has(ch)) {
isGsm7 = false;
break;
}
}
if (isGsm7) {
const units = text.length + extCount; // extension chars cost 2 septets (escape + char)
const segments = units <= 160 ? 1 : Math.ceil(units / 153);
return { encoding: "GSM-7", units, segments, perSegment: segments === 1 ? 160 : 153 };
}
// JavaScript strings are UTF-16, so text.length already counts UCS-2 code units
// (1 for BMP characters, 2 for astral plane emoji / surrogate pairs).
const units = text.length;
const segments = units <= 70 ? 1 : Math.ceil(units / 67);
return { encoding: "UCS-2", units, segments, perSegment: segments === 1 ? 70 : 67 };
}
Four messages under 160 characters, three different bills
Take four appointment reminders that all look safe to a naive text.length <= 160 check:
const msgA = "Your appointment with Dr. Patel is confirmed for Tuesday at 2:30 PM. Reply YES to keep this slot or call 555-0199 to reschedule. Ref: #48210";
const msgB = "Your appointment with Dr. Patel is confirmed for Tuesday at 2:30 PM. Reply “YES” to keep this slot or call 555-0199 to reschedule. Ref: #482";
const msgC = "Your appointment with Dr. Patel is confirmed for Tuesday at 2:30 PM. Reply [YES] to keep this slot or call 555-0199 {desk} [ext 4] ~urgent~ [floor 2] €50.";
const msgD = "Your appointment with Dr. Patel is confirmed for Tuesday at 2:30 PM. Reply YES to keep this slot or call 555-0199 to reschedule. See you! 👋";
for (const [label, m] of [
["A: Straight quotes", msgA],
["B: Curly quotes (“YES”)", msgB],
["C: GSM-7 brackets/euro", msgC],
["D: Trailing emoji (👋)", msgD],
]) {
console.log(label, "visibleChars:", [...m].length, countSmsSegments(m));
}
Running that produces:
| Message | Visible chars ([...m].length) |
Encoding | Billing units | Per-segment limit | Billed segments |
|---|---|---|---|---|---|
A: Straight quotes ("YES") |
140 | GSM-7 |
140 septets | 160 | 1 |
B: Smart quotes (“YES”) |
140 | UCS-2 |
140 code units | 67 | 3 |
C: Brackets & Euro ([]{}~€) |
154 | GSM-7 |
165 septets | 153 | 2 |
D: Trailing wave emoji (👋) |
139 | UCS-2 |
140 code units | 67 | 3 |
Why each one behaves the way it does:
-
Message B (
“YES”): Copy-pasting copy from Google Docs or Notion often introducesU+201CandU+201D. Neither is in GSM-7, so all 140 characters become UCS-2 (ceil(140 / 67) = 3segments). -
Message C (
[YES] ... {desk} ... €50): Every character is valid GSM-7, and the string is only 154 characters long. But the 11 extension characters ([,],{,},~,€) add 11 escape septets, pushing the payload to154 + 11 = 165septets—2 segments of 153 capacity. -
Message D (
👋): The message has 139 visible characters (138 ASCII characters + 1 emoji). Because👋(U+1F44B) is outside the Basic Multilingual Plane, it occupies 2 UTF-16 code units, making138 + 2 = 140UCS-2 units (ceil(140 / 67) = 3segments).
Quick assertions for your test suite
import assert from "node:assert/strict";
assert.deepEqual(countSmsSegments(msgA), { encoding: "GSM-7", units: 140, segments: 1, perSegment: 160 });
assert.deepEqual(countSmsSegments(msgB), { encoding: "UCS-2", units: 140, segments: 3, perSegment: 67 });
assert.deepEqual(countSmsSegments(msgC), { encoding: "GSM-7", units: 165, segments: 2, perSegment: 153 });
assert.deepEqual(countSmsSegments(msgD), { encoding: "UCS-2", units: 140, segments: 3, perSegment: 67 });
// 35 astral-plane emoji = 70 UCS-2 code units (fits in 1 segment)
assert.equal(countSmsSegments("👋".repeat(35)).segments, 1);
// 36 astral-plane emoji = 72 UCS-2 code units (spills into 2 segments of 67)
assert.equal(countSmsSegments("👋".repeat(36)).segments, 2);
Sanitizing smart punctuation before you send
If your templates don't intentionally need typographic punctuation, normalizing curly quotes and dashes before handing text to your SMS gateway keeps templates in GSM-7:
function normalizeForGsm7(text) {
return text
.replace(/[\u2018\u2019\u2032]/g, "'")
.replace(/[\u201C\u201D\u2033]/g, '"')
.replace(/[\u2013\u2014]/g, "-")
.replace(/\u2026/g, "...");
}
If you would rather not run code
We put the same detector into a free browser tool: paste a template, and it shows whether the string stays in GSM-7 or flips to UCS-2, highlights the exact character and Unicode code point (U+XXXX) that triggered UCS-2, counts 2-septet extension characters, and shows how many units remain in the current segment: SMS segment calculator.
The full breakdown of 3GPP TS 23.038 character tables, UDH headers, and carrier concatenation limits is here: SMS character limits, encoding, and segments explained. You can also browse the rest of our telecom and contact center calculators on the free tools page.
Disclosure: I work with the team at EaseDial, which publishes the calculator and guide linked above.
Top comments (0)