📖 如何準確測試藍牙耳機延遲？
📖 How to accurately test Bluetooth headset latency?
開始測試後，螢幕上的球體會週期性落地。您的任務是調整滑桿，讓「聽到聲音的瞬間」與「球體觸底視覺」精確同步。滑桿數值即為您的耳機在此裝置上的感知延遲偏移量。
After starting the test, the sphere on the screen will bounce periodically. Your task is to adjust the slider so that the "moment you hear the sound" is perfectly synchronized with the "visual impact of the sphere." The slider value is your perceived latency offset.
測試時請注意：將音量調至 60% 以上；在安靜環境進行；若使用藍牙耳機，請確認已切換至「低延遲模式」（各品牌名稱不同，常見如 Game Mode、LE Audio）。
Note during testing: Set volume above 60%; test in a quiet environment; if using a Bluetooth headset, ensure it is switched to "Low Latency Mode" (names vary, e.g., Game Mode, LE Audio).
📏 各連接方式的延遲參考範圍
📏 Latency Reference Ranges by Connection Type
- 有線（3.5mm / USB）Wired (3.5mm/USB)：< 10 ms，感知上近乎即時< 10 ms, perceived as almost instantaneous.
- 2.4GHz 無線接收器2.4GHz Dongle：20–40 ms，電競耳機主流選擇20–40 ms, the mainstream choice for esports headsets.
- 藍牙低延遲模式Bluetooth Low-Latency：40–80 ms，可接受40–80 ms, acceptable.
- 標準藍牙（SBC / AAC）Standard Bluetooth (SBC / AAC)：100–300 ms，玩即時遊戲會明顯感受到落差100–300 ms, noticeable lag in real-time gaming.
📐 方法與限制：結果取決於你的對位判斷，適合在同一副耳機上比較不同連線方式。完整說明見測量方法與限制。
📐 Method & limits: The result depends on your own alignment judgement; best for comparing connection types on the same headset. See Methods & Limits.
⚡ 自動測試的原理是什麼？
⚡ What is the principle of the Auto Test?
系統透過喇叭 / 耳機播放一個 1kHz 的短促嗶聲，同時啟動麥克風監聽。程式計算「聲音發出」到「麥克風收到」之間的往返時間（RTL, Round-Trip Latency）。這個數字是「系統音訊迴路延遲」的估算：它包含瀏覽器與作業系統的音訊處理、輸出裝置（有線／藍牙／2.4G 接收器）、聲音傳播與麥克風輸入的總和，並不等於耳機單向的輸出延遲，適合用來比較同一組設備在不同設定下的差異。
The system plays a short 1kHz beep through the headset while monitoring via the microphone. The program calculates the Round-Trip Latency (RTL) between "sound emitted" and "sound received." This is an estimate of the system audio loopback latency: it includes browser and OS audio processing, the output device (wired / Bluetooth / 2.4G dongle), sound travel and the microphone input, so it is not the one-way output latency of the headset. Use it to compare the same setup under different settings.
如何讓測試結果更準確？
How to make the test results more accurate?
- 將耳機麥克風（或外接麥克風）盡量貼近耳罩，縮短空氣傳遞距離Place the mic as close to the ear cup as possible to shorten air transmission distance.
- 系統音量調至 70% 以上，確保麥克風能清楚收到嗶聲Set system volume above 70% so the mic clearly captures the beep.
- 在安靜環境進行，避免環境噪音誤觸發偵測Perform in a quiet room to prevent background noise from triggering false detections.
- 若出現「Stereo Mix」，請改選實體麥克風，否則會量到極低的系統內部迴路延遲If "Stereo Mix" appears, select your physical mic instead. Otherwise, it measures the system's internal loopback latency.
測試失敗或數值異常怎麼辦？
What if the test fails or shows abnormal values?
若結果顯示 < 5 ms，通常代表偵測到了系統內部迴路而非耳機，請檢查輸入裝置選擇。若結果 > 400 ms，代表環境噪音干擾了偵測，請移至安靜處重試。
If the result is < 5 ms, the internal loopback was detected; check your input device. If the result is > 400 ms, environmental noise likely interfered; retry in a quiet place.
🏆 如何上傳到延遲排行榜？
🏆 How do I submit to the leaderboard?
自動檢測得到有效結果後，下方會出現上傳表單，填入耳機型號與連線方式即可上傳。只接受自動檢測的客觀量測值；手動對位屬於主觀感受，因此不開放上傳。上傳內容僅包含型號、連線方式、裝置類型與延遲數值。
After a valid auto-test result, an upload form appears below it — enter your headset model and connection type. Only objective auto-test measurements are accepted (manual sync is subjective). Only the model, connection, device type and latency are sent.
📐 方法與限制：這是系統音訊迴路延遲的估算，不等於耳機單向的輸出延遲，只適合比較同一套設備在不同設定下的差異。完整說明見測量方法與限制。
📐 Method & limits: This estimates system audio loopback latency, not the headset's one-way output latency; use it to compare one setup under different settings. See Methods & Limits.
🔊 左右聲道測試怎麼聽？
🔊 How to listen to the L/R Channel Test?
點擊「左聲道」按鈕，聲音應清楚出現在左耳；點擊「右聲道」應出現在右耳。若兩者相反，代表耳機左右插反或線材接錯。若其中一側完全無聲，代表驅動程式或硬體有問題。
Click "Left Ch" and the sound should appear in your left ear; click "Right Ch" for the right. If reversed, your headset is worn backward or wired incorrectly. If one side is dead silent, there's a hardware/driver issue.
正相與反相的差異？
Difference between In-Phase and Out-Phase?
「正相」播放時，兩耳聲音同步，結像應集中在頭部正中央。「反相」播放時，聲音會感覺「往外擴散」或「中空」。若您的耳機在正相時就有空洞感，代表左右聲道存在相位問題（常見於接線不良）。
In "In-Phase," both ears play synchronously, centering the audio image perfectly inside your head. In "Out-Phase," the sound feels "hollow" or "diffused." If your headset sounds hollow during In-Phase, it indicates a phase issue (common in poor wiring).
頻率掃描能告訴我什麼？
What does the Frequency Sweep tell me?
掃描從 20Hz 到 20kHz。留意您「聽不到聲音」的頻率段：多數成年人在 16kHz 以上即開始衰減。此外，若在某個中頻段聽到明顯共鳴或破音，代表耳機在該頻率有諧振問題。
The sweep runs from 20Hz to 20kHz. Note the frequency where you stop hearing sound; most adults start dropping off above 16kHz. If you hear obvious rattling or distortion in the midrange, your headset has resonance issues at that frequency.
📐 方法與限制：這些是訊號產生器，由你用耳朵判斷；煲機是否改變音質仍有爭議，請使用中低音量。完整說明見測量方法與限制。
📐 Method & limits: These are signal generators judged by ear; whether burn-in changes sound is disputed, so keep the volume moderate. See Methods & Limits.
🔥 煲機真的有效嗎？
🔥 Does Burn-in actually work?
煲機（Burn-in）是透過持續播音讓耳機振膜的懸邊材料逐漸軟化，使音質趨於穩定。對高分子材質懸邊的動圈耳機效果較明顯；對靜電或動鐵驅動器則幾乎無影響。
Burn-in is the process of playing continuous audio to soften the diaphragm's surround material, stabilizing the sound quality. This is most noticeable in dynamic drivers; it has almost no effect on electrostatic or balanced armature drivers.
為什麼要用粉紅噪音而不是音樂？
Why use Pink Noise instead of music?
粉紅噪音的能量在各頻段均勻分布，能對振膜進行全頻段的均勻鍛煉。白噪音則讓高頻段獲得更多能量。智能循環模式交替使用兩者，並加入全頻掃描與休息期，是最全面的煲機方案。
Pink Noise has equal energy across all octaves, providing a uniform workout across the entire frequency spectrum. White noise pushes more energy into high frequencies. Our Smart Cycle alternates between both and adds rest periods, making it the most comprehensive solution.
煲機需要多久？建議音量多大？
How long does it take? What volume is recommended?
建議 50–200 小時，以中等音量（不超過日常聆聽音量的 80%）進行。過大音量可能造成振膜永久性損傷。智能循環模式的 60 分鐘可視為「入門快速版」。
Generally 50–200 hours at a moderate volume (no more than 80% of normal volume). Excessive volume can cause permanent damage. The 60-minute Smart Cycle is a "quick break-in."
📐 方法與限制：這些是訊號產生器，由你用耳朵判斷；煲機是否改變音質仍有爭議，請使用中低音量。完整說明見測量方法與限制。
📐 Method & limits: These are signal generators judged by ear; whether burn-in changes sound is disputed, so keep the volume moderate. See Methods & Limits.
🎤 頻譜分析器能用來做什麼？
🎤 What can the Spectrum Analyzer be used for?
即時頻譜分析將麥克風收到的聲音視覺化。實用場景包括：搭配頻率掃描尋找耳機的弱頻段、找出房間內的噪音頻率（如電腦風扇多在 200–500 Hz），或錄音前檢查是否有電流雜訊。
It visualizes the sound captured by your mic. Use cases include: finding weak frequency bands of a speaker, identifying environmental noise frequencies (e.g., PC fans at 200–500 Hz), or checking for electrical hum before recording.
📈 頻率響應量測怎麼用？
How to use the FR measurement?
選擇「左聲道」或「右聲道」後開始，系統播放 15 秒掃頻並畫出響應曲線；「L-R 對比」會連續量測兩邊並顯示最大差異 (dB)。請保持麥克風位置與音量不變再切換對象比較。曲線在中頻 (200Hz–2kHz) 的起伏最能反映單體特性。
Pick Left or Right and start: a 15-second sweep is played and the response curve is plotted. "L-R Compare" measures both sides back-to-back and shows the max difference (dB). Keep mic position and volume unchanged between comparisons. Midrange (200Hz–2kHz) bumps best reveal driver character.
db 數值代表什麼？
What do the db values mean?
畫面上方的 Min / Avg / Max 是麥克風收到的相對音量估算（以 0–100 刻度呈現，非絕對 dBSPL）。環境安靜時應接近 0–10；正常說話約 40–60。所有數據均在本機運算，不會上傳。
The Min/Avg/Max are relative volume estimations of the mic input (on a 0–100 scale, NOT absolute dBSPL). In a quiet room, it should be 0–10; normal talking is 40–60. All data is processed locally and never uploaded.
🧠 AI 聲音辨識是怎麼運作的？會上傳我的聲音嗎？
🧠 How does AI sound recognition work? Is my audio uploaded?
我們使用 Google 開源的 YAMNet 聲音分類模型（可辨識 521 種聲音），透過 MediaPipe 在您的瀏覽器中直接執行。首次使用時瀏覽器會下載約 10MB 的模型與執行環境，之後會被快取。麥克風的聲音只在您的電腦上分析，不會錄音、儲存或上傳。
We use Google's open-source YAMNet sound classifier (521 sound classes), running directly in your browser via MediaPipe. On first use the browser downloads about 10 MB of model and runtime files, which are then cached. Your mic audio is analyzed only on your device — never recorded, stored or uploaded.
🎙️ 麥克風評分的各項數值代表什麼？
🎙️ What do the mic score metrics mean?
- 訊噪比 (SNR)：說話聲比背景噪音大多少，30 dB 以上很好，低於 20 dB 隊友會聽到明顯雜音。SNR: how much louder your voice is than the background. 30 dB+ is great; below 20 dB teammates will hear obvious noise.
- 說話音量：理想約在 -30 ~ -12 dBFS；太小聽不清，太大容易爆音。Voice level: ideally around -30 to -12 dBFS; too low is hard to hear, too high clips.
- 有效頻寬：約 4kHz 代表電話音質，常見於藍牙耳機切到通話模式 (HFP)；8kHz 為一般通話等級；12kHz 以上聲音最清晰自然。Bandwidth: ~4 kHz is phone quality, typical when a Bluetooth headset switches to its call profile (HFP); 8 kHz is voice-call grade; 12 kHz+ sounds clear and natural.
- 電流聲：50/60Hz 及其倍頻的嗡嗡聲，多半來自接地迴路或 USB 供電干擾。Mains hum: a 50/60 Hz buzz (and harmonics), usually from a ground loop or USB power noise.
測試時已關閉瀏覽器內建的降噪與自動增益，量到的是麥克風的「原始」收音；在 Discord 等軟體中開啟降噪後，實際效果通常會更好。
Browser noise suppression and auto gain are disabled during the test, so you see the mic's raw pickup; with noise suppression enabled in apps like Discord, real-world results are usually better.
📐 方法與限制：頻譜與評分受距離、房間噪音與系統增益影響，是參考而不是認證。完整說明見測量方法與限制。
📐 Method & limits: The spectrum and score depend on distance, room noise and system gain; they are a reference, not a certification. See Methods & Limits.