logdetector logic semiapplied

RFG_ADC_dataplotter.py

@@ -33,18 +33,18 @@ import numpy as np
 
 WF_WIDTH = 1000  # максимальное число точек в ряду водопада
 FFT_LEN = 1024   # длина БПФ для спектра/водопада спектров
-LOG_BASE = 10.0
-LOG_SCALER = 0.001  # int32 значения приходят в fixed-point лог-шкале с шагом 1e-3
-LOG_POSTSCALER = 1000
-LOG_EXP_LIMIT = 300.0  # запас до переполнения float64 при возведении LOG_BASE в степень
 # Порог для инверсии сырых данных: если среднее значение свипа ниже порога —
 # считаем, что сигнал «меньше нуля» и домножаем свип на -1
 DATA_INVERSION_THRASHOLD = 10.0
+LOG_DETECTOR_OFFSET = 0.0
+LOG_DETECTOR_SCALER = -0.001
+LOG_DETECTOR_BASE = 2.0
+LOG_DETECTOR_EXP_MIN = -149.0
+LOG_DETECTOR_EXP_MAX = 128.0
 
 Number = Union[int, float]
 SweepInfo = Dict[str, Any]
-SweepAuxCurves = Optional[Tuple[np.ndarray, np.ndarray]]
-SweepPacket = Tuple[np.ndarray, SweepInfo, SweepAuxCurves]
+SweepPacket = Tuple[np.ndarray, SweepInfo]
 
 
 def _format_status_kv(data: Mapping[str, Any]) -> str:
@@ -64,7 +64,8 @@ def _format_status_kv(data: Mapping[str, Any]) -> str:
             return f"{fv:.3g}"
         return f"{fv:.3f}".rstrip("0").rstrip(".")
 
-    parts = [f"{k}:{_fmt(v)}" for k, v in data.items()]
+    hidden_keys = {"pre_exp_sweep", "sweep_1", "sweep_2"}
+    parts = [f"{k}:{_fmt(v)}" for k, v in data.items() if k not in hidden_keys]
     return "  ".join(parts)
 
 
@@ -90,44 +91,6 @@ def _parse_spec_clip(spec: Optional[str]) -> Optional[Tuple[float, float]]:
         return None
 
 
-def _log_value_to_linear(value: int) -> float:
-    """Преобразовать fixed-point логарифмическое значение в линейную шкалу."""
-    exponent = max(-LOG_EXP_LIMIT, min(LOG_EXP_LIMIT, float(value) * LOG_SCALER))
-    return float(LOG_BASE ** exponent)
-
-
-def _log_pair_to_sweep(avg_1: int, avg_2: int) -> float:
-    """Разность двух логарифмических усреднений в линейной шкале."""
-    return (_log_value_to_linear(avg_1) - _log_value_to_linear(avg_2))*LOG_POSTSCALER
-
-
-def _compute_auto_ylim(*series_list: Optional[np.ndarray]) -> Optional[Tuple[float, float]]:
-    """Общий Y-диапазон по всем переданным кривым с небольшим запасом."""
-    y_min: Optional[float] = None
-    y_max: Optional[float] = None
-    for series in series_list:
-        if series is None:
-            continue
-        arr = np.asarray(series)
-        if arr.size == 0:
-            continue
-        finite = arr[np.isfinite(arr)]
-        if finite.size == 0:
-            continue
-        cur_min = float(np.min(finite))
-        cur_max = float(np.max(finite))
-        y_min = cur_min if y_min is None else min(y_min, cur_min)
-        y_max = cur_max if y_max is None else max(y_max, cur_max)
-
-    if y_min is None or y_max is None:
-        return None
-    if y_min == y_max:
-        pad = max(1.0, abs(y_min) * 0.05)
-    else:
-        pad = 0.05 * (y_max - y_min)
-    return (y_min - pad, y_max + pad)
-
-
 def _normalize_sweep_simple(raw: np.ndarray, calib: np.ndarray) -> np.ndarray:
     """Простая нормировка: поэлементное деление raw/calib."""
     w = min(raw.size, calib.size)
@@ -428,7 +391,7 @@ class SweepReader(threading.Thread):
         stop_event: threading.Event,
         fancy: bool = False,
         bin_mode: bool = False,
-        logscale: bool = False,
+        logdetector: bool = False,
     ):
         super().__init__(daemon=True)
         self._port_path = port_path
@@ -438,7 +401,7 @@ class SweepReader(threading.Thread):
         self._src: Optional[SerialLineSource] = None
         self._fancy = bool(fancy)
         self._bin_mode = bool(bin_mode)
-        self._logscale = bool(logscale)
+        self._logdetector = bool(logdetector)
         self._max_width: int = 0
         self._sweep_idx: int = 0
         self._last_sweep_ts: Optional[float] = None
@@ -454,8 +417,8 @@ class SweepReader(threading.Thread):
         xs,
         ys,
         channels: Optional[set[int]],
-        raw_curves: Optional[Tuple[list[int], list[int]]] = None,
-        apply_inversion: bool = True,
+        ys1: Optional[list[int]] = None,
+        ys2: Optional[list[int]] = None,
     ):
         if not xs:
             return
@@ -465,43 +428,26 @@ class SweepReader(threading.Thread):
         width = max_x + 1
         self._max_width = max(self._max_width, width)
         target_width = self._max_width if self._fancy else width
-
-        def _scatter(values, dtype) -> np.ndarray:
-            series = np.full((target_width,), np.nan, dtype=dtype)
+        def _build_sweep(values) -> np.ndarray:
+            arr = np.full((target_width,), np.nan, dtype=np.float32)
             try:
                 idx = np.asarray(xs, dtype=np.int64)
-                vals = np.asarray(values, dtype=dtype)
-                series[idx] = vals
+                vals = np.asarray(values, dtype=np.float32)
+                arr[idx] = vals
             except Exception:
                 for x, y in zip(xs, values):
                     if 0 <= x < target_width:
-                        series[x] = y
-            return series
+                        arr[x] = float(y)
+            return arr
 
-        def _fill_missing(series: np.ndarray):
-            known = ~np.isnan(series)
-            if not np.any(known):
-                return
-            known_idx = np.nonzero(known)[0]
-            for i0, i1 in zip(known_idx[:-1], known_idx[1:]):
-                if i1 - i0 > 1:
-                    avg = (series[i0] + series[i1]) * 0.5
-                    series[i0 + 1 : i1] = avg
-            first_idx = int(known_idx[0])
-            last_idx = int(known_idx[-1])
-            if first_idx > 0:
-                series[:first_idx] = series[first_idx]
-            if last_idx < series.size - 1:
-                series[last_idx + 1 :] = series[last_idx]
-
-        # Быстрый векторизованный путь
-        sweep = _scatter(ys, np.float32)
-        aux_curves: SweepAuxCurves = None
-        if raw_curves is not None:
-            aux_curves = (
-                _scatter(raw_curves[0], np.float32),
-                _scatter(raw_curves[1], np.float32),
-            )
+        sweep_1: Optional[np.ndarray] = None
+        sweep_2: Optional[np.ndarray] = None
+        if ys1 is not None and ys2 is not None and len(ys1) == len(xs) and len(ys2) == len(xs):
+            sweep_1 = _build_sweep(ys1)
+            sweep_2 = _build_sweep(ys2)
+            sweep = sweep_1 - sweep_2
+        else:
+            sweep = _build_sweep(ys)
         # Метрики валидных точек до заполнения пропусков
         finite_pre = np.isfinite(sweep)
         n_valid_cur = int(np.count_nonzero(finite_pre))
@@ -509,22 +455,62 @@ class SweepReader(threading.Thread):
         # Дополнительная обработка пропусков: при --fancy заполняем внутренние разрывы, края и дотягиваем до максимальной длины
         if self._fancy:
             try:
-                _fill_missing(sweep)
-                if aux_curves is not None:
-                    _fill_missing(aux_curves[0])
-                    _fill_missing(aux_curves[1])
+                known = ~np.isnan(sweep)
+                if np.any(known):
+                    known_idx = np.nonzero(known)[0]
+                    # Для каждой пары соседних известных индексов заполним промежуток средним значением
+                    for i0, i1 in zip(known_idx[:-1], known_idx[1:]):
+                        if i1 - i0 > 1:
+                            avg = (sweep[i0] + sweep[i1]) * 0.5
+                            sweep[i0 + 1 : i1] = avg
+                    first_idx = int(known_idx[0])
+                    last_idx = int(known_idx[-1])
+                    if first_idx > 0:
+                        sweep[:first_idx] = sweep[first_idx]
+                    if last_idx < sweep.size - 1:
+                        sweep[last_idx + 1 :] = sweep[last_idx]
             except Exception:
                 # В случае ошибки просто оставляем как есть
                 pass
+        '''
         # Инверсия данных при «отрицательном» уровне (среднее ниже порога)
-        if apply_inversion:
+
+
+        try:
+            m = float(np.nanmean(sweep))
+            if np.isfinite(m) and m < DATA_INVERSION_THRASHOLD:
+                sweep *= -1.0
+        except Exception:
+            pass
+        '''
+        
+        pre_exp_sweep: Optional[np.ndarray] = None
+        if self._logdetector:
             try:
-                m = float(np.nanmean(sweep))
-                if np.isfinite(m) and m < DATA_INVERSION_THRASHOLD:
-                    sweep *= -1.0
+                if sweep_1 is not None and sweep_2 is not None:
+                    s1_pre = (sweep_1 - LOG_DETECTOR_OFFSET) * LOG_DETECTOR_SCALER
+                    s2_pre = (sweep_2 - LOG_DETECTOR_OFFSET) * LOG_DETECTOR_SCALER
+                    s1_pre = np.clip(s1_pre, LOG_DETECTOR_EXP_MIN, LOG_DETECTOR_EXP_MAX)
+                    s2_pre = np.clip(s2_pre, LOG_DETECTOR_EXP_MIN, LOG_DETECTOR_EXP_MAX)
+#                    with np.errstate(over="ignore", invalid="ignore"):
+#                        sweep_1 = np.power(LOG_DETECTOR_BASE, np.asarray(s1_pre, dtype=np.float64)).astype(np.float32)
+#                        sweep_2 = np.power(LOG_DETECTOR_BASE, np.asarray(s2_pre, dtype=np.float64)).astype(np.float32)
+                    sweep_1 = np.power(LOG_DETECTOR_BASE, np.asarray(s1_pre, dtype=np.float64)).astype(np.float32)
+                    sweep_2 = np.power(LOG_DETECTOR_BASE, np.asarray(s2_pre, dtype=np.float64)).astype(np.float32)
+                    sweep_1[~np.isfinite(sweep_1)] = np.nan
+                    sweep_2[~np.isfinite(sweep_2)] = np.nan
+                    sweep = sweep_1 - sweep_2
+                else:
+                    sweep = (sweep - LOG_DETECTOR_OFFSET) * LOG_DETECTOR_SCALER
+                    sweep = np.clip(sweep, LOG_DETECTOR_EXP_MIN, LOG_DETECTOR_EXP_MAX)
+                    pre_exp_sweep = sweep.copy()
+                    with np.errstate(over="ignore", invalid="ignore"):
+                        sweep = np.power(LOG_DETECTOR_BASE, np.asarray(sweep, dtype=np.float64)).astype(np.float32)
+                    sweep[~np.isfinite(sweep)] = np.nan
             except Exception:
                 pass
-        #sweep = np.abs(sweep)
+
+        #print(sweep)
         #sweep -= float(np.nanmean(sweep))
 
         # Метрики для статусной строки (вид словаря: переменная -> значение)
@@ -565,17 +551,22 @@ class SweepReader(threading.Thread):
             "std": std,
             "dt_ms": dt_ms,
         }
+        if pre_exp_sweep is not None:
+            info["pre_exp_sweep"] = pre_exp_sweep
+        if sweep_1 is not None and sweep_2 is not None:
+            info["sweep_1"] = sweep_1
+            info["sweep_2"] = sweep_2
 
         # Кладём готовый свип (если очередь полна — выбрасываем самый старый)
         try:
-            self._q.put_nowait((sweep, info, aux_curves))
+            self._q.put_nowait((sweep, info))
         except Full:
             try:
                 _ = self._q.get_nowait()
             except Exception:
                 pass
             try:
-                self._q.put_nowait((sweep, info, aux_curves))
+                self._q.put_nowait((sweep, info))
             except Exception:
                 pass
 
@@ -584,6 +575,7 @@ class SweepReader(threading.Thread):
         ys: list[int] = []
         cur_channel: Optional[int] = None
         cur_channels: set[int] = set()
+
         buf = bytearray()
         while not self._stop.is_set():
             data = chunk_reader.read_available()
@@ -612,7 +604,6 @@ class SweepReader(threading.Thread):
                     cur_channels.clear()
                     continue
 
-                # sCH X Y  или  s CH X Y  (все целые со знаком). Разделяем по любым пробелам/табам.
                 if len(line) >= 3:
                     parts = line.split()
                     if len(parts) >= 3 and (parts[0].lower() == b"s" or parts[0].lower().startswith(b"s")):
@@ -621,16 +612,15 @@ class SweepReader(threading.Thread):
                                 if len(parts) >= 4:
                                     ch = int(parts[1], 10)
                                     x = int(parts[2], 10)
-                                    y = int(parts[3], 10)  # поддержка знака: "+…" и "-…"
+                                    y = int(parts[3], 10)
                                 else:
                                     ch = 0
                                     x = int(parts[1], 10)
-                                    y = int(parts[2], 10)  # поддержка знака: "+…" и "-…"
+                                    y = int(parts[2], 10)
                             else:
-                                # формат вида "s0"
                                 ch = int(parts[0][1:], 10)
                                 x = int(parts[1], 10)
-                                y = int(parts[2], 10)  # поддержка знака: "+…" и "-…"
+                                y = int(parts[2], 10)
                         except Exception:
                             continue
                         if cur_channel is None:
@@ -647,6 +637,8 @@ class SweepReader(threading.Thread):
     def _run_binary_stream(self, chunk_reader: SerialChunkReader):
         xs: list[int] = []
         ys: list[int] = []
+        ys1: list[int] = []
+        ys2: list[int] = []
         cur_channel: Optional[int] = None
         cur_channels: set[int] = set()
         words = deque()
@@ -670,88 +662,9 @@ class SweepReader(threading.Thread):
                 words.append(w)
                 i += 2
 
-            # Бинарный протокол:
-            #   старт свипа (актуальный): 0xFFFF, 0xFFFF, 0xFFFF, (ch<<8)|0x0A
-            #   старт свипа (legacy):     0xFFFF, 0xFFFF, channel, 0x0A0A
-            #   точка: step, value_hi, value_lo, 0x000A
-            while len(words) >= 4:
-                w0 = int(words[0])
-                w1 = int(words[1])
-                w2 = int(words[2])
-                w3 = int(words[3])
-
-                if w0 == 0xFFFF and w1 == 0xFFFF and w2 == 0xFFFF and (w3 & 0x00FF) == 0x000A:
-                    self._finalize_current(xs, ys, cur_channels)
-                    xs.clear()
-                    ys.clear()
-                    cur_channels.clear()
-                    cur_channel = (w3 >> 8) & 0x00FF
-                    cur_channels.add(cur_channel)
-                    for _ in range(4):
-                        words.popleft()
-                    continue
-
-                if w0 == 0xFFFF and w1 == 0xFFFF and w3 == 0x0A0A:
-                    self._finalize_current(xs, ys, cur_channels)
-                    xs.clear()
-                    ys.clear()
-                    cur_channels.clear()
-                    cur_channel = w2
-                    cur_channels.add(cur_channel)
-                    for _ in range(4):
-                        words.popleft()
-                    continue
-
-                if w3 == 0x000A:
-                    if cur_channel is not None:
-                        cur_channels.add(cur_channel)
-                    xs.append(w0)
-                    value_u32 = (w1 << 16) | w2
-                    ys.append(self._u32_to_i32(value_u32))
-                    for _ in range(4):
-                        words.popleft()
-                    continue
-
-                # Поток может начаться с середины пакета; сдвигаемся по слову до ресинхронизации.
-                words.popleft()
-
-            del buf[:usable]
-            if len(buf) > 1_000_000:
-                del buf[:-262144]
-
-        self._finalize_current(xs, ys, cur_channels)
-
-    def _run_logscale_binary_stream(self, chunk_reader: SerialChunkReader):
-        xs: list[int] = []
-        ys: list[float] = []
-        avg_1_vals: list[int] = []
-        avg_2_vals: list[int] = []
-        cur_channel: Optional[int] = None
-        cur_channels: set[int] = set()
-        words = deque()
-
-        buf = bytearray()
-        while not self._stop.is_set():
-            data = chunk_reader.read_available()
-            if data:
-                buf += data
-            else:
-                time.sleep(0.0005)
-                continue
-
-            usable = len(buf) & ~1
-            if usable == 0:
-                continue
-
-            i = 0
-            while i < usable:
-                w = int(buf[i]) | (int(buf[i + 1]) << 8)
-                words.append(w)
-                i += 2
-
-            # Бинарный logscale-протокол:
-            #   старт свипа: 0xFFFF x5, затем (ch<<8)|0x0A
-            #   точка: step, avg1_hi, avg1_lo, avg2_hi, avg2_lo, 0x000A
+            # Новый бинарный формат:
+            # - старт: FFFF,FFFF,FFFF,FFFF,FFFF,(CH<<8)|0x0A
+            # - точка: X,avg1_hi,avg1_lo,avg2_hi,avg2_lo,0x000A
             while len(words) >= 6:
                 w0 = int(words[0])
                 w1 = int(words[1])
@@ -761,24 +674,14 @@ class SweepReader(threading.Thread):
                 w5 = int(words[5])
 
                 if (
-                    w0 == 0xFFFF
-                    and w1 == 0xFFFF
-                    and w2 == 0xFFFF
-                    and w3 == 0xFFFF
-                    and w4 == 0xFFFF
-                    and (w5 & 0x00FF) == 0x000A
+                    w0 == 0xFFFF and w1 == 0xFFFF and w2 == 0xFFFF
+                    and w3 == 0xFFFF and w4 == 0xFFFF and (w5 & 0x00FF) == 0x000A
                 ):
-                    self._finalize_current(
-                        xs,
-                        ys,
-                        cur_channels,
-                        raw_curves=(avg_1_vals, avg_2_vals),
-                        apply_inversion=False,
-                    )
+                    self._finalize_current(xs, ys, cur_channels, ys1=ys1, ys2=ys2)
                     xs.clear()
                     ys.clear()
-                    avg_1_vals.clear()
-                    avg_2_vals.clear()
+                    ys1.clear()
+                    ys2.clear()
                     cur_channels.clear()
                     cur_channel = (w5 >> 8) & 0x00FF
                     cur_channels.add(cur_channel)
@@ -789,13 +692,14 @@ class SweepReader(threading.Thread):
                 if w5 == 0x000A:
                     if cur_channel is not None:
                         cur_channels.add(cur_channel)
-                    avg_1 = self._u32_to_i32((w1 << 16) | w2)
-                    avg_2 = self._u32_to_i32((w3 << 16) | w4)
                     xs.append(w0)
-                    avg_1_vals.append(avg_1)
-                    avg_2_vals.append(avg_2)
-                    ys.append(_log_pair_to_sweep(avg_1, avg_2))
-                    #ys.append(LOG_BASE**(avg_1/LOG_SCALER) - LOG_BASE**(avg_2/LOG_SCALER))
+                    avg1_u32 = (w1 << 16) | w2
+                    avg2_u32 = (w3 << 16) | w4
+                    avg1 = self._u32_to_i32(avg1_u32)
+                    avg2 = self._u32_to_i32(avg2_u32)
+                    ys1.append(avg1)
+                    ys2.append(avg2)
+                    ys.append(avg1 - avg2)
                     for _ in range(6):
                         words.popleft()
                     continue
@@ -806,13 +710,7 @@ class SweepReader(threading.Thread):
             if len(buf) > 1_000_000:
                 del buf[:-262144]
 
-        self._finalize_current(
-            xs,
-            ys,
-            cur_channels,
-            raw_curves=(avg_1_vals, avg_2_vals),
-            apply_inversion=False,
-        )
+        self._finalize_current(xs, ys, cur_channels, ys1=ys1, ys2=ys2)
 
     def run(self):
         try:
@@ -824,9 +722,7 @@ class SweepReader(threading.Thread):
 
         try:
             chunk_reader = SerialChunkReader(self._src)
-            if self._logscale:
-                self._run_logscale_binary_stream(chunk_reader)
-            elif self._bin_mode:
+            if self._bin_mode:
                 self._run_binary_stream(chunk_reader)
             else:
                 self._run_ascii_stream(chunk_reader)
@@ -857,7 +753,7 @@ def main():
         "--spec-clip",
         default="2,98",
         help=(
-            "Процентильная обрезка уровней водопада спектров, % (min,max). "
+            "Процентильная обрезка уровней водопада спектров, %% (min,max). "
             "Напр. 2,98. 'off' — отключить"
         ),
     )
@@ -898,19 +794,16 @@ def main():
         "--bin",
         dest="bin_mode",
         action="store_true",
+        default=True,
         help=(
-            "Бинарный протокол: старт свипа 0xFFFF,0xFFFF,0xFFFF,(CH<<8)|0x0A; "
-            "точки step,uint32(hi16,lo16),0x000A"
+            "Бинарный протокол: старт FFFFx5,(CH<<8)|0x0A; "
+            "точки X,avg1_hi,avg1_lo,avg2_hi,avg2_lo,0x000A (sweep=avg1-avg2)"
         ),
     )
     parser.add_argument(
-        "--logscale",
+        "--logdetector",
         action="store_true",
-        default=True,
-        help=(
-            "Новый бинарный протокол: точка несёт пару int32 (avg_1, avg_2), "
-            "а свип считается как 10**(avg_1*0.001) - 10**(avg_2*0.001)"
-        ),
+        help="Лог-детектор: после инверсии ((sweep-OFFSET)*SCALER) и затем BASE**sweep",
     )
 
     args = parser.parse_args()
@@ -943,8 +836,8 @@ def main():
         q,
         stop_event,
         fancy=bool(args.fancy),
-        bin_mode=bool(args.bin_mode),
-        logscale=bool(args.logscale),
+        bin_mode=bool(getattr(args, "bin_mode", False)),
+        logdetector=bool(getattr(args, "logdetector", False)),
     )
     reader.start()
 
@@ -957,7 +850,9 @@ def main():
 
     # Состояние для отображения
     current_sweep_raw: Optional[np.ndarray] = None
-    current_aux_curves: SweepAuxCurves = None
+    current_sweep_1: Optional[np.ndarray] = None
+    current_sweep_2: Optional[np.ndarray] = None
+    current_sweep_pre_exp: Optional[np.ndarray] = None
     current_sweep_norm: Optional[np.ndarray] = None
     last_calib_sweep: Optional[np.ndarray] = None
     current_info: Optional[SweepInfo] = None
@@ -982,6 +877,7 @@ def main():
     contrast_slider = None
     calib_enabled = False
     norm_type = str(getattr(args, "norm_type", "projector")).strip().lower()
+    logdetector_enabled = bool(getattr(args, "logdetector", False))
     cb = None
 
     # Статусная строка (внизу окна)
@@ -996,10 +892,8 @@ def main():
     )
 
     # Линейный график последнего свипа
-    line_avg1_obj, = ax_line.plot([], [], lw=1, color="0.65")
-    line_avg2_obj, = ax_line.plot([], [], lw=1, color="0.45")
     line_obj, = ax_line.plot([], [], lw=1, color="tab:blue")
-    line_calib_obj, = ax_line.plot([], [], lw=1, color="tab:red")
+    line_calib_obj, = ax_line.plot([], [], lw=1, color="gold")
     line_norm_obj, = ax_line.plot([], [], lw=1, color="tab:green")
     ax_line.set_title("Сырые данные", pad=1)
     ax_line.set_xlabel("ГГц")
@@ -1138,7 +1032,7 @@ def main():
         freq_shared = np.arange(fft_bins, dtype=np.int32)
 
     def _visible_levels_matplotlib(data: np.ndarray, axis) -> Optional[Tuple[float, float]]:
-        """(vmin, vmax) по центральным 90% значений в видимой области imshow."""
+        """(vmin, vmax) по текущей видимой области imshow (без накопления по времени)."""
         if data.size == 0:
             return None
         ny, nx = data.shape[0], data.shape[1]
@@ -1163,8 +1057,8 @@ def main():
         if not finite.any():
             return None
         vals = sub[finite]
-        vmin = float(np.nanpercentile(vals, 5))
-        vmax = float(np.nanpercentile(vals, 95))
+        vmin = float(np.min(vals))
+        vmax = float(np.max(vals))
         if not (np.isfinite(vmin) and np.isfinite(vmax)) or vmin == vmax:
             return None
         return (vmin, vmax)
@@ -1217,17 +1111,22 @@ def main():
                 y_max_fft = float(fr_max)
 
     def drain_queue():
-        nonlocal current_sweep_raw, current_aux_curves, current_sweep_norm, current_info, last_calib_sweep
+        nonlocal current_sweep_raw, current_sweep_1, current_sweep_2, current_sweep_pre_exp, current_sweep_norm, current_info, last_calib_sweep
         drained = 0
         while True:
             try:
-                s, info, aux_curves = q.get_nowait()
+                s, info = q.get_nowait()
             except Empty:
                 break
             drained += 1
             current_sweep_raw = s
-            current_aux_curves = aux_curves
             current_info = info
+            s1 = info.get("sweep_1") if isinstance(info, dict) else None
+            s2 = info.get("sweep_2") if isinstance(info, dict) else None
+            current_sweep_1 = s1 if isinstance(s1, np.ndarray) else None
+            current_sweep_2 = s2 if isinstance(s2, np.ndarray) else None
+            pre = info.get("pre_exp_sweep") if isinstance(info, dict) else None
+            current_sweep_pre_exp = pre if isinstance(pre, np.ndarray) else None
             ch = 0
             try:
                 ch = int(info.get("ch", 0)) if isinstance(info, dict) else 0
@@ -1296,31 +1195,46 @@ def main():
             else:
                 xs = np.arange(current_sweep_raw.size, dtype=np.int32)
             line_obj.set_data(xs, current_sweep_raw)
-            if current_aux_curves is not None:
-                avg_1_curve, avg_2_curve = current_aux_curves
-                line_avg1_obj.set_data(xs[: avg_1_curve.size], avg_1_curve)
-                line_avg2_obj.set_data(xs[: avg_2_curve.size], avg_2_curve)
-            else:
-                line_avg1_obj.set_data([], [])
-                line_avg2_obj.set_data([], [])
-            if last_calib_sweep is not None:
-                line_calib_obj.set_data(xs[: last_calib_sweep.size], last_calib_sweep)
-            else:
+            if current_sweep_1 is not None and current_sweep_2 is not None:
+                line_calib_obj.set_data(xs[: current_sweep_1.size], current_sweep_1)
+                line_norm_obj.set_data(xs[: current_sweep_2.size], current_sweep_2)
+            elif logdetector_enabled:
                 line_calib_obj.set_data([], [])
-            if current_sweep_norm is not None:
-                line_norm_obj.set_data(xs[: current_sweep_norm.size], current_sweep_norm)
+                if current_sweep_pre_exp is not None:
+                    line_norm_obj.set_data(xs[: current_sweep_pre_exp.size], current_sweep_pre_exp)
+                else:
+                    line_norm_obj.set_data([], [])
             else:
-                line_norm_obj.set_data([], [])
+                if last_calib_sweep is not None:
+                    line_calib_obj.set_data(xs[: last_calib_sweep.size], last_calib_sweep)
+                else:
+                    line_calib_obj.set_data([], [])
+                if current_sweep_norm is not None:
+                    line_norm_obj.set_data(xs[: current_sweep_norm.size], current_sweep_norm)
+                else:
+                    line_norm_obj.set_data([], [])
             # Лимиты по X: 3.3 ГГц .. 14.3 ГГц
             ax_line.set_xlim(3.3, 14.3)
             # Адаптивные Y-лимиты (если не задан --ylim)
             if fixed_ylim is None:
-                y_series = [current_sweep_raw, last_calib_sweep, current_sweep_norm]
-                if current_aux_curves is not None:
-                    y_series.extend(current_aux_curves)
-                y_limits = _compute_auto_ylim(*y_series)
-                if y_limits is not None:
-                    ax_line.set_ylim(y_limits[0], y_limits[1])
+                y_candidates = [current_sweep_raw]
+                if current_sweep_1 is not None and current_sweep_2 is not None:
+                    y_candidates.extend([current_sweep_1, current_sweep_2])
+                elif logdetector_enabled and current_sweep_pre_exp is not None:
+                    y_candidates.append(current_sweep_pre_exp)
+                y_concat = np.concatenate([np.asarray(v, dtype=np.float32) for v in y_candidates])
+                y0 = float(np.nanmin(y_concat))
+                y1 = float(np.nanmax(y_concat))
+                if np.isfinite(y0) and np.isfinite(y1):
+                    if y0 == y1:
+                        pad = max(1.0, abs(y0) * 0.05)
+                        y0 -= pad
+                        y1 += pad
+                    else:
+                        pad = 0.05 * (y1 - y0)
+                        y0 -= pad
+                        y1 += pad
+                    ax_line.set_ylim(y0, y1)
 
             # Обновление спектра текущего свипа
             sweep_for_fft = current_sweep_norm if current_sweep_norm is not None else current_sweep_raw
@@ -1410,8 +1324,6 @@ def main():
         # Возвращаем обновлённые артисты
         return (
             line_obj,
-            line_avg1_obj,
-            line_avg2_obj,
             line_calib_obj,
             line_norm_obj,
             img_obj,
@@ -1453,8 +1365,8 @@ def run_pyqtgraph(args):
         q,
         stop_event,
         fancy=bool(args.fancy),
-        bin_mode=bool(args.bin_mode),
-        logscale=bool(args.logscale),
+        bin_mode=bool(getattr(args, "bin_mode", False)),
+        logdetector=bool(getattr(args, "logdetector", False)),
     )
     reader.start()
 
@@ -1472,10 +1384,8 @@ def run_pyqtgraph(args):
     # Плот последнего свипа (слева-сверху)
     p_line = win.addPlot(row=0, col=0, title="Сырые данные")
     p_line.showGrid(x=True, y=True, alpha=0.3)
-    curve_avg1 = p_line.plot(pen=pg.mkPen((170, 170, 170), width=1))
-    curve_avg2 = p_line.plot(pen=pg.mkPen((110, 110, 110), width=1))
     curve = p_line.plot(pen=pg.mkPen((80, 120, 255), width=1))
-    curve_calib = p_line.plot(pen=pg.mkPen((220, 60, 60), width=1))
+    curve_calib = p_line.plot(pen=pg.mkPen((220, 200, 60), width=1))
     curve_norm = p_line.plot(pen=pg.mkPen((60, 180, 90), width=1))
     p_line.setLabel("bottom", "ГГц")
     p_line.setLabel("left", "Y")
@@ -1533,7 +1443,9 @@ def run_pyqtgraph(args):
     width: Optional[int] = None
     x_shared: Optional[np.ndarray] = None
     current_sweep_raw: Optional[np.ndarray] = None
-    current_aux_curves: SweepAuxCurves = None
+    current_sweep_1: Optional[np.ndarray] = None
+    current_sweep_2: Optional[np.ndarray] = None
+    current_sweep_pre_exp: Optional[np.ndarray] = None
     current_sweep_norm: Optional[np.ndarray] = None
     last_calib_sweep: Optional[np.ndarray] = None
     current_info: Optional[SweepInfo] = None
@@ -1548,6 +1460,7 @@ def run_pyqtgraph(args):
     spec_mean_sec = float(getattr(args, "spec_mean_sec", 0.0))
     calib_enabled = False
     norm_type = str(getattr(args, "norm_type", "projector")).strip().lower()
+    logdetector_enabled = bool(getattr(args, "logdetector", False))
     # Диапазон по Y: авто по умолчанию (поддерживает отрицательные значения)
     fixed_ylim: Optional[Tuple[float, float]] = None
     if args.ylim:
@@ -1600,7 +1513,7 @@ def run_pyqtgraph(args):
         freq_shared = np.arange(fft_bins, dtype=np.int32)
 
     def _visible_levels_pyqtgraph(data: np.ndarray) -> Optional[Tuple[float, float]]:
-        """(vmin, vmax) по центральным 90% значений в видимой области ImageItem."""
+        """(vmin, vmax) по текущей видимой области ImageItem (без накопления по времени)."""
         if data.size == 0:
             return None
         ny, nx = data.shape[0], data.shape[1]
@@ -1624,8 +1537,8 @@ def run_pyqtgraph(args):
         if not finite.any():
             return None
         vals = sub[finite]
-        vmin = float(np.nanpercentile(vals, 5))
-        vmax = float(np.nanpercentile(vals, 95))
+        vmin = float(np.min(vals))
+        vmax = float(np.max(vals))
         if not (np.isfinite(vmin) and np.isfinite(vmax)) or vmin == vmax:
             return None
         return (vmin, vmax)
@@ -1667,17 +1580,22 @@ def run_pyqtgraph(args):
                 y_max_fft = float(fr_max)
 
     def drain_queue():
-        nonlocal current_sweep_raw, current_aux_curves, current_sweep_norm, current_info, last_calib_sweep
+        nonlocal current_sweep_raw, current_sweep_1, current_sweep_2, current_sweep_pre_exp, current_sweep_norm, current_info, last_calib_sweep
         drained = 0
         while True:
             try:
-                s, info, aux_curves = q.get_nowait()
+                s, info = q.get_nowait()
             except Empty:
                 break
             drained += 1
             current_sweep_raw = s
-            current_aux_curves = aux_curves
             current_info = info
+            s1 = info.get("sweep_1") if isinstance(info, dict) else None
+            s2 = info.get("sweep_2") if isinstance(info, dict) else None
+            current_sweep_1 = s1 if isinstance(s1, np.ndarray) else None
+            current_sweep_2 = s2 if isinstance(s2, np.ndarray) else None
+            pre = info.get("pre_exp_sweep") if isinstance(info, dict) else None
+            current_sweep_pre_exp = pre if isinstance(pre, np.ndarray) else None
             ch = 0
             try:
                 ch = int(info.get("ch", 0)) if isinstance(info, dict) else 0
@@ -1715,28 +1633,36 @@ def run_pyqtgraph(args):
             else:
                 xs = np.arange(current_sweep_raw.size)
             curve.setData(xs, current_sweep_raw, autoDownsample=True)
-            if current_aux_curves is not None:
-                avg_1_curve, avg_2_curve = current_aux_curves
-                curve_avg1.setData(xs[: avg_1_curve.size], avg_1_curve, autoDownsample=True)
-                curve_avg2.setData(xs[: avg_2_curve.size], avg_2_curve, autoDownsample=True)
-            else:
-                curve_avg1.setData([], [])
-                curve_avg2.setData([], [])
-            if last_calib_sweep is not None:
-                curve_calib.setData(xs[: last_calib_sweep.size], last_calib_sweep, autoDownsample=True)
-            else:
+            if current_sweep_1 is not None and current_sweep_2 is not None:
+                curve_calib.setData(xs[: current_sweep_1.size], current_sweep_1, autoDownsample=True)
+                curve_norm.setData(xs[: current_sweep_2.size], current_sweep_2, autoDownsample=True)
+            elif logdetector_enabled:
                 curve_calib.setData([], [])
-            if current_sweep_norm is not None:
-                curve_norm.setData(xs[: current_sweep_norm.size], current_sweep_norm, autoDownsample=True)
+                if current_sweep_pre_exp is not None:
+                    curve_norm.setData(xs[: current_sweep_pre_exp.size], current_sweep_pre_exp, autoDownsample=True)
+                else:
+                    curve_norm.setData([], [])
             else:
-                curve_norm.setData([], [])
+                if last_calib_sweep is not None:
+                    curve_calib.setData(xs[: last_calib_sweep.size], last_calib_sweep, autoDownsample=True)
+                else:
+                    curve_calib.setData([], [])
+                if current_sweep_norm is not None:
+                    curve_norm.setData(xs[: current_sweep_norm.size], current_sweep_norm, autoDownsample=True)
+                else:
+                    curve_norm.setData([], [])
             if fixed_ylim is None:
-                y_series = [current_sweep_raw, last_calib_sweep, current_sweep_norm]
-                if current_aux_curves is not None:
-                    y_series.extend(current_aux_curves)
-                y_limits = _compute_auto_ylim(*y_series)
-                if y_limits is not None:
-                    p_line.setYRange(y_limits[0], y_limits[1], padding=0)
+                y_candidates = [current_sweep_raw]
+                if current_sweep_1 is not None and current_sweep_2 is not None:
+                    y_candidates.extend([current_sweep_1, current_sweep_2])
+                elif logdetector_enabled and current_sweep_pre_exp is not None:
+                    y_candidates.append(current_sweep_pre_exp)
+                y_concat = np.concatenate([np.asarray(v, dtype=np.float32) for v in y_candidates])
+                y0 = float(np.nanmin(y_concat))
+                y1 = float(np.nanmax(y_concat))
+                if np.isfinite(y0) and np.isfinite(y1):
+                    margin = 0.05 * max(1.0, (y1 - y0))
+                    p_line.setYRange(y0 - margin, y1 + margin, padding=0)
 
             # Обновим спектр
             sweep_for_fft = current_sweep_norm if current_sweep_norm is not None else current_sweep_raw