[audio] Compute frames_to_microseconds() in 64 bits (#21089)

esphome/components/audio/audio.cpp

@@ -23,7 +23,8 @@ AudioStreamInfo::AudioStreamInfo(uint8_t bits_per_sample, uint8_t channels, uint
 }
 
 uint32_t AudioStreamInfo::frames_to_microseconds(uint32_t frames) const {
-  return (frames * 1000000 + (this->sample_rate_ >> 1)) / this->sample_rate_;
+  // In 64 bits: frames * 1000000 overflows 32 bits above 4294 frames, about 90 ms at 48 kHz.
+  return (static_cast<uint64_t>(frames) * 1000000 + (this->sample_rate_ >> 1)) / this->sample_rate_;
 }
 
 uint32_t AudioStreamInfo::frames_to_milliseconds_with_remainder(uint32_t *total_frames) const {

tests/components/audio/test_audio_stream_info.cpp

@@ -0,0 +1,20 @@
+#include <gtest/gtest.h>
+
+#include "esphome/components/audio/audio.h"
+
+namespace esphome::audio::testing {
+
+TEST(AudioStreamInfo, FramesToMicroseconds) {
+  AudioStreamInfo info(16, 2, 48000);
+  EXPECT_EQ(info.frames_to_microseconds(48), 1000U);
+  EXPECT_EQ(info.frames_to_microseconds(1), 21U);  // 20.83 us, rounded
+}
+
+TEST(AudioStreamInfo, FramesToMicrosecondsAboveFourThousandFrames) {
+  // frames * 1000000 no longer fits 32 bits above 4294 frames, about 90 ms at 48 kHz.
+  AudioStreamInfo info(24, 2, 48000);
+  EXPECT_EQ(info.frames_to_microseconds(4800), 100000U);
+  EXPECT_EQ(info.frames_to_microseconds(48000), 1000000U);
+}
+
+}  // namespace esphome::audio::testing