strict receiver target
Clean physical playback with exact packet accounting and final-mmap PCM verification.
MEASURED RECEIVER PERFORMANCE
Same hardware.
Clear boundaries.
The canonical performance record for Bama Box: physical DAC output, sustainable Arm receiver density and separate x86 resource evidence, each shown at its actual measurement boundary.
Bama BoxRAVENNAGStreamer
1 ms
1404
3 ms
33205
5 ms
4024—
Sustainable stereo-stream density · captured or final-mmap PCM boundary · 48 kHz AES67
REAL HARDWARE OUTPUT
1 ms configured receiver target sustained through PCM5122/I²S.
One Bama Box stereo stream sustained through ALSA mmap/DMA, I²S and the TI/Burr-Brown PCM5122 with a maximum runtime ALSA queue of 34 frames, or 0.71 ms.
SENDERmacOSRTP/L16 + PTPd2 GM
→NETWORKDirect 1 GbEno intervening switch
→RECEIVERBCM27114× Cortex-A72
→OUTPUTPCM5122 / I²S34-frame max ALSA queue
practical physical profile
At a 1 ms configuration, the internal ALSA queue remained above 2 ms.
practical physical profile
The tested bridge path had a 2.3 ms configured floor before DAC delivery.
EMBEDDED RECEIVER DENSITY
One Cortex-A72 platform. Three receiver stacks.
Reference platform: Raspberry Pi 4, quad-core Arm Cortex-A72, Debian 12, software PTP/software timestamps. Every stream is stereo, so 40 streams represent 80 audio channels.
| Receiver stack | Output / validation boundary | 1 ms | 3 ms | 5 ms | Recorded resources |
|---|---|---|---|---|---|
| Bama Box | ALSA null mmap · final playback PCM checked | 14 stereo28 channels | 33 stereo66 channels · +65% vs RAVENNA | 40 stereo80 channels · +67% vs RAVENNA | 0.875 / 2.780 / 3.355 cores 24.0 / 43.7 / 51.0 MiB RSS |
| Tested stock RAVENNA/Linux | Direct RAVENNA capture device | 0 | 20 stereo40 channels | 24 stereo48 channels | 1.996 / 2.650 cores not separately reported |
| GStreamer 1.22 | Synchronized tmpfs PCM capture | 4 stereo8 channels | 5 stereo10 channels | Inconclusive | 0.4804 / 0.6700 cores 23.33 MiB RSS at 1 / 3 ms |
Density configurations are receiver-software timing profiles, not analog or network-to-DAC end-to-end latency measurements. GStreamer’s 5 ms soak missed 2 of 480,000 payloads despite zero lost/late counters, so no sustainable figure is claimed.
X86 BACKSTAGE POTENTIAL
Receiver efficiency beyond embedded endpoints.
A cloud KVM/Linux-container benchmark exercised the portable receiver path with the UDP fallback boundary available in that environment. It shows the potential for mixers, console engines and backstage servers; direct packet access depends on the OEM system security configuration.
6×lower CPU load
11×lower memory
16 stereo AES67 streams · 32 channels · compared with FFmpeg on the same x86/KVM environment
MEASURE YOUR PRODUCT TARGET
Plan a joint benchmark Move from reference data to your hardware.
We can reproduce the receiver comparison using your SoC, timing source, NIC and audio interface.