Appearance

A little motion, or a quiet place to read.

Following your device’s motion preference.

← RESOURCE INDEX

WORKING NOTE

forum report

v1.5b3: host IN stalls after patch switch — MIDIFACE 8x8 full-speed, PCM81 on DIN 8

2026-10-18

Report for the ck2 v1.5 beta thread.

Setup: Cirklon 2 host port -> Miditech MIDIFACE 8x8 (direct, no hub) -> DIN 8 -> Lexicon PCM81. Cirklon routes host port 8 to its USB-device port 8 (Cirklon2 Port 8 on macOS). Only sender is PCM81 Control.

Reproduced 2026-10-18: switching patches in the editor, not slider dragging. After the switch the PCM81 stops replying. App: "No edit-buffer reply. Check MIDI ports and PCM81 SysEx On." Independent CoreMIDI probe (app disconnected), one whole edit-buffer request, 2.5 s: 0 bytes, 0 partials. A lone F7 then the same request: still 0 bytes.

Host OUT is working: pressing Connect sends the editor dump and the PCM81 panel shows "Midi data dump complete", but the Mac receives no reply. The PCM81 is alive and accepting data; the loss is on the host IN path.

Replugging the 8x8 at the Cirklon host port recovered the IN path briefly. A 40-request probe then logged 6 complete 1,421-byte edit-buffer replies, followed by partial frames of 1,418, 1,412 and 1,418 bytes and a final stuck buffer of 1,065 bytes. A separate capture after a later replug saved a 969-byte partial whose first 11 bytes match a new frame and whose byte 11 is 04 (the newly selected Dyno TSC / Glide>Hall patch). It is a clean prefix, missing the final 452 bytes including checksum/F7. Replug recovers a burst, then the IN path wedges mid-frame again. The stream debug page never shows an error.

Stream debug page at failure:

TX s:1 B:0     us:1  rst:00  errs:00  sus:00  I:00 ctr:0000 O:00 sof:0000
RX BUFF POS: h00t00 h00t00 h00t00 h00t00 h00t00 h00t00

No error/suspend/reset and head==tail on every displayed stream; the page does not show the stall.

Device descriptor details recorded in the report:

  • 2 interfaces: AUDIO/AUDIOCONTROL and AUDIO/MIDISTREAMING.
  • EP 0x01 OUT, EP 0x82 IN, both bulk, wMaxPacketSize 64, bInterval 0.
  • Full-speed bulk, 8 virtual cables on the pair. On the HS DWC2 host this uses split transactions even directly attached.

Question: does the b3 split-transaction fix (the microframe-position one) cover a directly attached full-speed bulk device, or only the MioXL-on-hub path? If not, I can test a build and provide a Cynthion/USB capture.

Screenshots

No edit-buffer reply

PCM81 Control connected to Cirklon2 Port 8 showing No edit-buffer reply and an offline PCM81.

Open the original app screenshot

Cirklon stream debug page at failure

Cirklon stream debug page at failure showing zero error, suspend and reset counters, with matching receive buffer heads and tails.

Open the original stream debug screenshot

Probe logs