Index: trunk/libffado/src/motu/motu_avdevice.h
===================================================================
--- trunk/libffado/src/motu/motu_avdevice.h (revision 659)
+++ trunk/libffado/src/motu/motu_avdevice.h (revision 660)
@@ -164,5 +164,5 @@
     struct VendorModelEntry * m_model;
     signed int m_iso_recv_channel, m_iso_send_channel;
-    signed int m_bandwidth;
+    signed int m_rx_bandwidth, m_tx_bandwidth;
 
     Streaming::MotuReceiveStreamProcessor *m_receiveProcessor;
Index: trunk/libffado/src/motu/motu_avdevice.cpp
===================================================================
--- trunk/libffado/src/motu/motu_avdevice.cpp (revision 659)
+++ trunk/libffado/src/motu/motu_avdevice.cpp (revision 660)
@@ -191,5 +191,6 @@
     , m_iso_recv_channel ( -1 )
     , m_iso_send_channel ( -1 )
-    , m_bandwidth ( -1 )
+    , m_rx_bandwidth ( -1 )
+    , m_tx_bandwidth ( -1 )
     , m_receiveProcessor ( 0 )
     , m_transmitProcessor ( 0 )
@@ -481,5 +482,4 @@
     //   * DATA_PREFIX = 0.16 us
     //   * DATA_END    = 0.26 us
-
     // These numbers are the worst-case figures given in the ieee1394
     // standard.  This gives approximately 0.5 us of overheads per packet -
@@ -492,17 +492,19 @@
     // 192 kHz for the traveler) but now do this based on the actual device
     // state by utilising the result from getEventSize() and remembering
-    // that each packet has an 8 byte CIP header.  Note that m_bandwidth is
-    // a *per stream* bandwidth - it must be allocated for both the transmit
-    // and receive streams.
-    signed int max_event_size = event_size_out>event_size_in?event_size_out:event_size_in;
+    // that each packet has an 8 byte CIP header.  Note that bandwidth is
+    // allocated on a *per stream* basis - it must be allocated for both the
+    // transmit and receive streams.  While most MOTU modules are close to
+    // symmetric in terms of the number of in/out channels there are
+    // exceptions, so we deal with receive and transmit bandwidth separately.
     signed int n_events_per_packet = samp_freq<=48000?8:(samp_freq<=96000?16:32);
-    m_bandwidth = 25 + (n_events_per_packet*max_event_size);
+    m_rx_bandwidth = 25 + (n_events_per_packet*event_size_in);
+    m_tx_bandwidth = 25 + (n_events_per_packet*event_size_out);
 
     // Assign iso channels if not already done
     if (m_iso_recv_channel < 0)
-        m_iso_recv_channel = m_p1394Service->allocateIsoChannelGeneric(m_bandwidth);
+        m_iso_recv_channel = m_p1394Service->allocateIsoChannelGeneric(m_rx_bandwidth);
 
     if (m_iso_send_channel < 0)
-        m_iso_send_channel = m_p1394Service->allocateIsoChannelGeneric(m_bandwidth);
+        m_iso_send_channel = m_p1394Service->allocateIsoChannelGeneric(m_tx_bandwidth);
 
     debugOutput(DEBUG_LEVEL_VERBOSE, "recv channel = %d, send channel = %d\n",