Troubleshooting: audit and watch

What a finding from audit or watch means when the cause is not obvious, and what to do when a reading looks wrong or slow.

audit or watch says C003

The same fault, found on a desk that already has it: a show applied before apply checked for it, a patch made by hand at the desk, or a card pulled after the show went in. The three things to check are the ones for apply says the console has nothing in a bay.

$ atoo audit --state mod.state.json
Audit findings: 1 error.

[ERROR] C003 Patch Uses An Empty Expansion Bay
  CH 40 "Podium" is patched from MOD 1, but /$syscfg/opt_mod reports 'NONE': there is nothing in that bay, so no audio passes. The desk accepts and shows this patch without complaint.
...

It is an error, so audit exits 1. A live run asks the desk about a bay only when something on it is patched from or routed to that bay, so a desk that uses neither pays nothing for the check.

A capture that was taken without the bay’s reading cannot answer the question, and the audit reports that instead of passing quietly. The line comes just above the socket summary:

Expansion bay MOD (/ch/40/in/conn/grp) is in use but /$syscfg/opt_mod was not read: C003 (Patch Uses An Empty Expansion Bay) could not check it.

A channel is silent but its patch is right: C004

A WING channel has two input patches, a main one and an alternate, and a switch between them. The alternate is how virtual soundcheck works: the show is recorded to the WLIVE card and played back into the channels from there. A channel left on its alternate afterwards hears the card, not the stage — and its main patch, which is what the desk’s patch screen and every other check show you, still looks exactly right.

$ atoo audit --state alt.state.json
Audit findings: 1 warning.

[WARNING] C004 Channel On Its Alternate Input
  CH 12 "Kick In" is listening to its alternate patch from CRD, not its main patch (AES50-A 1). After a virtual soundcheck this is a channel that does not hear the stage; switch it back on the desk if the show is live.
...

It is a warning, not an error, because during a virtual soundcheck it is exactly what you want; audit exits 0. Switch the channel back to its main input on the desk. A capture taken before this check existed has no reading of the switch, and the rule then says nothing either way.

During a virtual soundcheck this warning is noise. In Watch, the Virtual soundcheck switch (or atoo watch --virtual-soundcheck) shows C004 and C005 at info, still naming every channel, with Virtual soundcheck is on, so this is expected in the row. Turning it off makes any channel still on its alternate a warning again at once, which is the check you want before doors. atoo audit has no such switch: it reads the desk once.

atoo apply puts every strip it builds back on its main input, so after an apply this finding means somebody switched the strip since. The desk’s global alternate switch is reported separately, once for the desk, as C005 Global Alternate Switch Engaged: it moves every strip set to follow it without any strip’s own switch changing, so C004 cannot see it. Switch it off on the desk; apply never writes it.

A channel is on, named and up, but you hear nothing: R008

Most checks start at a bus or a Main and ask whether it reaches the PA. R008 starts at the channel. A channel can have a source patched, a name and its fader up, and still go nowhere: no send to a bus, no send to a Matrix, no route to a Main. Nothing on the strip looks wrong, and a scene recall or a stray tap on the Main button is all it takes.

$ atoo audit --state cut.state.json
Audit findings: 1 warning.

[WARNING] R008 Strip Reaches Nothing
  CH 3 "Hat" is in use, unmuted and above the floor, but it has no bus send, no Matrix send and no Main route switched on, so its signal goes nowhere. Switch a send or a Main route on, or mute the strip if it is not wanted.
...

Switch a send or a Main route on for it, or mute the strip if it is meant to be out of the mix. It is a warning, so audit exits 0.

What it leaves alone, on purpose:

• A strip that is muted, or whose fader is at the floor. Both are silent already, and a fader at the floor has its own L002 finding.
• A strip that sends somewhere that goes nowhere. The bus (R002), Main (R005) or Matrix (R006) that reaches no output is reported once, as itself; the channels feeding it are not listed again.
• The talkback strip, and an aux a monitor output listens to.
• A strip it could not read in full. If any of a strip’s sends or routes was not read, the check says nothing rather than guess.
• A strip you have given a plan. With --plan, and in Watch with a show, a channel the plan routes somewhere is checked by CV016 Planned Routes All Off instead, so the same fault is not said twice.

One thing it cannot see: a channel can also leave the desk through a user signal (a direct tap to a recorder, say) with no send or route. Atoo does not read those slots, so a strip that goes only to a recorder that way is reported. If that is deliberate, the strip is working as set up.

A channel feeds itself and howls, or could: R009, R010

A channel or aux does not have to take its signal from a microphone. Its input can be a bus, a Main or a Matrix: the desk’s own mix patched back in, for an effects return, a recording feed or a mix-minus. That is safe until the strip also sends to, or routes to, the thing it takes. A channel patched from Bus 3 that sends to Bus 3, or an aux patched from Main 1 that is routed to Main 1, hears itself and howls, loud enough to hurt people and drivers. It can go through other busses, Mains, Matrices and other strips too.

$ atoo audit --state feedback.state.json
Audit findings: 2 error, 1 warning.

[ERROR] R009 Feedback Loop (2)
  AUX 3 "Return" takes MAIN 1 (patched from MAIN 1, its left side); AUX 3 routes to MAIN 1. The signal comes back round to what the strip takes, so this feeds back as soon as there is audio in it. Switch off the send or route that closes it, or patch AUX 3 from another source.
  CH 12 "FX Return" takes BUS 3 "Reverb" (patched from BUS 5, its left side); CH 12 sends to BUS 3. The signal comes back round to what the strip takes, so this feeds back as soon as there is audio in it. Switch off the send or route that closes it, or patch CH 12 from another source.

[WARNING] R010 Feedback Loop Held Closed
  CH 13 "Verb Return" takes BUS 4 "Plate" (patched from BUS 7, its left side); CH 13 sends to BUS 4 (switched off). This is silent only because of the closed leg marked above, and feeds back the moment it is opened. Remove the send or route, or patch CH 13 from another source.
...

The finding names the loop one step at a time, so you can see which send or route closes it. R009 is an error: every step is open, so it feeds back the moment there is audio in it, and audit exits 1. Switch off the send or the route it names, or patch the channel from something else.

R010 is a warning: the loop is there, but one step is off (a send switched off with a level set) or pulled to the floor, so it is silent today and howls the moment someone opens it. In Watch the warning is replaced by the error when the last step is opened, and cleared when the loop is taken apart.

A strip that takes a bus and sends somewhere else is not a loop, and a send that is off with no level (the desk’s own state) is not a step: the check only calls it a loop where the path really comes back to what the strip takes. A route to a Main has no level in the read, so an off route is not counted. The check cannot see a strip fed through a user signal, or one on its alternate input (C004 says that); it says nothing for them.

What the patch means is inferred. Each bus, Main and Matrix has two input sockets, left and right, as it has two output slots, so BUS 5 as a channel’s input is the left side of Bus 3 and BUS 6 its right. The finding quotes the socket it read (patched from BUS 5, its left side) so you can see at a glance if that is not what the desk shows.

A channel is on and named, but nothing is patched to it: C006

The other end of the same path. R008 asks where a channel’s signal goes; C006 asks whether it has one. A channel that is named, or has sends on, with nothing patched to its input carries silence, and its overview strip looks finished.

$ atoo audit --state nosource.state.json
Audit findings: 1 warning.

[WARNING] C006 Strip In Use Has No Source
  CH 3 "Hat" is in use (named) but nothing is patched to it, so it has no source. Patch an input to it on the desk, or pull its fader down and clear its name and sends if the strip is not wanted.
...

Patch an input to it on the desk, or clear it if the strip is not wanted. It is a warning, so audit exits 0.

It does not wait for the fader: a show built from your conventions starts every fader at the floor, and a strip with no source is as dead at the floor as it is at unity. It stays quiet for a strip on its alternate input (C004 already reports it) and for the talkback strip. With --plan, a patch the plan states is checked by CV005, so C006 leaves that strip to it. A capture taken without the patch readings cannot answer, and the check says nothing then.

A channel is silent and nothing on it is muted: C007

A WING input socket has a mute of its own, separate from the strip’s. A muted socket silences every channel patched to it, and nothing on the strip shows it: the strip’s mute is off, its fader is up, its meter is flat. It is easy to leave on after a line check, and it can happen mid-show.

$ atoo audit --state sockmute.state.json
Audit findings: 2 warning.

[WARNING] C007 Strip's Input Socket Muted (2)
  CH 3 "Vox" is fed from AES50-A 3, which is muted at the socket. Nothing on the strip shows it; unmute the input on the desk.
  CH 5 "OH" is a stereo source on AES50-A 5-6, and AES50-A 6 is muted at the socket while the other is not, so one side of it is silent. Unmute the input on the desk.
...

Unmute the input on the desk’s input page. A stereo source is two sockets, so one leg can be muted while the other is not; the finding says which. It is a warning, so audit exits 0: a socket mute is something an engineer does on purpose, and this one stays only because it cannot be seen from the strip.

It reads the local inputs (LCL) and the three AES50 ports (A, B, C), which costs no extra request. A channel patched from a card, a module, USB or StageConnect is not checked: those sockets are not read, so the rule says nothing for them rather than guess. A channel on its alternate input is reported by C004 instead, the talkback strip is left out, and with no input sockets read the run says that C007 could not be checked.

A stage box is on the patch but no audio comes from it: C008, C009, C010

A WING reports the state of each AES50 port: nothing on it (-), linked (OK), a link error (ERR) or a stage box whose firmware needs updating (UPD). A stage box whose link has failed silences every input on it and every output to it at once, and nothing on a channel shows it: the patch is as you set it, the fader is up and the meter is flat. C008 reports a port that something depends on, and C009 for a stage box in error that nothing is patched to. C010 says that the desk wants the stage box’s firmware updated (UPD).

$ atoo audit --state stagebox.state.json
Audit findings: 1 error, 1 info.

[ERROR] C008 Stage Box Link Error
  Stage box AES50 port A reports a link error and can not carry the inputs of CH 1 "Kick" and CH 2 "Snare".

[INFO] C010 AES50 Stage Box Firmware Needs Updating
  AES50-B (S32) reports UPD: the desk flags the stage box's firmware as older than it expects. Audio can still pass, but preamp remote control and dual-port A/B need current stage-box firmware; update the box. The meters are not being read, so Atoo cannot say whether it is passing audio.
...

C008 is an error. A port that reads ERR, or - (no stage box there at all), while a channel takes its input from it or an output on it carries a source, is a patch that cannot carry audio, and audit exits 1. The finding names the port, the channels and the outputs on it. Its title says which: Stage Box Link Error for ERR, Stage Box Not Connected for -, as the church desk with its stage box B unplugged reads:

[ERROR] C008 Stage Box Not Connected
  Stage box AES50 port B appears to not be connected and can not carry outputs AES50-B 25-32.

Check the cable and the stage box; for -, connect the stage box, or patch those channels and outputs somewhere else. A stage box that is simply not connected, with nothing patched to it, is not an error.

C009 is a warning. It says a stage box is in error with nothing patched from it or to it yet: nothing in the show is lost, but the link will not carry audio.

C010 is information, never a warning. UPD means the desk flags the stage box’s firmware as older than it expects, or not fully matching the console’s version. The link still passes audio: a healthy Behringer S32 on AES50-A of a full-size WING reads UPD while it carries the show. What needs current stage-box firmware is preamp remote control from the desk and dual-port A/B, so update the box when you can. C010 names the device the desk reports on the port and, in Watch, what the meters show: how many of the strips patched from that port are carrying signal, so the box is passing audio, or that none is just now, which may only be a quiet stage. It never lists every channel on the port as silent; that is for a port that is down (C008).

In Watch C008 and C009 appear when the link drops, and C010 while the desk reads UPD; each clears when it reads OK. A port that reads OK is silent, and so is a run that did not read the link: audit and Watch read it whenever they read the output sockets, and say that C008, C009 and C010 could not be checked when they did not. A channel on its alternate input is reported by C004 instead.

What ERR and UPD mean. The desk’s own list has those words. Atoo reads ERR as a link error; UPD is stage-box firmware the desk wants updated (see C010). R002 follows sockets, not links, so a bus patched to a dead stage box’s output still counts as routed there; C008 is the check that the link is up.

An unplugged stage box that only receives outputs (monitor wedges, no inputs) reads -, and its outputs are read only when the show’s plan declares them. Give the run the plan (--plan, or a show in Watch) and C008 sees a source patched to the missing box.

A whole group of channels is silent and every one is up: CH004

A DCA has a mute button as well as a fader. Mute the Band DCA and every channel in it goes silent, while each channel’s own mute is off and its fader is up. CH002 reports a DCA pulled down; CH004 reports one that is muted, with the channels it is silencing.

$ atoo audit --state dcamute.state.json
Audit findings: 1 info.

[INFO] CH004 DCA Muted, Silencing Active Members
  DCA 2 "Band" is muted and is silencing 2 active member(s): CH 2 "Bass", CH 3 "Gtr".
...

Unmute the DCA on the desk if that was not the idea. It is a note, so audit exits 0: pressing a DCA’s mute is deliberate and the DCA’s own button shows it lit, so this says what the press is silencing and does not call it a fault. (A mute group, CH003, is reported the same way.) A DCA that is muted and also pulled to the floor is two findings, CH004 and CH002, because unmuting one does not bring the band back while the other holds. It says nothing for a DCA with no channel in use under it, and nothing when the DCA’s mute was not read.

With --plan (and in Watch, with a show), a DCA the plan lists is judged against the plan instead, so a mute is reported once:

$ atoo audit --state dcamute.state.json --plan show.plan.json
Audit findings: 1 warning, 1 info.

[WARNING] CV017 Muted Against Plan
  DCA 2 "Band" is muted on the desk; the plan leaves it unmuted.

[INFO] S001 Findings Withheld From This Report
  1 finding(s) are not listed above because a plan was supplied: 1 CH004 (1 about a parameter the plan states). Re-run the same audit with no plan to see every finding the rules produced.
...

CV017 is a warning because the plan said otherwise. A DCA the plan lists as muted is as you asked, and neither rule reports it. A DCA the plan does not mention is still CH004’s.

A condenser is silent, or 48V is on where it should not be: CV019

Phantom power is set per input socket, and the plan states it for every channel: 48V for a condenser, none for a dynamic mic or a line source. atoo apply writes it. With the plan, audit and Watch compare the desk with it, on the socket each channel is patched to now. Two faults go unreported without it. A condenser with 48V off is silent and looks like any other channel. 48V on at a source the plan says must not have it can damage a ribbon mic. (C001 covers only 48V on a socket nothing is patched to.)

$ atoo audit --state touched.state.json --plan show.plan.json
Audit findings: 3 warning.

[WARNING] CV019 Phantom Contradicts Plan (3)
  CH 1 "Kick": the plan has 48V on at AES50-A 1; the desk has it off, so a condenser patched there is silent.
  CH 2 "Snare": the plan has no 48V at AES50-A 2; the desk has it on, which can damage a source that does not want it.
  CH 3 "OH": the plan has 48V on at AES50-A 6; the desk has it off, so a condenser patched there is silent.

Physical sockets read: in AES50-A 1-6; out LCL 1.

The last is a stereo pair: each of its two sockets is its own source with its own 48V, so the plan can power one and not the other. Switch the 48V on or off on the desk’s input page, or, if the desk is right, say so in your conventions or the input list and rebuild.

It is a warning either way, so audit exits 0. It says nothing about a channel the plan states no phantom for, a channel patched from a group that has no 48V (a card, a module, USB, StageConnect), a socket that was not read, or a channel that is not patched. In Watch it is one of the rules the show’s plan adds (Watching from the command line), and it clears when the desk is put back. A socket this rule judges is not also reported by C001.

A channel reaches the same Main or Matrix by two routes: R012, R013, R014

When one channel reaches the same Main or Matrix by two routes, the two copies add together. Level builds up where nobody expected it, and if one route has more processing latency than the other (an insert, a delay) the copies comb-filter against each other and the channel sounds hollow. The common ways to do it by accident: a channel routed to Main 1 directly and through a group bus that is routed to Main 1; a channel sent to a Matrix directly and through a bus or Main that feeds it; a channel sent to two busses that both reach one Main. audit and Watch say so, naming each route.

$ atoo audit --state split.state.json
Audit findings: 2 warning.

[WARNING] R012 Signal Splits And Recombines (2)
  CH 1 "Kick" reaches MAIN 1 twice: directly, and through BUS 1 "Drums". The copies sum there, so level builds up, and if one route adds more latency than the other (an insert, a delay) they comb-filter. Atoo has no plan it could read for this show, so it cannot tell a parallel bus from a mistake: run it with the show's plan (open the show, or pass --show or --plan); if this is a mistake, switch one of the routes off.
  CH 2 "Snare" reaches MAIN 1 twice: through BUS 1 "Drums", and through BUS 2 "Drums Comp". The copies sum there, so level builds up, and if one route adds more latency than the other (an insert, a delay) they comb-filter. BUS 2 "Drums Comp" is muted, so the signal through it is silent now and joins the moment it is unmuted. Atoo has no plan it could read for this show, so it cannot tell a parallel bus from a mistake: run it with the show's plan (open the show, or pass --show or --plan); if this is a mistake, switch one of the routes off.
...

R012 is a warning. The desk works and the show can go on, so audit exits 0, but the level is probably not what you meant. One finding per channel and per Main or Matrix, with the routes one hop at a time. Only a route that is switched on and above the floor counts: a send that is off, or pulled all the way down, is not a route. A muted bus, Main or channel on a route still is one: it is silent now, and the finding mentions it, because the copies add up the moment it is unmuted.

Two busses into one Main is how parallel compression is built (a “Drums” bus and a “Drums Comp” bus), so the check stands down when the plan asked for it. With --plan it is quiet for a channel and Main when the plan includes every route of the split: the channel’s sends and Main routes, and the busses’ own Main destinations. To ask for it, state the purpose of the parallel bus in your conventions or instructions and rebuild. When the plan includes every route, R013 notes it at INFO, so a rule that went quiet is never silent about it:

$ atoo audit --state parallel.state.json --plan show.plan.json
Audit findings: 1 info.

[INFO] R013 Signal Split Asked For
  CH 1 "Kick", CH 2 "Snare" reach MAIN 1 by more than one route, and the plan includes every one of those routes, so it is read as asked for (parallel processing) and R012 is quiet about it.
...

A bus whose role in busses.csv is parallel or fx (a captured show has one) is said by that role, one note per bus, quoting the cell: Parallel Bus Asked For, or Read As Effect Sends for an effect return. On the church desk’s captured show:

[INFO] R013 Read As Effect Sends (4)
  "busses.csv, Bus 13 "Delay Long", role: fx." CH 1 "Piano1" reaches MAIN 1 by more than one route, one of them through BUS 13 "Delay Long", read as an effect return beside the dry signal; R012 is quiet about it.
  ...

If the plan includes some of those routes but not all (the plan has the two busses, and the conventions never gave the channel a direct Main route that is also on), R012 still speaks and names the routes the plan does not include, for example the plan does not include the route CH 2 "Snare" to BUS 2 "Drums Comp". With no plan, nothing can have been asked for, so every split is reported except an effect send (below), and the finding adds: Atoo has no plan it could read for this show. That is the message to expect if you ran audit without --plan or --show, or started Watch without opening the show (Watch also notes this once, in an INFO, W005): run it again with the show’s plan. If you rebuilt the show while Watch was running, Watch is still using the plan it started with, and an INFO (W004) tells you, and the fix is to stop and start Watch. To make a split go away once the plan is in hand: switch one route off on the desk, or state the purpose of the parallel bus in your conventions or instructions and rebuild. In Watch both rules run on every pass and need nothing extra from the desk.

A reverb or delay bus is not a second copy (R014). On a WING an effect is a bus that inserts it: a channel sent to a “Reverb” bus that runs a reverb, routed to Main 1, reaches Main 1 directly and through that bus, but the bus carries the reverb’s return, not the dry channel again. So with or without a plan, a route through a bus with a reverb, delay, chorus, flanger, phaser or pitch shifter inserted on it (and switched on) is read as an effect send and left out of the split. R014 notes it at INFO, naming the effect it went on:

$ atoo audit --state verb.state.json
Audit findings: 1 warning, 3 info.

[WARNING] R012 Signal Splits And Recombines
  CH 1 "Kick" reaches MAIN 1 twice: directly, and through BUS 1 "Drums". The copies sum there, so level builds up, and if one route adds more latency than the other (an insert, a delay) they comb-filter. Atoo has no plan it could read for this show, so it cannot tell a parallel bus from a mistake: run it with the show's plan (open the show, or pass --show or --plan); if this is a mistake, switch one of the routes off.
...
[INFO] R014 Signal Split Through An Effect Send
  CH 5 "Vox" reaches MAIN 1 through BUS 11 "Reverb" as well as by another route. Read as an effect send: BUS 11 "Reverb" runs VSS3 in FX4 (a reverb), so it carries the effect's return rather than a second dry copy, and R012 is quiet about it. If BUS 11 "Reverb" is meant to carry the dry signal too, set the effect fully wet or take it off.
...

A bus with no such effect on it, such as a “Drums Comp” bus for parallel compression, is still two dry copies and still R012, until the show’s plan asks for it. So is a bus running a doubler (DOUBLE), which mixes a doubled copy with the dry voice, or a compressor or EQ. A channel with two dry routes and a reverb send gets R012 for the two, and the finding says which route it left out as an effect send. Atoo goes on the effect model, not on how wet the effect is set: a reverb bus set half dry still reads as an effect send.

A bus is heard on one side only, or is not the width the plan says: R011, CV020

Every bus, Main and Matrix has two output slots, a left one and a right one. When the object is stereo it puts a different signal on each, so if everything it is patched to uses the same slot, whatever is panned to the other side goes nowhere. The commonest case is a stage wedge on a stereo bus: the wedge hears the left side, and a send panned right is lost. audit and Watch say so.

$ atoo audit --state stereo.state.json
Audit findings: 1 warning.

[WARNING] R011 Stereo Output Patched On One Side
  BUS 3 "Drummer" is stereo but is patched on its left side only (LCL 3), among the output sockets read (LCL 1-4), so anything panned right is not heard on an output. If a mono feed is meant, set it mono and this goes away; if not, patch its right side as well.

Physical sockets read: in none; out LCL 1-4.
...

R011 is a warning. A stereo bus into one speaker is a one-sided mix, the kind you notice at soundcheck, so the audit reports it. It is a warning and not an error, because the desk works and the show can go on, and audit exits 0. If one speaker is what you meant (a stage wedge on a stereo bus, or a mono PA on Main left), set the bus mono on the desk, so the jack carries the whole mix, and the warning goes away. If not, patch its other side to a second output. It speaks only for a bus, Main or Matrix that is in use, so a freshly reset desk, which patches the left side of its first six busses to the local outputs, says nothing. It says nothing for an object patched on both sides, patched nowhere (R002, R005 and R006 say that), or whose width was not read, and nothing at all when the output patch was not read (the “could not be checked” note names it).

With the plan, CV020 compares the width. The plan decides every bus’s width, mono or stereo, and atoo apply writes it. A bus the plan makes mono that the desk has as stereo is a warning of its own, and then R011 stays quiet, because the plan says which jacks that bus is patched to.

$ atoo audit --state stereo.state.json --plan show.plan.json
Audit findings: 1 warning, 1 info.

[WARNING] CV020 Bus Width Contradicts Plan
  BUS 3 "Drummer": the plan has this bus mono; the desk has it stereo, so an output patched to its left side carries only the left of the mix and anything panned right is lost.

[INFO] S001 Findings Withheld From This Report
  1 finding(s) are not listed above because a plan was supplied: 1 R011 (1 about a parameter the plan states). Re-run the same audit with no plan to see every finding the rules produced.
...

Set the bus’s width on the desk, or, if the desk is right, say so in your conventions and rebuild. The plan states a width for a bus and nothing else, so a Main’s or a Matrix’s is not compared. CV020 is a warning, so audit exits 0; in Watch it is one of the rules the show’s plan adds (Watching from the command line), and it clears when the desk is put back.

Trim, gain, a high-pass, hi-cut, tilt or EQ is not what the plan says: CV008

With --plan, audit compares the processing the plan states for a channel or aux with what the desk holds: the trim, balance, invert and head-amp gain, the high-pass, hi-cut and tilt, the EQ’s bands, and a channel’s PreSend EQ. It does the same for the EQ and tilt the plan states for a bus, a Main or a Matrix. A filter the plan asks for is asking for it to be on, so a filter that is off is reported on its own, before its frequency.

$ atoo audit --state filters.state.json --plan filters.plan.json
Audit findings: 3 warning.

[WARNING] CV008 Processing Contradicts Plan (3)
  CH 2 "Vox": plan says the high-pass is on at 120Hz; live console has it off. (plan provenance: Processing: 'Vocals get a high-pass at 120Hz')
  CH 2 "Vox": EQ band 2: plan says frequency 2500Hz but 1800Hz live (tolerance 2%). (plan provenance: Processing: 'Vocals get a high-pass at 120Hz')
  CH 3 "Bass": plan says the high-pass is 40Hz, live console reads 80.2Hz (tolerance 2%). (plan provenance: Processing: 'Bass gets a high-pass at 40Hz')
...

In that capture the Kick’s high-pass reads 59.7Hz against a plan of 60Hz, and says nothing. A frequency or a Q is a stepped control: the desk holds the nearest step of what was written, so Atoo allows 2% (and 0.02 for a value at or near zero, such as a tilt of 0dB), the same tolerance apply uses when it reads a write back. A hi-cut written as 4000Hz that reads 4012Hz is the next step, and a Q of 1.4 that reads 0.998 is ten steps away and is reported. The tilt moves in coarse steps (a desk reads 1.25dB and -0.25dB), so a plan that asks for 1.1dB is reported, with the numbers on both sides.

The trim and the balance are compared within 0.5dB. The head-amp gain is a coarser control (as few as 20 steps on a local input, about 2.4dB each), so it is compared within 3dB, and it is read at the socket the strip is patched to, since the gain belongs to the input, not the strip:

$ atoo audit --state trim.state.json --plan trim.plan.json
Audit findings: 2 warning, 2 info.

[WARNING] CV008 Processing Contradicts Plan (2)
  CH 2 "Snare": plan says trim is 0dB, live console reads 4.5dB (tolerance 0.5dB).
  CH 2 "Snare": plan says head-amp gain on LCL 2 is 20dB, live console reads 9.5dB (tolerance 3dB).
...

An EQ band is one finding naming every value that differs (frequency, gain, Q, and the shape of the two shelf bands); the EQ’s own switch, model and mix are another. Stating any band implies the EQ is on, as apply writes it; a plan that says the EQ is off asks for it bypassed.

A bus, a Main, a Matrix and a channel’s PreSend EQ are compared the same way, band by band, and are named as the plan names them. A channel’s PreSend EQ is its second EQ: three bells and a switch. A bus’s, Main’s or Matrix’s tilt is one finding of its own, and it has no switch of its own to report.

$ atoo audit --state eq.state.json --plan eq.plan.json
Audit findings: 3 warning.

[WARNING] CV008 Processing Contradicts Plan (3)
  CH 1 "Kick": PreSend EQ band 1: plan says gain 3dB but 0dB live (tolerance 2%).
  BUS 1 "Drums": EQ band 2: plan says frequency 2500Hz but 1800Hz live (tolerance 2%).
  MAIN 1 "PA": EQ band high: plan says gain -3dB but 0dB live (tolerance 2%).
...

It says nothing for a value that was not read (a capture taken before this check has none of these values, and a live run with --plan reads exactly the ones the plan states, at no extra request), for an aux strip’s filters (an aux has no filter node, and no PreSend EQ), or for what a plan does not state. These are warnings, so audit exits 0. Watch does not run this one: a filter is set up, not something that moves during a show. The compressor is not compared.

An effect is on the strip but you cannot hear it: E001, E002

A WING effect is heard only through an insert: the strip’s pre- or post-insert names an FX slot (FX1-FX16), and the slot holds the effect. There is no FX send and return. So the insert button can be lit on a vocal channel while the slot it names is empty, and a reverb can be loaded in a slot that no strip inserts. Either way the mix sounds dry, and nothing on the strip itself looks wrong.

$ atoo audit --state fx.state.json
Audit findings: 1 warning, 1 info.

[WARNING] E001 Insert On An Empty FX Slot
  CH 3 "Lead Vox"'s pre-insert is on and names FX4, which has no effect loaded. Nothing is processing this strip; load the effect into FX4 or point the insert at the slot that has it.

[INFO] E002 FX Loaded But Not Inserted
  FX5 (DE-S2) has an effect loaded, but no channel, aux, bus, Main or Matrix inserts it, so nothing is heard through it.
...

Here the de-esser went into FX5 but the insert still names FX4. Point the insert at FX5 on the desk and both findings go. The E001 warning leaves audit exiting 0, because audio still passes.

A third rule, E003, reports an insert that names a loaded slot but is switched off, meaning the effect is bypassed. It is only INFO, since engineers bypass effects on purpose. A capture taken before these checks existed has no reading of the inserts, and all three rules then stay silent.

An effect your plan inserts (an FX bus’s from a bus preset, or one a capture found on any strip) is checked against the plan instead: CV015 Effect Insert Contradicts Plan is a WARNING when its insert is switched off, names another engine, or the engine holds another effect than the plan’s (Bus presets, Capture a desk). audit runs it when given the plan, and Watch runs it from the open show’s plan, with more plan checks: a channel cut off from every planned route (CV016), a mute the plan does not have (CV017), 48V that is not what the plan states (CV019, above), a bus whose width is not the plan’s (CV020, below), and the plan’s sends and Main routes (CV006, CV007, CV018) (Watching from the command line; the last three are below). On that bus E001 and E003 then stay quiet in audit, so the problem is said once.

The band cannot hear talkback: T001

Talkback is two settings on a WING, and it takes both. Assign picks the strip the talk mic comes through (CH40 or AUX8). Then TALK A and TALK B each pick which busses, Matrices and Mains they reach when pressed. Talkback that is assigned but has no destinations on either button looks set up and reaches nobody:

$ atoo audit --state talk.state.json
Audit findings: 1 warning.

[WARNING] T001 Talkback Goes Nowhere
  Talkback is assigned to CH40, but neither TALK A nor TALK B is routed to any bus, Matrix or Main, so pressing either reaches nobody. Pick TALK A's destinations (and TALK B's, if it is used) on the desk's talkback page.
...

On the desk’s talkback page, turn on the monitor busses TALK A should reach, and the finding goes. One destination on either button is enough. It is a warning, so audit exits 0. Talkback left unassigned is the factory setting and says nothing. A read only asks where TALK A and B go when talkback is assigned, so an unassigned desk pays for one address and no more.

The talk mic’s own strip is left out of the mix checks. A talk mic is normally kept out of the mix: fader down, in no DCA. So when talkback is assigned, audit does not report that strip as “fader at floor” (L002), “sends nowhere” (R008), “nothing patched” (C006) or “socket muted” (C007). It says once that it did this, so the exemption is never silent:

[INFO] T002 Talkback Strip Left Out Of Mix Checks
  CH 40 "TALKBACK" is the strip talkback is assigned to (CH40), so it is treated as the talk mic: 'fader at floor' (L002), 'sends nowhere' (R008), 'nothing patched' (C006) and 'socket muted' (C007) are not reported for it.

Talkback that misses one monitor mix: CV021, CV022

T001 only catches talkback that reaches nobody. Talkback can reach the PA and two of three wedges and still leave the drummer in the dark, and the desk cannot say which busses are monitor mixes: that is your plan. With --plan (or --show, or in Watch), audit takes a monitor mix to be a bus the plan feeds to an output socket and to no Main, unless every send the plan lists into it is post-fader. It then checks that TALK A or TALK B reaches each one. The plan here builds three monitor mixes, and TALK A reaches two of them:

$ atoo audit --state talk-monitors.state.json --plan plan.json
Audit findings: 1 warning, 1 info.

[WARNING] CV021 Talkback Misses A Monitor Mix
  Talkback is assigned to CH40 and reaches other mixes, but not BUS 2 "Gtr Wedge", a monitor mix the plan builds: neither TALK A nor TALK B is routed to it, so whoever listens to that mix cannot hear you. Switch B2 on in TALK A's (or TALK B's) destinations on the desk's talkback page.

[INFO] CV022 Monitor Mixes Talkback Was Checked Against
  Talkback reaches 2 of the 3 monitor mixes the plan builds: BUS 1 "Vox Wedge", BUS 2 "Gtr Wedge", BUS 3 "Drum IEM". A bus is taken for a monitor mix when the plan feeds it to an output socket and to no Main, and does not send into it post-fader only; a mix the plan builds another way is not checked.
...

There is one CV021 row for each mix it misses, so each goes when its own switch is turned on, on either button. Turn on B2 on TALK A’s page and the warning is gone; the CV022 note stays, now saying 3 of the 3.

• CV022 is how you check the choice. Which busses count as monitor mixes is inferred from the plan, so the note names the ones it took. A monitor mix it did not take (a bus the plan feeds only post-fader, say) is not checked, and is not in the list.
• A warning, not an error. The show goes on without talkback to one wedge, but whoever listens to it cannot hear you, and you find out when they do not answer.
• Silent when talkback is OFF, when the plan builds no monitor mix, for a mix whose switches were not read, and when talkback is routed nowhere at all, which is the single T001 row above.
• Only busses. Talkback destinations are busses B1 to B16, Matrices and Mains; a monitor mix the plan builds on a Matrix is not checked.
• It costs nothing extra. The switches are read whenever talkback is assigned, for T001, and a Watch pass is no longer.

Talkback and a band input on one strip: CV014

With --plan, audit also checks that talkback’s strip is free. When talkback is assigned to CH40 and the plan puts an input on channel 40, the talk mic and that input share one strip. The same goes for AUX8 and aux 8. On this desk talkback does reach a monitor bus, so T001 stays quiet, and channel 40 carries the plan’s Tom 3:

$ atoo audit --state tom3.state.json --plan plan.json
Audit findings: 1 warning, 1 info.

[WARNING] CV014 Talkback Strip Holds A Plan Input
  Talkback is assigned to CH40, and the plan puts 'Tom 3' on ch 40, so the talk mic and that input share one strip. Move the input off ch 40, or set talkback's assign to OFF or the other strip on the desk's talkback page.

[INFO] T002 Talkback Strip Left Out Of Mix Checks
  CH 40 "Tom 3" is the strip talkback is assigned to (CH40), so it is treated as the talk mic: 'fader at floor' (L002), 'sends nowhere' (R008), 'nothing patched' (C006) and 'socket muted' (C007) are not reported for it.
...

The INFO is the exemption above, applied to a strip that is really a tom. It goes once channel 40 is empty again.

Move the input to another channel in the input list and rebuild, or change the assign on the desk. If channel 40 in the plan is the talk mic, name it so: an entry whose name or source starts with “talk” (Talkback, Talk Mic), or is TB or T/B, is taken as the talk mic and says nothing.

A send or route you did not plan: CV018, CV006, CV007

With the plan, audit and Watch also check the desk’s sends and Main routes against what the plan built. Three rules, two WARNINGs and an INFO:

• CV018 Send Or Route Not In Plan, the INFO, is a send or route that is on and that the plan does not have: a channel sending to any Matrix above the floor, or a bus routed to a Main its plan entry does not list. It is how a wedge summed into Main is caught. A channel sending to a bus the plan gave it no send to is not reported: adding a send is an ordinary act.
• CV006 Bus Send Contradicts Plan is a send the plan states that is off, in the other pre/post mode, or (in audit only) at another level.
• CV007 Main Route Contradicts Plan is a channel’s route to a Main the plan lists but the desk has off, or one the plan does not list but the desk has on.

$ atoo audit --state touched.state.json --plan show.plan.json
Audit findings: 8 warning, 3 info.

[WARNING] CV006 Bus Send Contradicts Plan
  CH 2 "Snare": plan says send to Drum IEM mode is 'pre' (PRE), live console reads POST. (plan provenance: ...)

[WARNING] CV007 Main Route Contradicts Plan
  CH 3 "Rack Tom": plan lists no direct route from this channel to Main 1, live console reads on. (plan provenance: ...)

[INFO] CV018 Send Or Route Not In Plan
  BUS 4 "Drum IEM": routed to MAIN 1 on the desk; the plan lists no route from this bus to it.

(the other findings are left out here)

Physical sockets read: in AES50-A 1-15, LCL 1-9; out AES50-A 1-8, LCL 1-2.
...

A desk left exactly as atoo apply set it raises none of them. If the send or route is what you want, leave the row (a WARNING or an INFO, and audit exits 0) or put it in your conventions and rebuild, so the plan has it. Two things to know:

• Watch compares the switch and pre/post, not the send level. A level is what you ride at soundcheck. audit --plan still reports a send more than 0.5 dB from the plan’s.
• The plan cannot say that a channel feeds a Matrix. A send from a strip to a Matrix therefore always reads as outside the plan, until you switch it off or take it to the floor.

If every planned send of a channel is off, that is one row, CV016, naming each; CV006 and CV007 do not repeat it. A send or route that did not answer is not reported.

audit or watch says rules “could not be checked”

Coverage: no output sockets read -- R002, R003, R004, R005, R006, R011, CV009, C008, C009 and C010 could not be checked.

That is watch’s one-line form; audit says the same in a sentence after its Physical sockets read: line. If no input sockets were read either, C001 (phantom power on an unpatched socket) and C007 (a strip’s input socket muted) join the list.

Nothing is broken. The output rules are all about the output patch — which bus, Main or Matrix feeds which physical jack — and C008, C009 and C010 read each stage box’s link along with it. This run never read it, so they declined to answer rather than guess.

That happens when both of the sources that know what jacks the desk has come up empty: no --plan naming the sockets it patches, and a console whose model string Atoo does not recognize, so it cannot look up the panel. It is the normal case for --state captures taken before those sockets were being read at all.

Give it the sockets and the rules come back.

atoo watch --console 192.0.2.16 --output-sockets LCL=8

or point the run at the plan for the show, which names every socket it patches, stage boxes included.

A partial read is still a read. The rules only stand down when no output socket was read. A stage box, a module or a card the desk reports is read whatever you named; if you named LCL and the mix goes out on USB or StageConnect, they will run and they may well fire — the coverage line tells you what was covered, and that is the thing to check first.

audit says “No findings” on a console that is visibly wrong

Kick and snare are crossed at the stage box. The desk has Kick on LCL 2 and Snare on LCL 1, and it sounds wrong from the first hit. audit still passes:

$ atoo audit --state swapped.state.json
No findings -- every rule in this audit passed.

Physical sockets read: in LCL 1-2; out LCL 1.

Without --plan, audit checks that the desk works, not that it is your show. Every rule it runs alone is structural: a bus nothing feeds, 48V on a socket with nothing patched, a mix that reaches no loudspeaker. Two channels on two different sockets, both reaching Main, break none of those. Nothing on the desk says which socket the kick mic is plugged into; only your input list says that.

Give it the plan, and it compares the desk against what you asked for:

$ atoo audit --state swapped.state.json --plan swapped.plan.json
Audit findings: 2 warning.

[WARNING] CV005 Patch Contradicts Plan (2)
  CH 1 "Kick": plan says patched to LCL 1, live console reads LCL 2. (plan provenance: Input list: LCL 1)
  CH 2 "Snare": plan says patched to LCL 2, live console reads LCL 1. (plan provenance: Input list: LCL 2)

Physical sockets read: in LCL 1-2; out LCL 1.

Both exit 0, because a WARNING does not fail the run. Against a live desk the command is the same with --console in place of --state, or atoo audit --show DIR on a bundle, which picks up the bundle’s plan.json by itself.

If a plan is already there and the finding still does not appear, check these, in this order:

• The plan does not state it. The comparison covers what a plan states: patch, mute, fader, DCA and mute-group membership, bus sends, routes to Main, and trim, balance, invert, head-amp gain, the filters and the EQ. It does not cover the compressor, so a wrong compressor passes.
• The finding was withheld. With --plan, a finding about a strip the plan never mentions is left out, unless its rule matters on any desk. S001 Findings Withheld From This Report, at the end of the report, counts what was withheld and why. Run it again without --plan to see all of them.
• That part of the desk was never read. The Physical sockets read: line at the foot of the report lists the sockets this run covered. A patch on a socket outside that list is not checked, and neither is anything left out with --channels, --buses and the other count flags. See the section above.
• No rule covers it. Some things are left out on purpose. Listenback is one, and word clock against the expansion bays is another.

audit or watch reports a duplicate name that is not on the desk

[WARNING] N001 Duplicate Channel/Aux Names
  2 channel/aux strips share the name 'Acoustic': CH 4, CH 7.

— and the desk plainly shows one channel called “Acoustic” and one called “Acoustic 2”.

A WING strip normally displays the name of the socket it is patched to, not its own (the desk’s “Link Customization to Source” switch, clink, which is on from the factory). Rename a channel on the desk and the name goes to the socket; whatever was on the strip stays there and goes stale. Repatch that channel to a different socket and the two disagree.

Atoo reads that switch and prefers whichever name the desk is actually showing. So a duplicate it reports is a real collision — two strips the desk really is showing the same name for.

One case that is deliberately not a collision: two strips fed from one socket, a festival split or a monitor feed. That is one name in one place, and C002 Shared Physical Input reports it as what it is.

audit or watch looks like it has hung

Look for the read N/… progress line before you stop it. Every live read prints one on stderr, and while it moves, the run is working. This one was captured on a WING Compact by a version whose read counted to 300; today’s counts to about 323:

  read 120/300 (40%) 2s -- /io/in/USR/44 (node)

Five seconds or less is normal for a WING Compact, read a whole object at a time. With --per-leaf, one request per address, about two minutes is normal.

No progress line at all means you passed --no-progress, or the desk never answered, in which case stderr reports it and atoo ping is the next step.

A line that moves slowly, or sits on one address, means the desk is not answering some addresses. Each silent one is asked three times and costs about 3 seconds at the default --timeout, though several silent at once wait side by side, so a handful costs about 3 seconds between them. The read counts them at the end and names the first ten:

  read: 6112 of 6112 addresses in 0.9s
  3 address(es) went unanswered, about 0s of the time above. Those values are unknown -- not defaults.
  unanswered: /ch/5/fdr, /bus/2/name, /io/out/LCL/3/grp

Those values are reported as unknown, never as defaults. --raw-log FILE records every request and reply, line by line, so it survives a read you stop with Ctrl-C. A (node) on an address means a whole-object request went unanswered; the read then falls back to that object’s values one at a time, so nothing is lost but time.

When three addresses of the same kind (/fx/{n}/mdl, say) go unanswered while the desk answers everything else, the read stops asking for the rest of that kind, with an INFO: stopped asking for … line. If that line names something your desk does have, such as a fader or a send level, the link is slow or lossy: run the read again, and report it in an issue.

Ways out, in this order:

Try	What it does
--timeout 0.3	cuts each silent request from 3s to 0.9s; changes nothing for a desk that answers
--per-leaf	only if the (node) requests go unanswered: skips them, at about two minutes a read (Checking a desk)
drop --per-leaf, if you passed it	about 323 requests instead of 6,143: ~5s or less instead of ~2 minutes
fewer objects (--channels 16, --matrices 0)	last resort: a rule that goes quiet because you narrowed the read is still a rule that went quiet, though the report says what it covered

See also

• Troubleshooting — every symptom, and the page it is on
• Checking a desk — atoo audit
• Watching from the command line — atoo watch
• Reading findings — what a finding’s code tells you
• Rules reference — every rule audit and Watch can report