superstockneo

Yes this appears to be a PID tuning issue.  I cant really make any suggestions for PID values as I have no personal experience with the EvoX throttle so you will need to experiment yourself.  Based on other cars with similar throttles, my guess is most likely derivative gain needs to be increased a lot.   
To tune e-throttle PID I typically set up a time plot with TP main and E-throttle target on the same graph, set up the E-throttle target table so that it quickly steps between 20% and 80% with a small pedal movement.  With the time plot running live (engine off) move the pedal and watch how well TP follows target, how much oscillation, overshoot, undershoot etc.  Change only one P, I or D gain at a time and test each change.  
Target set up:

From memory the gain and update rate table examples given by me above are from my Evo road car (pretty much dead stock evo 7) the lambda probe is very close to the turbine outlet probably at the most 500mm from the exhaust port, in this car I have the reactivation delay set to 3 samples.  In our WRX11 test car which also has the probe right after the turbine - but has much longer manifold primaries etc (probably more like 1500mm from port to probe) I have reactivation delay set to 5 samples.  The update rate table is similar in both cars, the subaru could actually do with less than 1Hz at idle as it takes about 3 secs to respond to a change of PW but 1Hz is currently the lowest it can do. 
So what this means is in my evo with the reactivation delay set to 3 samples, if CLL was temporarily disabled due to over-run fuel cut or similar around idle where my update rate is 1Hz (1 sample per second), then CLL would be re-enabled 3 secs later.  To give an example at the other end of the table, if I hit the RPM limit at 7000RPM where my update rate was 10Hz, then CLL would be disabled for only 0.3s.    

I have a map i can share you, it gets RPM / Fuel / Power Steering / Coolant / Fuel consumption (L/100km) all working properly. It doesn't address everything but it's a great start.

I'm hoping the team at Link will start to work on this because I'll be building swap kits for the 86 chassis and I really don't want to use other ECU's as I'm quite proud of the capabilities of the Link G4X series.

Just an update. I've reverted the idle base position table to span a single axis. and i have reverted to open loop idle. Was stuck in a 2 hour traffic jam yesterday with IAT hitting 69degC (kinda hot because i'm running an open pod). Seems to be running pretty good with barely noticeable undershoots or overshoots during AC operations. Only the neutral decel overhangs. But i did managed to max out the speed lockout. Open loop seems to run better, but I dont have a log for this. Will take one with i've got the time. 

Try reducing the timing in the areas just above ide and at lower map reading than idle to like 0-5 deg and see if that helps.

Thanks @koracing. I'll definitely have a go at that. Previously I did bring the timing in the region below 2000 down, it's roughly around 14-11degs now. My target idle ign is set at 10deg. I could try running a lower timing. But during neutral decel will in fall into idle control mode as i believe speed lockout is still activated. i've set my lockout to be at 20kmh if i recall correctly. unfortunately reducing clutch delay did not do anything. Maybe as another workaround i'd revert back to open loop idle for the time being. the issue is just prominent during hot conditions like traffic and such. or idle stationary when ect is around 80deg. i'm not sure if this is related to other tuning matters like fueling etc.

Hi @koracing, appreciate the input. I've tried messing about with the idle actuator integral gain but seems to not getting any changes. i've also tried messing about with increasing base positions. it reduces undershoots but still slow to respond and gets high overshoots during neutral decel. Spoken to @Adamw offline and he believes it could be firmware related. Thanks Adam for looking into this. 

some bloke on the Subaru forums plotted a few Defi calibration curves
http://bbs.22b.com/forums/showthread.php?8223-Defi-Sensor-Calibration-Curves
 
Pressure:
0psi=.580v
5psi=.753v
10psi=.837v
15psi=.978v
20psi=1.112v
25psi=1.250v
30psi=1.392v
35psi=1.525v
40psi=1.670v
45psi=1.811v
50psi=1.954v
55psi=2.095v
60psi=2.239v
65psi=2.372v
70psi=2.504v
75psi=2.647v
80psi=2.780v
85psi=2.926v
90psi=3.063v
95psi=3.200v
100psi=3.326v
105psi=3.454v
110psi=3.596v
115psi=3.739v
120psi=3.882v
125psi=4.000v
130psi=4.150v
135psi=4.280v
140psi=4.420v

Temp:
DegC Voltage
25 = 4.548
30 = 4.481
35 = 4.402
40 = 4.3115
45 = 4.212
50 = 4.1045
55 = 3.986
60 = 3.858
65 = 3.726
70 = 3.583
75 = 3.429
80 = 3.2785
85 = 3.119
90 = 2.9625
95 = 2.8105
100 = 2.6595
105 = 2.509
110 = 2.342
115 = 2.2025
120 = 2.0575
125 = 1.9205
130 = 1.775
135 = 1.657
140 = 1.542
145 = 1.435
150 = 1.329
155 = 1.2355
160 = 1.1565
165 = 1.0755
170 = 1.0045
175 = 0.942
180 = 0.8855
185 = 0.8515
190 = 0.814
195 = 0.795
200 = 0.776

If idle "hangs" in neutral decel, it's likely it's not entering the idle control range very quickly - you can help this by setting the timing table values a bit low in the region it's hanging in as that should help drop the idle quicker into the proper range for control.
Did reducing the clutch delay not do anything?

The firmware side may be the best solution, but my advice is more of a work around/make it work how it is set of advice.

I had previously discussed this with the poster before I saw it on the forum here.  There appear to be 2 issues going on here to me, one is the e-throttle is getting a lot of integral wind up for some reason which makes it slow to respond to step changes in target.  Secondly, I dont think the AC offset is always being applied as it should.  I think this is because the deadband is set larger than the AC idle up so the offset gets ignored until the status goes active.  I need to spend some more time playing around with it on the bench and thinking through all the logic, it may need a change in firmware.  

One thing you can try is reducing the clutch delay from .5 seconds to .25 or some smaller value.  I see your AC output is a virtual aux, but I didn't see if it was actually controlling the clutch or not via some other logic or output.  Also your integral gain for the idle actuator is .05 - have you tried making this higher or lower?

Your base position for idle is below the position it stabilizes at in either case (ac on or ac off) where normally I prefer the base postion to be just slightly above the nominal idle duty/percentage. 
I would up the 0 row to 3% in the warm position based on your datalog (perhaps .5% higher across the board).
 
I don't really like the way the ac offset table works as it doesn't typically change the idle position up and down both the way I would like with closed loop working.
If you aren't getting the results you want with the above, I would try to do the following:
1. Zero out your ac idle offset table
2. Add a vertical axis to your base position table for AC request and have values 0 and 1 for two rows (current table values will be in row 0)
3. Set all the values in the "1" row to 1 to 1.5% higher than your 0 row.  

Not to take anything away from you post, I love to see these ideas coming through - but Im just mentioning this in case you werent already aware of the "quick trim" function that has been around since the G4, it will do this same basic calculation and apply the result to the fuel table for you (you still need to manually type in the measured value though).  If you have a log open and the fuel map on the same page, click on the cell in the fuel map that you want to adjust, then hit the M key, it will grab the target lambda from the lambda target table, you enter the measured lambda from the log and it will apply the correction to that cell for you.

On 2nd thoughts gp idle isn't a table just a value and slapping CAN DI on the axis of the Idle base position table wouldn't give you the ability to remove the extra throttle at idle so 2nd E-Throttle table is probably the best option. Slap RPM on one axis and have 0% APS be something bigger than 0 at Engine speeds above like 2000rpm and 0% below ~2000rpm. Note the value you put in the e-throttle table isn't the exact value you're going to get, you're going to get some combination of the idle target and the E-Throttle target so have a play with it.
As a side note I recently had a play with one of my cars using antilag but only enabled under conditions that I want pops and bangs as Antilag gives you the ability to control cut percentage, ignition trim, fuel trim and it disables overrun fuel cut while active but I couldn't get it to work on my car due to what I suspect was too much vacuum/not enough air.

Yes you can put the CAN DI in as the activation for the 4d/5d ignition and Fuel tables so that they are activated at the same time.
You could do a 2nd E-Throttle table or if you are using the E-Throttle for idle control you could use the GP Idle trim setup to increase the throttle position when you are off the accelerator pedal.

Leave overrun fuel cut off but setup a gp rpm limit as shown here, it should act pretty much the same as the overrun fuel cut shown but won't cut when CAN DI 1 is active.
 

Try the following:

The typical method of hysteresis would be (temperature > lower value AND output already on) OR temperature > higher value.
So use a GP Output for each feeding into virtual Aux's and then a GP Output that takes both the Virtual Aux statuses and feeds it into the actual output or feed the Engine Fan output into a Virtual Aux, use a GP Output into a Virtual Aux for the IAT and then a 2nd GP Output that looks at the engine fan status or engine fan Virtual Aux and the IAT GP Output Aux.

The update rate is to take care of the gas transport delay and the sensor response.  So effectively how long it takes for the ecu to be able to detect the change in mixture after it has adjusted the fuel.  i.e no point in making corrections 4 times a second if it takes 1 second for the change to reach the sensor.  Gas flow, exhaust size and sensor location etc play a big part.  To get an idea you can do a quick step test at various conditions.  Turn off CLL.  Log lambda and lambda target, change the lambda target by say 15%, then watch how long it takes for the measured lambda to change by a similar 15% - this is your combined transport delay and sensor response time.  Idle will usually be about 1Hz (Actually often needs to be slower but that is the slowest we allow at present), most engines will be ok with a 10Hz update at WOT, in between those points is where you need to spend the time observing and optimizing.  
It is more technically correct to have the update rate table referencing something mass flow related - for example inj DC or "air per cyl estimated" since this is what effects exhaust flow and hence transport delay but I usually find the default RPM based table to work fine.  You can customise the update rate table axes by setting the trim limits tables to 3D if you really want to.
For the gain you want to it be as high as you can go without causing an oscillation or large over correction.  If the update rate is too high this will also be observed as overshoot and or oscillation.  
An example set up from our WRX below just to give you an idea.  This has the probe right at the turbine exit but the primaries are very long on these engines - probably 1M long.  

I'm interested to be a child myself. I've heard about "ghost cam", I kinda get the idea of playing around with ignition maps. But how does one go about properly setting it up? is it ideal to have the trims on a 4D or the main map? Does idle ignition need to be disabled in order for this childish feature to work properly? Do you need to adjust fueling as well, like leaning it out?
Would appreciate to have your input @Adamw. Maybe an example. He he he.