The Baconator's latest victim

[Kevin Bacon 5]

Yeah I'm pleased with the DTA overall. It can control shed loads of things. I've just scratched the surface there! I know some folk struggle with DTAs and hate them, but I've always got good results from them :D As Wigs said, if you want turn-key, stick to the standard management.

 

Yeah I'm using a PLX wideband mate and I do it all on my own. I don't really have anyone local to me who could help unfortunately. I use a combination of logging and closed loop at the moment. I have got the fuel / timing 'on the fly' adjustment box but I haven't got round to using it yet. My bum dyno is telling me it's all going well though :D

[Rob_B 0]

Cool. I wish I could help you out, it's fascinating.

I think I need to invest in some wideband monitoring, for piece of mind.

Glad the MBE doesn't have as many functions, too much to mess up!!

[Kevin Bacon 5]

Cheers mate. I did consider going up and seeing Mr Walker at Emerald and tie that in with a mini Norfolk meet, but he only maps his own ECUs. Understandable I guess. I do like reading his workshop report in PPC every month :D

 

The only other mapper I get on with who does DTA is Vince at Stealth, but he's lost focus on standalones and concentrates more on remapping OE ECUs now, so I'm kinda stuck doing it myself. I don't trust any other mappers really and don't want to fork out £500 to confirm my instincts :lol:

 

Wideband is a nice to have actually and they're well priced these days.

 

I really like this one for the OE look and Stack's reputation for reliability - http://www.merlinmotorsport.co.uk/p13961/STACK-WIDEBAND-LAMBDA-GAUGE/product_info.html

 

This is another great WB kit at a cool price - http://www.efi-parts.co.uk/index.php?productID=204

 

Yeah you can easily wreck an engine with a standalone but fortune favours the brave :D

[VAG-hag 0]

Hats off Kev & cheers for sharing.... I did start to try & digest all the figures but Saturday night may not be the ideal time!

 

Out of interest have you sold the 12v T setup? Or do you plan to? Hope to see your car again sometime 8)

[Kevin Bacon 5]

Hi mate

 

Still got some turbo stuff kicking around. Engine block, turbo, dump valve, wastegate, chargecooler etc etc..... was, or am going to sell it, yeah. Just haven't got round to it yet :D

[Rob_B 0]

I know a man who can probably map your dta for you. He has mapped my MBE and my mates emerald with no problems.

I was looking at your cam control PID loop and wondering why you were using PD rather than PI? When I'm doing closed loop PID systems at work I tend to not bother with D and work with PI. Usually get good results with just those two.

That is for proportional valves mainly.

What sort of control are you after with that loop?

Edited January 14, 2012 by Rob_B

[Kevin Bacon 5]

They're just the numbers that seemed to stop the cam advance oscillating. Cam advance is very hard to control smoothly. Even oil viscosity completely changes the behaviour. Since that map I've found 300 P and 30 I works better but I think I need more P as sometimes the target is never reached. Trouble is, as you know, the more P you pile in, the bigger the adjustment and therefore the more chance of error. I'll keep playing with it.

 

I don't fully understand PID though, so any advice is greatly appreciated :D

[Rob_B 0]

The idea with PID is you set a target value and then you the PID to maintain that value.

So the Proportional figure is what you would use to set how far above and below your target value you can go, then the Integral time is used to control the speed of reaction.

So for tight control I would set a low P and low I. This would make the valve move small amounts and often. A longer I and larger P with give a smoother operation but potentially less control.

If the thing you are controlling moves quickly I would try and lower my PI values and see if it reacts quick enough to get to my target value and stabilise.

[Kevin Bacon 5]

Cheers Rob. That's probably the best explanation of PIDs I've heard yet! What is D's job?

 

After some more testing, 300P and 30I does seem to be close. A fellow DTA friend of mine advised me not to make I any more than 10% of P, otherwise it winds up too much creating big errors. I have witnessed that behaviour when experimenting with the numbers.

 

I think I might have a faulty cam solenoid. This morning when it kicked in (80 deg water temp) it locked onto max advance (50 deg) and wouldn't budge. 50 deg advance is great around 2000rpm when you tickle the throttle, big increase in torque, but at idle it will just stall. Luckily after a few minutes it reduced to the target value. As the engine was stood for so long, I think I'll get a new solenoid.

 

The front end and side covers have to come off to replace it, so I may aswell stick the 268 cams in whilst I'm there :D

[Jim 2]

Well it'd be rude not to mate :)

[Rob_B 0]

D is derivative. It should be used to control how far past the set point the process value should go, but tends to add complication when tuning the loop. I usually stay away from it but they only way to find out is to play with it.

Do you want someone who can map dta to play with it then?

[Kevin Bacon 5]

Nah it's OK, they would only do what I'm doing, but on a dyno, which will cost me ££s.

I'll keep playing with it and will get professional help when I've conceded defeat!

 

The other way to do it is to not us a control loop and just run base PWM % instead, but that can causes issues as well.

[Rob_B 0]

When you use pwm without a control loop do you find it's not very stable?

[VR6Joni 0]

Kev, can you adjust the time constant for the pid control. If your not reaching your setpoint its too long or your p and I value are well short of what they ought to be. If you can run pid not just pi I'd highly recommend it as it prevents overshoot and oscillation.

[RedTyphoon 0]

Never had luck with PIDs. Ive had used closed loop for lambda corrections,but found it to be unstable on idle speed. For cam adjusting is used PWM base map. Not sure if its correct. What do you think Kev?

BTW.I have set frequency of cam adjusting solenoid at 200hz.

Edited January 15, 2012 by RedTyphoon

[Kevin Bacon 5]

When you use pwm without a control loop do you find it's not very stable?

 

The only area I've found a lack of a loop to be a minor irritation, is idle speed. With fixed PWM numbers, the same number could give you an 800rpm idle, or a 1000rpm idle, due to engine and ambient temp conditions.

 

The best way to control idle, I've found, is to use a base PWM which gives an idle a few hundred rpm faster than desired, use a PID loop to pull it down to the target and then make sure the "minimum" base PWM number is 2-3% below that which gives the right speed. This way you don't get wild oscillations.

 

Over the weekend I did the same with the cam advance and it's getting better. Just need to keep working on it!

 

Kev, can you adjust the time constant for the pid control. If your not reaching your setpoint its too long or your p and I value are well short of what they ought to be. If you can run pid not just pi I'd highly recommend it as it prevents overshoot and oscillation.

 

Thanks Joni, I will give that a try. You can adjust how often the PID loop looks at the error, if that's what you mean? On the cam it's updating every 0.01 second and it's adjustable up to 2 seconds.

 

Never had luck with PIDs. Ive had used closed loop for lambda corrections,but found it to be unstable on idle speed. For cam adjusting is used PWM base map. Not sure if its correct. What do you think Kev?

BTW.I have set frequency of cam adjusting solenoid at 200hz.

 

A control speed that fast will give you a narrower adjustment band. I tried 200hz at first too, but the cam didn't start to move until at least 80% PWM. I now use 80Hz and the cam starts to move at 62%, so gives a wider adjustment band. You could go lower still, say 50/60hz, but the solenoid might start to get noisy and sluggish. In fact, looking at the PWMs VW use to move the cam, they must use a low frequency.

 

So looking at your map there, I don't think your cam is even going to move at all until your 78KPA column between 2300 - 4000rpm.

 

The display I use with my DTA (further back in my thread) means I can run monitor my cam target and cam adjustment and make sure the two agree, which they do for the large part :D

 

Be careful not to open the cam much too soon (50 degrees max advance) and too soon (before oil gets up to temp) or you'll get some very weird behaviour. When it's setup right you feel a big increase in torque :D

Edited January 16, 2012 by Kevin Bacon

[Rob_B 0]

By adjusting the sampling time of the error you can get better control. You just have to make sure that your sampling time is longer than it takes for the loop to make an adjustment and settle back down.

If the sampling time is too short then the loop will make an adjustment before the last adjustment has taken effect which leads to large oscillations

[Kevin Bacon 5]

Do you know of any PID modelling software where you can create a dummy loop and see what happens to it when you start fiddling with the PIDs? I only have a vague understanding of how it works and I want to understand it better, but nearly all online guides get far too deep into the maths, which I'm crap at!

 

This is probably the most layman's explanation I've seen so far - http://www.csimn.com/CSI_pages/PIDforDummies.html

 

You were right, PI is the most common loop to use. Probably why my 300P, 30I isn't far off.

[VR6Joni 0]

Kev the maths is quite simple, however you need to know the factoring used between measured values and the output which will be hidden away in the software. And then you need to know the rate of change of the cam timing which no doubt will be not be linear but proportional to the current position and actual position difference. Then you could model it in a spread sheet quite easily. I Guess what I'm saying is unless you can find out / work out those numbers simulation is impossible.

 

Here is the basic maths that I use for my PID software loops:

 

P = Set point - Measured Point (the easy one)

 

I = P + P-1n + P-2n + P-3n + P-3n etc. etc.(where P-1n = previous sample and P-2n 2 samples previous to current, etc.)

This may be divided by n+1 to give uniform factoring between P and I. Just another unknown to add to the equation.

How many previous sums are totalled together is generally not clear and obviously determines how great the (I) constant should be. Alternatively the constant entered for I can also be the number of samples/time used giving the same effect.

 

D = Measured Point - Measured point X many samples ago.

Obviously X is another unknown. Typically X is the same as how many samples were used to calculate P.

 

Then there are your multiples:

lets call them p, i, and d.

 

The loop runs in a cycle updating every set number of samples, the output is then adjusted accordingly.

And will be something along the lines of.

Current output + (P x p) + (I x i) - (D x d)

 

Hope this helps for you guys messing with the PID control loops.

[Rob_B 0]

Kev that is about as simple as I've seen an explanation. Still takes a couple of read throughs for me though.

Joni, I can't compete with that info! All of my info is based on personal experience with no training on the subject, just trial and error! Lol!

I like that info though, will come in useful

[Kevin Bacon 5]

Thanks for that Joni!

 

What I struggle with is relating all those equations to a physical object moving. I'm fine with the mechanics, electrics and using the software, but understanding the relationship between the two, I struggle. Unless something is tangible which I can touch and see working, my brain rejects it :lol: I can see and hear the effects the PID loops are having, but I just don't what it's actually doing at a low level.

 

I don't think I'll ever fully understand it, but thanks for trying! :D

 

I guess there is no better way to control a loop, otherwise something better than PIDs would have come along by now?

 

The DTA has "Use Modified PID" check boxes for lambda and cam control and these do seem a damn site more stable than the 'classic' PID I have to say.

[Kevin Bacon 5]

OK, so some more experimenting with the cam and it's pretty much there!

 

The PID loop is now stable enough to engage at a much lower temperature. I've been monitoring it and the actual advance tracks the target advance within a couple of degrees. You don't 'feel' an overshoot until you're talking 10 or so degrees, so a 2-3 deg overshoot isn't noticed at all.

 

I'm sure I could refine it still further but for now, I'm happy with it. The key to it was using the "Modified PID (Honda, Toyota)" option. Using the classic PID, I just could not get a stable response at all rpms and acceleration rates. I don't know what DTA did to modify the PID exactly, but what ever it is, it works very well.

 

I don't have a duff cam variator. I was a plum and set the "Minimum PWM" to 60%, so the PID loop couldn't retard the cam far enough, quickly enough. 60% does nothing with hot oil, but with cold oil, 60% whacks it round to max advance (52 deg), so I reduced the minimum PWM, which sorted it.

 

 

These advance numbers give some very strong torque low down :D An LSD is definitely required! What I like about the cam adjustment is you can tailor it exactly to your preferences, to either flatten the torque curve, or make it feel peakier where you want it. I went for a fat midrange. :D

 

Edited January 24, 2012 by Kevin Bacon

[RedTyphoon 0]

Great info,as always. :)

What I dont understand is,how can you make low down torque under 2500 rev? But I have to admit the torque is immense. Had to learn how to drive it in low gears. Now Im preparing bits for low boost turbocharging. Thats just for a start Im afraid. :)

Edited January 24, 2012 by RedTyphoon

[Kevin Bacon 5]

A boosted R32 in a FWD car is insanity imo :lol:

 

How can you make torque under 2500rpm? It's what the engine was designed to do :D High compression, high flow head, very long intake runners, variable cams etc. The beauty of the R32 lump is you get that and also the 5 - 7K rpm hit as well.

[RedTyphoon 0]

Here you are Kev