MED 9.1 basics

Pages: 1 2 3 [4] 5 6 ... 26

« previous next »

Author

Topic: MED 9.1 basics (Read 317966 times)

prj

Hero Member

Karma: +1072/-480
Offline

Offline

Posts: 6034

	Re: MED 9.1 basics « Reply #45 on: March 09, 2014, 11:58:40 AM »

Quote from: nyet on March 08, 2014, 03:59:05 PM

1) The FR does not mention methods for tuning the PID near the MAP limit.

Pretty much the normal method applies, all you have to do is go really aggressive on the I-limiting.
That said, you should not run PID at MAP limit if at all avoidable.

Quote

If you truly think there is nothing of value outside the FR, then what is the point of any discussion of anything?

I don't think there is nothing of value outside the FR. I think that one should at the very least read both the FR (at least the schematics) and the forum.


	Logged

PM's will not be answered, so don't even try.
Log your car properly - WinOLS database - Tools/patches

Basano

Full Member

Karma: +90/-3
Offline

Offline

Posts: 192

	Re: MED 9.1 basics « Reply #46 on: March 10, 2014, 07:43:18 AM »

Finally!

It’s responding as expected Cheesy

If you recall, I was targeting the initial overshoot. I lowered the values in KFLDIMX in the region of the initial ramp-up (decreased them a lot, but after all this I now understand why) and WGDC now tapers off a lot more as actual boost approaches requested boost.

Attached are the before and after comparisons. It’s the size of the initial boost overshoot that has diminished.

KFLDIMX – stock

hPa / RPM		1000		1250		1500		1750		2000		2250		2500		2750		3000		3500		4000		4500		5000		5500		6000		6500
0	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	5.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000
300	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	16.8750	¦	15.0000	¦	14.0650	¦	17.1900	¦	21.2500	¦	21.7200	¦	22.1900	¦	23.1250	¦	25.0000	¦	26.8750
400	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	22.5000	¦	20.0000	¦	18.7500	¦	21.2500	¦	25.0000	¦	25.6250	¦	26.2500	¦	27.5000	¦	30.0000	¦	32.5000
500	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	28.1250	¦	25.0000	¦	23.4400	¦	25.3150	¦	28.7500	¦	29.5300	¦	30.3150	¦	31.8750	¦	35.0000	¦	38.1250
1000	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	56.2500	¦	50.0000	¦	46.8750	¦	45.6250	¦	47.5000	¦	49.0650	¦	50.6250	¦	53.7500	¦	60.0000	¦	66.2500
1200	¦	76.0000	¦	76.0000	¦	76.0000	¦	76.0000	¦	76.0000	¦	76.0000	¦	67.5000	¦	60.0000	¦	56.2500	¦	53.7500	¦	55.0000	¦	56.8750	¦	58.7500	¦	62.5000	¦	70.0000	¦	77.5000
1500	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	84.3750	¦	75.0000	¦	70.3150	¦	65.9400	¦	66.2500	¦	68.5950	¦	70.9400	¦	75.6250	¦	85.0000	¦	94.3750
1600	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	90.0000	¦	80.0000	¦	75.0000	¦	70.0000	¦	70.0000	¦	72.5000	¦	75.0000	¦	80.0000	¦	90.0000	¦	95.0000

KFLDIMX – v10

hPa / RPM		1000		1250		1500		1750		2000		2250		2500		2750		3000		3500		4000		4500		5000		5500		6000		6500
0	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	0.0000	¦	5.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000	¦	10.0000
300	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	19.0000	¦	16.8750	¦	15.0000	¦	14.0650	¦	17.1900	¦	21.2500	¦	21.7200	¦	22.1900	¦	23.1250	¦	25.0000	¦	26.8750
400	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	25.3350	¦	22.5000	¦	20.0000	¦	18.7500	¦	21.2500	¦	25.0000	¦	25.6250	¦	26.2500	¦	27.5000	¦	30.0000	¦	32.5000
500	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	31.6650	¦	28.1250	¦	25.0000	¦	23.4400	¦	25.3150	¦	28.7500	¦	29.5300	¦	30.3150	¦	31.8750	¦	35.0000	¦	38.1250
1000	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	63.3350	¦	56.2500	¦	50.0000	¦	46.8750	¦	45.6250	¦	47.5000	¦	49.0650	¦	50.6250	¦	53.7500	¦	60.0000	¦	66.2500
1200	¦	76.0000	¦	76.0000	¦	76.0000	¦	76.0000	¦	76.0000	¦	63.3350	¦	56.2500	¦	56.2500	¦	56.2500	¦	53.7500	¦	55.0000	¦	56.8750	¦	58.7500	¦	62.5000	¦	70.0000	¦	77.5000
1500	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	63.3350	¦	56.2500	¦	40.0000	¦	40.0000	¦	56.2500	¦	66.2500	¦	68.5950	¦	70.9400	¦	75.6250	¦	85.0000	¦	94.3750
1600	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	95.0000	¦	63.3350	¦	56.2500	¦	40.0000	¦	40.0000	¦	56.2500	¦	70.0000	¦	72.5000	¦	75.0000	¦	80.0000	¦	90.0000	¦	95.0000

Quote from: prj on March 04, 2014, 03:53:22 AM

Lower KFLDIMX to fix boost overshoot.
Load is irrelevant at this point as it is above diagnostic limit.

Thank you, exactly right.

Initially I wasn’t sure how aggressive the changes should be but after working through this with all of you, I understand what I’m changing and why it’s had this result.

I’m going to hold off tweaking the derivative. I’m really happy with this and I don’t actually want to dampen the response any further!


	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #47 on: March 10, 2014, 08:48:38 AM »

IMO your approach is not a good one, it only works if you are trying to kill overshoot in one specific rpm range. Don't believe me? Delay the start of your run rpm wise and see what happens when your boost peak is at 4 or 5k rpm

You need more Q2. The i limit should remain more or less rpm independent. You can tune DRL to compensate for any extra DC you need as flow/rpm go up.

Q0 is for transient response, Q1 is long term, Q2 is for pre-emptive transient response.

Do not use Q0/Q1 for doing Q2's job.


« Last Edit: March 10, 2014, 08:51:19 AM by nyet »	Logged

ME7.1 tuning guide
ECUx Plot
ME7Sum checksum
Trim heatmap tool

Please do not ask me for tunes. I'm here to help people make their own.

Do not PM me technical questions! Please, ask all questions on the forums! Doing so will ensure the next person with the same issue gets the opportunity to learn from your ex

Basano

Full Member

Karma: +90/-3
Offline

Offline

Posts: 192

	Re: MED 9.1 basics « Reply #48 on: March 10, 2014, 02:46:34 PM »

Hmmmm, I was pondering that.

Yes, I had just been looking at one scenario - a WOT pull from low down. I tried a few additional runs and went WOT at different starting rpm points (2000 rpm, 3000 rpm, 4000rpm etc)

Initial overshoot did increase (compared to the 2000 rpm point where I'd been focusing) as I moved to different starting points in the rev range.

WOT at 2500 RPM

WOT at 4000 RPM

But I'd say that very roughly, the biggest initial overshoot was low down and whilst initial overshoot still occurs higher up, it's not quite as large. Maybe. I could be fooling myself Huh

I originally saw a peak boost of 2650 mbar when I went WOT at 2000 rpm and peak of 2500 mbar when I went WOT at 4000 rpm. Specified was 2350 mbar.

I did look at Q2 and likewise it's spread across the rpm range. So each point in the rpm range would need to be addressed. That's quite a task Sad


« Last Edit: March 10, 2014, 02:59:40 PM by Basano »	Logged

Basano

Full Member

Karma: +90/-3
Offline

Offline

Posts: 192

	Re: MED 9.1 basics « Reply #49 on: March 11, 2014, 02:40:52 PM »

Still a bit hung up on this.

If KFLDIMX specifies the 'steady state' values, at what point is the system considered 'steady state' Sad

I actually did have a go at looking at the FR.

There are basically two distinct operating modes:

1. !B_lddy: Quasi steady-state operation with PI control which gives a relatively weak control action. Derivation of the control parameters is carried out via oscillation testing on an engine dynamometer using the Ziegler-Nichols tuning method.

2. B_lddy: Dynamic performance with PID control which gives a strong control action. Derivation of the control parameters is carried out via oscillation testing on an engine dynamometer.

These operating states are distinguished via the control error, i.e., a positive deviation above a threshold activates the dynamic control intervention and it is only withdrawn when the deviation changes sign (i.e. the actual value exceeds desired value). The transient is managed with the aim of not causing overshoot over the entire region in the quasi steady-state mode.

So if quasi steady state is only entered when actual exceeds desired, does that mean KFLDIMX (steady state) is only referred to when actual exceeds desired i.e overshoot Huh

And yet changing KFLDIMX seems to affect the WGDC before the overshoot occurs?


« Last Edit: March 11, 2014, 02:49:33 PM by Basano »	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #50 on: March 11, 2014, 03:30:50 PM »

My usage of "steady state" is for the general PID context, not FR or ME specific.


	Logged

phila_dot

Hero Member

Karma: +173/-11
Offline

Offline

Posts: 1709

	Re: MED 9.1 basics « Reply #51 on: March 11, 2014, 04:06:49 PM »

IIRC steady state is when lde < UMDYLDR


	Logged

weijie

Jr. Member

Karma: +0/-0
Offline

Offline

Posts: 46

	Re: MED 9.1 basics « Reply #52 on: March 25, 2014, 12:21:23 AM »

Quote from: Basano on March 04, 2014, 02:27:07 AM

Now my changes.

Same again, here are the maps and I’ve marked what I’ve changed.

I spent much, much too much time worrying about the relationship between KFMIRL and KRMIOP. I read some amazing posts and I’m not going to go into it any further right now since it’s been covered elsewhere far better than I could do! Suffice it to say, KFMIRL values are raised slightly, KFMIOP axis is rescaled to match the increased values in KFMIRL and values in KFMIOP on the changed axis are interpolated accordingly. I believe it’s mathematically OK.

KFMIRL - v1

4.0009	11.4375	10.9453	10.7109	10.2188	9.7031	9.3047	8.8828	8.6250	8.5313	8.5078	8.5313	8.5078	8.4375	8.2734	8.0625	7.8281
5.9998	15.7266	15.2813	15.0703	14.6250	14.1094	13.6875	13.1719	12.8438	12.6563	12.5859	12.5625	12.4922	12.3516	12.1406	11.9063	11.6250
8.0002	19.9922	19.6172	19.4297	18.9844	18.5391	18.0703	17.4844	17.0391	16.8047	16.6875	16.5938	16.4531	16.2891	16.0313	15.7500	15.4219
13.9999	32.7188	32.6016	32.5547	32.3672	31.8281	31.1719	30.3047	29.6719	29.2500	28.9219	28.6641	28.3594	28.0313	27.6563	27.2813	26.8359
19.9997	45.4219	45.3047	45.2344	45.0469	44.6953	44.1094	43.1484	42.3516	41.7656	41.2031	40.7109	40.2188	39.7500	39.2578	38.7891	38.2500
30.0003	66.6563	66.3750	66.2344	66.0000	65.8594	65.3203	64.3359	63.3984	62.5313	61.4531	60.5391	59.7422	59.1563	58.6172	58.1016	57.5391
39.9994	88.1250	87.5391	87.2578	86.7422	86.2969	86.0156	85.1016	83.8359	82.5703	81.2578	80.1797	79.3359	78.7969	78.3047	77.9063	77.5313
50.0000	109.9922	109.4766	109.1953	108.7031	108.1406	107.5781	106.2891	104.4609	102.7969	101.3672	100.3359	99.5859	99.2109	99.0469	99.0703	99.3281
60.0006	132.3516	131.9766	131.8125	131.4375	130.9688	130.3125	128.8828	126.7500	124.6875	122.9531	121.7578	121.0313	120.8672	121.1719	121.8047	122.8359
69.9997	154.9453	154.7344	154.6172	154.3594	154.0313	153.5625	152.5078	150.6328	148.3828	146.1797	144.6563	143.6953	143.5781	144.1641	145.3594	147.1875
80.0003	179.6841	179.4398	179.2805	179.0032	178.9535	179.1020	178.9808	177.8739	175.6823	172.8974	170.8025	169.3654	169.1139	169.7420	171.2305	173.4012
90.0009	211.6003	212.1133	212.3823	212.7985	213.0964	212.5937	211.7293	210.0845	208.1243	205.7168	203.7935	202.5029	202.1585	203.0555	204.5441	206.2987

KFMIOP - v1

% / RPM	500	700	800	1000	1240	1520	2000	2520	3000	3520	4000	4520	5000	5520	6000	6520
10.0078	3.3279	3.5660	3.6819	3.9017	4.1397	4.3228	4.5212	4.6539	4.7150	4.7333	4.7348	4.7516	4.8019	4.8965	5.0125	5.1498
15.0000	5.6625	5.8685	5.9708	6.1752	6.3995	6.5964	6.8466	7.0282	7.1259	7.1793	7.2113	7.2632	7.3441	7.4707	7.6111	7.7774
20.0156	8.0154	8.1833	8.2703	8.4579	8.6685	8.8852	9.1873	9.4131	9.5474	9.6359	9.7015	9.7900	9.9030	10.0571	10.2234	10.4172
35.0156	15.0833	15.1321	15.1581	15.2313	15.4663	15.7715	16.2064	16.5375	16.7740	16.9830	17.1585	17.3630	17.5674	17.7979	18.0283	18.2968
50.0156	22.1634	22.2382	22.2778	22.3740	22.5143	22.7722	23.2239	23.6282	23.9563	24.3332	24.6750	25.0076	25.2869	25.5615	25.8224	26.1215
65.0156	29.2191	29.3457	29.4083	29.5181	29.5868	29.8599	30.3360	30.7983	31.2408	31.7978	32.2876	32.6996	32.0002	33.2718	33.5144	33.7845
80.0156	36.2335	36.4670	36.5814	36.7966	36.9568	37.1140	37.5504	38.1317	38.7222	39.3753	39.9200	40.3412	40.6006	40.8249	40.9988	41.1514
100.0078	45.4575	45.7199	45.8481	46.0892	46.3287	46.5317	47.0551	47.8561	48.6252	49.3210	49.8322	50.1999	50.3677	50.4395	50.4135	50.2960
120.0000	54.4983	54.7012	54.8004	54.9927	55.2216	55.5099	56.1401	57.0435	57.9056	58.6594	59.1934	59.5367	59.6100	59.4757	59.2041	58.7906
160.0000	72.0999	72.1705	72.2092	72.2989	72.3951	72.5596	72.9223	73.5772	74.4243	75.3539	76.0423	76.5123	76.5807	76.3221	75.7678	74.9619
200.0000	86.9569	87.0487	87.1186	87.2400	87.2264	87.0242	86.8844	87.1235	87.7141	88.4918	89.0605	89.4246	89.5158	89.3507	89.0238	88.5631

noob question, how did you scale the axis in KFMIOP?


	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #53 on: March 25, 2014, 10:11:58 AM »

Quote from: phila_dot on March 11, 2014, 04:06:49 PM

IIRC steady state is when lde < UMDYLDR

That is "static" vs "dynamic"... also there is a hysteresis filter which prevents it from bouncing around (iirc 10 to UMDYLDR)


	Logged

phila_dot

Hero Member

Karma: +173/-11
Offline

Offline

Posts: 1709

	Re: MED 9.1 basics « Reply #54 on: March 25, 2014, 04:00:04 PM »

Quote from: nyet on March 25, 2014, 10:11:58 AM

That is "static" vs "dynamic"... also there is a hysteresis filter which prevents it from bouncing around (iirc 10 to UMDYLDR)

ME7.1 Mbox specific:

Yes exactly, D drops to zero and P drops to 2% when B_lddy is clear. This is "steady state" and I is the main control.

B_lddy is cleared if lde == 0 and set if lde > UMDYLDR.


	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #55 on: March 25, 2014, 05:11:24 PM »

Quote from: phila_dot on March 25, 2014, 04:00:04 PM

ME7.1 Mbox specific:

Yes exactly, D drops to zero and P drops to 2% when B_lddy is clear. This is "steady state" and I is the main control.

B_lddy is cleared if lde == 0 and set if lde > UMDYLDR.

Agreed, except I think its lde<10, not lde==0

In any case, just trying to clarify my usage of "steady state", which is for ALL conventional PIDs, including simple ones that do not have parameterized co-efficients.... that is to say, even w/o a case that explicitly turns off the D and/or P components, steady state is defined as when lde is at (or near) zero, and no longer changing, such that P and D are zero *even if they have non zero co-efficients!*


	Logged

phila_dot

Hero Member

Karma: +173/-11
Offline

Offline

Posts: 1709

	Re: MED 9.1 basics « Reply #56 on: March 25, 2014, 05:47:29 PM »

The factor for lde is 10, so lde < 10 and lde == 0 is equivalent.

I was going off the code.


	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #57 on: March 25, 2014, 06:13:14 PM »

Quote from: phila_dot on March 25, 2014, 05:47:29 PM

The factor for lde is 10, so lde < 10 and lde == 0 is equivalent.

I was going off the code.

Whoa, interesting!

Thank you for the clarification. Another reason to be wary about reading the FR :/

If in doubt, disassemble Sad


	Logged

Basano

Full Member

Karma: +90/-3
Offline

Offline

Posts: 192

	Re: MED 9.1 basics « Reply #58 on: April 08, 2014, 03:11:53 AM »

Hi guys,

I’ve been pretty quiet for a while and haven’t really done very much more than what I’ve posted. In fact, the car is actually back to stock since it went in for a service at the main dealer and I swapped from my modified ECU back to the original unmodified ECU (no problems encountered, just unplugged, swapped and cleared any fault codes).

However, I’ve been able to spend some more time on this over the last two weeks. Using IDA and the FR, I think I’ve located the following ram variables:

0x7FD71C   pvdkds_w   BoostPressureActual
0x80255C   plsol_w   BoostPressureSpecified
0x7FD830   nmot_w   EngineSpeed
0x7FEE58   rl_w      EngineLoad
0x80252E   rlmx_w      EngineLoadSpecified
0x802554   ldtv_w      WastegateDutyCycle
0x80254E   ldrkp_w   LDR-control parameter for P component
0x80254A   ldrkd_w   LDR-control parameter for D component
0x80254C   ldrki_w      LDR-control parameter for I component
0x802546   ldptv_w   LDR, duty cycle from the P controller
0x802548   ldrdtv_w   LDR, duty cycle from the D controller
0x802544   lditv_w      LDR, duty cycle from the I controller
0x80253A   ldimn      Current value for the minimum value limit I component LDR
0x802542   ldimx      Current value for the maximum value limit I component LDR

Now even with the stock ECU, there is an initial boost overshoot as you can see in the graph.

(Sorry about the graph, I know it’s a very busy picture with a lot going on. RPM and Boost on left axis, P, I & D on right axis)

Also in the graph I’ve logged ldptv_w, ldrdtv_w and lditv_w. I wanted to see how the P, I and D components contribute.

Previously in this thread I’ve targeted KFLDIMX to clamp down on the initial overshoot. I think this is a bit unwieldy since I’m basically then running at the I limit?

What I’m going to try now is the Derivative component. In the graph I can see how it increases (gets more negative) until it reaches its most negative value just before the actual boost exceeds the specified boost. I propose to increase this so it’s larger (gets even more negative).

To accommodate the whole rpm range, I’m thinking of copying the entire row of values in KFLDRQ2 from row hPa=100 to row hPa=200 and then using new values in row hPa=100 (their value would be row hPa=200 plus 20%). Make any sense?

As always, comments appreciated!


	Logged

nyet

Administrator
Hero Member

Karma: +607/-168
Offline

Offline

Posts: 12268

	Re: MED 9.1 basics « Reply #59 on: April 08, 2014, 03:27:14 PM »

Yes. Adjust Q2.

Also consider using ECUxPlot.


	Logged

Pages: 1 2 3 [4] 5 6 ... 26

« previous next »

Jump to:

Navigation

Personal Tools

November 22, 2024, 08:05:02 AM

Welcome, Guest. Please login or register.
Did you miss your activation email?

Navigation

Personal Tools

Search