204 DSG intermittent momentary power loss

I wonder if this clipping is to do with the sampling rate of the ECU topping out. 60 teeth on the disc on the end of the shaft, at 1400rpm, it needs to sample to find the on and off edges of the signal, so it makes sense that the sampling resolution might be, say, 20 samples per tooth. At 1400 rpm that’s 28khz.
I always love some detective work and theories!

There's other things wrong with that signal - cam speed is 1/2 of crank speed in all 4-stroke engines so it's not anyway matching the physical realm :geek:
 
That's quite a fascinating capture.

The Camshaft speed was "latched" at 1413 rpm, during the glitch the Engine RPM went below threshold, and immediately above the threshold. Thus Camshaft speed was "latched" again at 1599 rpm (by the way, record high value seen this far).

Did the recording miss the 1599 rpm's - perhaps not?

View attachment 210967
Yes I agree, it is a strange step and I can't think of a logical explanation for it.... o_O
 
There's other things wrong with that signal - cam speed is 1/2 of crank speed in all 4-stroke engines so it's not anyway matching the physical realm :geek:
Yes, that might actually be just VCDS scaling. E.g. the raw (binary) value for the IDE00021 Engine RPM is actually 4 times the RPM which is in VCDS divided by 4, which would actually give resolution of 0.25 RPM (this was actually reported at full resolution in some (earlier) versions of VCDS).
 
Yes, that might actually be just VCDS scaling. E.g. the raw (binary) value for the IDE00021 Engine RPM is actually 4 times the RPM which is in VCDS divided by 4, which would actually give resolution of 0.25 RPM (this was actually reported at full resolution in some (earlier) versions of VCDS).

Yeah, might have to verify this ;)

Then there's these ones... cam speed is unfortunately zero but I wonder if l1l2 in ENG113715 refers to the software levels? As in post 204 DSG intermittent momentary power loss

ENG113715,ESM_Eng_spd_l1l2_diff,0.75, /min
ENG113716,ESM_Eng_spd_unchecked,827.50, /min
ENG113717,ESM_Eng_spd_unchecked_cam,0.00, /min
ENG113718,ESM_Eng_spd_unchecked_crk,825.00, /min
ENG113719,ESM_Engine_speed,825.00, /min
 
Well, the new crankshaft sensor is fitted. I took it on the test loop that I've been using where I have always managed to get it to glitch at least 4 times.... and I got NO GLITCHES!!!! The data for the crank sensor is now nice and smooth so it looks like we may have fixed it!!
Thankyou all (especially @mmi LEGEND!) SO SO much!!! :D:D:D
 

Attachments

  • LOG-01-IDE00021_&81.CSV
    735.3 KB · Views: 5
I got NO GLITCHES!!!! The data for the crank sensor is now nice and smooth so it looks like we may have fixed it!!
Jumping.gif
Indeed, nice and smooth RPM curves now. Congrats!!
Thanks for sharing all the data - definitely a good package for future reference.

Sounds like the investment into VCDS paid off right away!
 
View attachment 211019
Indeed, nice and smooth RPM curves now. Congrats!!
Thanks for sharing all the data - definitely a good package for future reference.

Sounds like the investment into VCDS paid off right away!
No, thank YOU MMI! Great advice and guidance along the way!
Yeah it seems you can't do anything without VCDS (or equivalent) nowadays so the investment will definitely pay for itself in time, if not already as you say! :D
 
Yes, that might actually be just VCDS scaling. E.g. the raw (binary) value for the IDE00021 Engine RPM is actually 4 times the RPM which is in VCDS divided by 4, which would actually give resolution of 0.25 RPM (this was actually reported at full resolution in some (earlier) versions of VCDS).

Yeah, might have to verify this ;)

Can confirm camshaft RPM caps to about 1450 rpm also on CAN level. Both come out from CAN exactly as shown in VCDS - so actually the cam rpm should be scaled by 0.5 to get the correct speed. However, VCDS does the right thing and just shows what comes out without touching the values in any way.

Code:
    SG_ IDE00405_crankshaft_rpm m6430830 : 39|16@0+ (1,0) [0|6000] "1/min" 0x62203E
    SG_ IDE00406_camshaft_rpm m6430831 : 39|16@0+ (1,0) [0|6000] "1/min" 0x62203F

Poor resolution graph but shows the point. CAN internal 100hz rpm offset by 200rpm to not overlap crank rpm.
1692978815819.png
 
investment will definitely pay for itself in time
I’m in the already paid, many times -camp.

Just imagine VW doing a scan but refusing to do anything as there’s no error codes. Then on to an indie and couple of hours to even repeat the issue, let alone figuring out what to look at and perhaps eventually fixing. Count in the time taken to find the indie, taking van there, perhaps rentals while the van sits in the shop. It piles up pretty quick.

Imho it only takes ONE job where VCDS helps some, basically any job apart from a simplest coding change, and it’s covered.
 
Well, the new crankshaft sensor is fitted. I took it on the test loop that I've been using where I have always managed to get it to glitch at least 4 times.... and I got NO GLITCHES!!!! The data for the crank sensor is now nice and smooth so it looks like we may have fixed it!!
Thankyou all (especially @mmi LEGEND!) SO SO much!!! :D:D:D
Hey there, this thread is helping me out no end. I beleive I will need to replace my crankshaft sensor, experiencing exactly the same issue as you did. Getting to the bottom of them with the expert help of mmi.
If you remember would you mind letting me know the exact part number of the crankshaft sensor and details of how to replace? I beleive I have the same CXEB 204PS engine as you on a 2017 T32 4 motion.
Many thanks!
 
Hey there, this thread is helping me out no end. I beleive I will need to replace my crankshaft sensor, experiencing exactly the same issue as you did. Getting to the bottom of them with the expert help of mmi.
If you remember would you mind letting me know the exact part number of the crankshaft sensor and details of how to replace? I beleive I have the same CXEB 204PS engine as you on a 2017 T32 4 motion.
Many thanks!
Hi, Apologies I didnt get a notification email so have only just seen your message. I'm not sure the exact part number of the crankshaft sensor that I got - I got it from Parkers as they have a branch local to me. They only listed 1 for the van, but it's apparently common with hundreds of other vehicles. Fitting is really easy. If you take the undertray off, and then look adjacent to the oil filter, you can see the sensor recessed in to the engine. Its one plug and one 4mm (from memory) internal hex head bolt to remove and replace. Takes 5 minutes to change once you've got the tray off. Good luck! I hope it resolves your issue!!!
 
No worries at all and thanks. i dont have a parkers near me as too far south.

These come up for me on eurocarparts/gsf.


if you can let me know if either look correct then i’ll have a go myself. quoted £80 for part from a local mechanic plus an hours labour, so if easy might as well fo myself.
 
No worries at all and thanks. i dont have a parkers near me as too far south.

These come up for me on eurocarparts/gsf.


if you can let me know if either look correct then i’ll have a go myself. quoted £80 for part from a local mechanic plus an hours labour, so if easy might as well fo myself.
It looks exactly the same as the Hella part from GSF in the links you sent. Yeah you can do it yourself, no special tools or access required
 
Ben, many thanks. Part changed for an OE one. Nearly 100% sure issues are gone after a long drive. So you can be almoist sure that this thread has helped. Many thanks and best wishes!
 
  • Like
Reactions: mmi
I wonder if this clipping is to do with the sampling rate of the ECU topping out. 60 teeth on the disc on the end of the shaft, at 1400rpm, it needs to sample to find the on and off edges of the signal, so it makes sense that the sampling resolution might be, say, 20 samples per tooth. At 1400 rpm that’s 28khz.
This is very plausible, you can count a fast pulse train as a smoothed analogue signal into and analogue to digital converter and those will have sample rates. Another option that occurs to me is quantisation. When you map an analogue signal to the digital domain you only have so many steps to map it to, so for an 8 bit signal you have 256 values. So what may be happening is that there is a deliberate choice to have fine resolution in the area it's needed at the expense of data outside that range - if the action the ECU would take over 1400 rpm is a fixed strategy all it needs to know is "engine over 1400" - you can then have finer resolution in the area you need to take action based on the measurements.

These trade offs are very common with microcontroller control systems, especially when making your needs fit into a commodity part will make the costs significantly lower.

In recent years there's been a largely hidden but massive change in electronics around microcontrollers. You can now get simple ones from Paduk for 3 cents programmed in bulk, which is astonishing. At that price even basic analogue circuits are now shifting to using a microcontroller, a smoothing capacitor and a drive MOSFET as it's cheaper than building it from a couple of discrete transistors and resistors.
 
This is very plausible, you can count a fast pulse train as a smoothed analogue signal into and analogue to digital converter and those will have sample rates. Another option that occurs to me is quantisation. When you map an analogue signal to the digital domain you only have so many steps to map it to, so for an 8 bit signal you have 256 values. So what may be happening is that there is a deliberate choice to have fine resolution in the area it's needed at the expense of data outside that range - if the action the ECU would take over 1400 rpm is a fixed strategy all it needs to know is "engine over 1400" - you can then have finer resolution in the area you need to take action based on the measurements.

These trade offs are very common with microcontroller control systems, especially when making your needs fit into a commodity part will make the costs significantly lower.

In recent years there's been a largely hidden but massive change in electronics around microcontrollers. You can now get simple ones from Paduk for 3 cents programmed in bulk, which is astonishing. At that price even basic analogue circuits are now shifting to using a microcontroller, a smoothing capacitor and a drive MOSFET as it's cheaper than building it from a couple of discrete transistors and resistors.

Having thought this a bit (pun intended) I think the simplest would be to use pulse edges to trigger interrupts on microcontroller and just calculate time differences between interrupts - the difference basically tells you the angular speed directly and very accurately. In addition, some zero position detection is needed, like a different tooth pattern at a known crank position.

High frequency clock is probably readily available in the mcu anyway for measuring time differences. The result signal should be low or at least fairly constant noise as well as the actual pulse levels are not super important, just the transitions.

By using ADC the sampling frequency would indeed have to be relatively high to get crank position down to sub 1° accuracy - accentuated by the fact that at least 2x sampling frequency is needed as compared to highest frequency of interest to avoid aliasing / frequency folding of the high frequencies to the lower bands. High frequency translates to lot of samples to process - added cost.

It is indeed amazing that these days mcu’s can replace even simple analog electronics cost effectively. However, we have to remember T6 ECUs have been on a drawing table like 2014 or earlier, and even then pulling lot of design and code from historical incarnations - it’s not like manufacturers could just do all from scratch for every generation.

Fascinating subject! ;)
 
Last edited:
Hi,
I would jump to open post.

I have a similar but different issue.

During quite fast driving, when I reach about 3200-3400rpm I have momentary power loss. It's kind of wobbling.

I would like to do a dynamic log with temp.
Could you help me to properly set VCDS to log temperatures?

Thx.
Marek
 
Back
Top