高清日本在线成人免费视频-高清视频一区二区三区-高清一级-高清一区二区三区-高清一区二区三区视频

 
實戰(zhàn) 6:一次一位


The car had an intermittent drivability complaint, bogging down under acceleration with no MIL flagging. The customer insisted that it wasn’t just intermittent; it was very intermittent. First things first: I connected the scan tool to see if any DTCs had been logged. It was here that this investigation took a sharp turn.

OBD socketThe scan tool oddly came back with No Communication Possible, and whenever I see this I automatically think, what have I done wrong? It’s only natural to think this because computers don’t make mistakes, do they? I ran back through everything but it was still the same.

I checked other modules and quickly noticed No Communication on a lot of them. This is typical of diagnostic line failure, and already a mental game plan was beginning to form. Before doing anything else, I made the customer aware of what I had found and that this was a vital investigation in itself before I could begin to address his original complaint.

If I am dealing with a diagnostic line failure, then a cabling short to battery negative or positive is the most likely cause. I know from past dealings with this make of vehicle that nearly all connected modules run to one pin on the OBD socket, which isn’t going to help things because now my suspected fault could be anywhere on the car.

It was time to hook up the oscilloscope to get a clearer picture of the fault. Monitoring the diagnostic line electrically is a little awkward on this model, to say the least, so I opted for a known connection at the back of the radio which actually could help the diagnosis.

connecting the scope

In this picture I’m connecting to a module away from the OBD socket and my next move will depend on what I capture. Because I’ve already decided that other modules apart from the engine are affected, I’m anticipating that every capture I get here will be the same for the rest of the diagnostic line and its connected modules. Initially I look for a voltage level present, usually near battery voltage, and then as the scan tool attempts a handshake I look for a pulsed ground in a code sequence. Because each module shares the diagnostic line, only one will respond to a certain code given by the scan tool, leaving all others ignoring it. Enough theory — let’s have a look.

For a diagnostic line problem, I like to take a wide view and then zoom in later. This helps to see the scan tool handshake, which can be lengthy until the module responds, and also any abnormalities that might be lurking.

 

waveform 1

On the face of things there is not a lot wrong. We see a good handshake structure, with good high and low voltage levels present. It also tells me that I’m not dealing with an open circuit problem between the scan tool and the modules. What isn’t clear at the moment is the integrity of the module’s response, but this could be due to the PC struggling to display the fast dialogue reply with the long time base. Let’s zoom in for a better look.

waveform 1a

 

Now things don’t look so good. The highlighted section which is supposedly our module reply is an immediate cause for concern. There’s something trying to break through but it certainly isn’t binary. It looks more like a sawtooth pattern. The same signal “arch effect” seen in the module response can also be seen in the rising edge of the handshake. This is potentially a big problem and worth pointing out.

waveform 1b

In binary signals, there is always a threshold of operation, and the voltage doesn’t have to be full on or full off for a module to recognise a valid binary bit. There is a crossover point where a module understands the difference between high and low states. If noise spikes don’t cross this threshold, then the module won’t see them and detect the wrong state. Here I’ve chosen a threshold of 5 volts as an example.

binary thresholds


waveform 1c

In the capture 1b you can see that, because the rising signal edge has an “arch effect” as it returns to rest, it will have an effect on the rise time and thus the overall handshake time from the scan tool. I recorded the rise time here to be around 23 ms (0.023 seconds), which I don’t consider normal, but the time between state changes of 1.007 seconds wasn’t enough to prevent the module from responding, however wrongly, to the handshake.

So what’s causing this? At the moment I’m not considering a hard wiring fault, but rather leaning towards a module problem. The described “arch effect” in the signal appears very controlled, neat, and clean. This type of control is indicative of a resistive-capacitive (RC) circuit, but has no place in binary signalling. This is an issue at an electronic level.


 

 

 

waveform 1dOn closer inspection of a falling binary edge, we can see a similar characteristic. I would have expected the binary structure to be much more definite than shown at this time scale.

 

 

 

 

 

 

 

At this point, I took a step back and jotted down some ideas in diagram form. This has helped me in the past to spot problems that were not originally anticipated. This time, the most likely cause I could come up with was as follows:

capacitance leak

I’m suspecting a capacitance leak into the diagnostic line, and that the capacitance is already connected to ground potential, or certainly close to it. Here’s my reasoning.

The diagnostic line voltage at rest is derived from a high-resistance source inside the connected modules, including the scan tool that is now also connected to the line. Only a very small amount of current is needed to pull this voltage down to 0 V; that’s all the scan tool does. With the line voltage at rest, our suspect capacitance will already be charged. As the scan tool handshake pulls the line voltage down, the capacitance is also forced to dump its stored charge to ground, but not without putting up a fight first when we see the small exponential curve at the base in extract 1d. Now the capacitance is at near ground potential for the duration of the binary low state. When we see the binary low state released in extract 1c, our suspect capacitance charges up and once again we see the characteristic exponential curve. Notice how the two binary edges are affected to different degrees. This is because the scan tool ground command comes from a low-resistance source pulling the signal down at the beginning, while the diagnostic line has a higher resistance pulling the signal back up at the end.

Of course, none of this theory is any use whatsoever, as I’m still left with the question: where’s the problem?

ISO connector

The engine module is a bit of a pain to gain access to under the wiper scuttle trim, so I started the search within the passenger cabin. I stripped the underside of the driver’s dash panel to have a look around. A Convenience Unit lives behind here, and then I planned moving on to the instrument panel. Then it dawned on me that the radio was obviously a module too...

I pulled out the power ISO connector and retried the scan tool: Connection Established.

Incredible! It was staring me in the face. The whole diagnostic system was now back in operation.

 

 

 

waveform 2

 

Capture 2 was a segment of the handshake now: perfect.

waveform 2a

 

Capture 2a was taken with the engine module sending a full data stream to the scan tool. With the radio connected (fault present) the signal didn’t stand a chance of developing a full “bit”, because our stray capacitance stored enough energy to keep the diagnostic line potential quite high. We come back to time constants of RC circuits, where a capacitor is allowed to charge up from a power supply and then discharge into a load.

electron flow

 

The scan tool handshake is long in comparison to the full data stream, and it managed to create a defined binary bit without letting the stray capacitance interfere too much.

I isolated the diagnostic line connection from the radio and informed the customer of the situation. He didn’t seem too bothered that his radio was missing a diagnostic line.

After all this, there were no DTC’s Logged in the engine control unit.

 

其它輔導(dǎo)資料 >

 

  400 999 3848 | sales@hkaco.com
  廣州虹科電子科技有限公司版權(quán)所有 | 科學(xué)城潤慧科技園C棟6層
汽車診斷示波器
汽車示波器
模塊
軟件
附件
資料庫和培訓(xùn)
  輔導(dǎo)資料
  波形庫
  培訓(xùn)課程
  應(yīng)用案例和文章
技術(shù)支持
產(chǎn)品系列 公司介紹 培訓(xùn)課程 新聞快報 聯(lián)系我們  
主站蜘蛛池模板: 国产日韩欧美一区二区三区在线| 98精品全国免费观看视频| 亚洲精品国产福利在线观看| 久久精品国产99国产精偷| 成人久久网站| 欧美桃色| 国产精品亚洲片在线花蝴蝶| 亚洲人成在线免费观看| 女人a级毛片| 在线视频免费观看a毛片| 国产精品乱码免费一区二区| 成人午夜在线观看国产| 麻豆免费视频网站| 香蕉网站在线观看| 91成人免费福利网站在线| 久久精品黄色| 亚洲综合精品一二三区在线| 免费看黄色小视频| 国产日产欧美精品一区二区三区| 国产成人精品免费视频大全麻豆| 亚洲人成亚洲人成在线观看| 一级片毛片| 日本黄在线| 亚洲免费视频网站| 成人性开放大片| 欧美爱爱网站| 国产成人精品日本亚洲专一区| 美女视频黄a视频全免费网站一区 美女视频黄a视频免费全过程 | 欧美亚洲人成网站在线观看刚交| 精品国产不卡一区二区三区| 欧美日韩国产在线观看一区二区三区| 一区二区三区四区日韩| 小明看看在线视频| 中日韩美中文字幕| 亚洲18色| 亚洲国产一二三| 欧美日韩亚洲一区二区精品| 国产成人v爽在线免播放观看| 国产精品自拍在线观看| 美国黄色片视频| 黄色欧美|