Widebands That You Would or Wouldn't Buy

[toalan]

Something seems off with the stated response times.

From my own work testing the dynamic response times of the just the Bosch LSU 4.9 sensor 17025 model, it is in the order or 50-75ms. It takes approximately 50-75ms for the the feedback line on a 4.9, black wire, to show change when AFRs changes suddenly, the wideband controller tracks the feedback signal to apply the right pump current such that the feedback signal is within certain bounds, most controllers use a PID approach and the net result is 100-200ms response times typically from the instant the gas hit the sensor to when the controller will start showing a change.

I know bosch offers a special high speed 4.9 sensor that has an advertised response time of~25ms, so even with that sensor you should not be getting 20ms response times if the tests were grounded in real world scenarios.

20ms from when the gas hits the sensor to change in controller output is in the realm of possibility if you are blasting the sensor with very high pressure gases. I regulate my gas for dynamic testing to ~40PSI, the peak pressure wave is 40PSI for a very short instant. Sensor response time is limited by the diffusion of O2 ions into and out of the small exhaust gas sample chamber inside the sensor, with high enough pressures you can brute force much higher diffusion rates and I suppose much faster response times.

Anyways, it is all food for thought, it is nice to see more people with a technical background on wideband controllers and sensors joining in on the discussion.

[DelSolid]

where is the gas going after it hits the o2? Is it venting to atmo though the holes that the o2 sensor is connected to?

Yes, there are 5 open holes around the peripheral of the sensor boss, the gas flows out those. The gas flow is low, only about 2L/min so there is almost zero pressure buildup/drop.

[DelSolid]

From my own work testing the dynamic response times of the just the Bosch LSU 4.9 sensor 17025 model, it is in the order or 50-75ms. It takes approximately 50-75ms for the the feedback line on a 4.9, black wire, to show change when AFRs changes suddenly, the wideband controller tracks the feedback signal to apply the right pump current such that the feedback signal is within certain bounds, most controllers use a PID approach and the net result is 100-200ms response times typically from the instant the gas hit the sensor to when the controller will start showing a change.

I have seen the exact same behavior when the traditional Bosch PID method is used.

I know bosch offers a special high speed 4.9 sensor that has an advertised response time of~25ms, so even with that sensor you should not be getting 20ms response times if the tests were grounded in real world scenarios.

The sensor used is the Bosch 0 258 017 025

20ms from when the gas hits the sensor to change in controller output is in the realm of possibility if you are blasting the sensor with very high pressure gases. I regulate my gas for dynamic testing to ~40PSI, the peak pressure wave is 40PSI for a very short instant. Sensor response time is limited by the diffusion of O2 ions into and out of the small exhaust gas sample chamber inside the sensor, with high enough pressures you can brute force much higher diffusion rates and I suppose much faster response times.

On my test rig, the regulated feed pressures are set to 6psi and flow meters are adjusted to limit the flow to 3 L/min for both gasses.

A 1 bar pressure sensor is set into the housing to make sure a substantial pressure change is not seen at the sensor when changing gasses. But this sensor was also used to try to approximate the solenoid switching and gas transport delay time of the test rig by setting one gas pressure & flow substantially higher than the other and comparing the solenoid activation signal with the arrival of the pressure wave signal on the sensor. We found this to be typically between 8-11 mS but teasing that out of the data was cumbersome, ultimately we left it in the quoted response time.

[toalan]

The quote of 50-75ms is not particular to PID control topology. Basically you run your controller until you reach some kind of steady state where readings are stable, you scope the feedback line (black line) and you observe changes on that line in response to the a pulse of exhaust gas. Based on the change in the feedback line you can tell when the sensor is responding to the gas.

The time between when the gas is turned on to when you see changes in the feedback line is not dependent on the control methodology, you can basically turn the controller off for that brief period of time and it would not change anything. In practice though you can not feasibly turn off the controller while observing the feedback line. If you do not turn off the controller then the lambda delta on the pulse gas has to be large and the gas has to flow very fast so to overwhelm the controller such that the natural response of the sensor is observable. If the lambda delta on the pulse gas is small or the flow of the gas is slow then the controller will be able to keep the feedback line constant, not observable on a scope, and you will end up with no useful data.

Also controllers which do not utilize PID control like the innovate stuff and my spartan controller, it is practically impossible to do this test on as there is no steady state signal to serve as a basis for seeing changes in the feedback line. On those controllers the voltage on the feedback line constantly changes, so a simple observation on a scope would not yield anything useful. In a PID system there is a steady line or a stable voltage on the feedback line so it is easy to observe changes on the feedback line.

The above can be applied to almost any controller/sensor without knowing about the intricacies of the controller or the sensor. For the stuff I designed myself, I do not need to rely on the scope method as a controller has to track the feedback line anyways so I can just run a debugger and insert extra code to track when the feedback line changes and I can turn off my controller momentarily so to expose the natural response of the sensor. Using both methods I determined the response time is 50-75 ms.

[DelSolid]

The quote of 50-75ms is not particular to PID control topology. Basically you run your controller until you reach some kind of steady state where readings are stable, you scope the feedback line (black line) and you observe changes on that line in response to the a pulse of exhaust gas. Based on the change in the feedback line you can tell when the sensor is responding to the gas.

I'm sorry, I should have been clearer. The response is of course not dependent on the PID method itself, I was using that term loosely to describe the Bosch reference design, both hardware and software.

However, It is dependent on the circuitry used to interface the sensor to the controller IC and the iteration between them, specifically the filtering required.

Do you still have your bench test setup so you do response testing?

[toalan]

I had a closer look at the picture of the your testing rig, initially I thought it was only a partial picture because a certain part which I always assume necessary was not there, however it is clear to me now that the picture you posted is the complete testing rig.

You have an issue with your testing rig.

[DelSolid]

I had a closer look at the picture of the your testing rig, initially I thought it was only a partial picture because a certain part which I always assume necessary was not there, however it is clear to me now that the picture you posted is the complete testing rig.

You have an issue with your testing rig.

Which is ????

the bubbler? the plenum? the heater?

I'm always looking to improve.

[toalan]

In the Bosch 4.9 datasheet section 4.3.1

[DelSolid]

In the Bosch 4.9 datasheet section 4.3.1

Well, 4.3.1 in my Bosch Technical Customer Information sheet (Y 258 K01 036-000e 13.07.2007) refers to vibration testing of the 4.9 sensor, which I doubt is what you are referring to.

4.3.1 Sinusoidal vibration test acc to IEC 68-2-6 test Fc
Test equipment: electrodynamic vibrator
Test:
50 to 160 Hz with amplitude ≤ 0.3 mm
160 to 2000 Hz at constant acceleration of ± 300 m/s2.
Frequency change velocity: 1 octave/min.
Test duration: 24 h to be performed in all 3 perpendicular planes.
Ambient temperature: 25 ± 3 °C.

So why do you just say it?

[toalan]

We probably have a different version of the 4.9 datasheet.

You do not install a wideband sensor such that the opening on the tip is coincident with the exhaust stream. To do so means that the sensor will constantly be subjected to much higher exhaust pressure than if the sensor was installed normally (~90 degrees WRT the exhaust stream) which will destroy accuracy and reduce sensor life. Of course the benefit is faster response times, the sensor response time is limited by the diffusion rate of the O2 ions, you can do things to increase the diffusion rate; over heat the sensor, cut the metal shroud off of it, install the sensor so that the openings as in the direct path of exhaust gases, etc... unfortunately any of those ways will either destroy accuracy and/or dramatically reduce sensor life.