Widebands That You Would or Wouldn't Buy

[HelmutVonAutobahn]

Can you please clarify the voltages and signals with the wire color? Are you talking about the wire color on the bosch sensor or some cable that connects to the sensor? There is no orange wire on the bosch 17025, do you mean yellow wire?

Yeah, I guess that is more yellow. From Bosch nomenclature:

Black = UN
YELLOW = VM // the one I called orange.
GREEN = IA
RED = IP
WHITE = H-
GREY = H+

1) I screwed up in my post, for some reason I thought the 62 ohm reason was inline with the pump cell, it is not. The 62 ohms has no impact on the available voltage to drive the pump cell. My bad.

Yes, it is in series with the pump cell. At least in section 1.1 of the LSU4.9 spec.

This wouldn't give you a number, it would give you a "can it see the pulse this short", maybe.

That was my idea. Reality is, if it's fast, what speed of actual event can it resolve, in the real world?
"t63=20ms" doesn't mean a lot to most people. "Can resolve individual cylinder events up to X rpm" means something real.

The lean spike in the AFR tracks the cycle speed as well as the resolution of the logged channels allows.

Maybe if the CAN data rate were higher...

[DelSolid]

The lean spike in the AFR tracks the cycle speed as well as the resolution of the logged channels allows.

Maybe if the CAN data rate were higher...

yeah, i know, i know.

I wish that data was recorded from the 500hz analog output line but it is what it is.

[Fred]

Zip it and it will. It's just not in the white list. How about the log file and info on the structure so I can use a tool of my choice (and making)?

[DelSolid]

Here is a csv file of the data. Should import into anything.

[toalan]

I feel kinda responsible for this mess, now

The heater is a variable voltage DC design. One of the big problems with most AFR/LAMBDA/UEGO controllers is noise induced by the PWM control of the heater. Up to 2 Amps is constantly being switched on and off which causes the ground plane to jump around and cause problems with both the measurements and analog output signals. In the new AEM unit the heater is driven by a variable DC voltage, exactly as spec'd in the sensor datasheet rather than being hit with a 12V/2A pulse train. That 12V PWM pulse train impresses on the sensor signal wires through resistive, capacitive, and inductive coupling. This is especially problematic when there is a long cable run between the controller and sensor. Using a DC signal to drive the heater removes the A/C coupling from the sensor cable run, resulting in a noticeably cleaner signal.

Using a variable voltage DC power supply for the heater circuit has been done before. I did it with my first wideband in ~2008, tech edge did it with their earlier widebands, and PLX still does it with their wideband last time I check in 2012/2013. It has been done specifically for issues noise issues, though I would strongly contend that the noise culprit is not the wire or length of it but inside the sensor as the ceramic heater is in very close proximity to the sensing parts of the sensor and there is strong noise coupling.

I moved away from that type of design as I had means to deal with the heater switching noise, I have 2 different methods in dealing with the heater noise, the first method is in my SLC design, it is also in the open source SLC Free design for anyone's inspection. I specifically chose the cypress PSOC microcontroller because it had integrating ADCs, setting the heater switching frequency to be an integer multiple of the ADC integration period will theoretically result in perfect heater noise rejection, in practice it is very close to perfect. Using the natural noise rejection of an integrating ADC requires no filters at all, and generally integrating ADCs have much superior noise performance than other topologies across all frequency ranges. The other method I use I can not talk about but it is a better method primarily because it is not dependent on ADC topology.

When you use a variable voltage supply for the heater, you face the issue of heat generation. Now I am not going to assume anything about your design, but for PLX in the model that I tested used a switching power supply to output variable DC voltage that switches at ~200khz and the thermals around the switching regulator is in my opinion too high to be worthwhile, they have a much large space to work with than in your design and they use an aluminium case. The other issue you face is economics in a broad sense, nothing beats the efficiency; thermals, space, cost, and reliability of a plain vanilla FET switching the heater with a PWM signal.

[DelSolid]

Yeah, it is definitely possible to do a switching FET for the heater that doesn't cause all kinds of big issues but I am still amazed to see that a few designs out there still do a really poor job of it. We did it because we were hyper focused on all the sources of noise in the system and not as much on BOM cost, and in the end we figured a small increase in mfg cost was worth the benefit to the signal stability. In our situation, the packaging of the extra components was the royal pain in the ass. The number of painstaking iterations made to the PCB layout to fit everything on a single board within a 52mm gauge was ridiculous. I used to think our CJ based design was tight and it had the luxury of 2 stacking boards.

[HelmutVonAutobahn]

IIRC, PLX was running their PWM at 200kHz with no high-side driver. Just a 1k pull-up. So, that FET probably got hot enough to roast weenies over.

[toalan]

My impression of the PLX thermal problem was the switching frequency, the frequency was just too high, each time the power supply switched you momentarily run a FET in ohmic region and you generate a lot of heat.

[HelmutVonAutobahn]

The two issues are directly related. The high frequency, coupled with the R/C node formed by the pull-up resistor and the gate capacitance means that the gate, and, therefore Rds, are in the midrange WAY too long. Or, more properly, for much too high a % of the time. At least a proper high-side driver would "square up" the gate drive signal. In any case, 200kHz seems excessive.

[toalan]

Yes could be the FET driver issue, but that is such a simple issue to solve for pennies worth of parts, I give them the benefit of the doubt that it was not that.

A normal switching power supply is maybe 85% efficient using value oriented parts, the sensor heater needs roughly 12 watts on average, so that is around 2 watts worth of heat to dissipate. It is a considerable amount of heat to deal with considering the typical size of a wideband controller.