Widebands That You Would or Wouldn't Buy

[Evolution-VII]

I'm very inclined in buying the SLC Free B with LSU 4.9. I'm impressed with the work done by Alan, I never heard about it until I read the recommendations here.

This product is really good? More accurate than Innovate and AEM units?

[HelmutVonAutobahn]

This product is really good? More accurate than Innovate and AEM units?

For now, I would put the current Innovate products on the no-buy list. Unless I have completely misinterpreted what I have seen, there is just something wrong, over there, engineering-wise. This is only my opinion. But, based on the cut-and-paste issue; and accumulation of errors, I have little or no confidence in them, going forward. This is a shame. Because, the LM-1 was a really great product.

The AEM UEGO that I showed is, basically, a plain-vanilla implementation of the BOSCH CJ125 reference design. Kinda boring. But, probably reliable and as accurate as anything that relies on the sensor trim resistor for calibration. It does lack the ability to verify calibration with a "free air cal". But, other than that, it seems good for what it does.

I have not played with the SLC yet. But, in conversing with Alan, here, he seems to have a solid grasp of the sensors and engineering practice. So, I am inclined to go that way. Plus, I don't think you would get a chance to talk to the actual design engineer for the AEM or Innovate products.

[toalan]

If they just copied and pasted the MTX power circuit then that makes sense why the switcher made it's way onto the LC2, driving 7 segment displays does increase the power requirements significantly.

The innovate design is very different that most other designs, they run the sensor in such a way that it is totally outside the norms of what the bosch datasheet specifies. I give them credit for thinking outside the box, but perhaps it is too much outside the box. Part of the issue is that innovate measures pump cell impedance to determine sensor temperature, where as bosch specifies that you need to measure nernst cell impedance, if the sensor temperature is not precisely measured and controlled then sensor life will erode quickly. I do not think that the temperature measurement is the primary cause of error 8s. To make sense of things, one has to consider that the LM1 was a very well regarded product, everything after the LM1 was very error 8 prone, the brains behind innovate left the company shortly after the LC1 came out. My guess is that the LC1 had minor issues with the firmware, but nobody who knew a damn was with innovate any more, so the firmware issues were never fixed and continued to propagate into newer products. When I say minor, i mean it was probably some mistake that even the best firmware programmers make, something like a "=" instead of a "==", the results are not minor but the code mistake was.

Going back to the PLX design, I think I mentioned that they switch the heater at 200khz, that was wrong, they using a switching regulator topology to output an varying analog voltage to power the sensor heater. The switching regulator operates at 200khz, the actual heater never gets switched.

[HelmutVonAutobahn]

I don’t think there is any problem determining the sensor temperature from the pump cell impedance. As long as you know what the correct target value is. I get the feeling that they don’t; because you can see, on a bench supply, that an LC-1 / LC-2 draws much less current than an AEM 30-2310 or PLX SM-AFR with the same sensor in free standing air.

Also, the best-known remedy for E-8 is to install a heat-sink on the sensor. This would make sense if the sensor were regulated to a lower than spec temperature. High and variable EGTs would heat the sensor up past its set point. Then, the temp would drop back to the lower value. The heat-sink reduces the temperature swing.

As noted, this problem didn't seem to exist in the LM-1. So, it is not endemic in their control methodology. It is just something that got copy-n-pasted into successive product generations. Same goes for the lack of “instant mode” output on all of the post-LC-1 products. It worked fine in the LM-1. It just looks like “hobbyist” level engineering.

I think that there were some design changes between the LM-1 and LC-1 that only work with the super-short sensor cable arrangement of the LC-1 ( hence the faster response times ) . Then, those changes were cut-n-pasted into designs with long sensor cables ( LM-2 MTX-L and LC-2 ), and were “unstable”. Thus, requiring reducing the output speed from about 150 s/s to 12 s/s. It could also be a problem of using a switching regulator in the analog section. If you look at their patent, it is OBVIOUS that you need a VERY clean power rail and reference to keep the system accurate. The common ripple from a switcher is probably PLENTY to upset the switching thresholds. That’s why I was shocked to see the switcher in the LC-2.



On the PLX, I don’t see an inductor on the PCB to form a switching regulator for the heater ?

[toalan]

It is very tough to say what the real issue is, but certainly you are correct that the innovate design is more noise sensitive as it relies on comparators on the feedback, nernst cell, signal to keep the feedback signal controlled, it is the reason why they can not couple an AC signal to the feedback line to measure nernst cell impedance as it would trigger the comparators. That is not to say the the normal method of feeding the feedback signal to an ADC is not noise sensitive, but it is fairly straight forward to add significant noise immunity; IIR, FIR, or plain vanilla averaging.

My PLX unit, first gen smAFR, has a through hole inductor on it.

[HelmutVonAutobahn]

PLX seems to have dropped the inductor on the GEN2 and GEN3 (same) PCBs that I have.
I think I posted pics a few pages back. That explains the high switching frequency, though.

On the Innovate design, yeah, a single transient, of just a few mv, will trigger the comparator and spoil an entire measurement cycle. I'm pretty sure that is why they have to average together 10-12 measurement cycles to get a single output value now. Hence, the measly 12 samples/sec rate; when the measurement cycle rate is 120-140 samples/sec.

[Hentai]

I would add the NGK AFRM to the list of would buys

Things I like

its a small unit compared to the afx
the update rate on display is changable
you can switch to Lambda
you can see high and low AFR
same length harness as the afx
Wider reading range ~ .61 to 1.37

thins I don't like about it
the unit is not rated as high for dealing with tempature as the afx is
the display is not as bright as the afx
the unit feels like it would be easier to damage


Things I noticed different
-different ntk sensor
-harness has a fuse in it
-comes with a different o2 bung now
-manual tells you more about the errors the display will show
-analog output is 0.4 to 4.5 volts instead of 0 to 5 volts
-wideband connector is not a 8 pin like on the afx harness
-controller connector is not the same afx one, it feels crappy
-AFX handle larger voltage differences 11v to 28v, AFRM can only deal between 12v to 16v and it shows easily

If I could get an afx that displayed lambda I would still

[toalan]

I did not know they averaged it, but thinking about, that is all about what they can do to try to address the susceptibility to noise inherit in their design. In Innovate's defense, an average such that the output changes @~100ms is not a huge detriment, the sensor response time is at best between 50-75ms to a sudden change in exhaust gas composition, this is limited by the diffusion rate. So in practice even with a perfect wideband controller with no averaging you will be looking at 50-75ms at best.

[HelmutVonAutobahn]

Yep, they are averaging. If you scope the IP line on the sensor you can see it switching at 100Hz-150Hz. The LM-1 and LC-1 will update their analog outputs for each measurement cycle.
I have seen the LC-1 respond to transients in about 20ms-25ms. Or, about 3 full measurement cycles @ 140Hz. The duty cycle variance between measurement cycles is much more consistent on the LM-1/LC-1 than the MTX-L/LC-2. The Bosch sensors ( LSU4.2/LSU4.9 ) seem to have a limit of about 5ms-10ms. The NTK sensors are MUCH slower.

So, the LC-1 has a t63 time of about 15ms. and a t90 of about 21ms

The MTX-L and LC-2 have a t63 of ~150ms and a t90 of ~178ms.

Compared that to the Bosch CJ1xx chips which have a t63(best) of about ~80ms and a t90 of ~185ms

With the LM-1/LC-1, you could resolve transition events like lean spikes on shift points or accelerator-pump shot profiles; even misfires, in some cases. I don't think you can buy anything today that can do that.


In the NTK stuff. I am not a fan of the NTK sensors. They have a few issues. The biggest one is that ,while relatively insensitive to temperature changes, they are VERY pressure sensitive. That is, very small changes in exhaust gas back-pressure cause significant measurement changes. The Bosch sensors are the opposite way around. They are relatively insensitive to pressure changes; but, are very temperature sensitive. Since the controllers can regulate sensor temperature, the overall accuracy is better with the Bosch sensors.

The NTK sensors are MUCH easier to control. The L1H1/L2H2 do not even require active temperature control at all; just a constant voltage. So, many controller manufacturers use them, mostly, for this reason. They are NOT more accurate. And, they are MUCH slower to react.

[Fred]

With the LM-1/LC-1, you could resolve transition events like lean spikes on shift points or accelerator-pump shot profiles; even misfires, in some cases. I don't think you can buy anything today that can do that. :(

I have an old LC-1 in my truck that I intend to sell ASAP. I'll weld a second bung next to the first and side by side test Alan's unit with it, first. I can definitely see tip in profiles and spikes from momentary sync losses, but to what extent, I'm not sure. Will upload some screen is from LC-1 and SLC-OEM tonight.