Widebands That You Would or Wouldn't Buy

Discuss MegaSquirt, VEMS and other non-free hardware and software here.
HelmutVonAutobahn
LQFP112 - Up with the play
Posts: 109
Joined: Fri Nov 29, 2013 12:10 am

Re: Widebands That You Would or Wouldn't Buy

Post by HelmutVonAutobahn »

OK. Looks like I'll give the Spartan II a shot :)
User avatar
bmotorsports
TO220 - Visibile
Posts: 9
Joined: Fri Sep 12, 2014 11:13 pm
Location: Richmond, VA
Contact:

Re: Widebands That You Would or Wouldn't Buy

Post by bmotorsports »

HelmutVonAutobahn wrote:I think the AFR500 V1 is the same as the NGK Powerdex ? Generally, any wideband that uses an NTK sensor is going to be limited to about 80ms, at best, in response speed. So, the AFR500 V2 may be significantly faster, with the Bosch sensor. The NTK sensors are durable but, sluggish.
The Ballenger Motorsports AFR500V1 is the same as an NGK Powerdex AFX. The Ballenger Motorsports AFR500V2 is essentially the same controller with these customer requested improvements:

Lambda Display Option
Wide Range Option
  1. 0.411-1.373 Lambda
  2. 6-20 Gas AFR
  3. 2.66-8.88 Methanol AFR
Bosch LSU 4.9 Option
Methanol Display Option
Faster analog output
All harnesses, control units, accessories, etc (except the LSU 4.9) are backwards and forwards compatible
Options are selectable at any time via internal jumpers

The AFR500V2 can control the 4mA NTK sensor varieties, the 6mA NTK varieties, the Bosch LSU 4.2, & the Bosch LSU 4.9. It's worth noting that nearly all other control systems ONLY utilize the 4.9, which is typically not the best choice for leaded fuel, various race fuels, & methanol applications.

Most notes about the LSU speed vs NTK speed are based on slow NTK controls with extreme averaging routines focusing on stability over speed. The NTK sensors are durable as hell and certainly faster than 80ms on a free air to lambda 1 sweep.

There are some posts here with rail to rail testing https://www.hptuners.com/forum/showthre ... band/page2 . The relevant images would be the sweep tests moving from 5v to 4v (free air to lambda 1). We are looking at < 10ms for the 4.9 with the X-Series controller & the AFR500V2. We are also looking at < 10ms on the AFR500V2 with the NTK Production Grade (4mA) & NTK Calibration Grade (6mA) Sensors. These were quick tests I put together to benchmark the V1 vs V2 and people kept asking about the new "fastest" wideband according to their chart. We found the X-Series unit to be VERY fast and we benchmarked the AFR500V2 very well against it. The AFR500V1 was up to 10x slower on the analog line, focusing much more heavily on steady-state averaging than responsiveness. The AFR500V2 analog is now both dramatically faster and significantly faster than the display.

I am hoping to perform more thorough & better controlled testing in the future but at this moment, we are in peak season and we are working our butts off just to keep up :)
Ballenger Motorsports
sales@bmotorsports.com
toalan
Wideband Wizard
Posts: 241
Joined: Tue Jun 10, 2008 2:53 am
Location: Toronto Canada
Contact:

Re: Widebands That You Would or Wouldn't Buy

Post by toalan »

I contend the test rig they use is very flawed, with that said. Based on the results of the chart http://aemelectronics.com/files/info/AE ... _Chart.jpg, I would estimate that my Spartan 2 unit is between zeitronix and daytona sensor unit, so maybe 50ms if Spartan 2 was subject to that same test.

In my own test rig which I contend better represents a real world installation of a wideband sensor, Spartan 2 benchmarks around 150ms

Based on my experiments and testing, an O2 sensor installed into the exhaust of a real engine gives at best a response time of 50ms to 75ms, that is the sensor itself, the wideband controller will add it's own delays. So in the case of my 150ms time of Spartan 2; 50-75ms is due to the sensor and the remainder is due to my controller.

The limiting factor in response time is the diffusion rate of exhaust gasses through a diffusion barrier into the sample chamber of the wideband sensor. The gas that makes it through the diffusion barrier is the sample of exhaust gas that gets reduced/oxidated by the pump cell to derive the lambda.

My Priority when designing Spartan 2 was reliability, it is why I use all high temperature -40C to 125C automotive grade components.

My 2nd priority was cost, high temp automotive grade parts are not cheap, anyone who is in electronics knows that most components are not available in automotive grade or high temperature grade, components that are both high temp and automotive grade occupy the highest price tier (not counting exotic stuff like for medical, military, space). The only way to use such high quality components is to reduce the part count of my wideband controller, and Spartan 2 is lowest part count wideband controller on the market. I spend my money on the best components, but since there are so few components on Spartan 2 it is practical for me to use the best quality components.

My 3rd priority is accuracy, in my design there are very few external analog components. The most important external analog component is the 61.9 pump current measuring resistor, a 1% tolerance in this resistor directly translates to a 1% error in the measured pump current and pump current is what lambda/AFR is derrived from. In Spartan 2 the 61.9 measuring resistor is 0.1% tolerance. I chose to use PSOC1 for Spartan 2, one of the great features of PSOC is that it has programmable analog, you can build amplifiers and all types of crazy analog peripherals using code. I used the PSOC analog blocks to create an instrumentation amplifier as a front end to the ADC, which amplifies the voltages across the 61.9 pump current measuring resistor before it gets sampled by the ADC on PSOC. Other designs will also have a front end amplifier before the ADC, but those are typically made from a single op amp rather than using multiple op amps to create an instrumentation amplifer

My Last priority was response time. I say this because in my design fast response time is a natural consequence of my design choices. PSOC has integrating ADCs, I set the integration period to be an integer multiple of the heater switching frequency. The heater on the LSU 4.9 produces noise inside the sensor everytime it is switch on/off, this noise is very invasive and overpowering. By using an integrating ADC with the correct integration period, you will have a theoretically perfect rejection of heater noise, and in practice it is essentially perfect rejection. In power systems, it is equivalent to using an integrating ADC with an integrating period of 1/60 second to measure accurately the average voltage/power/current of a 60hz electrical circuit. Also integrating ADCs are generally much more noise immune to noise across all frequency ranges WRT popular ADC topologies like sample and hold ADCs. The alternative to using my method is to use a heavy filter, either as an external circuit or on the microcontroller using a heavy IIR/FIR/Averaging filter which will have a detrimental effect on response time as the filter greatly delays the signals coming from the sensor to the controller. This method I use can be argued that is has a bigger contribution towards accuracy as noise performance of spartan 2 is much better than any comparable product on the market, but I think the contribution so fast repsonse times is of greater significance.

Products that use PID control of pump current, Spartan 2 should be faster them as spartan 2 also use PID pump control but without the need for heavy filters to remove noise which slows response time dramatically. I know the AEM and Innovate units do not use PID pump current control, so their response time is faster than Spartan 2. PLX uses PID Pump current control and is faster than Spartan 2, but IMHO that product is full of very bad design choices made just to hit fast response numbers before.
HelmutVonAutobahn
LQFP112 - Up with the play
Posts: 109
Joined: Fri Nov 29, 2013 12:10 am

Re: Widebands That You Would or Wouldn't Buy

Post by HelmutVonAutobahn »

All good stuff.

So, you don't use a PID pump current. And you don't use a constant current PWM ( like Innovate and AEM ) ?
Something in between ?

This part, though...
Based on my experiments and testing, an O2 sensor installed into the exhaust of a real engine gives at best a response time of 50ms to 75ms, that is the sensor itself, the wideband controller will add it's own delays.
Doesn't jive with the ability of the old LC-1s to pick out individual cylinder AFRs from a collector at low RPM, by synching the scope to the #1 spark. I have watched this be done. And you can see the adjustments made to the idle mixture of individual carbs, up to maybe 2000RPM. That requires a response time of about 25ms. But, that was the old LSU4.2 sensor.

The trace that DelSolid posted on page 20 showing their WB responding to the exhaust gasket leak would have required a complete system latency of less than 20ms, including the sensor. So, in the real-world, it seems like the huge flow/velocity of the exhaust gas changes out the gas at the chamber entry really fast ?

I recommend an experiment. With the sensor in pure CO2, feed the pump a squarewave of +/- 1mA and, watch for the latency in the Nernst voltage signal. That SHOULD be the ultimate limit for response time.

Could also do a tri-state test to determine the gas diffusion speed. i.e. pump down the chamber, go HiZ on the pump, and see how long the Nernst voltage takes to recover.
toalan
Wideband Wizard
Posts: 241
Joined: Tue Jun 10, 2008 2:53 am
Location: Toronto Canada
Contact:

Re: Widebands That You Would or Wouldn't Buy

Post by toalan »

If you could transport exhaust gas instantaneously into the sample chamber, the nernst response to it is almost instanteous, I have measured it to be less than 25us. My gut tells me that it is about 5us.

The issue to me is the transportation of the exhaust gas into the measurement/sample chamber.

I actually have done that Hiz test, though it was not to monitor the response time of the nernst voltage as a proxy to gas diffusion speed. Looking over my mental notes, the recovery is very fast.

Looking at the sensor construction, it seems to me that Bosch goes out of their way to reduce the flow/velocity of exhaust gas entering the metal part of the sensor. It probably has to do with controlling carbon buildup. If you ever cut the metal shroud of the sensor and test it, the sensor is super fast. I would hazard to guess that anyone experiencing faster than say 50 ms has installed their sensor in such a way that the gas blows directly into the face of the sensor, if a sensor is installed in such a manner then you might as well cut the metal shroud off the sensor, you will get faster response times. Both ways will guarantee that in hours or days your sensor needs to be replaced.
toalan
Wideband Wizard
Posts: 241
Joined: Tue Jun 10, 2008 2:53 am
Location: Toronto Canada
Contact:

Re: Widebands That You Would or Wouldn't Buy

Post by toalan »

HelmutVonAutobahn wrote:All good stuff.

So, you don't use a PID pump current. And you don't use a constant current PWM ( like Innovate and AEM ) ?
Something in between ?
Spartan 2 used PID pump control. Spartan 2 should be a bit faster than the average wideband that uses PID pump control. My prior post, I gave the impressions that it is as fast as a PID pump control wideband could be, but thinking back over it, that is not the case. I forgot that I use a 4.16v internal bandgap ref for the analog parts of the PSOC, typically most widebands will be able to drive the pump cell to 5v but mine can only drive to 4.16v. So that leaves some performance on the table. The reason I used the 4.16v is because it is an internally generated and compensated bandgap voltage and I gain ADC accuracy by forgoing full 5v operation.
HelmutVonAutobahn
LQFP112 - Up with the play
Posts: 109
Joined: Fri Nov 29, 2013 12:10 am

Re: Widebands That You Would or Wouldn't Buy

Post by HelmutVonAutobahn »

I don't understand why would you ever want to drive the pump harder than what is required to offset the current oxygen surplus/deficit ? Going from free-air to lambda 1.0 requires cutting the pump drive current to zero. Do you go negative ?

Since the current is the actual measurement value. What is to be gained by overdriving it ? Especially, if you can't convert the overcurrent to an actual AFR value ?

With a pump cell impedance of around 60ohm, the maximum current of, maybe, 4ma gives a maximum voltage of 240mv. Maybe 300mv on the rich side.

Putting 5v ( 2.5v really ) into the 60ohm pump cell would dump 42mA through the pump. That would probably damage it. Not to mention reducing the Zirconium Oxide. That probably happens at less than 2v ?

Reading the AEM and Innovate patents, they both use a switched constant current of about +/- 5ma which gives a drive voltage of about +/- 300mv. Which matches what I see on the scope ( ~600mv p-p ). Although, the AEM is wired ass-backwards.

It would be saver to drive it from +/- 1v rails. Then, you couldn't damage even a cold sensor with overvoltage. And, you'd still have plenty of headroom.

Or, I am missing something?
toalan
Wideband Wizard
Posts: 241
Joined: Tue Jun 10, 2008 2:53 am
Location: Toronto Canada
Contact:

Re: Widebands That You Would or Wouldn't Buy

Post by toalan »

You also have the calibration resistor in parallel with the 62 ohm resistor.

But you are right, 4.16v is more than enough. I still have the conviction inside of me that 5v operation is more desirable, though I can not think of why.
HelmutVonAutobahn
LQFP112 - Up with the play
Posts: 109
Joined: Fri Nov 29, 2013 12:10 am

Re: Widebands That You Would or Wouldn't Buy

Post by HelmutVonAutobahn »

You also have the calibration resistor in parallel with the 62 ohm resistor.
Right. That's about another 40ohm. I didn't think about that because the Innovate and AEM designs don't go through that pin. Still about 1v p-p. Maybe 1.2v

Maybe you need the extra headspace for the amps ? Is their performance near the rails is not so great ?

It will still be interesting to see how fast a CLEAN analog PID can go. The PLX is reasonably fast. But, by the time you apply enough filtering to get a usable signal, they are pretty slow.

I would hazard to guess that anyone experiencing faster than say 50 ms has installed their sensor in such a way that the gas blows directly into the face of the sensor,

Looking at a Bosch sensor, it looks to me, like it is designed to have gas sucked out the tip hole via the Venturi effect. And have the sample gas replaced by flow into the peripheral holes. Blowing straight into it would put equal pressure on all of the holes. Thus, reducing/killing flow.

I'd say, the fastest installation position is at 90deg to the flow. With the peripheral holes flush with the tubing wall. Or, maybe extended, slightly, towards the center, to catch higher velocity.

Plus, I've seen the individual cylinder thing done w/ the stock sensor position ( LC-1 ). Not sure how one could physically install a sensor inline with the pipe. Maybe at a 90deg bend?

I think, more likely, the exhaust gas velocity/turbulence, especially under load, is just so high that it overwhelms the baffling.
The issue to me is the transportation of the exhaust gas into the measurement/sample chamber.

I actually have done that Hiz test, though it was not to monitor the response time of the nernst voltage as a proxy to gas diffusion speed. Looking over my mental notes, the recovery is very fast.
You don't want to transport it "into the chamber" you want to deliver it to the diffusion orifice. Basically, at the surface of the element. That's where the diffusion-vs.-pump-current fight happens. The "very fast" recovery indicates that this is not the bottleneck. As is reinforced by the data from running the element "bare-back".

The ultimate limit for the sensor is probably within 1ms of when the test gas hits the surface of the element.
Daemos
TO220 - Visibile
Posts: 14
Joined: Sat Apr 18, 2015 4:27 am

Re: Widebands That You Would or Wouldn't Buy

Post by Daemos »

How is the heater control of the new innovates? Is it now on par with AEM?

The only frustrating thing I have with the AEM units is that you need to wire in your own serial out (not a difficult thing to do) but it would be nice if they had a snap on clip like the innovates do.
Post Reply