New Member - have question about flow

[superlen]

Tony,

My ECU is highly configurable and I can add a virtual dash gauge on the gui and new sensor calibration faster than I can place an order, unpack the PTS DM, & install the software. Well, realistically, it may be a close tie. Once I have my box built I might need a hour or so to have CFM, as well as grams/sec @STP, and the corresponding AFM/MAF reading all showing on the screen and being logged to an excel file. I'm sure there will be fiddling with the system in getting it calibrated and testing and such, but for me its just easier to use what I have. Also, if I understand the PTS DM it's only 10bit. My ADC is configurable up to 16bits so even with the 40" samples I have coming, compared to the 16" ones in the PTS, I'll have 25 times more resolution available to me (65535 counts vs 1024). Overkill I'm sure, and I'll still need to make sure the orifice size keeps the signal up in the meaty range of the sensor and not in the weeds in order to use the effective range of my ADC. I haven't calculated the size yet & I'm a bit concerned that in order to get the signal large enough, I'll have to shrink the orifice down so much that my three vac system will not pull enough to get the flow I need. I *think* I'll have enough, but if not I can always sub in 16" sensors at any time. If I don't get sidetracked tomorrow, I'll have a better feel for this.

Sadly, I have over 500 pieces of the 250kpa sensors in inventory that are used on another product, but totally useless for this project.

Len

[superlen]

John,

Thanks for the pic. That's kinda what I was thinking only like Tony mentioned, I think horizontal would work better for my AFM/MAF measurements. I could make it vertical like yours, yet add an additional horizontal port in the left side. Then just use either it or the top depending on what I'm playing with.

I'm out of pics of my potential flow bench parts, but here's a quick snapshot of the virtual dash screen during the ECU gui development. I'll just add one gauge for the delta P & another for the CFM.

Edit: I just added a CFM gauge & updated the picture.

So your two shop vacs gave you 12" + 16" orifice DP max?

Len

[Tony]

Also, if I understand the PTS DM it's only 10bit. My ADC is configurable up to 16bits so even with the 40" samples I have coming, compared to the 16" ones in the PTS, I'll have 25 times more resolution available to me (65535 counts vs 1024).

Thats not how it works out in practice.
Ten bits is plenty of resolution when the noise level coming out of the pressure transducer (due to random pressure fluctuations) is much greater than each step of resolution.
What you need to do is average many hundreds (or thousands) of ten bit readings to get extremely fine pressure resolution way beyond just 1,024 theretical steps.

Think about how you might measure the averge depth of the ocean when there are random ten foot waves.
If you could take several hundred readings to plus or minus one foot resolution and average the results, you would have a pretty accurate reading of average ocean level.

Now someone comes along and says it would be much better to be able to measure those ten foot waves to 0.1 inches resolution, and average that ?
In practice it will make no real difference to the final result. You still need to average enough of those ten foot waves to reach some kind of smooth average value that can be displayed for ocean level.

You trade off a very high sampling speed to directly gain resolution, by averaging a very lage number of samples.

If you look at your raw data coming straight out of your pressure transducer and analog to digital converter, you will see vast amounts of noise due to air turbulence and air roar plus any acoustic motor noises.
Ten bits is more than sufficient to capture all of that.
Much more important than bit resolution is sampling speed, and having a software averaging algorithm to smooth out the chaos.

Having sixteen bit resolution, where the lower eight bits are just totally random due to noise is not helpful, especially if thedata measurement and conversion speed is very slow.
Much faster data acquistion of fewer bits resolution will give a much better result in a flow bench.
All this is not at first intuitively obvious, but quite a bit of thought and practical testing has gone into optimising the PTS digital manometer to make it as good as it is.

[jfholm]

Len,
that is nice looking. I had mine calibrated for a 12" DP. Your vacuum motors look to be a little stouter than the two shop vacs I used. I think you should not have any trouble testing at a higher test pressure and delta p

John

[superlen]

Tony,

No doubt on all you said, it's nice to meet other designers that have a similar background. I know there will be tons of air noise & I have familiarity with pulsing air streams. On my ECU for instance, in the raw MAP data I can pick out individual cylinder events & many ECU designers even go all out & time the ADC to sample only during the lull so to speak. I don't go that far as I haven't found it to be necessary after the averaging is done.

I currently sample all the channels at 12bits just to save a little conversion time. I think of ADC inputs like a triangle of speed, resolution, & accuracy. As you move towards one, you pull from the other two. I filter the different channels of course before the numbers are fed into the fueling algorithm. As for speed, If I wanted I can get continuous 12bit conversions every 4-5uS into the micro (10bit a little more/16bit a little less of course) & I'm running a 32bit ARM core @100mhz so my filter algorithms have plenty of horsepower. I won't need that much sampling speed of course, & I don't even need that in the ECU, but it's nice to have the headroom. I'll probably start with a few kilohertz on my DM project and see what the numbers look like.

John,

Thanks for the feedback on your design's air flow. My instinct was telling me Ill have enough air, but my instinct was based on a few days of googling flow meters, and probably not the most acute. It's great to get feedback from people who have good direct experience like yourself and others.

Ok, I'm off to go break down some sheet goods. I haven't received my plans yet so I'm off to get this cart assembled so I can drag the horse behind it. LOL

Len

[Tony]

These silicon strain gauge pressure tranducers have an amazingly good high frequency response, and its the scatter of individual sample readings that gives the software averaging routine something to really chew on.
Far from being a curse, this broad band noise is really a blessing we can convert to fantastically good pressure resolution.

I suggest not using any analog low pass filtering between the transducer and the analog to digital conversion. Do all your smoothing in one really big hit with a continuous rolling average routine.

It sounds like you have the electronic side of this pretty well sewn up.

[superlen]

Ok, so I got a little done today. I was hoping to be a little farther along but I had to help a friend get his daughters car back together and had to run into town for a part. I needed a trip anyway as I needed one piece of 3" PVC. Other than that everything was lying around in the barn.

Taking advice from the crowd I just built the left half of the bench. Two chambers, both 16" x 16" x 16". The WxD I chose arbitrarily at 16", the height was set by the fact that I had a few pieces of plywood from the old cabinets in the house lying about that were counter top height. I sliced off the toe kick and cleaned up the other edge and poof, old countertops -> FlowBench sides. It pays to be a packrat sometimes.

step1.jpg

[superlen]

Here are a few more pics/details.

I still need to route the ports & also drill pilot holes and screw everything together. I just assembled with a good dose of glue and the nail gun this morning.

Comments and suggestions appreciated. Also if anyone knows the diameter to cutout on the front to match Bruce's 8" deck plates that would be handy tomorrow when I cut the access ports. I was just going to undersize and use my own plywood plate for now, but if I had the correct dimensions it would save me re-cutting the port later.

The orifice plate I made is 2.000" diameter and is beveled 45 degrees to not quite a knife edge. I left maybe 20 thousandths. It's also 4" square as that is the stock I had on hand. According to the numbers it should flow around 340cfm max at 40" across it. I'm not exactly what the DC is going to be. My plan is to order an orifice from Bruce, but for now this will let me test the basic operation.

Other than the 3" piece of pvc to screw into the blower, this was all just ratholed parts.

Thanks for looking.

Len

[Brucepts]

Also if anyone knows the diameter to cutout on the front to match Bruce's 8" deck plates that would be handy tomorrow when I cut the access ports. I was just going to undersize and use my own plywood plate for now, but if I had the correct dimensions it would save me re-cutting the port later.

8-1/2"

[superlen]

Thanks Bruce.