Hot wire anemometry

[Vintagent]

This is probably on one of the previous 2 forums, but I couldn't find it too easily anywhere.

What do people think of hot wire anemometry as a technique for finding volume flow rates? I'm really keen on getting a bench together and I'm looking at making it reasonably good to start with, then developing it (rather than making one, then having a Mk 2 version, as it were!)

I've had a bit of experience of benches with orifice plate measurement, but the pressure drop and associated flow decrease which it caused was one of the bigger problems that were encountered. Being that there's going to be lots of little troubles, it's best to avoid the bigger stuff if possible

[98-1074649673]

No prob on bringing something current from the old forums, if anyone wants to discuss something from the "past" feel free to do so . . .

Look at the Ford MAF sensor it does just that. Someone on the board has a setup to read the MAF. Pretty neat setup. Its in another post on the forum. I played with one to use on my small engine dyno to read the flow into a running engine. I can flow it on my bench to get the readings at various volts and then run an engine and see what the engine is seeing at various loads.

Should work quite well . . . ugh something else to build!

[84-1074663779]

The problem I have with this, is that the output from the flowmeter does not correspond to anything you can really rely upon. It will need to be first calibrated on someone elses bench, to develop a voltage versus flow curve. Then assuming that absolutely nothing ever changes, you then can duplicate all the errors in the other guys bench.

So you set it all up and you find that at a certain cylinder head valve lift there is 1.034 volts or something. Then you either apply a complex multi order polynomial equation to convert it direct to CFM, or you eyeball a curve drawn on a piece of graph paper.

The whole thing is prone to errors creeping in, and not terribly convenient to use. The idea that you can just feed the electrical output into a digital display that then says 125.184 CFM, looks impressive but might be less so if you knew the flow was actually closer to 130 CFM, and yesterday it would have been closer to 120 CFM.

The beauty of an orifice plate, or pitot bench is that the output pressure readings follow a predictable square law curve. If you can prove the flow at one or more points, you can pretty well rely on the results everywhere over the whole flow range for that particular sized flow path.

A water manometer is simple, reliable, linear, and repeatable. As a retired electronics engineer, I am painfully aware of the errors that can creep into more seemingly sophisticated methods of measurement.

I use an orifice style bench with a ready excess of available blower pressure. The only really irritating feature is converting percentage flow to CFM. Soon, I hope to make up eight flow scales calibrated directly in CFM displayed behind my flow manometer.

For ease of use there will be eight scales that correspond to the eight orifice plates installed in my turret. The individual flow scales will be mounted on a drum that revolves behind the flow manometer. Changing ranges should then be pretty easy, and If I want exactly 300 CFM, I do not have to first work out how much flow percentage it should be on which orifice.

I can just wind up my blower speed to the 300 CFM line on orifice number #8 or whatever. It should be the next best thing to a digital readout. This has not been done yet, but will probably be my next flowbench upgrade.

[Vintagent]

Exactly the sort of response I was hoping for! It's always good to hear both sides of the argument...

My line of thinking was that if you can get a hot wire probe setup working well, then it is ridiculously easy to (basically) duplicate this into a turbulence probe which you can then position wherever you like downstream (usually) of a component and then find how laminar/turbulent the flow is at this point. Not only this, but you get a velocity profile which has a great deal of uses. Also, there's only one other step to find the energy contained within this turbulence (if that's what you're interested in finding), again with a variety of uses.

I think that if this route were to be followed, then it would be best to 'lift' one from an engine management system and start from there. I'll have to talk to some of my friends who are into the electronics design side of engines to get a better understanding. I'll keep you informed...

[84-1074663779]

Fascinating stuff Vintagent.

Measuring turbulence is something I have never tried to do, or even thought about. I suppose turbulence could be regarded as random variations in pressure and velocity throughout a flowing air-stream ?

Hotwire anemometers do respond pretty fast, but might not be sensitive enough to do what you want. The wire actually has to heat and cool, and thermal inertia is going to always limit how fast it can respond.

One thing that does respond fast to air movement is the ordinary microphone ! For something a bit more robust, a strain gauge type pressure transducer can withstand rated pressure and will respond to pressure variations from zero (static pressure) to beyond human hearing easily.

An ordinary MAP sensor might work, I have no idea.

Ask Bruce about this. He was having pressure fluctuations cause annoying inconsistencies in his anubar setup when used with a sensitive pressure transducer. The cure was an acoustic muffler to kill the noises in the pressure line to the transducer.

The human ear cannot detect frequencies lower than about 20Hz, and I suspect air turbulence might mostly be down there in that range somewhere in the inaudible region. But it might form the basis of a turbulence measuring and analysis instrument that could be made far more sensitive than a hot-wire setup.

A small probe connected to a pressure transducer and some electronics might do it for you ? At least initially it might be possible to see something useful on an oscilloscope.

A hot-wire setup might be fast enough, but there are other things to try as well.

[Vintagent]

Tony,

All good stuff and worth keeping in mind. The reason I thought of using hot wire probes to measure turbulence is that this is how I've seen it done in many technical papers, so all of the problems have been solved at some point- just need to find how :;):

Whilst on the subject, the turbulence contained within the exhaust gases can contain up to 50% of the energy released from the combustion process. If you reduce turbulence, this % gets cut down and losses due to pumping are reduced too. Net effect is an increase in either engine efficiency or power output!

As an aside, I did my degree in mechanical engineering and my final year project was the development of an exhaust valve. All the reaserch work (both theoretical and practical) has lead me to want to do more, hence thinking about my own flowbench !

Cheers

Paul

[Mouse]

[84-1074663779]

There is a bit more to it than just the drifting zero.

The pressure transducers themselves usually use a strain gauge, it is a mechanical device. The output is not linear, but actually follows an S shaped curve. The errors might only be a few percent from straight line but they are there. Also there is a temperature dependence that shifts both the zero point, and the gain, and that is not linear either. Also the output is ratiometric, output varies as a proportion of the applied excitation voltage, so that need to be stable as well. They also suffer to a small degree from hysteresis. Increasing pressure indicates slightly differently to decreasing pressure. An accidental overpressure event, can permanently deform the flexure, altering the characteristics. Also dropping or rough handling can permanently alter the characteristics.

All things considered a $2.50 silicon strain gauge pressure transducer is a wonderful thing, but I would be surprised if it can hold long term accuracy without frequent lab checks against a dead weight pressure reference. They are cheap and simple, but far from easy to use, if accuracy and repeatability are important.

Now that same $2.50 could make a really nice water manometer that suffers none of the above problems. It's just that I worked in a nationally accredited instrumentation laboratory for many years where these issues were very important. And I am an old cranky cynical bastard.

Anyone can build a water manometer, from Timbuktu to China, and it will read perfectly accurately, and stay accurate till the end of time no matter what you do to it.

[Mouse]

[84-1074663779]

You are quite right, if you go about it properly, those sensors can work quite well, but they will not take abuse.

The average hobbyist buying a low cost pressure transducer and hooking it up to a digital voltmeter wanting to scale it with a few resistors so it will then read directly in pressure, is in for a difficult time.

The sensors I mentioned are the basic strain gauge sensor, no compensation or internal electronics at all. They are available from Sensym and National Semiconductor, and probably other sources.

As soon as you add electronics, and start compensating for the gross errors it is a completely different thing altogether, and it is no longer a $2.50 sensor.

It is possible to pay several hundred dollars for a qualified pressure transducer with a calibration chart.

The point I am trying to make is simply that an accurate water manometer takes no trouble to make. Anything electronic is either going to be expensive or not very good, or maybe both.

For example the Sensym SX15 reads 0 to 15 psi, and quoting directly from the spec sheet, full scale output at 15 psi can range from 180mv to 360mv. So right away there is the possibility of huge scale errors.

Temperature coefficient of span can range from -2400 to -1900 ppm per degree centigrade. It can be compensated, but it will have to be done individually for each pressure transducer.

Output voltage at zero pressure typically -40mV, but can be from 0mV to -80mV.

Non linearity can be +/- 0.5% of full scale, repeatability (hysteresis) 0.5% of full scale. And on it goes.

So what is done is that the same identical sensor is fitted into a package with some electronics, and the worst of the ugly errors are tuned out. It then becomes a much more expensive sensor, but the actual measurement part is identical to the cheapies.

It can then be hooked up to a digital display, but realistically even three digit accuracy is going to be very optimistic. On the other hand a one metre tall water manometer with good lighting can be read to 1mm accuracy fairly easily.

[98-1074649673]

I have both the water manometers verticle and Inclined on the bench now plus the flow processor and they are reading the same. I only use the water meters now for reference. There is no way the digital setup is leaving my flowbench (I don't even want to give it up for an upgrade)

The processor setup will see changes my inclined meter will not show. Now I am only using a 0-2" Dwyer meter so its not showing that small of changes. I'm not arguing (not really arguing on your parts, healthy discussion) the point either of u are trying to make, I'm looking at real world testing here and I know what I'm going with . . . my shop is far from laboratory quality LOL

[84-1074663779]

Hehe, yes Bruce know exactly what you mean.

Not really an argument. I certainly agree that a digital readout is going to show up very small changes during back to back testing. It is certainly a much nicer way to go about it no dispute from me on that.

I am just thinking about the guy that wants to build his first orifice flowbench for pennies. There are far more problems to overcome than just accurately measuring the pressure differentials. At least you know where you are with water - what you see is what you get.

Even if you go out and buy a brand new mega-dollar Superflow bench, all you are going to get are fluid manometers anyway (chuckle).

[Anthony]

I was reading with interest your discussions and as part of a cylinder head modification project i have to build a flow bench. I used a Bosch MAF and a Honeywell pressure trancducer(0-28inch) and i have a calibration table for each. I found that on inital testing they seened fairly accurate and the main disadvantage is have to use 5 volt regulators and 12V supplies and the wiring is a bit messy.Anyway the whole idea of calibration is a bit new. I was hoping someone could explain how exactly a flow bench is calibrated, is there even a need for me to calibrate it as i have tables and curves. Thanks for any info

[Mouse]

Tony,

All your points are quite valid, and yes, trying to read a pressure sensor with a voltmeter is indeed an excerise in futility.

You do leave one vital ingrediant out however, and that is software compensation. The FP1 contains a state of the art microprocessor that is individually programmed to compliment the individual calibration of each pressure sensor in the FP1. The result is an accuracy for each sensor that very well may exceed the sensors expectations.

With the special attention to the mechanical, electrical and software of the FP1, I believe that the FP1 will lead the way and set a new standard for the way future pressure processors are designed and built.

John

[Mouse]