Airflow meter

[Angela]

This is a brain dump - be warned: I've been sitting and having a coffee. I have a roast using Steve's MD2 and was surprised it was quite good a few days after roasting (Sorry Steve ) What I mean is that I tried a roast using his MD4 profile and got nowhere. So I started thinking... always a dangerous thing...

What is different between Steve's roaster and mine. Well, OK, he has shrouded thermocouple placed elsewhere in the roast chamber and uses an air-flow setting of 91.

Airflow means a volume of air in a unit of time. Instead of volume we could measure mass and calibrate for mass moved per unit time. But the air has a varying density day by day, hour by hour. That means the mass of air is a variable, moving through our roasting system, we cannot control.

Or could we? What if we could use a flow meter before each roast session and calibrate for the mass flow. Then when someone says they have roasted at 'calibrated 91' we could all set out airflow parameter to 91, using a calibration tool, and follow a profile and expect to get a similar outcome regardless of the local weather conditions.

The glass filled nylon impeller has a moisture content on manufacture that reduces in early use. The shrinkage changes the impeller geometry and reduces air-flow slightly. Whether the impeller re-absorbs some moisture on a day by day basis I do not know. I think not enough to cause any significant change

If this idea has legs with some of you, I wonder if Kaffelogic could produce a simple add-on optional accessory to allow air-flow calibration? I imagine a small propeller in a venturi coupled to an Arago's disc and a scale designed to temporarily fit atop the roaster in place of the chaff catcher, would be enough to do the job. What thinkest thou, Oh Exalted Ones?

And its Christmas Eve - Happy Christmas! In the mad wilds of a decaying UK we are in lockdown number 3 with airlifted food parcels arriving hourly to stave off Brexit-induced starvation. Think yourselves lucky you Kiwis and Ozzies.

[Angela]

Assumption: The air impeller drives a constant volume of air per minute for any constant impeller speed. As air density varies with weather, the mass of air per unit volume through our system varies.

On high air pressure days more air mass passes through the system than on low pressure days over a given time.

Higher air-mass means more energy needed to heat the air.
Lower air-mass means less energy to needed heat the air.

On high pressure days energy pumped, by the heater, into the air should be controlled (lowered) by the PID to prevent too much energy in the system.

On low pressure days energy pumped, by the heater, into the air should be controlled (raised) by the PID to provide enough energy to the system.

Subjective result: There is a difference between roasts on low pressure days versus high pressure days.
Why?

Question: Do we need to standardize mass flow through the system? See first post.

[fnq]

Happy holidays and kudos to Angela for thinking over a xmas coffee.

I have similarly wondered about having to slightly customise my machine parameters Versus say Steve;s shared profiles.

My thoughts.:

"Calibrated at 91 " type methods stilled aren't easily applied across different machines. My machine's 91 wind speed is different to others 91 - probably as they come off the production line, but almost certainly when they are used with varying voltages. The voltages vary across markets ( those sold in Europe cf those in NZ) but also in power availability at each site... My power is sometimes too compromised to run say D roast profile.

Air density at different elevations above sea level could also make for some differences ( no idea how quickly or differently air thins out)

My other area for thought provocation has been in determining a set temp for First crack.. I have finally succumbed to defeat in now thinking , the bean mass, the exo and endo thermic traits of different beans, the thermal inertia of different profiles and power together make for bean behaviour variance. But it is good to think outside the box at times. Cheers Darryl

[Angela]

"Calibrated at 91 " type methods stilled aren't easily applied across different machines. My machine's 91 wind speed is different to others 91 -

Air density at different elevations above sea level could also make for some differences ( no idea how quickly or differently air thins out)

The difference in machines - yours to mine, for example - is why I suggest we need a method to ensure they work the same over time. I.e. they are 'calibrated'.

Electric supply differences between us should be taken care of by the PID; that's what it is for. But within limits - and the machine checks that you are within the limits before it roasts.

According to Mrs Google, air pressure lapse rate is around 34 millibars per 1000 feet (~300m), for standard air temperature and pressures. On days when you have a column of cold dry air above you, the value will be more. On wet days it will be less.

[kaffelogic]

Happy New Year Angela. Sorry about the delay in replying, it has been that time of year.

First a bit of background.

The calibration of a roaster's air speed is done to accommodate manufacturing variation in air vent geometry and impeller mount height. The purpose is to adjust the actual RPM of the motor by a linear constant so that all roasters deliver the same volume of air per second. Fan speed should really be specified in m3/s, but in practice it is in normalized RPM. A further adjustment is made for ambient temperature. This is a simple gas law adjustment intended to deliver the same volume of heated air. This is the density adjustment visible in the logs. Factory calibration is done in a controlled environment at close to 20 deg C.

There are of course many other factors including potential non linear behaviour of the system. Foremost of these is likely to be barometric air pressure. Fitting an air pressure sensor is a future upgrade option that will connect to the USB port and this will enable another adjustment to be applied automatically. How effective this will be is still an open question.

Current options would be measure the barometric air pressure and adjust fan calibration number accordingly, or perform factory recalibration for every roast. The latter is not impractical because we make the recalibration equipment available to our service techs and we can make it available to interested users as well. Angela, I'm giving you access to our service tech zone on the forum so you can have a look and see what you think.

[Angela]

Welcome back from your holidays, Chris. Shame about the rain but at least the cherries were good this year!

I see you are well ahead of me. The commercial anemometer plus some 3D printed parts is a great idea. Thanks for the link. I will investigate getting one printed. I envisaged a cheaper mechanical anemometer working on the principle of the old car speedometer at around $10 but I suppose digital at $150 is progress, for some?

Your on-board pressure sensitive device will not be able to assure calibration though. I have the feeling that the new impeller, that I installed, has tightened up in the past year or so; resulting in a change of air throughput and loads of lid-lifting.

You suggested to me once that nylon has a moisture content on manufacture and later shrinks.
(I found the fact very enlightening. I recall I felt much the same after a Swiss tunnelling engineer, I once met, explained rocks shrink after a tunnel is bored through mountains. Sort of obvious when you come think about it - a bit like the way cheese behaves as it dries.)

In use, then, the chamber will reduce in volume and the impeller will likely dish - I would be surprised if those two movements cancel each other out, thus leaving a net change in the once calibrated air flow after some time of use.

Any board compensation for atmospheric pressure, whilst being a great addition for future control boards, will not take account of this geometry change as a new roaster settles down. And does relative humidity in the environment change the impeller geometry over time? Or is the change a once only event after nascent-moisture is driven off?

In the interim, before new boards, it may be interesting to experiment with a pre-roast calibration; have any of your team done this?

PS I'm surprised members of the Kaffelogic proletariat don't appear to be interested in this topic.

[kaffelogic]

Hi Angela

Interesting discussion.

Anemometer
The use of a cheaper anemometer was trialed, but the problem is that we do the calibration using moving beans which introduces variation in the airspeed. This requires the ability to perform statistical calculations on the data (a 15sec moving mean is calculated, with criteria of standard deviation, range, and duration applied). This means a data connection to a device with the ability to run software. The Testo product chosen is comparatively inexpensive when compared to other connected anemometers.

Going out of calibration
From tests conducted I do not believe that the Nano 7 has a tendency to go out of calibration. Originally I was concerned to see some impellers acquiring an upwards curl after a period of use, but I am now confident that this curl is completely eliminated by centrifugal forces once the impeller is at operating speed. In other words, the impeller takes the same shape when spinning, regardless of any curl it might have at rest. As a general rule they tend to stay calibrated unless they have been disassembled/reassembled.

One thing we do know about is that a small number of roasters, after assembly or reassembly, followed by calibration, followed by shipping, slip out of calibration. They then need to have between 0.03 and 0.06 added to their calibration numbers. This process is not fully understood, but might be a settling of pressure fitted components. It's not very common, and there does not seem to be any evidence of it occurring a second time for the same unit. In the handful of known cases the customers have all succeeded in making the calibration adjustment themselves and have reported no further problems. In your case a further tweak might be useful, but we can discuss that. As a rule it is not necessary.

Air pressure compensation
The air flow in the Nano 7 has not been mathematically modelled, so we are working on observations of what is potentially a complex system.

I believe that air speed will not be dramatically affected by air pressure variations. This is because the impeller RPM is PID controlled. Any variation in the work done to maintain the RPM is accommodated by the motor PID. Within limits a constant RPM will move a more-or-less constant volume of air even though the ambient air density and pressure might vary. So re-calibration prior to every roast might have a negligible effect.

I would expect a much greater effect to be the difference in thermal mass and thermal conductivity of air at different density. This affects the difference between probe temperature and bean surface temperature. Higher pressure air will have a higher density and therefore greater thermal mass and conductivity, reducing the difference between probe temperature and bean surface temperature. This will effectively make the bean surface temperature higher (with equal air speed).

So the best way to compensate for barometric air pressure variation would be with an adjustment to the profile temperature curve itself using a temperature conversion envelope, and not an adjustment to fan RPM. Adjustment values would need to be worked out heuristically.

When we come to the point where we are in a position to fit an air pressure sensor we will need to validate all of this, but since you are in a position where you have already given it a considerable amount of thought I am interested in your response.

Cheers
Chris

[Angela]

Hi Chris,

I saw you response and have been working towards a reply; forgive the delay. You bring a lot and I'll deal with your headings if I may.

Anemometer
The use of a cheaper anemometer was trialed, but the problem is that we do the calibration using moving beans which introduces variation in the airspeed.

That begs the question, why use beans in the chamber when checking airflow? You could certainly do that once, then remove the beans and check for the new airflow value without beans and then calibrate to that, surely? If experiment showed that 'with beans' wasn't a simple analogue for 'without beans' the you'd need to experiment to produce a comparison table. After that you'd know which 'with beans' results equated to 'without beans' across the setting range.

Wind tunnels use baffles to produce a laminar flow. Perhaps the bottom tube could have box baffles introduced.
I did a quick check on the Testo price in the UK; $160NZ and coupled with 3D print costs averaging $16NZ/hour counts that out for amateur users like me.

Going out of calibration
From tests conducted I do not believe that the Nano 7 has a tendency to go out of calibration. Originally I was concerned to see some impellers acquiring an upwards curl after a period of use, but I am now confident that this curl is completely eliminated by centrifugal forces once the impeller is at operating speed. In other words, the impeller takes the same shape when spinning, regardless of any curl it might have at rest. As a general rule they tend to stay calibrated unless they have been disassembled/reassembled.

That wasn't my experience. Some months ago I noticed lid lifting. First at cool-down and later during the roast and it became an increasing problem during roasting. Beans also began to get tipped on profiles that had previously been good. The change wasn't sudden by grew slowly. Initially I though it was the screen in the chimney getting clogged and cleaned after each roast. I even tried using a small coin to hold the lid down. Eventually I recalibrated and dropped 6 points.

If you say centripetal forces act to flatten the disc I would argue that it might be the centre of mass being below the disc that is causing the disc to warp upwards in the first place.

Air pressure compensation

I believe that air speed will not be dramatically affected by air pressure variations. This is because the impeller RPM is PID controlled.

I agree entirely. A constant rpm delivers a different mass/time as the air pressure changes.

I would expect a much greater effect to be the difference in thermal mass and thermal conductivity of air at different density.

Great! Now we're singing from the same hymn-sheet.

So the best way to compensate for barometric air pressure variation would be with an adjustment to the profile temperature curve itself using a temperature conversion envelope, and not an adjustment to fan RPM. Adjustment values would need to be worked out heuristically.

That is a good step. But if you are going to put in the work then you'd might as well go one further step and complete the puzzle as I see it.

Air isn't dry. Your value for specific heat capacity of air you use in your calculations will probably be for dry air or perhaps for a Standard Atmosphere. But air has varying levels of moisture; a few grams in 1 Kg of air.

Do you need to add an on-board hygrometer?

At the temperatures in the roaster, that water vapour is super heated steam - with its own specific heat and latent heat issues. There might not be much of it but it certainly has more entropy than a comparable mass of dry air.

The issues you face are also faced by air-conditioning system designers. They design systems to produce a set temperature and controlled humidity. Their systems need to work with all weather conditions.

This issue of air density and humidity has had me confused for ever. For we have a system controlled by a PID. One might reasonably expect the PID to alter the total energy input to suit the medium being heated.
Except that empirical results show differences in roasts on different days with changed weather conditions. In my mind I cannot square the circle.

Confused, of Seaford

[Angela]

Re-reading my comment it suddenly hit me that this statement is key -"There might not be much of it but it [ water vapour] certainly has more entropy than a comparable mass of dry air."

The PID is totally unable to control the thermal transfer to the bean. All it can do is set up the air temperature in the chamber and the mass flow.

Greater entropy in moist air will give a larger energy transfer to the bean compared to dry air.

Circle squared!

P.S. as an edit and afterthought:- increased air pressure will also increase the thermal energy transferred to the bean - regardless of the PID. At higher pressure there are more molecular collisions on the surfaces restraining the volume (beans) -- more collisions more energy transferred.
Circle cubed!

[Angela]

PPS

I'm on a roll today! If what I suggest is correct, we have the basis for a mathematical model of the system.

Modelling the energy transfer to the bean would allow one to see the differing effects of air pressure and humidity on the roast process. Chris, you live in a university town; a message to the careers guidance counsellor of said university asking them to circulate a request for part-time help from a post grad physicist or engineer might raise an impecunious student's interest in helping in return for some dosh; (or for something to put on a CV and for a reference).

Modelling may already have been done by others, but an academic would have access to research papers behind paywalls to find out.