Public Lab Research note

using soap bubbles for pump calibration

by mathew | June 04, 2015 01:32 04 Jun 01:32 | #11949 | #11949

What I want to do

I want to see if @davidmack's suggestion that a graduated cylinder with a soap bubble traveling along it can be used to calibrate the flow rate of a pump.

My results

My goal is to set a DIY formaldeyhide testing kit to 0.3L/minute of airflow. I chose a 250ml graduated cylinder, with the intent of watching the bubble traverse 200ml of its height in 40 seconds. I drilled a 3/8" hole in the base of the graduated cylinder and screwed in a threaded 1/8" brass barbed fitting, first to tap some threads in the cylinder, and then a second time with some teflon tape around the threads to seal the threads. I didn't use aquarium sealant as @davidmack suggested because the graduated cylinder is polypropylene and glues won't stick to it readily. to test the airtightness I filled the cylinder with water and waited a few minutes to see if there were any leaks.


I then filled a tray with an unmeasured mixture of a lot of dish soap and water. This bubble solution was terrible and wouldn't hold a bubble. I then googled "bubble solution" and made a mixture of 1/4 cup soap, 1/2 cup water, and a teaspoon of sugar, and of course forgot to save where I got that recipe.

I found that it was important to evenly coat the inside of the cylinder with bubble solution, otherwise the bubble in the cylinder didn't move straight down, and would get hung up on the walls and become lop-sided. I mitigated this situation by dipping some newsprint in the bubble solution and putting it down the tube at a few different points. I then let the bubble solution settle to the bottom of the cylinder and poured it out. This process needed to be repeated every half hour or so.


Running a test

I opened a formaldehyde tube and put it into the system and drew air into the cylinder, through the tube, and into the pump, the configuration that @nshapiro laid out.


IMG_20150603_154542.jpg (these images credit:@warren)

To prepare the cylinder I dip the top of the cylinder in bubble solution, and then draw a little air into it until the bubble is into the cylindrical part of the graduated cylinder, turn the air pump off, and wait for the bubble to settle into a flat shape. This is an important step to get a flat bubble.



I always give the bubble some travel distance above the mark I'm measuring from so the pump has time to warm up. Then I start the pump and wait for the bubble to hit the mark I'm measuring from, start a stopwatch, and stop it when the bubble gets to the second mark. With the 250ml graduated cylinder, these two points are at 250ml and 50ml. I didn't put in a dropout, or a sealed chamber with a break in the line to prevent bubbles from entering the test line, as @davidmack suggested. I just turned the pump off before it got to the inlet and then popped the bubble with a stick.

Sources of error

My different measurements of the same flow setup have a 300ms (milliseconds) range. For instance, 39.98 seconds and 39.65 seconds. were the high and low of 5 different trials. Given that human response time is somewhere between 100-500ms, this 300ms range is inevitable. In a 40-second time trial, it introduces roughly 1% variability.

Other sources of variability are that the graduated cylinder is only marked to 2ml accuracy, and is rated for its current volume at 20 degrees Celsius. I conducted these tests at 22 degrees. Also, the extra fitting for the cylinder may cause some changes in airflow.


The bubble method seems to work pretty well! Next week I'll have a high-precision NIST-traceable pump to compare it to, and will therefore have a better understanding of its accuracy. While not the fastest means of measuring, I was able to tweak a small flow control valve, and over about 25 tests and an hour of testing, I set the pump to the 300ml/minute flow rate I sought, to within a few percentage points of accuracy.


Looks like you had problems breaking the tip of the tube. What process did you use?

I am really curious to find out how much flow variation there is between formaldehyde tubes. Breaking the tube in different places on the taper will vary the restriction. I have only observed 2 measurements so far so my sample size is small. I believe that the flow rate changed over the first minute or so and then became stable. I could propose that the chemical reactions in the tube caused this variation or some settling of the packing material is occurring.

A pipette might provide a more accurate bubble measuring setup. It has an opening in both ends, is available in multiple volumes, and is available cheaply as a disposable item.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Looks like a nice setup! For bubbles, the cheap ready made option would be any liquid bubble--the kid's toy to blow bubbles. @danbeavers I think a pipette is a great idea for this low flow measurement. This is a very old, proven method for measuring flow but like all primitive methods it takes some tinkering to perfect. I also prefer to use an actual stop watch--not a phone app, which can be slower.

RE: breaking the tip at the tube consistently shouldn't effect the flow if you have a strong enough pump. If you have a pump that normally runs at 5 lpm without a valve and then you add a valve to restrict it to 0.3 lpm I think you'll find the flow is fairly consistent even with moderate loading. The high differential pressure across the valve creates a critical flow situation. Not all pumps designs can be restricted though--some will over heat. What is important though, is to warm up the pump before you start testing or measuring the flow. Often the pumps need a few minutes to settle into a constant temperature. Some pumps never reach a constant flow and will ever so slightly cycle in flow like a sine wave.

A sampling protocol might look something like:

  1. Turn on pump and let run for at least 5 min
  2. Attach dummy tube (a used tube) and set flow rate
  3. Attach sampling tube and record start time
  4. Measure sample flow and record start flow rate (don't bother adjusting--should be close enough to previously measured)
  5. (optional) before the end of the sample, measure sample flow and record end flow rate.
  6. Remove tube and record end time
  7. Turn off pump

then sample volume = (start flow rate + end flow rate) / 2 x (end time - start time)

for flow field measurements I would use a glass rotameter that was calibrated in the lab with the graduated cylinder technique. the glass rotameters aren't cheap but they're less expensive than the automated methods and they're quite durable (and never require charging!). There are many makers of these and the better ones are sold by Matheson (order without the valve):

They do make tube tip breakers though. I used to use the Drager version and while the $50 cost is shocking, if you're doing a lot of testing it makes it much easier and safer. It's basically a shallow hole with a ceramic edge that you stick the glass end in and twist the tube to score the glass. Then you stick the end of the tube in a deeper, crooked hole and snap off the end.

Reply to this comment...

the pump's output seems fairly consistent without a warm-up.

I wanted to see if the setup retained its configuration from yesterday, so I ran three measurements from a cold start, and then three more at the tail end of a 40 minute warm-up run. the three measurements before the warm up were 40.06, 39.98, 40.13. The three measurements after the warm-up were 40.23, 40.05, and 40.32-- which is to say any change in flow is lost in the noise of this measurement method.

Reply to this comment...

Can i get a 250ml pipette? My choice of graduated cylinder volume was based on a desire to make the time period I measured long enough that my reaction time would be about 1% of the time period. I could use a smaller, more accurately graded glass pipette but I worry I'm increasing the accuracy of my volume measurement at the expense of the accuracy of my timed measurement.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

I broke the tip off the tube with a pair of pliers. I do need a proper tip breaker-- my method left a good mess.

Reply to this comment...

@davidmack I don't know how much vacuum a 5lpm pump will produce but it needs to generate "below 7psia" to get sonic or choked flow. An aquarium pump is only going to produce about 2 ft of water pressure and 7psia is around 200" of water. See comments in

@mathew The "accepted" method is to score and then break. :-) The "traditional" method is to use a triangle file. The Kitagawa tubes are packaged with an abrasive chip. The largest pipette I have seen from is 100mL for around $3 ea. in quantity. I really did not look very far. Another option would be a turkey baster that you calibrate. If you have an accurate scale then you can use 1mL = 1gm of water.

I still suggest that the least expensive and accurate enough flow measurement is My mechanical engineer friend says a .0198" orifice and about 3" of water difference would do the trick.

Reply to this comment...

I ran a test in our kit packing area and found 26ppb of formaldehyde. while running the test, a Rotameter from VWR arrived in the mail. I hooked it up in line with the graduated cylinder and it read .25L/m, and my 200ml draw of air through the graduated cylinder took 40.33, 40.38, and 40.62 seconds on the three tests I ran. So, both methods are not in agreement. one says i'm getting .25L/m and the other .3L/m.


@danbeavers We ordered 20 tubes here at the office and they didn't come with anything like a tip breaker. I have a triangle file though. I drilled a 7/64" hole into a piece of brass tubing to make a tip breaker. It worked much better than pliers, and all the glass pieces are contained by the tube, but still left a nasty rough edge. I'll try adding a score with a triangle file to the next tube I break.


Reply to this comment...

hmmm.... the level of the rotameter goes and stays at .3L/m when I hold it down to the table at a level point. the vibration of the pump and levelness seem very crucial to an accurate reading. in terms of a DIY kit, I'm inclined to think that the rotameter is more likely to introduce calibration issues.

Reply to this comment...

rotameter on the table reading .25L/m


rotameter on a block of foam for vibration isolation, reading .3L/m


Reply to this comment...

I think the rotameter is going to be easier to use than the critical orifice/differential pressure method, especially with a fluid manometer. You just need to make sure it's vertical and read perpendicular--the better rotameters have a mirror backing to aid in getting a good reading straight on. Read it at the widest point of the ball--the center.

The problem with the combo valve/rotameter is you are measuring the flow under vacuum since the rotameter is between the valve and the pump. Under vacuum the reading is going to be higher than at atmospheric pressure--so the difference in reading makes sense. To get a more accurate reading you need one end of your flow meter to be open to the atmosphere. You can apply the ideal gas law to correct to standard temp and pressure but unless you're at high altitude simply leaving one end open to atmospheric is good enough.

If you buy the valve and rotameter separately, you can set it up differently:

pump <- valve <- rotameter <- tube <- reference flow meter (attached at the beginning and end of sampling)

Regardless, these types of roatmeters are not going to be super accurate and are not perfectly linear--some readings will be closer to actual than others on the scale. The cylinder method will be more accurate.

also, I know you're trying to make this really cheap but I also like to put a filter between the tube and the rotameter to protect the equipment from dust. The better valve sand rotameters can last decades.

Reply to this comment...


@mathew That heart shaped object is the abrasive chip. When @nshapiro and I did my first measurement Nick said the scoring chip should be in the box but I could not find it. Days later I was inspecting the box and out came the scoring device. It was apparently stuck in the box because of the little zip lock bag. So completely empty the box. My friend found this needle valve for $10:

@DavidMack I think the driving forces are cheap and accurate which are conflicting requirements. That makes the users proficiency important. Vertical orientation, reading at the meniscus/ball center, accurate timing, no leaks all will be important.
I think there is no need for a filter since the measurement tube will filter any particles.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

@danbeavers I couldn't find that little heart. It may have been lost when the box arrived.

@davidmack I ran the VWN rotameter with the valve all the way open and one end open to the atmosphere. the pump came with a built-in filter. My results weren't much better. reading from the middle of the bead the flow was still .275L/M and not .3, as the 250ml graduated cylinder tests suggest. To check my graduated cylinder i put a air hose barb into a 500ml graduated cylinder and it clocked in at 79.5 (79.42, 79.58, 79.71) seconds to traverse 400ml of the cylinder, verifying the results achieved with the 250ml graduated cylinder.

The good news here is that over 3 days with no adjustment and about 3-4 hours of run time, the pump's flow hasn't changed appreciably.

But there's bad news: now that I've opened two of the Kitagawa tubes, I can verify that flow changes between the different tubes. The second tube consistently posted 86.5 (three tests 86.71, 86.27, 86.58) seconds over 400 ml in the 500ml graduated cylinder test, an 8% difference in flow rate. The rotameter still reads .275L/m with the second tube, and so that 8% change in flow doesn't appear measurable with it.

I'm inclined to return to my earlier attempts that @Danbeavers has been offering suggestions on-- I'm going to drill a small control orifice into a brass rod and put it at the intake. we'll see if that improves the consistency of airflow between different Kitagawa tubes.

Reply to this comment...

ok, looking at a photographic record of the flow rate between the two rotameter measurements it does seem like the change in flow is noticeable.

Flow with tube 1:


Flow with tube 2:


Reply to this comment...

@mathew were the two tubes that you are comparing opened by different means? perhaps if we can just standardize the tip breaking better we wouldn't have to resort to a tip collar. I also wonder how much a intake regulator will further strain the vacuum motor/diaphragm.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

@mathew What do you mean by " small control orifice into a brass rod and put it at the intake"? I hope you are thinking manometer . . .

What I suggest is pump to needle valve to T with one end on the manometer tube and the other going to the orifice and from the orifice to a second T then the other end to the manometer and formaldehyde tube.

There is a method of making a pin hole camera that Kodak published many years ago that might be pretty easy and will produce an almost ideal orifice. My concern was connecting the orifice in line with the tube. I think it would be hard to get a thin orifice by axially drilling a rod. Perhaps carefully soldering 2 brass tubes to either side of a brass orifice?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


I believe you have the VWR series FR2000 rotameter, which the accuracy is specified as +/- 5% of full scale, so that's +/- 0.05 lpm. see

So if the rotameter is reading 0.275 and you're measuring it as 0.300 lpm with the graduated cylinder, that's within the rotameter specifications: 8% low but just 2.5% of full scale. These rotameters are fine for setting the flow but for the best results you still need to measure the actual flow and use the actual flow rate to calculate results.

RE: the tubes causing different flow rates, it would be really helpful to have a vacuum gauge on the pump (temporarily) to evaluate this further--how much back pressure difference is there between tubes?

I agree you want to oversize the pump and then choke it down. That should help get a consistent flow even with mild variations in tubes. For critical orifices, the ideal choice is glass, see

I might have some data somewhere on what size would yield 300 ml/min but the opening is going to be tiny, probably less than 1 mm in diameter. I don't think you're going to be able to drill that yourself. You might be able to find one cheaper in brass though, like an orifice for a gas appliance pilot light. Or maybe you could make one from a hypodermic needle?

Getting the orifice sized just right is going to take a lot of trial and error so using a valve is probably the easier option. Plus the cynic in me thinks the system flow will need adjusting now and then.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

I bought and tried a 50ml pipette and made a dropout jar.


The jar kind of works-- some condensation still seems to get through, but never bubbles.

The 50ml pipette is very difficult to use, as I expected, the size makes measurement hard because the bubble moves too fast.

it also requires a stand, takes too long to attach and detach to wet, and is hard to clean. A 100ml model would only solve a few of these issues-- and I can't find anything larger than that.

Reply to this comment...

ran some tests on the different Kitagawa tubes that have been used.

I put the end into the drilled-out brass tube I've been using to break them, and use a steel nail file to scratch a circular break line onto the edge right against the brass break tube. this produced clean, circular breaks, but breaking both ends in the same place resulted in two different diameter openings (see tube 3, below). IMG_20150605_161952.jpg

My hypothesis is that the tube with the narrowest opening would have the lowest airflow. the three used tubes I have do not confirm this hypothesis. each tube had three flow tests in a 250ml graduated cylinder (averaged here).

tube 1 left side opening: 1.7mm, right opening: 1.3mm 40.5 seconds to traverse 200ml = .296L/m

tube 2 left opening: 1.4, right opening 1.8 43.5 seconds to traverse 200ml = .276L/m

tube 3 left opening: 1.25, right opening: 1.9 44.8 seconds to traverse 200ml = .268L/m

putting a valve in front of the tube doesn't seem to normalize the flow. I'm inclined to believe @davidmack that we can't get around measuring the flow of each tube, or excepting some error range around the tubes' differences. Getting clean, consistent breaks doesn't lead to consistently sized openings. The size of the openings does not appear to be the primary driver of flow change.

On the bright side, the tube's don't seem to change flow-- so I don't think they necessarily have to be flow tested DURING sampling. they may be flow testable when off-site. that may work in a library mailing situation.

Reply to this comment...

I ended up with pump calibration kit...did i kill the momentum on this project? signs point to yes @gretchengehrke

I think mat did a youtube video

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

@eustatic, no worries -- Mathew got this pretty far (complete, perhaps?). The video is embedded in this page: From that header ("Filters for reliability..."), scroll down until you see a video with a label on it for calibration.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

@eustatic-- you definitely didn't kill momentum on the project! Once we had the calibration locked in, we needed something to test for. The Kitagawa tubes let us down.

Reply to this comment...

Login to comment.