For sale: birectifier ($1800USD)
Recently, I started to tackle new proofing techniques based on pycnometry to create something useful for distillers before they shell out $20k for a U-tube densitometer.
The first thing to tackle was adjusting proof for various temperatures that readings could be taken at. The solution here is the Alcodens software which hopefully I will find the time to go more in depth on.
The next thing to tackle was the quality of the pycnometers themselves. Chinese pycnometers are $20 while American made examples, which barely seem to even be available can be $800. What is the difference? And was there any nuance to the exact pycnometers used by organization like NIST or the TTB to construct all the alcoholimetry tables mid century that we still rely on.
I slowly found that the quality of the taper is everything. If there is a slight rocking in the taper, it acts like a pump in the capillary tube and can throw off your measurement by multiple micro liters (0.001). The tapers can be lapped with diamond paste to improve them or larger volume pycnometers can be used to reduce the impact of the error.
Each percentage point of ethanol has roughly a 0.0015 (not quite correct but useful enough here) effect on specific gravity. Error from taper issues, bubbles, etc. are about the same in absolute terms no matter how big the pycnometer bulb and we can see how they influence measurements as the sample size grows.
| weight w/ error | perfect weight | SG | ABV impact |
| 1.0000 | 1.0000 | 1.0000 | 0.0 |
| 0.9990 | 1.0000 | 0.9990 | 0.7 |
| 0.9980 | 1.0000 | 0.9980 | 1.3 |
| 0.9970 | 1.0000 | 0.9970 | 2.0 |
| 0.9960 | 1.0000 | 0.9960 | 2.7 |
| 0.9950 | 1.0000 | 0.9950 | 3.3 |
| 0.9940 | 1.0000 | 0.9940 | 4.0 |
| 0.9930 | 1.0000 | 0.9930 | 4.7 |
| 0.9920 | 1.0000 | 0.9920 | 5.3 |
| 0.9910 | 1.0000 | 0.9910 | 6.0 |
| 0.9900 | 1.0000 | 0.9900 | 6.7 |
| 10.0000 | 10.0000 | 1.0000 | 0.0 |
| 9.9990 | 10.0000 | 0.9999 | 0.1 |
| 9.9980 | 10.0000 | 0.9998 | 0.1 |
| 9.9970 | 10.0000 | 0.9997 | 0.2 |
| 9.9960 | 10.0000 | 0.9996 | 0.3 |
| 9.9950 | 10.0000 | 0.9995 | 0.3 |
| 9.9940 | 10.0000 | 0.9994 | 0.4 |
| 9.9930 | 10.0000 | 0.9993 | 0.5 |
| 9.9920 | 10.0000 | 0.9992 | 0.5 |
| 9.9910 | 10.0000 | 0.9991 | 0.6 |
| 9.9900 | 10.0000 | 0.9990 | 0.7 |
| 100.0000 | 100.0000 | 1.00000 | 0.0 |
| 99.9990 | 100.0000 | 0.99999 | 0.0 |
| 99.9980 | 100.0000 | 0.99998 | 0.0 |
| 99.9970 | 100.0000 | 0.99997 | 0.0 |
| 99.9960 | 100.0000 | 0.99996 | 0.0 |
| 99.9950 | 100.0000 | 0.99995 | 0.0 |
| 99.9940 | 100.0000 | 0.99994 | 0.0 |
| 99.9930 | 100.0000 | 0.99993 | 0.0 |
| 99.9920 | 100.0000 | 0.99992 | 0.1 |
| 99.9910 | 100.0000 | 0.99991 | 0.1 |
| 99.9900 | 100.0000 | 0.99990 | 0.1 |
An error that is a 1% deviation in repeatability for a 1.0 ml sample throws ABV off by 6.7 percentage points, but as the samples size grows to 100 ml, that same error in absolute terms (from sloppy taper, etc.) has an impact of only 0.1 percentage points. For the birectifier we will commonly be using a 10.0 ml pycnometer.
Sadly, this is not the end of the story because error is going to compound again. We are likely to introduce more error by having an inaccurate temperature reading. And then we are going to error again because our scale calibration weight is only good for 0.0005. All in all, I bet with a 100 ml pycnometer, not even exquisitely lapped, we can measure ABV under 0.5 percentage points. This may be quick and viable enough for TTB proofing. For the birectifier, when working quickly, with a 10.0 ml pycnometer, we may keep our error down to 1.0 percentage points which will be accurate enough for in house work to support useful decisions.
NIST put out two great videos anyone getting into distillery lab work without much of a science background should be aware of.
The first video is about the determination of liquid density. Curiously the video does not feature a hydrometer at all and demonstrates the Lang-Levy micropipette method, U-tube densitometry, and then finally the pycnometer. The Lang-Levy method is the very interesting thing here and it pretty much is not documented anywhere else on the web. The Lang-Levy method is intended for volumes less can 1.0 ml, but I suspect jumbo versions could be constructed for alcoholimetry that may be faster and more accurate than pycnometry (no taper and easier to avoid bubbles. I just ordered a box of them to experiment).
The second video is about the volumetric transfer of liquids. It gives a good overview of the differences between volumetric and gravimetric (think scales) metrology. One of the ideas that was new to me, and where a lot of volumetric error can creep in is that solvents like ethanol are wetting. Basically they cling to material like plastic in ways that water alone does not.
The video on analytical balances may be relevant because they are a big part of pycnometry.
The NIST video series is much better than anything I’ve found on youtube.
If you have any insights, please let me know.