Refractometer calibration

[Horseflesh]

We're always told not to trust the specific gravity scale in a refractometer, and that you need to calibrate it by referring to a hydrometer. Having too much time on my hands currently, I made a solution with DME and did a bunch of dilutions and measurements. Here are the results.

The chart is sorted by the "authoritative" value, the temperature corrected hydrometer reading in SG. The colored column shows the delta between that value and the refractometer SG reading. I also captured the refractometer brix reading.

A few values were discongruous, for example sample #5 showed a difference of 1 unit though the surrounding values were all 0. In some of these cases, I took new readings, and the re-tests did seem to be more in line with the trend.

If you eliminate the outliers, you can make a chart like this:

It seems like you can fit a somewhat linear curve to the data in the beginning, but as gravity increases, you get into a non-linear correction.

From approximately 45 to 75 gravity units, the indicated correction is less than one-half a unit, meaning that I guess I could basically trust the SG value on the instrument if it falls in that range. Below ~45 units I should add 1 point to the indicated value; above 75 units, I should subtract one, and above about 90 units, I should subtract 2.

I guess the good news is, for my instrument, if I just trust it I am unlikely to be off by more than 2 units, and likely just 0-1. That isn't bad, I am not launching a rocket.

BUT, I don't yet understand how different kinds of wort might change these values. The test solutions were made with light DME only. If worts of different origins don't follow this kind of curve, then the value of a refractometer seems dubious.

The data also shows a "wort correction factor" as calculated on the Brewer's Friend spreadsheet. The last couple of values, at very low gravities, are way out of line with the rest, which makes me suspicious. If you eliminate any value over ~1.10, the average correction factor is 1.054. If we used that as the authoritative correction across the whole range, the correction results are similar, but the breakpoints are a little different.

Bottom line, I am more confused than when I started. I might go back to my hydrometers for pre-fermentation readings!

 

[mabrungard]

I wish I could see your chart.

One thing to consider is if your DME was actually 'dry'. Any moisture that has absorbed into the DME could throw your calculations and assumptions off. In my experience, DME can be pretty bad about absorbing moisture, even when it looks perfectly dry.

Another way to assess and measure specific gravity of a fluid is via a picnometer (a really precise volumetric flask). The problem is that they are really expensive. That will be a more reliable method than refractometer or hydrometer.

 

[kh54s10]

So the biggest difference between a hydrometer and a refractometer is 2 gravity points? First thought is that pretty close to the range of just reading the scales. Second thought is that I don't really care if my OG is 1.050 or 1.052, easily close enough for me.

Another thing is that it appears that your sample temperatures range from 78 to 88. How much error did this introduce.

Where the problem with refractometer vs hydrometer comes into play is AFTER fermentation, when alcohol skews the refractometer reading.

 

[Horseflesh]

@mabrungard, the image sources are here. If the site doesn't work for you I will upload the images here instead.

http://i.imgur.com/p3vGvSQ.png
http://i.imgur.com/rY6gc3F.png

The DME was about as dry as it came from the manufacturer. The bag was open, but was stored in a larger vacuum-sealed pouch. But even if it was damp, that would just make it harder to make a solution of a specific density. For this test that wasn't necessary. I put about 2 oz DME into 8 oz water, and whatever I got was OK as a starting point.

@kh54s10 -- good observation on the temperature. I was doing the work while sitting outside on two afternoons and you can see the sample temperatures change with the varying heat of the day. While it is straightforward to correct a hydrometer reading based on sample temperature that is another place where you might end up with a rounding error. The refractometer automagically corrects for temperature, but you do need to wait 30 seconds or so for equilibration.

I had one interesting observation that you cannot see in the images. Each refractometer SG/brix value is an average of 3-5 readings. It was common for a string of readings to begin at a lower value and creep up during the session. It wasn't by much, just 0.1-0.2 brix, but enough to see. Then, sometimes I'd get an identical reading 3-4x.

"So the biggest difference between a hydrometer and a refractometer is 2 gravity points?"

That appears to be true for THIS refractometer, in the density range I tested, with wort made from light DME. YMMV!

 

[ba-brewer]

Where the problem with refractometer vs hydrometer comes into play is AFTER fermentation, when alcohol skews the refractometer reading.

Yes you cant get a direct reading once fermentation begins but with a calculator you can get decent results. I take progress reading with refractometer while fermenting then do a final reading with a hydrometer before kegging. The final calculated refractometer gravity is usually within 1 or 2 points of the hydrometer. Very dark beers seem to be a bit more off. If you are only looking for a stable gravity then no need to calculate.

I also check my starting gravity with both a hydrometer and refractometer and usually get reading about the same value.

This calculator seem to give the best correction for me.
http://seanterrill.com/2012/01/06/refractometer-calculator/

 

[ajdelange]

The reason you are cautioned against trusting refractometer readings is that refractometers are calibrated for sucrose. Wort is not sucrose and neither is the DME you used for your experiment. That is why there is a clear trend in error as a function of DME concentration. If you wanted an instrument calibrated for DME I think you can probably figure out what you should do. Fit a curve through that data and add or subtract corrections from it to your readings. The instrument manufacturer would take a different approach. He would prepare a table of RI vs concentration for this DME, fit a curve to that and program that into his instrument which is exactly what he did except that he got the Bx/SG vs RI data most probably from the ICUMSA (International Commission for Uniform Methods in Sugar Analysis) formula.

I'm a little puzzled about the plot as compared to the table though. The table shows SG readings to a resolution of 0.001 whereas the plot shows differences of less than 0.001. What am I missing?

The correlation between temperature and error is small r = 0.33. If error and temperature are completely uncorrelated we would see (based on 20 measurements) correlation coefficients this large in 8% of experiments which suggests that perhaps there is a correlation but does not suggest it very strongly. Anyway, I'm not sure I have the right error values as I noted above.

So it seems you have demonstrated the basic flaw in using refractometers. While, in your case, the errors are not dramatic in other cases they can be. I have seen, based on actual worts, discrepancies of as much as 2 °P.

Just to get you thinking: the hydrometer scale is also based on sucrose.

And I very loudly applaud your effort here. The Brewing Science forum needs more guys doing some brewing science.

 

[Horseflesh]

I'm a little puzzled about the plot as compared to the table though. The table shows SG readings to a resolution of 0.001 whereas the plot shows differences of less than 0.001. What am I missing?

Good catch. I was being sloppy with sig figs. In a few cases I saw my refractometer readings as being somewhat clearly between 2 SG values. So, in a few cases I would have a value like 1.0405 in the average() function, in a formatted cell that yielded a visibly rounded value like 1.040. Excel operates off the true value though, not the formatted value. Below is a chart based on the correct number of sig figs.

The new chart (direct link) also includes a linear fit. I tried breaking the data into linear and non-linear ranges but with so few data points I didn't see anything meaningful.

Now what I really want to know is how this compares to actual wort from a typical brew of mine. If I can get within 1-2 SG units of the true value with this calibration data, great. If not... then I guess my refractometer is relegated to checking for the end of fermentation, where the value doesn't matter. Though, in that case I would be able to scratch an expensive electronic refractometer off my Christmas list...

 

[doug293cz]

...

Just to get you thinking: the hydrometer scale is also based on sucrose.

...

Just to be clear, a °Plato scale on a hydrometer is based on sucrose. An SG scale doesn't care what the solute is, but SG to extract concentration computations are almost always based on sucrose (e.g. SG to °Plato.) And some hydrometers have even other scales.

Brew on

 

[ajdelange]

The new chart (direct link) also includes a linear fit. I tried breaking the data into linear and non-linear ranges but with so few data points I didn't see anything meaningful.

Given the precision (0.001) of the measurements I'd say that's a pretty convincing linear fit.

Now what I really want to know is how this compares to actual wort from a typical brew of mine.

One way to find out! It requires very little additional effort to grab a refractometer reading after taking a hydrometer measurement. In may cases I think you will find that the two will agree within half a °P. But then occasionally they will not and disagree by more. The problem I would have with relying on a refractometer is that you don't know when those occasions are going to occur.

If I can get within 1-2 SG units of the true value with this calibration data, great.

The problem is that this calibration data is for a bag of DME and while it may be typical of the usual worts it will not be typical of the occasional worts.

Though, in that case I would be able to scratch an expensive electronic refractometer off my Christmas list...

I would.