The ale yeast pitching rate generally recommended by commercial brewers is one billion cells, per liter of wort, per degree Plato. Assuming a 25% loss in viability prior to re-pitching results in the rule of thumb of 0.75 billion/L-°P. Since yeast products designed to inoculate at this level are not available on the homebrew scale, [...]]]> Background

The ale yeast pitching rate generally recommended by commercial brewers is one billion cells, per liter of wort, per degree Plato. Assuming a 25% loss in viability prior to re-pitching results in the rule of thumb of 0.75 billion/L-°P. Since yeast products designed to inoculate at this level are not available on the homebrew scale, many home brewers will make a Directly pitching a vial or pack of commercial yeast without a starter results in about half of the recommended pitching rate. The effects of under-pitching are varied, but generally held to be undesirable. Wyeast Laboratories, for example, states that under-pitching can cause “excess levels of diacetyl, [an] increase in higher/fusel alcohol formation, [an] increase in ester formation, [an] increase in volatile sulfur compounds, high terminal gravities, stuck fermentations, [and an] increased risk of infection”.

Among home brewers, however, the effects of under-pitching are less universally agreed upon, with some brewers maintaining that under-pitching has no impact, or at least no negative impact, on the final product. Many others pitch multiple yeast products, or make starters prior to brewing, in the belief that it will result in better beer. To test these assertions, a controlled experiment needed to be conducted.

Experimental Setup

In order to provide a reasonable baseline for comparison, I brewed six gallons of an American Amber Ale, targeting a BU:GU ratio of about 0.5. The idea was to come up with a middle-of-the-road representation of a style that would be accessible to most craft beer drinkers. After chilling, the wort was split into two plastic bucket fermenters, with one pitched at 0.73 billion cells/L-°P (“control”), and the other at 0.29 B/L-°P (“under-pitched”), which is roughly the pitching rate that would result from using a month-old smack pack in a five gallon batch. Since the cell counts used to calculate the pitching rates are based on slurry volume rather a true count, the associated error is high – I would suggest ±10%. More complete information on the beer brewed and the yeast propagation procedures can be found in the experiment proposal and yeast ranching posts, respectively.

Seventeen sample sets of three bottles each were distributed to home brewers from Oregon to Florida. One (presumably) broke in transit and was not delivered, and there was one non-respondent. The 15 effective sets resulted in feedback from 37 individual tasters. Twenty-three received “Set 1″, which contained two controls and one under-pitched sample, with the remaining fourteen having “Set 2″, with two under-pitched beers and one control. All sets were labeled only as “A”, “B”, and “C”, resulting in a blind triangle test. Aside from being recruited via a topic on a brewing forum, no effort was made to establish the participants’ credentials with regard to brewing or judging beer. I’d like to think they can therefore be considered a representative sample of the (online) home brewing community.

The chief difficulty associated with collecting impressions from such a widely divergent group of respondents is converting them into unambiguous numerical data. My first thought was to have each taster complete a standard BJCP Scoresheet, but there were several obvious problems with that method. First, filling out the scoresheet can be a daunting, time-consuming task, especially considering that most respondents were not BJCP judges and would probably be completing it for the first time. Second, since by design it addresses only a single beer, a minimum of three sheets per “judge” would be required. Finally, since it lacks room for additional, comparative questions, at least one additional sheet, for a total of four, would be required. In order to minimize the amount of paper and the time commitment required, I laid out a simple feedback form, loosely based on the BJCP scoresheet, which was distributed to all participants.

In order to obtain objective, internally consistent data, two different methods were used. First, the volunteers were asked to perform two simple, quantifiable tasks: identify the control and under-pitched beers; and express a preference for one or the other. In addition to providing these definitive, binary answers, each respondent was also able to share more detailed impressions about appearance, aroma, flavor, and mouthfeel. The descriptive adjectives used were then compiled by means of a simple frequency count. In order to simplify the dataset as much as possible, some descriptors were combined – first, variations on the same root word (“malt” and “malty”, e.g.), then words with substantially similar meanings (“hot” and “solventy”, e.g.).

Personal Observations

My personal tasting notes for the two beers should be taken with a grain of salt, since they don’t represent a blind tasting, but the photographs are so dramatic that I thought they should be included. The photos show, left and right, the under-pitched and control beers. The first photo was taken two minutes after the beers were poured; the second after they had been drunk over the course of about 35 minutes.

The under-pitched beer is slightly lighter in color (10.5-11 SRM instead of 12), and has minimal head retention or lacing when compared to the control beer. The aroma is malty with a slight hot alcohol character, whereas the control has a perfumy, floral hop aroma, with the malt more subdued. The under-pitched beer has a vaguely spicy or vegetal off-flavor, particularly in the aftertaste, and a slightly solventy finish. The control also has a peppery or spicy taste (which I would attribute to the use of Munich malt), but a lingering citrus flavor predominates. The mouthfeel of the under-pitched beer is thin and astringent compared to relative fullness of the control, although it’s difficult to make a fair comparison due to the difference in head retention. The control may have a very slightly more compact, “stickier” yeast sediment.

Results

As a means of gauging fermentation performance, a refractometer reading was taken every twenty-four hours after pitching. The difference in the time required to start and finish fermentation is striking.

The control beer not only exhibited faster fermentation to begin with, but reached terminal gravity approximately twice as fast as the under-pitched fermenter. This provides a clear rationale for the use of higher pitching rates in commercial breweries, where fermenter time is extremely valuable. It could also point to a potential advantage for the higher pitching rate in allowing the yeast to out-compete any contaminating microbes.

Of the 30 tasters who attempted to differentiate the beers, thirteen were able to do so, with nine tasters correctly identifying the beers. This seems to support a hypothesis that there is a difference between the beers – if the three samples were truly indistinguishable, ten respondents could be expected to differentiate the beers, and five identify them.

Since the results follow a binomial distribution (a given sample is either identified or not, with no middle ground), the probability that these results are purely due to chance can be assessed. Given a random distribution, 13 of 30 respondents (or more) would be expected to differentiate the samples about 16.6% of the time. Based on these results, one can conclude, with 83% confidence, that the two beers do in fact taste different.

But what, specifically, are the differences between the beers? In addition to the more subjective descriptors I’ll get to in a bit, we can make a reasonable inference from the 13 tasters who correctly differentiated the samples. If the beers tasted different, but those differences revealed nothing about their identities, we would expect half of the 13 to guess correctly; in fact, the probability that nine or more would guess correctly is only 13.3%. This leads me to believe that at least some of the flavors generally associated with under-pitching – increased esters, fusel alcohols, diacetyl, acetaldehyde, etc. – are in fact present.

Twenty-four of the participants expressed a preference for one beer over the others, with the control being preferred sixteen to eight. When weighted appropriately for the number of samples, the control beer was preferred by 54.9% of tasters. Interestingly enough, this would seem to suggest that while the beers almost certainly are different, there is no consensus about which is better. Simply put, nearly half of people prefer under-pitched beers. An argument could be made, though, that only the opinions of the participants who actually differentiated the samples should be considered. And the data would seem to bear that out. Five of the differentiating tasters preferred the under-pitched beer; however, fully eight of the thirteen tasted Set 2, and when weighted accordingly they overwhelmingly prefer the control, 65.8% to 34.2%.

It is also worth noting that the first half of the respondents (those who tasted the beers after four weeks in the bottle or less) also preferred the control about two to one. So it’s entirely possible that additional conditioning time can reduce the off-flavors that result from under-pitching, but another controlled experiment, with the date of the tasting as a variable, would be needed to satisfactorily answer that question.

Finally, we come to the subjective tasting results. I think the above image largely speaks for itself, so I won’t elaborate too much further. The ten most common descriptors for the control beer are:

    1: Malty
    2: Hoppy
    3: Bitter
    4: Fruity
    4: Sweet
    6: Estery
    6: Smooth
    8: Thin
    9: Solventy
    9: Clean
    9: Dry
    9: Cidery

For the under-pitched beer, they are:

    1: Bitter
    2: Malty
    3: Hoppy
    4: Astringent
    5: Fruity
    5: Estery
    5: Solventy
    8: Clean
    9: Smooth
    9: Thin

Based on the relative frequencies of some words, I think one can reasonably conclude that the under-pitched beer was perceived to be more bitter, more astringent, more solventy, less sweet, and – bizarrely – cleaner than the beer using the standard rate. Obviously, the increased perception of negative characteristics makes a persuasive case for the use of higher pitching rates.

Summary

• No difference in attenuation was observed, but fermentation in the control finished twice as quickly.
• Under-pitching negatively impacts head retention and lacing.
• There is a 43% chance that an “average” home brewer will be able to distinguish between under-pitched and standard-pitched ales.
• There is a 30% chance that he will be able to identify which beer is which.
• Overall, home brewers exhibit no strong preference for either beer.
• Among tasters who can differentiate the two beers, the standard pitching rate is preferred nearly two to one.
• The control beer was described as malty, hoppy, bitter, fruity, and sweet.
• The under-pitched beer was described as being more bitter, more astringent, more solventy, less sweet, and cleaner than the control.

Summary of the Summary

Using a starter makes better beer.
 
 
 
Download the full dataset:
pitchrate_experiment.ods | pitchrate_experiment.xls

Thoughts on Flash

Flash was created during the PC era – for PCs and mice. Flash is a successful business [...]]]> Why does Apple only stand up and say, “this is what we’re doing, and this is why” every year or two? When they do, it usually makes the hairs on the back of my neck stand up.

Thoughts on Flash

Flash was created during the PC era – for PCs and mice. Flash is a successful business for Adobe, and we can understand why they want to push it beyond PCs. But the mobile era is about low power devices, touch interfaces and open web standards – all areas where Flash falls short.

It probably isn't necessary to stir Iodophor.

The basic idea behind these homebrew stirplates is to control the [...]]]>

It probably isn't necessary to stir Iodophor.

The basic idea behind these homebrew stirplates is to control the speed of a motor (computer case fans being a cheap and accessible source) by varying the supply voltage. The best way to do it would actually be using pulse width modulation via a microcontroller, but I didn’t have one on hand, and energy efficiency probably isn’t a major concern for most people interested in powering a motor drawing around a couple watts. The simplest way to provide voltage adjustment (though not regulation, which I’ll get to in a second) is simply to put a resistor in series with the motor, which is what a lot of home brewers do. The problem with that (aside from the engineer in me hating the kludginess) is that electronic components are sold on profit margins more frequently associated with Wal-Mart stores, and are therefore made of plastic. If you have one lying around, or pick one up at Radio Shack, it will typically be rated 0.5 W, sometimes less. We’re going to be sinking around 12 V • 150 mA = 1.8 W into it. It will melt.

So, the bare minimum takeaway here is to use a potentiometer rated for at least 2 W continuous power. A better, or at least more elegant, solution is to incorporate some actual voltage regulation. Which brings us to one of the most useful components ever invented, the LM317 adjustable voltage regulator. The LM317 is capable of providing anything from 1.25 to (V_in – 1.7) V with good regulation (±1%), using only two resistors to set the output voltage:

V_out = 1.25(1 + R₂/R₁)

The basic schematic for the LM317. Note the potential problem with the value of R1.

Technically, an LM317 can have a minimum load current as high as 10 mA, although most will be much lower. That puts the maximum value of R1 at about 120 Ω (10.4 mA). Note that the datasheet specifies 240 Ω because it’s lifted from the specs for the LM117, which tops out at 5 mA. It’s also worth noting that I’m using a 220 Ω resistor simply because I had a 2.5 kΩ pot and didn’t feel like buying another. Do as I say, not as I do.

Circuitry aside, the rest of the build is pretty easy. I had a plastic kitchen container that happens to be the perfect size for my 1 gallon starter jugs. I bought a pair of neodymium rare earth magnets, which are simply adhered magnetically to the hub of the case fan. At 4 pounds of lift each, they aren’t going anywhere. If the fan was attached directly to the container, the magnets would make contact with its surface, so you’ll need some sort of spacer. I used roughly 5 mm lengths cut from a plastic drinking straw. The magnets and the power supply both came from American Science and Surplus, and the only other thing I needed to buy was a stir bar: $6 on eBay. If you’re going to use your stirplate with a convex-bottomed vessel, it might be worth noting the type of stir bar that works for me. It’s 1″ x 3/8″ with a ring.

The stirplate internals. Blue LEDs optional but awesome.

• Project box or other plastic container
• 12 VDC power supply (at least 300 mA to allow for power spikes on startup)
• PC case fan
• Mounting hardware for fan
• LM317T
• TO-220 heat sink
• 120 Ω resistor
• 1 kΩ potentiometer
• Knob for potentiometer
• 0.1 μF ceramic disc capacitor
• 1 μF electrolytic capacitor
• 2 rare earth magnets
• Magnetic stir bar

Plus wire, solder, perf board, etc. If you’re a tinkerer you probably already have some of this stuff on hand. Even if you had to buy everything online and pay shipping, it would only cost about $30. Tayda Electronics is a great source for components. Avoid Radio Shack like the plague.

Now that I’ve gotten off my lazy ass and put this thing together, I’ll be updating my aeration experiments with one last trial, to test my assertion that frequent shaking and a stirplate are equivalent.

Three of the four words on this tag are lies.

Apparently Republicans are sore losers.

The pitching rate of yeast is generally accepted to be one of the most important factors in fermentation performance and the resultant beer character. The often quoted “optimal” pitching rate is 0.75 billion cells per liter of wort, per degree Plato for ale, and 1.5 billion/L-°P for lager. However, at the homebrew level, commercial yeast [...]]]> Background

The pitching rate of yeast is generally accepted to be one of the most important factors in fermentation performance and the resultant beer character. The often quoted “optimal” pitching rate is 0.75 billion cells per liter of wort, per degree Plato for ale, and 1.5 billion/L-°P for lager. However, at the homebrew level, commercial yeast cultures are not available with cells counts adequate for fermentation of a typical (i.e., 20 L at 13°P) batch of ale, let alone a high-gravity lager, and multiple “smack packs” or vials of yeast can easily exceed the total cost of the remaining ingredients. The method most commonly proposed for increasing cell counts is to use a stirplate and flask, with a sponge, cotton, aluminum foil, etc. on or in the neck. There are three reasons typically given:

1. Keeping the yeast in suspension mechanically increases the attenuation of wort sugars.
2. Removing the toxic CO₂ from solution promotes the growth of healthy yeast.
3. Allowing O₂ to diffuse into the starter head space maximizes yeast reproduction.

Some sources also suggest that the reduced pressure in a starter without an airlock will have a beneficial effect, but I find this claim dubious at best. The static head associated with the water in an airlock – assuming a 5 cm column – is given by:

     P = ρgh = (1000 kg/m³)(9.8 m/s²)(0.05 m) = 490 Pa

which is roughly 0.5% of atmospheric pressure at sea level. This is well within the error of the experimental measurements, and so any effect present is not expected to be observable.

Experimental Method

A total of five starters were fermented out, and the volume of the resultant yeast slurry measured:

     A: A control starter, with airlock.
     B: A control starter, with foil wrapped over the neck in place of an airlock.
     C: An airlock starter, which was swirled frequently to simulate a stirplate.
     D: A foil starter, which was also swirled.
     E: A foil starter, which had air added via an aquarium pump and airstone, as often as foaming would allow.

For a starter medium, ordinary granulated table sugar, which is effectively pure sucrose, was used. Sucrose was chosen not only for its low cost, but also because using DME or wort from a mash would necessarily introduce some amount of trub (hot break and cold break), which would be measured along with the slurry and could introduce some non-systemic experimental error. The use of sucrose should also allow the yeast to consume 100% of the sugars in solution, eliminating the “fermentability” of the media as a variable. Finally, using sucrose means that a refractometer can be used to take gravity measurements directly, without applying a “wort correction factor” or removing a hydrometer sample with a statistically significant volume. The starters were each made by dissolving 200 grams of sugar into 2 liters of water, which was then boiled for 5 minutes. The resulting starter solutions averaged 10.7 Brix. In order to provide the nutrients that would otherwise be lacking in an all-sugar starter, two packets of yeast (14 g) were boiled for 5 minutes in water, then topped off to 250 mL total volume. Prior to pitching, this suspension was shaken, and 50 mL added to each starter.

Into each starter, one 7 gram packet of Red Star bread yeast was added dry. Bread yeast was chosen primarily for its low cost; as a strain of S. cerevisiae, its performance in a starter should be essentially identical to any ale yeast. No effort was undertaken to control the fermentation temperature in the starters, other than placing them in a room with a household thermostat set for 68°F (20°C). To accelerate fermentation as much as possible, the starters were placed directly in front of a heating vent, and observed air temperatures ranged from 20.4°C to 25.3°C. While not particularly well-regulated, the fermentation environment therefore emulates one which would be typical for a homebrewer. After fermentation was complete, as determined by identical refractometer readings on two consecutive days, the starters were placed in a cooler at 0°C for at least 24 hours, after which the majority of the liquid was poured off and a final gravity measured via hydrometer. The slurry was then resuspended and poured into a 250 mL graduated cylinder, and the jug rinsed once with tap water. The slurry was returned to the cooler and allowed to settle for another 72 hours before measurement. This does not provide perfect isolation of dense slurry, but again, in the absence of more sophisticated equipment (a centrifuge, e.g.) it emulates the methods available to a typical homebrewer.

Finally, the slurry volumes were converted into approximate cell counts by assuming 100% viability and a cell density of 3.8 billion/mL. This inevitably introduces a great deal of uncertainty, but true cell counts are not achievable without a cytometer. From a practical perspective, a variation in pitching rate of 20%, or even more, is probably of negligible importance in brewing.

Observations

The control starters (A, left, and B, right) at high krausen, 18 hours after pitching.

Starters C and D were fermented next. To simulate the effects of a stirplate, these starters were agitated in a vigorous circular motion for approximately 15 seconds, approximately every 15 minutes, 12-18 hours a day. Clearly this is not a perfect analog for a stirplate starter, but it was observed to be sufficient to keep yeast from collecting in the bottom of the glass jug between periods of agitation. Again, marked differences in the surface appearance of the starters were apparent. Within a few minutes of swirling, bubbles began to appear at the surface of D (the foil starter); C remained clear between periods of agitation, although airlock activity resumed quickly. When swirled, C also produced fewer and coarser bubbles than D. After being measured, the slurry from D was inadvertently left in the graduated cylinder for an additional 12 days; the value of 3.8 billion/mL was obtained by assuming this allowed for full compaction to 4.5 billion/mL.

Finally, starter E was aerated using an “Elite 800″ model aquarium air pump and a plastic aquarium air stone. The pump is rated for 2.0 W and 2.5 psi. Two 0.45 micron syringe-type filters were used in series to ensure sterility. The starter was aerated continuously for eight hours after pitching, and thereafter for approximately one minute in fifteen, 12-18 hours a day. This was the maximum duty cycle that was possible without the vessel overflowing. Qualitatively, this seems to be much less foam than would be expected from a wort starter of similar volume and gravity. This is sensible, given that the sucrose medium lacks most of the proteins associated with malt-based wort.

Results

Starter B produced approximately 8% more yeast than A. Although a rigorous calculation of the experimental error was not conducted, it is almost certainly within the error bar of the graduated cylinder measurements (±0.9 mL), and so the results for the control starters offer significant support for the hypothesis that access to atmospheric oxygen increases cell growth in a yeast culture. At least part of the increase, however, can be attributed to an overall more thorough fermentation in B, which exhibited approximately 2% greater attenuation than A.

Interestingly, although the “stirplate” starters did produce significantly larger volumes of slurry, they were not nearly as large as other sources suggest. For example, MB Raines observed a four-fold increase over a starter which was shaken. The Mr. Malty™ calculator seems to assume a factor of 1.27 (for shaking) or 1.53 (for a stirplate) versus a “simple starter”. The experimental results, however, show an increase of 17% for the foil starters, and only 8% when using an airlock. It is also worth noting that C and D exhibited identical attenuation, which was statistically equal to the attenuation of A.

Finally, starter E, the aerated sample, produced slightly less slurry than D (about 5% less), which is still 12% more than the “simple starter”. Again, however, the increase is not in line with others’ results; Raines and Zainasheff report increases of 50% and 35%, respectively. The final refractometer reading for E was 0.3 °Bx higher than any other starter; I hypothesize that this is due to the liquid being saturated with air, and the fact that the hydrometer-measured gravity was also 0.5 “points” higher than the other starters would seem to bear this out. The attenuation of E was, however, roughly in line with starters A, C, and D – meaning B exhibited significantly higher attenuation than the other four samples. The reason for this discrepancy is unknown.

The slurry from B, after settling for 72 hours.

1. All other things being equal, a starter covered in foil will grow more yeast than one with an airlock.
2. The primary reason to use a stirplate is not the mechanical mixing of the starter, but the introduction of oxygen. Using an airlock significantly reduces the effectiveness of a stirplate.
3. Contrary to what other sources indicate, a stirplate does not produce several times as much yeast per unit volume.
4. Given that it can also be used to aerate the main batch of wort, an aquarium pump is probably a more cost-effective investment for a homebrewer than a stirplate.
5. In the case of a pure sucrose fermentation, refractometer estimates of final gravity correlate well with hydrometer readings, with a maximum discrepancy of 1.5 “points”.
6. Bread yeast (at least this brand) tends not to flocculate, and on that basis alone would be a poor choice for beer.
7. I feel sorry for 17 year olds. This stuff isn’t terrible, but I wouldn’t want to drink it.

 
 
Update: 06 Feb 2010

To try to determine how much impact using sugar in place of malt would have on the behavior of a starter, I made up a 2 L starter, using 215 g of DME. In all other respects it was treated the same as starter D. Call it F. F produced 88 mL of slurry. Assuming 5% of the volume is non-yeast solids, this equates to 318 billion cells. Adjusted for attenuation, that’s 93% more yeast per unit of sugar compared to the sucrose starter. Significantly, it is also essentially identical to the figure (314 billion) given by the MrMalty calculator, when set to “intermittent shaking”. However, for a stirplate starter the calculator predicts a total of 379 billion cells. From this I draw a few additional conclusions:

1. Access to free amino nitrogen can be a limiting factor in fermentations which are largely non-malt based.
2. Increasing cellular access to oxygen can to a limited extent compensate for low levels of FAN.
3. The MrMalty calculator is sufficiently accurate at predicting cell counts for brewing, but may over-estimate the effect of a stirplate.

 
 
Update: 04 Mar 2010

I fermented out one additional DME starter, this time treating it identically to the aerated starter, E. The resulting slurry measured 116 mL. Again, assuming 5% non-yeast solids, that’s 419 billion cells. Not only is this significantly more yeast – 32% more – than the agitated starter, it’s more than the MrMalty calculator predicts for any starter handling technique, including “continuous aeration”. In fact, it’s roughly the quantity needed for a 5 gallon batch of average-gravity lager. Some final conclusions:

1. Given the choice between a stirplate and aeration stone, the stone will make more efficient use of starter wort.
2. By aerating, homebrewers can grow substantial amounts of yeast (such as those required for lagers) without having to make inconveniently large starters.

 
 
Update: 11 May 2010

I’ve conducted another trial in this series, to test whether or not my stirplate would in fact result in more yeast than simple agitation. It turns that it did – 100 mL, or about 361 billion cells. That isn’t quite as much as the MrMalty calculator would suggest, but it’s close, and a substantial (14%) improvement over the swirled starter. Interestingly enough, though, it still isn’t as much yeast as the aerated starter. This suggests that while the stirplate is effective at turning over the wort, introducing oxygen via diffusion, it can’t quite reach the levels achieved by actual air injection. The stirplate was run at 8.00 V for 72 hours – unfortunately, I don’t have any way to measure the actual RPM of the stir bar.

One practical benefit I’m seeing from using the stirplate is that it dissipates the foam from the aeration stone, allowing the air pump to be run continuously. I’ll be measuring one more slurry to see if this results in a further increase in volume.
 
 
 
Download the full experimental data:
starter_experiment.ods | starter_experiment.xls

I’ll be in St. Louis from December 26th until January 1st. To anyone I won’t see, happy holidays!