Data taken into account in Brewers'Friend and BrewFather vs the data exposed by the work of Riffe and Spencer in 2018

[meaulnes2]

Reading A Homebrewing Perspective on Mash pH. Distilled-Water pH and Buffering Capacity of the grist published by Riffe and Spencer in 2018
I am surprised by the very large dispersion of the results, and above all, by the considerable distance taken by software like Brewer's Friend and BrewFather with this data.
For example, according to the work presented by Riffe and Spencer, for roast malts DI ph=4.64 and B=68.7 mEq/ph.kg whereas Brewers'Friend considers 4.34 and 33.
There is therefore a ratio of two for the buffer capacity.
For crystal malts, the pH differs by 0.4 pH for the darkest while the buffering capacity evolves from 30 to 80 with the color for the study cited, whereas Brewers'Friend considers a constant value of 33 mEq/ph.kg
Admit that there is reason to wonder!
Has somebody an opinion on this study and, more generally, on the credibility on Mash pH prediction by software ?

 

[Silver_Is_Money]

I've written extensively upon this subject within this forum, so I don't intend to sound like a broken record (for those who might remember what that even is) here, but: When the critically combined (as in simultaneous) data for both DIpH and BC is highly rare to begin with (and wherein one without the other is utterly useless), and where what scant hard data is extant in this regard does not come close to being in agreement, even for the very same malts, from one titration source (researcher or team) to the next, one must cherry pick and bias and cull and alter/normalize the already scarce data as they see fit. Or simply fabricate it (BC and pHDI) at will for every one of hundreds of malts and adjuncts, with none of them associated with any titration based analytical history, such as is the presently the ongoing trend...

For the case of Brewer's Friend, I think it may merely default all grists to somewhere around 33-34 as to aggregate BC.

Cherry picking, empirical method, quasi-empirical mishmashes of math modeling and made up stuff, and strict math modeling all seem to wind up within relatively the same ballpark as to prediction.

When I started developing 'Mash Made Easy' it was a hodgepodge assortment of empirical, quasi-empirical, and (presumed sound) math modeling, pretty much in that order. Then as MME evolved, it 'eventually' became my quest to purge as much empirical and quasi-empirical stuff from it as I could, such that now the order of these is reversed. But when I take a historical step back and look at the results of Mash pH prediction output, it likely matters little as to the underlying approach, as results are more often than not ballpark similar throughout. And the reason for this is bias. I more than strongly presume that it is much the same for other of such software. Even Riffe and Spencer hide their biases behind nominal attempts at data unifying, such as via cherry picking "fiduciaries" (akin to what I have referred to in the past as "Standard Candles"), and making straw grasping presumptions as to which of the researchers data is more correct, etc... But roll it up properly, package it, and it sells. Just remember to use carefully chosen terminology that does not outwardly imply bias while applying loads of what is precisely that. Heck, it even sells if not rolled up properly under the cover. The bottom line though is that accuracy is not really what the end user wants or desires from such software, as next to no one checks for it anyway, and 9 of 10 who do check apply bias upon bias (often without realizing it, or even in self denial of such) in so doing anyway. Feelings, the strong need to belong and to conform, and the like, ... are generally far more important as to which software one chooses as being best. That plus when it comes to software, as in much of business life, "you don't sell the steak, you sell the sizzle". And for some there may even be an impression that if you are not paying for it, it simply can't be as good as something that you must pay for. And the more you pay the better it must be...

 

[BQ&W]

sound like a broken record (for those who might remember what that even is)

Thanks for making me feel OLD!

 

[Silver_Is_Money]

My most recent shocker is the discovery that the revered Henderson-Haslebach equation is effectively quasi-empirical if the data and data ranges and ion pairs are not duly cherry picked. Before that the biggest shocker was that Kolbach's 'RA' is even more quasi-empirical. So even attempts at purging software of empirical means come up short of the goal line.

 

[Silver_Is_Money]

Another shocker is that at least one highly popular brewing software package (which merely incorporates a subset of mash pH prediction among the multiplicity of sizzle options) is displacing/replacing their sound Riffe math model based mash pH prediction groundwork with the option of selecting a highly popular "other" of such groundwork, seemingly at (by now you may have guessed it) popular demand. Might I remind everyone that science is not a popularity contest.

 

[meaulnes2]

Thank you for the comment. It confirms what I feel rather confusedly and I can only agree with you when you say that "science is not a popularity competition". What is missing here is, at least, the definition of a method that is rigorous, well defined and shared by all, for the evaluation of these quantities. I hope it will happen.

 

[meaulnes2]

BTW, looking at Mash Made Easy I am wondering what are the 2 input values Actual Grist Buffer Multiplier and Kolbach Ca Mg pH Shift Multiplier. More generally is there some guide for the use of this tool?

 

[Silver_Is_Money]

BTW, looking at Mash Made Easy I am wondering what are the 2 input values Actual Grist Buffer Multiplier and Kolbach Ca Mg pH Shift Multiplier. More generally is there some guide for the use of this tool?

Kai Troester (Braukaiser) observed that extant BC (buffering coefficient or capacity) data is derived in what is referred to as a "Congress Mash" utilizing pulverized malts and unmalted grains. He further observed that for grains that were crushed to various real world mill gaps that are nowhere near pulverization, the measured BC is noticeably lower than for pulverized, and he quantified as to how much lower BC is in relation to "mill gap" settings for crush. Thus MME uniquely offers (again via user choice) to allow one to modify BC with respect to mill gap, via the application of Kai's relationship. An included mill gap "calculator" worksheet section on the main page (running horizontally across the lowest level of the page, and asking for mill gap input), suggests the modification multiplier to be applied based upon ones mill gap, but (for end user flexibility purposes) the user must choose to manually enter this value into the "Actual Grist Buffer Multiplier" cell. The observation here is that the release of malt acids is nominally somewhat akin (or corollary) to the release of malt sugars within the mash. The finer the crush, the greater the yield or release.

MME does not automatically deviate from Kolbach's method of determining the downward pH shift induced by calcium and magnesium in the mash water. The user must request MME to deviate via altering the "Kolbach pH Shift Multiplier" value manually. The reasoning and logic behind this is twofold. First, a number of years ago AJ deLange noticed that Kolbach did not measure calcium and magnesium induced pH drop within the mash at all, but rather he quantified it post boil and cooling at "knockout". The real downward shift AJ said he was measuring "during the actual mash" from additions of calcium and magnesium was often only in the ballpark of 50%-60% of what Kolbach predicted for knockout (and which "other instruments" apply mistakenly to the mash). Then in June of 2015 two research Chemists, Roger Barth and Ramaaz Zaman published a paper titled "Influence of Strike Water Alkalinity and Hardness on Mash pH" in the "Journal of the American Society of Brewing Chemists 73(3):240-242" that totally blew Kolbach out of the water, and for some malts went even further than did AJ in measuring downward pH shift deviation for calcium, whereby for certain common malts the downward calcium induced pH shift within the mash proper was observed to be on the order of only as little as 1/4 of what Kolbach predicts, and for others about half of what Kolbach predicts, and values in-between. By default (at download) MME sets this multiplier value to 0.50, or 50%, in line with AJ deLange.

For both of these if you do not want to allow for the level of quasi-empirical nature which they introduce, simply set them to a value of 1 whereby to negate any impact. I don't recommend this, but the end user has this full flexibility. Along with the flexibility to enter any fractional valuations (within reasonable permitted ranges) that may assist in bringing MME output more into line with valid measured observation.

Also, MME uniquely permits manual overrides for malts, unmalted grains, and adjuncts DI_pH and BC, for those who actually expend the time and effort to measure/quantify these values for their own lots of malts, etc... via pH meter and titration respectively.

As to a full manual, I've never written one for MME.

 

[Silver_Is_Money]

The work of Roger Barth and Ramaaz Zaman has essentially killed any viably efficacious use for the merely derivative value that Kolbach called RA (Residual Alkalinity), at least within the mash proper.

 

[Silver_Is_Money]

I recall where AJ deLange once commented that while giving speeches on this subject to professional brewers he casually asked for raised hands with regard to those who attempt to apply active mash pH adjustment via computation, and generally only a smattering of hands went up. He stated that when he further inquired as to why this practice was not more common, the generalized reply was along the lines of "You home brewers seem to be far more concerned with that than we are".

 

[meaulnes2]

It was more or less what I guessed, but as usual detailed and enlightening answers from you. Thank you so much.

 

[Silver_Is_Money]

It was more or less what I guessed, but as usual detailed and enlightening answers from you. Thank you so much.

Additionally as to general MME instruction, under the "Malt/Grain Classification (drop down)" column, if a malt or grain or other adjunct finds no match to a "categorized" classification, it's best to presume the entry for it to be selected as "Not Categorized" (which results in a blank field within this column for said grist component). MME then applies cherry picked (biased) log based curves derived from cherry picked (biased/normalized) data whereby to compute a nominal (or default) DI_pH as well as a nominal (or default) BC. This relieves the end user of any need to search through hundreds of actual malt (etc...) drop downs while searching for minute specifics of brand identity as to the item. Whereby (as already discussed) such specific entries do not (for a majority of potential brand specific grist component entries) have commensurate specific/quantified DI_pH and BC data of benefit (that I'm aware of at least) to begin begin with. A similarly biased pair of log based math model curves is utilized for Caramel/Crystal malt computation of nominal DI_pH and BC. With all of this biased default (yet user overridable) input derived from user entry supplied malt color. Categorized entries other than these categorizations/classifications get "fixed" (albeit also cherry picked) DI_pH and BC defaults assigned. I'm trying to be fully open and honest in emphasizing the degree of biased cherry picking that goes into this.

Note: Even if there was or is extant hard data that all can agree upon as to pHDI and BC for a single unified lot (or blended crop) of specific branded and class defined malt/grain/adjuct, a highly genetically similar crop harvested from another region, another nation, another season, another year, another set of soil or climate conditions, differences in fertilizer, etc... will not likely generate a hard BC or pHDI match to it.