A D V E R T I S E M E N T


Author Topic: Sourdough starter quantity predictive model  (Read 209933 times)

0 Members and 1 Guest are viewing this topic.

Offline Numerator

  • Registered User
  • Posts: 115
  • Location: California
  • Just keep cooking
Re: Sourdough starter quantity predictive model
« Reply #460 on: August 13, 2021, 03:33:15 PM »
I put two dough balls in CF to compare.  The CF dough hardy rose at all after three days.  Maybe 9 oz vs the starting 8 oz.

I am looking for the sweet spot temp for rising about 50% overnight with 10% SD starter. 

The quest continues.
Dough Fermentation Tool
http://www.shadergraphics.com

Offline Numerator

  • Registered User
  • Posts: 115
  • Location: California
  • Just keep cooking
Re: Sourdough starter quantity predictive model
« Reply #461 on: August 26, 2021, 01:41:15 PM »
I am having trouble resolving the fermentation time based one the table.

I want to determine the fermenting times for a 35% starter.

Proof stage one at 55 degrees F for 24 hours.

Two hours remain.

Now I want to add another stage to finish the two hours.  For other yeast models, I would subtract the time and find the appropriate yeast percentage.
But for this chart, that would be significantly off the charts to yeast 65%.

How should I resolve the last two hours for another proofing stage here?
Dough Fermentation Tool
http://www.shadergraphics.com

Offline BeanAnimal

  • Registered User
  • Posts: 89
  • Location: Pittsburgh, PA
    • BeanAnimal's Reef
Re: Sourdough starter quantity predictive model
« Reply #462 on: August 26, 2021, 05:29:23 PM »
I am having trouble resolving the fermentation time based one the table.

I want to determine the fermenting times for a 35% starter.

Proof stage one at 55 degrees F for 24 hours.

Two hours remain.

Now I want to add another stage to finish the two hours.  For other yeast models, I would subtract the time and find the appropriate yeast percentage.
But for this chart, that would be significantly off the charts to yeast 65%.

How should I resolve the last two hours for another proofing stage here?

Each temperature or concentration series can be curve fit. The number may be off the chart, but in theory the math will work to the extent that within the reasonable predictability of the data in the chart, your resulting return number will be reasonable as long as the new temperature is also in the predictability range of the chart to begin with.

Taking a more pragmatic look at the data and curves themselves...

There just is not a lot of usable range for SD starter temperatures and percentages. Too warm or too cold and the data becomes erratic due to the high degree of variability on other factors. 

When you are on that far end of the concentration scale - rise times as a function of temperature are both short and close to be outside of the useful range of predictability. Any  higher temp at 35% and your two hours remaining becomes a faction of an hour and any lower temp is outside of the range of predictable values at that end of the scale.

This holds true for almost any reasonable case at the far end of the concentration scale and at both HIGH and LOW ferment temperatures.   In other words, the model can't return a value because the real world numbers were so unpredictable that they couldn't be neatly charted, even if the curve fits says the should be.
« Last Edit: August 26, 2021, 07:16:49 PM by BeanAnimal »

Offline Numerator

  • Registered User
  • Posts: 115
  • Location: California
  • Just keep cooking
Re: Sourdough starter quantity predictive model
« Reply #463 on: August 26, 2021, 08:00:37 PM »
Each temperature or concentration series can be curve fit. The number may be off the chart, but in theory the math will work to the extent that within the reasonable predictability of the data in the chart, your resulting return number will be reasonable as long as the new temperature is also in the predictability range of the chart to begin with.

Taking a more pragmatic look at the data and curves themselves...

There just is not a lot of usable range for SD starter temperatures and percentages. Too warm or too cold and the data becomes erratic due to the high degree of variability on other factors. 

When you are on that far end of the concentration scale - rise times as a function of temperature are both short and close to be outside of the useful range of predictability. Any  higher temp at 35% and your two hours remaining becomes a faction of an hour and any lower temp is outside of the range of predictable values at that end of the scale.

This holds true for almost any reasonable case at the far end of the concentration scale and at both HIGH and LOW ferment temperatures.   In other words, the model can't return a value because the real world numbers were so unpredictable that they couldn't be neatly charted, even if the curve fits says the should be.

Thank you, but I understand all this.  I am interested to know how to handle the case I mentioned.
« Last Edit: August 26, 2021, 08:56:19 PM by Numerator »
Dough Fermentation Tool
http://www.shadergraphics.com

Offline Heikjo

  • Lifetime Member
  • *
  • Posts: 1113
  • Location: Oslo, Norway
  • Sour dough = happy me
Re: Sourdough starter quantity predictive model
« Reply #464 on: August 27, 2021, 06:54:57 AM »
I am having trouble resolving the fermentation time based one the table.

I want to determine the fermenting times for a 35% starter.

Proof stage one at 55 degrees F for 24 hours.

Two hours remain.

Now I want to add another stage to finish the two hours.  For other yeast models, I would subtract the time and find the appropriate yeast percentage.
But for this chart, that would be significantly off the charts to yeast 65%.

How should I resolve the last two hours for another proofing stage here?
You want to make a dough with 35% starter and leave it in 55 F for 24 hours? I don't know how active your starter is, but that would not work for mine. I typically use 4% for 24-30 hours in 59 F and in that time it doubles.

I don't understand what you mean by two hours remaining. Two hours until what? If you move it from 55 F to RT for two hours, the dough won't even warm up to RT. 55 F is maybe on the verge of dough temperature I'd want when opening, but it should work fine. Just leave it an hour in RT before opening if it's too cold at 55 F.

If these numbers is something you get from Craig's model, remember that this is not an absolute science. Whatever numbers you get are starting points or indications. If you happen to hit the exact time it predicted, that's just luck. Especially with sourdough you have to assume that there are big differences from person to person. As BeanAnimal says, when you reach the lower and higher fermentation temperatures, the chart can be off by a lot. The chart predicts 58 hours with 4% starter at 59 F, but mine is done in 24-30 hours.

Here's my advice: Pick a number of hours you want to ferment at a specific temperature. Use the chart to find a number that seems to correspond somewhat and try that. Write down everything you do, the recipe, amounts etc., and see how it goes. If it didn't ferment enough, increase the amount for next time. If it fermented too much, decrease the amount for next time. Keep doing this until you get the result you are looking for.

If your starter is active, grows well and predictably, I would try a much lower amount than 35%. In your previous example, the dough tripled in 24 hours at 65F, which is too much for a pizza dough. Most prefer an increase of 1.7-2x. If you increase to 35% and ferment at 55 F, I expect it to peak much faster than 24 hours. Since it already fermented too much at 65 F, try making the same dough with 10% and ferment it at 55 F. If it's still too much, reduce to 8% next time etc.

As you keep experimenting, you can just forget about the chart and use your own references. His charts for baker's yeast works a lot better since it has more data and yeast is more consistent around the world compared to sourdough. In those charts, it can be easier to use it for different fermentation schedules. For sourdough I'd say it's mostly a starting point reference and is most reliable in the green part of the chart.
« Last Edit: August 27, 2021, 07:00:00 AM by Heikjo »
Heine
Oven: Effeuno P134H

A D V E R T I S E M E N T


Offline BeanAnimal

  • Registered User
  • Posts: 89
  • Location: Pittsburgh, PA
    • BeanAnimal's Reef
Re: Sourdough starter quantity predictive model
« Reply #465 on: August 27, 2021, 04:26:47 PM »
Thank you, but I understand all this.  I am interested to know how to handle the case I mentioned.

Not trying to be obtuse here (honestly) - but I just gave you the answer. Let me try to explain again in hopes it is a bit clearer.

The return data in the case you mentioned is going to be outside of the range of predictable or reasonable values, not because the math does not work, because the value returned by the math returns a number that does not fit reality as it is outside the range of predictable behavior of the starter.

LONG ANSWER

You need two equations.
1 - The equation fitted to each temperature and known concentrations/times. 
2 - The equation fitted to each concentration and known temperatures/times.

You can use an integration of both equations to extend the chart as far in any direction you wish.  Doing so will yield an obvious result that can be discerned from simply looking at the table visually.  When you use a temperature or concentration that is already at the edge of the predictable (GREEN) envelope then you are already at the edge or predictability. Adding a second stage is not possible, as the first stage has already pushed/touched the boundary of predictable numbers.

Nutshell - the problem is not the data - you are asking a question that can't be answered with the dataset, or any extension of the dataset. The starter itself is to unpredictable outside of a very narrow range, and even in that range the dataset is a very loose fit.

« Last Edit: August 27, 2021, 04:50:27 PM by BeanAnimal »

Offline BeanAnimal

  • Registered User
  • Posts: 89
  • Location: Pittsburgh, PA
    • BeanAnimal's Reef
Re: Sourdough starter quantity predictive model
« Reply #466 on: August 27, 2021, 05:09:08 PM »
I have not bothered to curve fit or build out the SD table (mostly for the reasons of unpredictability that I mentioned above)

I did do it for predictable yeasts (using Craigs data as a starting point).

In the first image below - choosing 44 degrees and 9 hours, solving the 44 degree curve equation gives me 7.9% cake yeast - well outside of the charted data.

Using that number I can solve the NEW temperature equation (again 44 in this case) and in theory should return the remaining time 44 (which is 3 hours)

The second image is the proof that the return data for two stages (one of them calculated with a 7% derived lookup value) matches a single stage lookup at 44 that IS on the chart.

This all works because the commercial yeast is much more predictable - but even so, at the extremes of the chart - when hours get to 2 or less or more than maybe 150 or so - the return values are pretty meaningless.

So - again, if really want to return values for the dataset you can, you just need to use the existing data to curve fit and solve for the values that you want. The point is that those values are going to be rather meaningless when the inputs are on the extremes of the chart or close to complete fermentation to begin with.  Likewise if they input values are reasonable and small to begin with, then the chart can predict them and more often that not your lookup values will ALSO be on the chart for stage additions.

Sorry for the image sizes

« Last Edit: August 27, 2021, 05:18:35 PM by BeanAnimal »

Offline TXCraig1

  • Supporting Member
  • *
  • Posts: 28213
  • Location: Houston, TX
  • Pizza is not bread.
    • Craig's Neapolitan Garage
Re: Sourdough starter quantity predictive model
« Reply #467 on: August 28, 2021, 09:07:03 AM »
I am having trouble resolving the fermentation time based one the table.

I want to determine the fermenting times for a 35% starter.

Proof stage one at 55 degrees F for 24 hours.

Two hours remain.

Now I want to add another stage to finish the two hours.  For other yeast models, I would subtract the time and find the appropriate yeast percentage.
But for this chart, that would be significantly off the charts to yeast 65%.

How should I resolve the last two hours for another proofing stage here?

BeanAnimal's answer is correct. The chart was designed to help find a starting point for a new recipe or for converting a recipe from baker's yeast to SD. After that, experience and testing may have to take over. In the specific case you mentioned, the answer may be to raise the proofing temp up the the low 80's (typical SD yeast maximum growth rate), or there may be no answer. It's very possible there is no solution to the problem because you can only affect one of the variables (T) and there is a point where you push past the yeast's optimal temp and fermentation slows. Because you can't add more yeast, only affect the temp, not every multi-stage problem has a solution.

At the end of the day, trial and error, testing and tweaking, is a much more important tool than this chart.
"We make great pizza, with sourdough when we can, baker's yeast when we must, but always great pizza."  
Craig's Neapolitan Garage

Offline Numerator

  • Registered User
  • Posts: 115
  • Location: California
  • Just keep cooking
Re: Sourdough starter quantity predictive model
« Reply #468 on: August 28, 2021, 11:18:09 AM »
BeanAnimal's answer is correct. The chart was designed to help find a starting point for a new recipe or for converting a recipe from baker's yeast to SD. After that, experience and testing may have to take over. In the specific case you mentioned, the answer may be to raise the proofing temp up the the low 80's (typical SD yeast maximum growth rate), or there may be no answer. It's very possible there is no solution to the problem because you can only affect one of the variables (T) and there is a point where you push past the yeast's optimal temp and fermentation slows. Because you can't add more yeast, only affect the temp, not every multi-stage problem has a solution.

At the end of the day, trial and error, testing and tweaking, is a much more important tool than this chart.

I agree, but I needed a way to address this in the tool.  I was getting bad numbers. It turns out I was projecting so far and crossed zero hours.   I just needed to clamp to zero to get meaningful results.


This is the SD fermentation for 55 degrees and the fit curve
Chart data is blue dots, curve dots are orange


Y is hours
X is percent SD
« Last Edit: August 28, 2021, 11:22:13 AM by Numerator »
Dough Fermentation Tool
http://www.shadergraphics.com

Offline TXCraig1

  • Supporting Member
  • *
  • Posts: 28213
  • Location: Houston, TX
  • Pizza is not bread.
    • Craig's Neapolitan Garage
Re: Sourdough starter quantity predictive model
« Reply #469 on: August 28, 2021, 03:13:34 PM »
I don't think you were getting bad numbers; I think you were applying a bad assumption to the model - the assumption being that there was a feasible solution to the problem when there isn't.

There is also an upper limit to sd% at some point, it will dissolve your dough quickly into goop.
"We make great pizza, with sourdough when we can, baker's yeast when we must, but always great pizza."  
Craig's Neapolitan Garage

A D V E R T I S E M E N T