Jump to content
  • Welcome to the eG Forums, a service of the eGullet Society for Culinary Arts & Letters. The Society is a 501(c)3 not-for-profit organization dedicated to the advancement of the culinary arts. These advertising-free forums are provided free of charge through donations from Society members. Anyone may read the forums, but to post you must create a free account.

Brown Sauce Texture


Recommended Posts

I've spent years working on various forms of glace, jus, coulis, and the brown sauces made from them. My control over texture gets better all the time, but perfection is elusive.

One thing I've figured out is that every liaison, traditional or modern, had good qualities and bad. The bad ones assert themselves when the quantities go up. Examples:

Reduced Gelatin: gets gluey; overthickens and gets especially sticky if allowed to cool on the plate

Roux / Beurre Monte: gets pasty, opaque, and masks flavors

Purified Starches (corn, arrowroot, tapioca, etc.): can get disconcertingly shiny and slick.

Xanthan Gum: thixotropic texture makes sauce behave oddly on plate; reckless overuse creates snot.

My solution has been to mix them; this allows you to get the benefits of different ones while keeping the concentration low enough that the drawbacks aren't assertive.

Lately I've been using a combination of natural gelatin and xanthan gum. It's almost great. The ability to coat food is lovely, and the mouthfeel and flavor release are good as I can ask for. The xanthan even creates an illusion of richness, which lets me keep dairy enrichers (cream, butter) out of the sauces. I like this not out of abstemiousness, but because dairy fat tends to mute flavors.

Unfortunately, one of xanthan's star qualities--its higher viscosity while at rest than while in motion (thixotropism)--weirds me out a little. It makes sauces appear gelatinous on the plate, even though they don't feel that way in the mouth. I'm finding myself wanting less of this.

I'm considering a couple of paths:

-Go to a mix of natural gelatin, xanthan, AND arrowroot. In doing so use less xanthan, get less jiggle

-Try a different gum altogether. I'm intrigued by some form of methylcellulose; its tendency to thicken at high temps and thin at low ones exactly counterbalances the qualities of gelatin, and may make a perfect match. I don't know which kind to try (there are dozens), don't know what it will be like, and don't know if it will be too hard to work with (I'm not up for bringing out the blender every time I need to thicken a sauce).

I'm open to other suggestions. I have a strong preference for individual ingredients rather than blends ... I don't want to become dependent on one companies proprietary recipe.

Thoughts?

Notes from the underbelly

Link to post
Share on other sites

So, for some positive integer n,
you have n candidate ingredients
for thickening, e.g., gelatin,
roux, arrowroot.

For some positive integer m, you
have m criteria for the qualities
of the final sauce, e.g., hot
texture, cold texture, flavor,
static viscosity, dynamic
viscosity. Suppose for each
criterion j = 1, 2, ..., m you
are able to give an evaluation on
a scale from, say, 0 to 100 where
0 means awful and 100 means best
could ever hope for.

For one trial, for each
ingredient i = 1, 2, ..., n you
use X(i) grams of ingredient i
and get results, for j = 1, 2,
..., m, Y(j) on your scale from 0
to 100.

For more succinct notation, let
us agree that X denotes the full
list of n numbers (X(1), X(2),
..., X(n)) and similarly for Y.
Let's regard X as having 1 column
and n rows and regard Y as having
1 column and m rows. We say that
X(i) is the i-th component of X
and that Y(j) is the j-th
component of Y.

So, there exists some
mathematical function f so that
given X we have Y = f(X). That
is, function f is a little like
an old telephone book (assuming
we have not used up all of those
starting fires in the fireplace)
where you look up person X and
get back phone number Y.

So, we want X so that Y = f(X) we
like the best.

Hopefully we can get Y(j) = 100
for each j = 1, 2, ..., m.
Otherwise, for each Y, maybe we
have u(Y) for how well we like Y.
So, we seek X to make u(f(X)) as
large as possible.

Suppose we have some X^1 where we
have done a trial and have gotten
results Y^1 = f(X^1).

Now suppose you do n experiments.
On experiment i = 1, 2, ..., n,
you increase ingredient i by,
say, 20% and get list of
ingredients X^1_i. Then X^1_i is
the same a X^1 except X^1_i(i) =
1.2 X^1(i). Then we get results
Y^1_i = f(X^1_i).

So, to review, we start with our
first trial with list of
ingredients X^1 = (X^1(1),
X^1(2), ..., X^1(n)) with results
Y^1 = (Y^1(1), Y^1(2), ...,
Y^1(m)). For each ingredient i =
1, 2, ..., n we have list of
ingredients X^1_i = (X^1_i(i),
X^1_i(2), ..., X^1_i(n)) with
results Y^1_i = (Y^1_i(i),
Y^1_i(2), ..., Y^1_i(m)) =
f(X^1_i).

So, now we have a lot of data to
process.

Now we drag out an assumption
that works quite well in
practice: We assume that f is
close to 'linear'. So, there are
mn numbers A(i, j), for i = 1, 2,
..., n, j = 1, 2, ..., m, so that
we can get approximately f(X) = A
(X - X^1) + Y^1, but here we have
some undefined material. To make
the definitions, first, A is the
mn numbers in m rows and n
columns where number A(i, j) is
the number in row i and column j.
Second, X - X^1 has i-th
component X(i) - X^1(i). Third,

A (X - X^1)

is a special multiplication
really the same as in linear
equations in high school algebra.
In particular, for each row j =
1, 2, ..., m we have

A(1, j) (X(1) - X^(1)) +
A(2, j) (X(2) - X^(2)) + ...
+ A(n, j) (X(n) - X^(n)) =
Y(j)

So, where do we get A(i, j)?
Sure, from X^1(i), X^1_i(i),
Y^1(j), and Y^1_i(j). In
particular

A(i, j) = (Y^1(j) -
Y^1_i(j)) / (X^1(i) -
X^1_i(i))

Yes, this is, for result j from
changing ingredient i, the change
in Y divided by the change in X.

So, we are looking for the X that
gives us the Y = f(X) we want.
With our linear approximation
using A, we have

Y = A (X - X^1) + Y^1

We are willing to accept more, so
we are also happy with

Y >= A (X - X^1) + Y^1

where we mean that each component
of the left side is greater than
or equal to the corresponding
component of the right side.

Or

A X <= Y + A X^1 - Y^1

or

A X <= (Y - Y^1) + A X^1

So, one approach is to pick a
desired Y and see if we can solve
for X. Or, more promising, we
pick Y so each component is only
slightly larger than the
corresponding component of Y^1
and see if we can get a value for
X and then keep trying,
increasing components of Y.

Of course, we want all components
of X to be greater than or equal
to 0.

A little more promising, for each
j = 1, 2, ..., m, we pick a
positive number W(j) and set

Z = W(1) Y(1) + W(2) Y(2) +
... + W(m) Y(m)

(Advanced readers: Think utility
functions, u, and Pareto) and
then ask for X to make Z as large
as possible while

A X <= (Y - Y^1) + A X^1

and while each component of X is
greater than or equal to 0. What
we have now is a well posed
problem in the field of
optimization called linear
programming.

So, we can get a solution using
readily available software, e.g.,
in some spreadsheet software.

If we look at the results and for
some j = 1, 2, ,,,, m we want to
do better on Y(j), then we just
increase W(j) and have the
software solve again.

Suppose we get solution X. Then
we can set X^2 = p X - (1 - p)
X^1 for some number p between 0
and 1. If we are conservative and
do not believe the linearity
assumption holds very well, then
we pick p close to 0 and let the
next iteration be a small step.
If we are optimistic, then we
pick p closer to 1 and let the
next iteration be a larger step.
Yes, p = 1 is a candidate.

With X^2 we return to the kitchen
and find Y^2 and then proceed as
above for one more iteration.

When an iteration does not change
X, we stop and accept the
resulting X and corresponding Y.

So, each iteration requires n + 1
batches in the kitchen.

So, get a supply of each of the n
ingredients, get a suitable
spreadsheet file, make sauces
with just a water base, and go
for it!

If after a few iterations the 20%
seems a bit drastic, then lower
it to, say, 10%, eventually
perhaps even as low as 1%.

Might do one iteration an
evening. In a week or so should
be done.

The final X will be your
proprietary "blend"!

Edited by project (log)

What would be the right food and wine to go with

R. Strauss's 'Ein Heldenleben'?

Link to post
Share on other sites

Actually, Escoffier advocated the use of arrowroot in meat jus and the like.

Some of the best sources of gelatin are turkey wings, pigs' feet (split and blanched)and gelatinous cuts of beef or veal. Pig's feet are great, but if you don't like them, turkey wings work just as well, and to a certain extent, so do chicken wings.

Then there's "Vegetable matter" as a thickener. It could be as simple as roasted tomato puree,(added at the begining) or root vegetables that slowly dissolve into the liquid.

But with natural gelatin (turkey wings, etc) some vegetable matter, and at the final, a bit of arrowroot, I'd say you'd have a pretty fine sauce.

Link to post
Share on other sites

This maybe the most correct way to make sauce known to man.

But I will bet any amount of money, that the great majority of good professional chefs could not read it and follow it verbatim. The may be doing it, but not because of the written formula.

I would venture that 85% of all amateur wanna be cooks, like myself would not even attempt to try it.

Again, we may be following someone's recipe based on these series of equations but we ain't doing them ourselves.

I thought you were joking when you first posted it.

You must be one smart person to know that stuff.

Could you maybe turn your talents to inventing a full proof airport scanner, that does not breakdown >?

Whoever does, is going to make a fortune.


So, for some positive integer n,
you have n candidate ingredients
for thickening, e.g., gelatin,
roux, arrowroot.

For some positive integer m, you
have m criteria for the qualities
of the final sauce, e.g., hot
texture, cold texture, flavor,
static viscosity, dynamic
viscosity. Suppose for each
criterion j = 1, 2, ..., m you
are able to give an evaluation on
a scale from, say, 0 to 100 where
0 means awful and 100 means best
could ever hope for.

For one trial, for each
ingredient i = 1, 2, ..., n you
use X(i) grams of ingredient i
and get results, for j = 1, 2,
..., m, Y(j) on your scale from 0
to 100.

For more succinct notation, let
us agree that X denotes the full
list of n numbers (X(1), X(2),
..., X(n)) and similarly for Y.
Let's regard X as having 1 column
and n rows and regard Y as having
1 column and m rows. We say that
X(i) is the i-th component of X
and that Y(j) is the j-th
component of Y.

So, there exists some
mathematical function f so that
given X we have Y = f(X). That
is, function f is a little like
an old telephone book (assuming
we have not used up all of those
starting fires in the fireplace)
where you look up person X and
get back phone number Y.

So, we want X so that Y = f(X) we
like the best.

Hopefully we can get Y(j) = 100
for each j = 1, 2, ..., m.
Otherwise, for each Y, maybe we
have u(Y) for how well we like Y.
So, we seek X to make u(f(X)) as
large as possible.

Suppose we have some X^1 where we
have done a trial and have gotten
results Y^1 = f(X^1).

Now suppose you do n experiments.
On experiment i = 1, 2, ..., n,
you increase ingredient i by,
say, 20% and get list of
ingredients X^1_i. Then X^1_i is
the same a X^1 except X^1_i(i) =
1.2 X^1(i). Then we get results
Y^1_i = f(X^1_i).

So, to review, we start with our
first trial with list of
ingredients X^1 = (X^1(1),
X^1(2), ..., X^1(n)) with results
Y^1 = (Y^1(1), Y^1(2), ...,
Y^1(m)). For each ingredient i =
1, 2, ..., n we have list of
ingredients X^1_i = (X^1_i(i),
X^1_i(2), ..., X^1_i(n)) with
results Y^1_i = (Y^1_i(i),
Y^1_i(2), ..., Y^1_i(m)) =
f(X^1_i).

So, now we have a lot of data to
process.

Now we drag out an assumption
that works quite well in
practice: We assume that f is
close to 'linear'. So, there are
mn numbers A(i, j), for i = 1, 2,
..., n, j = 1, 2, ..., m, so that
we can get approximately f(X) = A
(X - X^1) + Y^1, but here we have
some undefined material. To make
the definitions, first, A is the
mn numbers in m rows and n
columns where number A(i, j) is
the number in row i and column j.
Second, X - X^1 has i-th
component X(i) - X^1(i). Third,

A (X - X^1)

is a special multiplication
really the same as in linear
equations in high school algebra.
In particular, for each row j =
1, 2, ..., m we have

A(1, j) (X(1) - X^(1)) +
A(2, j) (X(2) - X^(2)) + ...
+ A(n, j) (X(n) - X^(n)) =
Y(j)

So, where do we get A(i, j)?
Sure, from X^1(i), X^1_i(i),
Y^1(j), and Y^1_i(j). In
particular

A(i, j) = (Y^1(j) -
Y^1_i(j)) / (X^1(i) -
X^1_i(i))

Yes, this is, for result j from
changing ingredient i, the change
in Y divided by the change in X.

So, we are looking for the X that
gives us the Y = f(X) we want.
With our linear approximation
using A, we have

Y = A (X - X^1) + Y^1

We are willing to accept more, so
we are also happy with

Y >= A (X - X^1) + Y^1

where we mean that each component
of the left side is greater than
or equal to the corresponding
component of the right side.

Or

A X <= Y + A X^1 - Y^1

or

A X <= (Y - Y^1) + A X^1

So, one approach is to pick a
desired Y and see if we can solve
for X. Or, more promising, we
pick Y so each component is only
slightly larger than the
corresponding component of Y^1
and see if we can get a value for
X and then keep trying,
increasing components of Y.

Of course, we want all components
of X to be greater than or equal
to 0.

A little more promising, for each
j = 1, 2, ..., m, we pick a
positive number W(j) and set

Z = W(1) Y(1) + W(2) Y(2) +
... + W(m) Y(m)

(Advanced readers: Think utility
functions, u, and Pareto) and
then ask for X to make Z as large
as possible while

A X <= (Y - Y^1) + A X^1

and while each component of X is
greater than or equal to 0. What
we have now is a well posed
problem in the field of
optimization called linear
programming.

So, we can get a solution using
readily available software, e.g.,
in some spreadsheet software.

If we look at the results and for
some j = 1, 2, ,,,, m we want to
do better on Y(j), then we just
increase W(j) and have the
software solve again.

Suppose we get solution X. Then
we can set X^2 = p X - (1 - p)
X^1 for some number p between 0
and 1. If we are conservative and
do not believe the linearity
assumption holds very well, then
we pick p close to 0 and let the
next iteration be a small step.
If we are optimistic, then we
pick p closer to 1 and let the
next iteration be a larger step.
Yes, p = 1 is a candidate.

With X^2 we return to the kitchen
and find Y^2 and then proceed as
above for one more iteration.

When an iteration does not change
X, we stop and accept the
resulting X and corresponding Y.

So, each iteration requires n + 1
batches in the kitchen.

So, get a supply of each of the n
ingredients, get a suitable
spreadsheet file, make sauces
with just a water base, and go
for it!

If after a few iterations the 20%
seems a bit drastic, then lower
it to, say, 10%, eventually
perhaps even as low as 1%.

Might do one iteration an
evening. In a week or so should
be done.

The final X will be your
proprietary "blend"!

edited for grammar & spelling. I do it 95% of my posts so I'll state it here. :)

"I have never developed indigestion from eating my words."-- Winston Churchill

Talk doesn't cook rice. ~ Chinese Proverb

Link to post
Share on other sites

For some explanation:

The problem posed in this thread is both challenging and general.

We can see some of the challenge because a chef, clearly somewhere from good to excellent, has worked for "years" without a good solution.

The generality is adjusting some n inputs to get desirable results on some m outputs, all at the same time.

I like the problem: Currently I have some reduced blond chicken stock, from poaching two frying chickens in a vegetable stock from 8 pounds of mirepoix, and have stock that gels at room temperature, has some chicken flavor, a lot of vegetable flavor, but is surprisingly sweet! Also, diluted to usual strength, the stock is nicely clear. However, the color is surprisingly red. So, net, I have vegetable flavor, chicken flavor, clarity, color, viscosity, and, a surprise, sweetness, and some of these are in conflict! And this is just a simple chicken stock! Also, the goo won't freeze easily and, overnight in plastic ice cube trays, just becomes stiffer goo but won't pop out of the tray!

So, the problem of this thread has to appear elsewhere in cooking and, say, blending, maybe wine, grape juice, orange juice, a new soft drink, etc. A broad area of applications should be getting a balance of flavors. More generally, we can seek inputs that do well on flavor, color, texture, and cost all at the same time.

So, with the m outputs, there are what we can call m objectives. We are not nearly the first to see such problems. In part we are in the topic of multi-objective optimization. This topic is, for example, one research interest of the current President of Carnegie-Mellon university.

The field of mathematical economics also encountered this problem under the topic of Pareto optimality.

The main idea for balancing the possibly competing m objectives is to have a utility function, the u in the notes above. The axiomatic theory of utility functions was heavily the work of J. von Neumann and plays a major role in mathematical economics.

Another approach to balancing the multiple objectives is to apply weights which are the W(j) in the notes above. The weights do not promise to do as well as the utility function without some additional assumptions; in practice the weights usually do nearly as well without the assumptions.

In problems in cooking, likely the function f is fairly close to linear, as hoped, over much of its domain but, then, quite non-linear as some of the input ingredients reach saturation or the law of diminishing returns. Also if two or more of the ingredients interact in some ways, which may be the case in this particular problem in sauce texture, we can expect more non-linearity.

But, however non-linear the function f is, it is almost guaranteed to have a quite accurate local linear approximation for a definition of local reasonably large in this particular problem. E.g., if we are putting in 50 grams of blond roux, then the effects of 49, 50, and 51 grams, with everything else held constant, should, for each result m, define three points all essentially on the same straight line (m straight lines in all).

So, we have some significant local linearity, and this fact is enormously powerful, in particular, greatly reduces the number of trials we need to do in the kitchen, that is, because we can, as in the mathematics, at each iteration in effect extrapolate from the n + 1 kitchen trials (efforts, batches).

But where function f is non-linear, we take a linear approximation, move cautiously from there (as in the parameter p), and try again. So, we are taking careful, small, linear steps on a non-linear surface.

These steps, this iterative process, has some known, good properties, e.g., as explored in some of the work of D. Bertsekas at MIT.

Our stopping criterion, without more assumptions, really only guarantees a locally best result, but in this problem likely we will get the globally best result or nearly so. However, there may be ties, that is, more than one "blend" of inputs that give essentially the same, best output. In this case, also considering cost should break the ties!

The solution outlined involves a lot of work in the kitchen. However, as is too well known, a lot in progress in good cooking can involve a lot of work in the kitchen! E.g., my little stock making had me washing my 12 quart Vollrath pot, several 5 quart bowls, 3 quart bowls, measuring cups, strainers, cotton filters, etc. some of them several times, hauling some gallons of vegetable trimmings to the compost pile, etc.! At least some crows got happy!

For a serious approach to getting all of texture, flavor, appearance, and cost in good shape all at the same time, the approach outlined may provide a practical solution, maybe the only practical solution, and maybe much less effort than other efforts attempted.

Also this approach is a well defined procedure: Follow the steps, get the results, without a lot of guesswork in the interim.

No, this approach is not good for getting the family dinner on the table some weekday evening in 20 minutes or less. But for getting a sauce good on static viscosity, dynamic viscosity, mouth feel, gloss, color, and flavor, all at the same time, the effort may be comparatively small and, in some contexts, very much worthwhile. And, with the final results, maybe with an extra 10 minutes the weekday dinner can have a fantastic sauce!

What would be the right food and wine to go with

R. Strauss's 'Ein Heldenleben'?

Link to post
Share on other sites

I'm intrigued by your methodology, Project, but I'm not sure it's the answer to this problem. For one, it flies stratospherically over my head ... and I'm a big geek.

Leaving that aside, I think if I were truly starting from scratch-- in other words, my memory wiped clean, and all the cooks and cookbooks and other resources erradicated from the earth--then this kind of approach might be truly efficient.

But since we likely have centuries of accumulated experience making brown sauces right here in the egullet community, I was thinking I'd hear, "dude, try half a percent of brand X methylcellulose!"

Or something like that.

I don't think what I'm trying to accomplish actually requires original research.

FWIW, I may have inadvertently exaggerated the amount of work I've put into this problem. I've been making brown sauces for years, but have only recently started experimenting with new ways to control texture.

Notes from the underbelly

Link to post
Share on other sites

Aloha Steve:

What I outlined is not that difficult!

For the mathematical notation, I just used what is available in just simple typing without trying to use subscripts and superscripts. If you wish, then regard the hat character as the start of a superscript and the underscore character, a subscript.

Beyond mathematics, actually the notation is almost ready even for common computer languages. So, this notation can't be more difficult to read than source code in a computer language. And there are not many lines of code to read!

For the linear programming, that is now sometimes covered in high school. Well? Perhaps not, but well enough for what I outlined.

Really, the computing required is within what has long been available in common spreadsheet programs.

What I posted is not intended to be a "recipe". Instead, the effort of this thread is to develop a new recipe, a "blend", that would then be a recipe that, if not proprietary, could be used many times by thousands of chefs, and my post was to aid in developing some new recipes.

Right: At least 85% of cooks, chefs, etc. would not go through the procedure I outlined. Neither will they be running a three star Michelin restaurant or even inventing a fantastic new sauce for some chain of 10,000 restaurants.

Getting something both good and new can be challenging, commonly is.

Don't have to be "smart to know this stuff"; just have to have studied some appropriate directions in applied mathematics.

That this mathematics might be able to make a contribution to some high-end topics in cooking could be good to know. So, now the world of cooking has been so informed! And, just think, eG has it, is maybe the first! Yes, for a while it looked like maybe eG was not going to have it, but at least for now it still does!

For airport scanners, part of the problem is taking available data and, then, seeing if further investigation is justified. For this, an advantage is that there is a LOT of data on what peaceful passengers are like. So, given a candidate passenger, tentatively assume that they are peaceful, using the huge amount of data see what the probability then is for getting data like this passenger, and if the probability is unreasonably small, maybe 1 in 1 million, then reject the tentative assumption and justify more investigation. Could do this in stages, and that is an old technique called sequential testing which actually is an application of a topic called stochastic dynamic programming.

But I doubt that there is any money in such work. A big reason is that would have to work with really big organizations, especially the US TSA, and that is a fast way to Excedrin headache 395,295,223 and going broke.

Thankfully for US national security, one really big organization actually has been from good up to excellent at making good use of new, advanced ideas, the US DoD. Yup, that's where I started my career in applied math. But the TSA is definitely not the DoD.

I'm working on something else, where I don't have to sell advanced, new technology to large organizations!

To say more might further irritate some thread readers, so will let this be enough answer to your questions!

What would be the right food and wine to go with

R. Strauss's 'Ein Heldenleben'?

Link to post
Share on other sites

You only need to bring the full brunt of linear programming to this problem if there are significant interaction effects which I'm not convinced there are.

Also, Edward: The best source of gelatine is purified gelatine. I don't know why people have such a hangup about using purified gelatine in savoury cooking. It's no different from using white sugar or kosher salt as opposed to resorting to "natural" honey, soy sauce, molasses or pickle juice.

PS: I am a guy.

Link to post
Share on other sites

Shalmanese:

I should interject, my uses of i, j, m, and n in the post require some editing but, sadly, the editing is now no longer permitted. I anticipated that editing would be permitted for 24 hours but apparently that is not true. Basically table A has one row for each measure of texture and one column for each ingredient. It is best to use alphabetical order, that is, have i, j go with m, n so that i = 1, 2, ..., m and j = 1, 2, ..., n. Then it is standard to use i for rows and j for columns. Then table A has m rows and n columns. Then there are n ingredients and m measures of texture; j indexes ingredients and i indexes measures of texture. Actually, there are not many changes, and none are difficult.

For your

"You only need to bring the full brunt of linear programming to this problem if there are significant interaction effects which I'm not convinced there are."

No, that is not correct: To see this, suppose there are n = 5 ingredients that affect texture and m = 4 measures of texture of interest, say, hot viscosity, cold viscosity, dynamic viscosity, and static viscosity.

Well, easily enough, without interactions, we can still expect that each of the 5 ingredients affects each of the 4 measures of texture.

So, the table A with 4 rows and 5 columns has no zeros. This basically pushes us into general linear programming because there can be some conflicts, that is, ingredient 1 can improve texture 2 but give too much of texture 3. Then to save texture 3, have to adjust some of the other four ingredients. So, the ingredients are fighting with each other.

The problem is we want to get good results on all four measures of texture with just one "mixture" of 5 ingredients. So, we have 4 'objectives' which in general, even without interactions, can be in conflict.

There is a cute result: At each iteration, since there are only 4 measures of texture, we will need at most 4 ingredients to do as well as we can, for any weights we select. How 'bout that!

If there are interactions, then we will see that in the function f being non-linear. Well, the function f likely is non-linear over part of its domain anyway due to saturation and/or law of diminishing returns. If the function f were linear, then we wouldn't need the iterations but would still need linear programming.

That is, basically the problem is, even without interactions, non-linear, multi-objective optimization but we attack it with iterations of locally linear, multi-objective optimization, which should work with or without interactions.

Of course, cooking has done such problems essentially only from experience and intuition. However it may be that in some food technology lab working for some big company such techniques would be of significant value or actually already in use.

Whatever is done now in cooking, these techniques were invented for problems of essentially this kind: Adjust some 5 inputs to do well on all of some 4 outputs at the same time. Whenever cooking finds such a problem challenging enough and/or wants to take it seriously enough, these techniques are among the best non-intuitive techniques there are.

What would be the right food and wine to go with

R. Strauss's 'Ein Heldenleben'?

Link to post
Share on other sites

Project -

I studied math but couldn't bring myself to read your post. I do know, (unfortunately without equation back-up), that a good way to get reduced stock ice cubes to come out is to have ice cube trays that nest. Just put warmish water in an empty one, then set the one with the cubes in it in that for a few seconds - they will release nicely (even from cheapo trays that don't release regular ice all that well).

Link to post
Share on other sites

Prof. This has done quite a bit of work in modeling meat stocks...I recall seeing the presentation in person, but can't seem to find the model available online. I do recall that it took into account the types of sugars yielded by simmering carrots with respect to time, cooking temperature, volume of liquid and a number of other variables. Perhaps a quick e-mail to him would help inform the approach proposed here?

Link to post
Share on other sites

Paul, I used to subscribe to the concept that more flavor is better, and, like you, I went to great lengths in my attempts to thicken classical sauces without roux.

After spending years experimenting and pondering the subject, I came to the conclusion that some masking isn't necessarily a bad thing. Also, on a similar note, after years of revering all things Blumenthal, I came to the conclusion that he's a bit of a twat.

I'm all for experimentation and moving forward, but I think this whole 'flavor masking' phobia is, to an extent, throwing the baby out with the bathwater. Properly prepared, brown sauce base should be a little too intensely flavored and the roux should play a role in dialing that intensity back.

In other words, let's not turn back the clock, but, at the same time, let's not run away from valid ingredients. There's a happy medium here. Instead of removing roux completely from the equation, I understand it's shortcomings and mitigate them by combining roux with xanthan and guar. I'm a little more wary of cream than I was 10 years ago, but I still judiciously cook with that as well.

One thing that you might want to consider is that thickening is about molecules bumping into each other. Generally speaking, the more variety in molecules (different shapes), the more clutter, the thicker the sauce. Gums have a very well documented synergy with each other, but I have no doubt that other thickeners provide a synergistic bump as well. It's additional measuring/labor, but the more thickeners, the merrier.

At the moment, though, I'm pretty happy with roux, xanthan, guar and, occasionally, arrowroot.

Edited by scott123 (log)
Link to post
Share on other sites

Also, Edward: The best source of gelatine is purified gelatine. I don't know why people have such a hangup about using purified gelatine in savoury cooking. It's no different from using white sugar or kosher salt as opposed to resorting to "natural" honey, soy sauce, molasses or pickle juice.

Well of course, and I have no problem using pure gelatine.

Thing is, Turkey wings, pigs feet, etc are protein and contribute greatly to flavour, and that's #1 what I'm after. As a protein, I can roast them with bones and mirepoix to develop a more intense flavour as well as colour. Adding pure gelatine to a finished stock doesn't give this possibility.

Then again, I just like feaking out my staff when the meat delivery includes a couple of split pig's trotters or calve's feet..........

Link to post
Share on other sites

Also, Edward: The best source of gelatine is purified gelatine. I don't know why people have such a hangup about using purified gelatine in savoury cooking. It's no different from using white sugar or kosher salt as opposed to resorting to "natural" honey, soy sauce, molasses or pickle juice.

Well of course, and I have no problem using pure gelatine.

Thing is, Turkey wings, pigs feet, etc are protein and contribute greatly to flavour, and that's #1 what I'm after. As a protein, I can roast them with bones and mirepoix to develop a more intense flavour as well as colour. Adding pure gelatine to a finished stock doesn't give this possibility.

Then again, I just like feaking out my staff when the meat delivery includes a couple of split pig's trotters or calve's feet..........

Turkey wings are great when you want to make a brown turkey stock and acceptable when you're making a brown chicken stock. Pigs feet are great for making a pork stock. But there's something to be said for the purity of flavors as well as the intensity. I find too much addins just make it taste like a generic meat stock rather than honing in on the inherent beefiness or chickeniness of a good stock.

PS: I am a guy.

Link to post
Share on other sites

In other words, let's not turn back the clock, but, at the same time, let's not run away from valid ingredients. There's a happy medium here. Instead of removing roux completely from the equation, I understand it's shortcomings and mitigate them by combining roux with xanthan and guar. I'm a little more wary of cream than I was 10 years ago, but I still judiciously cook with that as well.

I used roux for years; stopped using it not out of principle but because I thought other things worked better. Now I'm looking for EVEN better :)

Notes from the underbelly

Link to post
Share on other sites

Turkey wings are great when you want to make a brown turkey stock and acceptable when you're making a brown chicken stock. Pigs feet are great for making a pork stock. But there's something to be said for the purity of flavors as well as the intensity. I find too much addins just make it taste like a generic meat stock rather than honing in on the inherent beefiness or chickeniness of a good stock.

Also, the point of adding those things (or of adding pure gelatin) is getting more gelatin. I get plenty from my initial bone stock, and really don't want any more (for reasons mentioned in the original post ... I find too high a gelatin concentration unwelcome.)

With chicken stock, I do find it helpful to up the gelatin a bit. Easy enough with a few chicken feet thrown in.

How much flavor I'm looking for in the stock really depends on use. When I make veal stock, I'm more interested in mouth feel, savor, and (if it's a brown stock) roasted flavors. Because I'm going to use this as a base for very intensely flavored meat coulis, using different kinds of meat for different applications. The flavor goes in later, which I find works better ... you're not constantly evaporating the volatiles while trying to extract flavors and concentrate the non-volatiles.

Escoffier threw so much meat into his stocks because his methods required it ... after all that simmering only a small percentage of the flavor would be left. There are much more efficient approaches.

Edited by paulraphael (log)

Notes from the underbelly

Link to post
Share on other sites

Ummmm....no, for me the point of adding those items is (1) flavour, and (2) price. Meat has more flavour than bone, and going to a chinese butcher will get your dirt cheap meat "parts" as opposed to veal bones which can cost. I'm not saying I make a sauce with 75% turkey wings, but 3 or 4 wigs tossed in with a half-case of veal bones is a welcome addition.

Pigs feet have a neutral flavour, but for those who don't like pork, veal feet do the same jpb, albeit a bit pricier.'

Link to post
Share on other sites

Good point about price. I'd consider experimenting with alternate sources of gelatin, but right now I get veal bones for pennies above wholesale, which makes them cheaper for me than turkey wings, hocks, etc.. I realize that in many places veal bones are marked up tremendously.

Edited by paulraphael (log)

Notes from the underbelly

Link to post
Share on other sites

In other words, let's not turn back the clock, but, at the same time, let's not run away from valid ingredients. There's a happy medium here. Instead of removing roux completely from the equation, I understand it's shortcomings and mitigate them by combining roux with xanthan and guar. I'm a little more wary of cream than I was 10 years ago, but I still judiciously cook with that as well.

I used roux for years; stopped using it not out of principle but because I thought other things worked better. Now I'm looking for EVEN better :)

:smile: Fair enough. Should you ever decide to give roux another chance (combined with gums), though, I think you'll be pleasantly surprised.

If you feel absolutely intent on pushing the envelope, I would recommend experimenting with inulin. Inulin is a pretty big player in the confection/sugar free realm, but I think it's poised to play a role in savory products as well. It's basically non sweet corn syrup in a powdered form. It has a tiny bit of sweetness, but I think you could get around that by cooking the tomato paste (in brown sauce) less/adding it later in the process. It's molecular size is massive, so it should play beautifully with gums and gelatin, and there's zero masking of flavors.

In large amounts it can be laxating (it's basically the ingredient in beans that can give them a gassy quality), so if you're cooking for company, you need to be careful, but as a single component in a multi thickener approach, I don't think you'll have to worry about it.

Btw, while we're on the topic of laxation... just in case you weren't aware of it, xanthan, being pure fiber can cause laxation issues as well.

Trader Joes carries it. Most health food stores (and Walmart) have it under the brand name Fibersure. It's heinously expensive. If it works out for you, let me know- I have some bulk sources for it.

Link to post
Share on other sites

Interesting. I don't think I've come across inulin in any of the food discussions.

Do you think the laxation thing could really be an issue when using any of this stuff in hydrocolloid quantities? For instance, I'm using xanthan at well under half a percent.

What do you like about roux that can't be achieved with a purified starch?

Edited by paulraphael (log)

Notes from the underbelly

Link to post
Share on other sites

paulraphael:

With my math hat off and back to a beginning cook trying to learn, I'll swing at your roux versus starch issue: Yes, I read the part in Escoffier where he explained that he wanted to cook the sauce until everything in the flour except the starch was skimmed or some such, all that was left was the starch, and that it would be better just to use a source of starch to begin with.

Back when I was doing trials adding various thickeners to water to see what would happen, I concluded that corn starch and arrowroot had similar effects but that a flour-butter blond roux was much different. Yes, it is possible to use just flour, and I concluded that the effects were closer to those of corn starch than a blond roux. Yes, Chinese cooking is big on using corn starch. But I concluded that quite broadly for the sauces I do, the blond roux gives a much different and much more desirable texture than just starch.

Also, somewhat relevant, a sauce thickened with corn starch can break -- suddenly thin out from, as I recall, too many changes in temperature or too much in additions after the starch has done its thickening, and I've never had that problem with roux. E.g., my favorite roux thickened sauce gets, after the roux, milk, cream, egg yolks, and lemon juice but doesn't break!

Net, I had to conclude that Escoffier's conclusion that the only part of a roux desirable was the starch in the flour was wrong, that, in his effort to be modern and scientific, he had been taken in by some overly simplistic views of chemists who didn't cook!

What is going on at the molecular level need not be simple and maybe is not: E.g., my usual technique is to (1) have the stock bubbling, (2) make the roux and, with no delay at all, have it bubbling, (3) again with no delay at all, dump the stock into the roux all at once, (4) right away whip rapidly. This way, I get a great texture each batch!

But, once, I made the roux, let it cool off heat, did something else for maybe 30 minutes, heated the roux again, and continued. When I dumped in the bubbling stock, the roux did next to nothing! I was shocked! Why I don't know, but since then I make sure to have no time delays in the four steps above.

Also at times I have made a roux of flour and cooking oil instead of flour and (common US) butter, and the effects of the flour-butter roux are much different and better.

Net, I conclude that what a roux is doing is not as simple as kindergarten children holding hands on the way to recess!

Look, Ma, no calculus!

What would be the right food and wine to go with

R. Strauss's 'Ein Heldenleben'?

Link to post
Share on other sites

With my math hat off and back to a beginning cook trying to learn,

Look, Ma, no calculus!

Many thanks, I was able to follow ! :biggrin:

edited for grammar & spelling. I do it 95% of my posts so I'll state it here. :)

"I have never developed indigestion from eating my words."-- Winston Churchill

Talk doesn't cook rice. ~ Chinese Proverb

Link to post
Share on other sites

Interesting. I don't think I've come across inulin in any of the food discussions.

Do you think the laxation thing could really be an issue when using any of this stuff in hydrocolloid quantities? For instance, I'm using xanthan at well under half a percent.

What do you like about roux that can't be achieved with a purified starch?

Like I said before, I will use arrowroot for certain dishes, but when I turn to roux for classic sauces, I do so because I prefer it's toasty, slightly less processed flavor.

Re; inulin in discussions, a few years back inulin producers were trying to position it as a fat replacer, and, although I'm sure that there's still someone trying to sing that song, fortunately, the bulk of the industry has backpedaled from such silliness. So, in a sense, it does have a history of being discussed in savory applications, just in a somewhat twisted and misrepresented manner.

Re; quantities, if memory serves me correctly, xanthan traps liquids in a matrix, starch swells to many times it size with water, while inulin (like sugar) just dissolves. In other words, if you want thickening, you can't use it in hydrocolloid quantities. The molecular weight is close to 8 times that of sugar, so you do get more viscous solutions with the same amount of inulin as you would sugar, but, if you've ever made a simple syrup, you'll see that sugar isn't giving you much viscosity at all. 8 times very little is still not so much :smile:

And then there's the sweetness level. At 10 percent the sweetness of sugar, it's not very sweet, but that can add up the more you use.

Basically, you want enough to add some viscosity, but not so much that savory sauces become sweet or your dinner guests all start lining up for the loo. If you've never worked with inulin before, I'd make a 1:1 inulin/water syrup with it just to see the viscosity it's bringing to the table, but when it comes time to formulate... I think 10% should be a good ballpark (in conjunction with xanthan and another thickener such as gelatin).

Like xanthan, and, to an extent, starch, inulin has assimilation issues. Fibersure's 'dissolves in water' claim is a bit off the mark. It has a tendency to form hard candy-like clumps unless you agitate the mixing process carefully. Even with agitation, you may still get some clumping that will need to be dissolved with heat. I tend to make thick syrups (in the 3:1 inulin/water realm) in advance and use it in that form. I find it a lot easier to work with as a syrup.

Also, inulin isn't all that salt stable. I've tried using it for a Chinese take out style brown sauce and the next day, much like corn starch, it had completely broken down. Brown sauce isn't as salty as soy sauce based sauces, but I'd still add the inulin right before serving, just to be safe.

Lastly, I've thought about it a little more, and one more thickener comes to mind. Apparently when you combine different hydrocolloids, the end result is a thickener that's not quite so slimy as xanthan is by itself. I use xanthan and guar, but I hear acacia really works wonders. If you're budget allows it, this seems to get good reviews:

http://www.expertfoods.com/package_notStarch.php

Rather than give them a boatload of money to combine hydrocolloids, I've been hoping to find acacia and mix them myself. I've been keeping my eye out for small quantities of good quality acacia at a reasonable price, but, so far, it's been slim picking.

Edited by scott123 (log)
Link to post
Share on other sites
  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...