The Crossflow Comix

by Clark Smith

Don't blame me. I didn't come along til late in the game. Crossflow filtration, whose roots are really in the wine industry, is finally enjoying a return to favor. But it wasn't easy.

The enthusiasm with which crossflow suppliers invaded our shores in the 1980's overcame the skepticism of many progressive winemakers. However, this was short-lived and a decade later, the industry would be littered with the corpses of failed projects, and R&D divisions of filter companies fled to greener pastures, leaving behind an industry still largely ignorant of their potential and a small group of former allies now turned confirmed skeptics.

What happened? The answer reveals much about what wine really is and how wine producers perceive it, uniquely, as it turns out. To understand the mind set underlying this comic catastrophe, we should trace the origin of these projects.

The German Wine Boom and the Birth of Membranes

Sterile filters by Seitz got it all started. After the war, the German economy needed an easy-to-market new wine style and off-dry wines became the style. Germany's new sweet rieslings were easy to like, and became consumer darlings, the original Afighting varietals,@ that helped the industry recover.

Nominally sterile cellulose pads got the ball rolling in the late 1940's, followed in the 1950's with bubblepointable cartridges of polyamide and polysolfone. The idea spread to other beverages, medicine and electronics. Sterile filters taught plastics engineers how to control pore size. There was no limit to achievable tightness, but below 0.1 microns, the filters were unusable because dissolved molecules started to foul the membranes. Perfectly clear solutions wouldn't filter but engineers were working on that solution.

Crossing the Submicron Barrier

Around 1960, in Cambridge, Massachusetts, a new idea had taken hold. Molecular separation of a fluid stream had endless uses but how to avoid molecular fouling was still the problem. The vital hints on how actually to accomplish this came from the human body. Medical researchers looking to replace the function of failed kidneys had managed to create dialysis membranes which mimicked human tissue, but they didn't work. The researchers realized that blood flow through capillaries was keeping the membranes clean.

The concept was that instead of pushing the flow stream through the filter, you direct it across the filter, then back to the tank. The scrubbing is accomplished with a very turbulent fluid flow, and is squeezed by the filter into a recycled Aretentate.@ What goes through the filter is Apermeate.@ You need large areas of membrane surface and you waste most of the energy of the flow keeping the membrane surface clean. However, although it is expensive, it is also very lucrative. This new type of filtration was dubbed Atangential flow@ or Acrossflow.@ The M.I.T. researchers figured out how to lay down integrally-skinned asymmetric polyamide membranes that would do the job.

Two early markets emerged. The medical applications saved a lot of lives and provided high margins in a small market but the big market was desalination, whether drinking water from the sea, or any well or brackish stream. The basic RO membrane can convert seawater into drinking water, however. RO is expensive and was costing a lot of money to get the job accomplished. Purification of brackish and sea water has fueled crossflow development since the initial idea was formulated and the U.S. Navy has been in the forefront for obtaining any new ideas that the industry has developed.

Enter the Cheese Heads

By the early 1970's, more chemically robust (and therefore sterilizable) membranes opened up the potential to consider food processing applications for crossflow and dairy producers jumped on the idea. Cheese is made by curdling milk, but the liquid whey stream comprises around 90% of the milk invested, and is still rich in protein. Even though RO proved effective to squeeze the water out of this stream, the process was expensive and it concentrated salt in the retained product, also, proteins were big. Minneapolis companies developed looser membranes to give better flows, while still retaining protein and allowing salt to wash through. Ultrafiltration had arrived.

At the same time, the notion of thin film composites was developed through U.S. government funds, opening up the possibility that different pore sizes might support one another in exotic ways. Several different filter designs had achieved success and the survivors (see box) were spiral wound, flat sheet, hollow fiber, tubular and ceramic, each with its market niche.

By 1980, water purification was big business in locales where results were more important than money, such as Saudi Arabia, Catalina Island, Las Vegas, and on board ships. RO concentration had taken hold in the concentration of paints, paper, and fruit juice and ultrafiltration became an integral part of the manufacture of cheese, enzymes and pharmaceuticals.

Bringing It All Back Home

Flush from these early successes, crossflow engineers descended on the California wine industry. They perceived before them a rapidly growing industry spending a hundred million dollars a year to gear up the production of a high value fluid product. It was an industry full of earnest, intelligent decision makers with the same sanitary fittings and the same dedication to improvement they'd seen in Wisconsin dairies and countries which needed pure water.

Black Eyes of the >80's

The near absence of engineers in high level winery positions might have tipped off the newcomers from Millipore, Romicon, and Koch. These were the filtration experts and they thought in terms of flows rather than flavors. They had little appreciation for the goals of winemaking, where yield and efficiency are quite secondary to quality. Wine quality wasn't a well-stated problem so design did not take this into account. Since the engineers couldn't define it, they farmed it out, using winemakers and sensory panels for analysis and feedback. This often merely obscured innate design flaws, so that the ultimate failures were all the more spectacular.

A classic example was the universal failure of pilot scale testing. A twenty liter experiment might demonstrate the engineering concept of an application, but the wine never tasted any good. The obvious hope was that if we could just scale up to production size, the problems of oxidation, high membrane surface-to-volume and so on might just vanish. Some wineries started taking risks and so did some membrane companies.

Much of the effort in the early years went to replacing bentonite. The problem looked just like whey separation, which ultrafiltration had deftly conquered, and wine was more valuable. But in whey, the product is the retentate, the protein concentrate, and nobody really cares if a bit of protein slips past into the permeate. As it turned out, the unstable proteins are small ones. The result was wine stripped of flavor and still unstable!

Another group of researchers thought the wine industry might be the place to take on the great unsolved crossflow problem: clarification. At first glance, this seems unnecessary since sub-micron filtrate is always dead brilliant so the problem is solved! But most existing filter formats had small spacings to promote scrubbing and fouled easily. The wine industry was a particularly poor place to perfect a format because there the challenge of good flavor passage was added. Part of the difficulty these early innovators faced was that they were trying to outcompete traditional methods which already worked pretty well. It would not be until brand new capabilities were uncovered that crossflow would make any progress in the wine industry.

By 1990, the only place anybody was selling systems was for non-alcoholic wine production. The industry, however, wanted all possible distance from these producers, who were perceived as having traded away any defensible notion of what wine is, substituting, if you will, technology for spirituality, to offer a soul-less product to a few consumers who wanted it for all the wrong reasons. Nobody was impressed that it could be done and the marketplace as a whole concurred.

The 90's: Applications that Work, Offered as Services

Eventually, successful crossflow applications, developed on the whole by winemakers, began to emerge. Working independently, Barry Gnekow, Dr. Richard Carey, and the Heubline R&D group began using ultrafiltration on red wines, not for protein removal, but as an alternative to protein fining for tannin removal. The process reduced color but not aroma, and created a tannin or color concentrate, a saleable byproduct, in place of lees. The equipment also proved effective in removing browning and hard press character from whites.

In the mid-80's, the French water giant Degremont began working with wineries in Bordeaux to develop RO juice concentration as an alternative to chaptalization. This didn't work when cool weather stalled maturity, because the acid and green tannins were too concentrated. But when ripe fruit had been rained on, the prospect of removing it rapidly gained RO a premier reputation there.

After participating in the 1980's crash-and-burn, your author finally came up with something worthwhile in 1991, while reluctantly chained to an RO by a consulting client who wanted to produce non-alcoholic wine. Although the original intent was never actualized, I did manage to develop an idea which led to some patented applications for removal of small MW compounds from wine:

  1. Create flavorless permeate.

  2. Pull out the compound(s) of interest (acetic acid, ethyl acetate, ethanol, aldehyde, sulfide, etc.) by appropriate means (e.g. adsorptive resin, distillation or sparging),

  3. Recombine the purified permeate.

The main use of this to date has been the reduction of volatile acidity. This was an easy application to commercialize, because there was no competing technology and clients had nothing to lose by trying it. We were also very lucky when ATF ruled that although an RO is regarded as a still, the recombination aspect meant that no distillate was produced and a Distilled Spirits Plant license was not necessary. The main challenge, as always, was to find the right filter and success put us into hundreds of wineries.

Clients now had a way to get a Afree look@ at RO on real wine, and the results were surprisingly good. This made it possible for Vinovation to sell (along with our colleagues at the Spinning Cone) the more powerful idea of uncoupling harvest decisions from Brix, a sort of reverse on chaptalization. We extract excessive alcohol post-fermentation via the same 1,2,3 scheme, allowing, for the first time, California grapes to be picked at optimum flavor to make wines of normal alcohol.

Fearless Predictions: On-Site Integrated Systems

Now that crossflow has earned a place in wine production, I see the coming decade as one in which wineries will acquire their own equipment. VA removal is naturally a service business. It takes a quarter million dollars of equipment to reduce 4000 gallons of wine per day to half the starting level, not an investment most wineries will make for an occasional problem.

Most other crossflow applications, however, lend themselves well to onsite processing. A one percent alcohol reduction takes a tenth of the equipment of a typical VA task. It is also critical to dial in a Asweet spot,@ the alcohol level at which flavors come together and the style works. Every wine is different and some may have two or three such spots. Often the levels in between don't taste nearly as good so an in-house RO is also handy for juice concentration during harvest, should the need arise. It also provides a means for rapid, reliable cold stabilization at room temperature. Substandard winery water can also be purified in great quantities by even a small unit.

In-house ultrafiltration to reduce polyphenolics is an economical and compelling alternative to egg whites and other protein fining agents and ultrafiltration can be an incorporated option built into an RO system. Wineries serious about juice concentration might consider tandem capabilities, for example, a UF producing temporarily clarified juice for an RO to concentrate prior to solids recombination. The PCI tubular format (see below) is another alternative offering a cleanable RO membrane for juice, which can also be adapted for UF.

For most winery applications, the best format choice is probably spiral-wound, which gives vast amounts of surface area at low cost. When they work, they are excellent and when they finally foul, you replace them. The exception is when you are working with seriously turbid material, such as juice or lees. In these cases, preclarification is a must, usually with large grade diatomaceous earth.

The Holy Grail of Out-Competing D.E.

It isn't clear that this will ever happen. A crossflow system that clarifies without any molecular separation isn't necessarily possible. Paradoxically, the scrubbing principle works less well as pore size approaches particulate size. Looser membranes tend to foul and big red wines are the worst culprits. There really isn't much difference in size between a tannin-anthocyanin complex and a bacterium. The prospects are better for whites and light reds. D.E. is cheap and inert. We will probably keep using it as long as we can. However, Silicosis toxicity to workers and landfill issues may some day remove it as an option. The leading format to replace D.E. is the PCI tubular design, a half-inch replaceable membrane tube that is inserted in a porous stainless tube support. This format can handle toothpaste-thick lees and comes in a wide range of pore sizes. At this time, no one particular application has the edge and it may well be the one goal that never will be realized but there will be many who couldn't resist the quest.

Devil or Angel?

I can scarcely conclude this wine opera without taking stock of the philosophical terrain in which we winemakers now find ourselves.

The Central Debate about crossflow applications and other high tech wine production innovations is not anymore about whether they work, it is about whether we will go to hell if we use them. This speaks well for our industry, which few of us entered for the money.

These philosophical concerns take many forms. I have spent a day agonizing with two rabbis about whether R.O. permeate, which looks like water, should be considered to be wine. If so, Leviticus requires that no gentile shall look upon kosher permeate and clear plastic hose is out of the question.

At least they have a book. Most winemakers today search their souls only, for winemaking is a craft involving much intelligent compromise. Reverse osmosis submits wine to terrific pressures and Ashear forces,@ but it also opens doors to natural yeast fermentation, lower sulfites and longer hang times, all old ideas which have been considered too risky of late.

The nexus of the Central Debate these days is Paris, where the O.I.V. struggles to decide whether to recommend allowance of Aosmose inverse@ for use on wines. Philosophical questions which weave through the discussion include agricultural vs industrial, juice vs wine, chemical vs filtration, olfactive vs gustatative, corrective of natural vs fabricative defaults. That it works is taken more or less for granted, in the same way that procreative ability is a side issue in assessing one's daughter's suitors. 


Terminology and Equipment Design

The tangential flow family of filtrations is used to exploit pore sizes smaller than possible with traditional Adead-head@ filtration, whose limit is about 0.1 microns, or about 500,000 daltons molecular weight (MW). The domain below 500,000 MW is by convention divided approximately as follows:

MW range

Crossflow Clarification



   Tannin & Browning

   Protein Removal








Reverse Osmosis

   Volatile Acidity Reduction

   Alcohol Adjustment

   Juice Concentration


All of these formats employ the strategy of pumping the feed stream at high velocity across rather than against the filter surface. This continually scrubs the surface to remove fouling materials. The majority of the feed stream does not pass through the filter, but is retained upstream and returned to the tank. This stream contains all the high MW compounds and is called retentate. The portion which passes through the filter, containing only low MW components, is called permeate.

Format Type


Flat Sheet

An early format prone to bleed-through due to its proliferation of gaskets. Easy to sterilize, though, and still the only certified method for some pharmaceutical applications.


Hands down cheapest, most reliable, but the most prone to fouling, and the hardest to sterilize.

Hollow fiber: AOutside In@ - The most robust tube structure, wherein a bundle of tubes receives permeate from a retentate stream in the enclosing tube.

Economical in hard water purification. Turbulent flow to scrub the surface is hard to achieve.

Hollow fiber: AInside Out@ - Expensive due to poor capillary lifetime, pressurized from within.

Good ability to handle solids; concentrate to toothpaste. Poor maintenance record: filters would fail in woefully short lifetimes.

Tubular: essentially large ID hollow fibers, with a steel external support.

Expensive but robust. Difficult to achieve highly turbulent surface scrubbing. Nominal pore size therefore gives more separation than other formats.

Ceramic: an expensive but rugged format, prone to breakage if moved.

High temperature tolerance. Autoclavable. Most competitive when flows can be processed at high temperatures.


{Clark Smith's article "The Crossflow Comix" was originally published in the 

March/April 1998 issue of Vineyard & Winery Management magazine}



< Winemaking Fundamentals



Home  |  WineQuality Enhancement  |   The State-of-the-Art Estate Winery  |  Services at our Facility  |   Winemaking Fundamentals  |  CostSavings  |   Vineyard Enology  |  International Services  |   About Us  |  Contact Us  |   International Links

Vinovation, Inc
ph:(707) 824-7900  •  fax:(707) 824-7905  •  email: info@vinovation.com
office: 652 Petaluma Ave, Sebastopol, CA 95472
facility: 1365 Gravenstein Hwy. South, Sebastopol, CA 95472