Turbidity is used as a means of assessing the particulate level in a wine (visual clarity), and from this its suitability for bottling is determined. There are many potential suspended components in a liquid, such as silt, yeast, bacteria, amorphous and crystalline materials that cause turbidity. A commonly used threshold for sterile bottling is < 1 nephelometric turbidity unit (NTU): If a wine has an NTU < 1, it is deemed suitable for sterile bottling in terms of how it will present in the bottle and its likelihood of fouling filtration media, specifically “sterile” membranes and membrane pre-filters. If the pre-bottling wine NTU > 1, and the wine is to be “sterile” filled, then it is recommended that the wine receives extra prior filtration. This may be depth or cross flow filtration in the cellar, or depth filtration on line, depending on the severity of the problem and the cost to the owner of the wine.
The blockage of sterile filtration media, either rapidly, or slowly via an exponential decline, can occur during wine bottling, even though the wine meets a pre-bottling turbidity specification suitable for the chosen filtration media. This article will explore the relationship between turbidity and the filterability of wine.
The purpose of filtration is to clarify or purify a product to meet a given specification for consumer acceptance in terms of microbial and/or physical stability. In general terms, filters fall into two main categories: depth filtration and membrane filtration. The purpose of depth filtration is to retain the primary particulate load, whilst membrane filtration is used to provide a defined barrier. Many forms of filtration are available to the winemaker, each with its own advantages or disadvantages, and these can include earth, pads, lenticulars, tangential flow (cross-flow) and membranes.
Diatomaceous earth (DE) is inexpensive but performs poorly and is hazardous, with a strong impact on wine flavour and colour. Most pads are a blend of cellulose and DE (although some pure cellulose pads are available), are relatively inexpensive, yet contain DE, and so some of the problems associated with DE flow into pad usage, such as disposal and organoleptic impact. Lenticular filters are reformatted pads. Tangential flow (cross-flow) filtration is a system that can remove particulate, is typically automated, and is best suited to cellar filtration. Contrary to popular belief, cross-flow filtration is not a sterile process, since it cannot be integrity tested. Although the utility of cross-flow filtration is very high, it is not a replacement for lenticular filters per se. Membranes are a form of barrier filtration, typically being made from nylon or polyethersulfone (PES). They are challenged with miro-organisms to ensure that the required levels of organism removal are met. Membranes are typically absolute rated, meaning that their filtration efficiency is measurable and stated. Absolute-rated filtration media are those for which an efficiency of higher than 99.98% is stated, which corresponds to a beta ratio (the ratio of particulates larger than the stated porosity retained on the medium relative to the those that pass through) of 5000:1 or better.
New Depth Media
The most recent development in depth media for wine filtration is the 100 % cellulose materials from E. Begerow of Germany (Figure 1; Bowyer, 2012). The new medium possesses several advantages over standard depth media, such as increased physical strength, increased service life, minimal drip loss, less water required to condition, the ability to steam-sterilise, and no citric rinse requirement. Since there is no DE in the medium there is minimal possibility of organoleptic impact on the filtrate. In other areas of food production – for example olive oil filtration – this has been shown to be of great significance, and an additionally an increase of approximately 100 % in terms of throughput (Thomas, 2012) compared with a standard medium was recently observed.
Perhaps the greatest advantage of the pure cellulose medium, however, is in terms of colour adsorption. The absence of DE substantially reduces pigment adsorption to the medium, which noticeably improves wine visual quality parameters. This is especially important for wines containing less pigmentation, or where colour (hue) and colour density are deemed important, as they are for most red and rosé wines.
Figure 1: Standard (left) and cellulose (right) depth media. The beige colouration of the standard medium is due to the presence of diatomaceous earth, which adsorbs colour.
New Membrane Media
Nylon is a versatile polymer, but one that is not necessarily the best in terms of wine filtration applications, even though it is in widespread use. The membranes are symmetrical, meaning that they are of approximately uniform density from entry to exit of the medium. Nylon in its various forms (Figure 2, right) is formed from the reaction of amines and carboxylic acids, the result being a peptide bond like those found in proteins. In the peptide bond the secondary amino nitrogen is bonded to a hydrogen atom, and due to the difference in electronegativity (Bowyer, 2003) between N and H, bond polarisation occurs, generating a partial positive charge on the hydrogen atom. This serves as an anchor point for any localisations of negative charge on other molecules as they pass through or come into surface contact with the medium, increasing retention and adsorption of these materials through hydrogen bonding. This is of particular concern in rosé and red wines, as colour is a key quality parameter (for example, see Francis, 1995), and it is easy to understand that nylon membranes will be inherently colour-retentive.
Figure 2: Chemical structures of polyethersulfone (left) and nylon (right). The nylon structure indicates potential hydrogen bonding to an anthocyanin via the amide hydrogen atom.
Polyethersulfone (PES) is a newer membrane medium that is becoming more widely accepted due largely to the pioneering efforts of Parker-domnick hunter, who have advocated PES for wine filtration applications since 2000 and do not offer a nylon wine filtration membrane. PES offers significant advantages in terms of colour retention, physical robustness and flow rates compared with nylon. The PES polymer (Figure 2, left) critically lacks the ability to hydrogen bond, leading to its almost zero colour adsorption. PES is typically cast in an asymmetric manner (Figure 3), meaning that the membrane is rather like a thin depth medium, which generates significant benefits in terms of flow rate and loading capacity, both of which are higher than nylon.
Figure 3: Cross-section of a PES membrane illustrating the asymmetric nature of the medium. The tighter central region provides the microbial retention capacity, whilst the coarser outer regions provide loading capacity and support structure. Image courtesy of Parker-domnick hunter.
Relative Performance of Media Types
Wine (2012 Malbec) was passed in sequence through a series of filtration media, commencing with coarse grade pad, then tight pad, followed by two 0.45 mm membrane discs. Commonly, 0.65 mm membrane pre-filters would be used at bottling, however in this study a 0.45 mm disc was used in sequence to illustrate that colour binding, as opposed to particulate retention, was taking place. The used media are pictured in Figure 4, and the colour binding data are given in Table 1. A graphical representation of these data is provided in Figure 5.
Figure 4: Spent filtration media after passing 100 mL 2012 Malbec in sequence from left to right as pictured for each train.
Table 1: Colour binding results in the model cellar/bottling train for a 2012 Malbec (100 mL) passed sequentially through 47 mm discs of specified medium.
Figure 5: A graphical representation of the data given in Table 2, illustrating the sequential colour loss in a 2012 Malbec as 100 mL are passed through stated media in sequence.
The same methodology was applied to a 2012 rosé, and the data are presented in Table 2 and Figures 6 and 7.
Figure 6: Spent filtration medium after passing 100 mL 2012 rosé in sequence from left to right as pictured.
Table 2: Colour binding results in the model cellar/bottling train for a 2012 rosé (100 mL) passed sequentially through 47 mm discs of specified medium.
Figure 7: A graphical representation of the data given in Table 3, illustrating the sequential colour loss in a 2012 rosé as 100 mL are passed through stated media in sequence.
We have previously demonstrated that NTU is a poor indictor of wine colloidal loading and capacity to foul filtration media (Bowyer, Edwards and Eyre, 2012), and investigated the potential impact of CMC. In order to further evaluate the influence of colloidal loading to a wine in terms of impact on wine filterability index (FI), investigations were made with concentrate and exogenous tannin. A 2011 Shiraz (NTU 0.56, FI = 7.6) was treated with additions of concentrate (equivalent to 4 g/L residual sugar) and exogenous tannin (25 ppm), the sample left overnight and then tested. The wine returned values of NTU = 1.1 and FI = Fail (> 1000). FI data recorded for the two samples, pre- and post-addition of tannin and concentrate, are given in Figure 8.
Figure 8: FI graphical real-time plots (mass vs. time) of the initial wine (left) and the wine approximately 12 hours after additions of concentrate and tannin. The plot on the right was manually interrupted by the operator due to membrane disc blockage during the measurement (FI > 1000), resulting in the sharp mass rise which forces cessation of data collection.
Colour adsorptivity differences between the cellulose depth medium and standard depth medium were obvious, with approximately four times as much colour adsorbed onto the latter using a highly coloured wine. The nylon membrane medium likewise adsorbed significantly more colour than PES. In a bottling filtration model, the standard depth/nylon media combination was 60 % more colour retentive than the cellulose depth/PES media combination. While virtually no colour loss was recorded for the PES membrane passes, the nylon membranes showed sequential colour adsorption with each exposure, which was visually evident (Figure 4). The fact that both 0.45 mm nylon discs visually appear the same and exhibit very similar quantified colour adsorption indicates that pigmented material is actually being adsorbed and retained as opposed to simply being removed from solution due to large particulate size. The tight standard depth medium also adsorbed a significant amount of pigment from the wine, observable in the bright red colouration of the corresponding second depth disc (Figure 4).
When a lightly coloured wine (2012 rosé) was passed through the different media types the colour adsorption was much more evident due to the lower pigmentation in the initial sample. Total colour adsorption by the standard pad material was again approximately twice that of the cellulose medium. The nylon membrane discs again adsorbed approximately 7 times that of the PES discs. Interestingly, colour adsorption by the nylon discs visually approximates that observed for the 2012 Malbec (compare the nylon discs in Figures 4 & 6), indicating that nylon membranes are far more damaging to rosé in terms of colour adsorption than PES, since the amount of colour adsorbed on the nylon membranes is similar in both red and rosé wines, irrespective of wine type.
Colloidal additions to wine have the capacity to impact negatively on filtration media (Czekaj, López & Güell, 2000; Vernhet, Cartalade & Moutounet, 2003; Vernhet, Pellerin, Belleville, Planque & Moutounet, 1999), as evidenced by additions tannin and grape juice concentrate to wines destined for membrane filtration. Typically colloidal additions do not elevate wine NTU, further emphasising the importance and superior relevance of filterability (FI) measurements as opposed to turbidity (NTU) measurements when additions of this type are to be made. 2011 was a poor vintage in South Australia for some reds. These wines can prove problematic during bottling due to the presence of glucans released by Botrytis into the fruit. These polysaccharides are not degraded by pectolytic enzymes and are known to cause membrane fouling, but also have shown themselves capable of fouling depth media as well (Edwards and Eyre, 2013).
Clear differences exist in terms of colour adsorption between the different types of filtration media. Cellulose depth media out-perform standard depth media in terms of colour adsorptivity, and with respect to membranes PES is seen to be vastly superior to nylon in terms of colour adsorption. FI has proven to provide a vastly superior estimate of the likely impact of a given wine on filtration media than NTU. The reliance on NTU as a means of evaluating wine suitability for membrane filtration is likely to lead to significantly higher filtration costs for bottlers due to increased media loading. Wines containing elevated colloidal levels, or where colloidal additions have been made, can lead to media blockage and higher bottling costs. The move to mainstream use of filterability determinations as an adjunct to turbidity measurement will service both winemakers and bottlers alike due to the greater understanding of wine filterability provided.
Paul K. Bowyer, Greg Edwards, and Amelia Eyre
Dr Paul Bowyer is the Regional Manager (SA/WA & SIHA) for BHF Technologies (Blue H2O Filtration). Greg Edwards and Amelia Eyre are the Laboratory Managers at Vinpac Angaston and Vinpac McLaren Vale respectively.
Bowyer, P. K. (2003) Molecular polarity – it’s behind more than you think, The Australian and New Zealand Grapegrower and Winemaker, November issue, 89-91.
Bowyer, P. K. (2012) Brettanomyces removal with Becopad, The Australian and New Zealand Grapegrower and Winemaker, March issue (578), 62.
Bowyer, P. K., Edwards, G. and Eyre, A. (2012) NTU vs wine filterability index – what does it mean for you? The Australian and New Zealand Grapegrower and Winemaker, October issue (585), 76-80.
Czekaj, P., López, F. and Güell, C. (2000) Membrane fouling during microfiltration of fermented beverages, Journal of Membrane Science, 166, 199-212.
Edwards, G. and Eyre, A. (2013), Vinpac International, Angaston SA,personal communication.
Francis, F.J. (1995) Quality as influenced by colour, Food Quality and Preference, 6, 149-155.
Thomas, S. (2012), The Big Olive, Tailem Bend, SA, personal communication.
Vernhet, A., Cartalade, D. and Moutounet, M. (2003) Contribution to the understanding of fouling build-up during microfiltration of wines, Journal of Membrane Science, 211, 357-370.
Vernhet, A., Pellerin, P., Belleville, M.-P.., Planque, J. and Moutounet, M. (1999) Relative impact of major wine polysaccharides on the performances of an organic microfiltration membrane, American Journal of Enology and Viticulture, 50(1), 51-56.
Have you ever wondered why it is that some filters cost more than others? Or perhaps how two filters that physically look identical can be vastly different in price? Or even why it is that filters that seem to be very similar in porosity (e.g. 0.5μm vs. 0.45μm) can vary enormously in performance and price? In this short article we will explore some of the reasons for these apparent anomalies in the world of filtration products.
Filter Types and Construction
There is a wide variety of filters available on the market, and previous publications have discussed many of these differences in terms of wine filtration products (Bowyer, Edwards and Eyre, 2012; Bowyer, Edwards and Eyre, 2013; Bowyer and Edwards, 2014). Differences between filters are not always apparent, although with some understanding of certain factors these differences can be ascertained, and for this a technical data sheet is typically required. Filter construction varies according to the design function of the filter. While it is possible to filter pretty much any liquid with 1 filtration stage, this is not a cost effective enterprise. (For example, sterile wine filtration can be achieved by simply pushing the wine through a 0.45μm integrity-tested membrane, which is fine if you don’t mind installing a new wine membrane every 5 minutes) For this reason filters are constructed according to purpose.
Nominal vs. Absolute
This terminology refers to the manufacturing tolerances employed during construction of the filter. A nominal filter will have broader manufacturing tolerances, and so the variations in effective porosity are greater than for an absolute filter. These differences can be accurately defined, and for absolute filters this is referred to as an efficiency rating or a β-ratio. The β-ratio of a filter is the ratio of particles captured to that passed for a given grade. For example, if a 1μm filter retains 1000 particles of a size greater than 1μm but allows 1 through for every 1000 retained, the filter is said to have a β-ratio of 1000. This can be converted to a % efficiency by using the formula below:
Efficiency = 100(β – 1)/β
|1000||99.9||For some manufacturers|
Table 1: Conversion data for β-ratio and efficiency.
An efficiency of 99% sounds pretty good, right? This value is perhaps acceptable for a nominal filter, but not for an absolute filter, and this is where the technical data sheet becomes invaluable. Some manufacturers do not make it clear whether their filters are nominal or absolute. Others do not present this information in a format that is easy to understand. Yet others consider an absolute filter to have a β-ratio of 1000 (99.9 % efficiency), but for most this figure is 5000 (99.98 %).
As an example, consider the filters in Figure 1. The upper filter is an Amazon SupaGard, a nominal depth filter often used for water filtration. The lower filter is an Amazon SupaSpun, an absolute depth filter, also used for water filtration. Both filters are made of spun-bonded polypropylene of the same porosity rating and have a polypropylene core. Both filters physically look the same and have similar void volumes, yet their efficiencies and performance vary due to the different ways in which they are fabricated, and this is reflected in the price difference between them. The correct filter for the job depends on the application, desired outcomes for the process and the budget of the customer, and it is BHF’s job to determine which is fit for purpose.
Figure 1: A comparison of two visually indistinguishable depth filters, one nominal (upper) and one absolute (lower).
To complicate matters further, no standard method exists for determining filter performance. Moreover, β-ratio will change throughout the life of a filter, since as it loads up during usage the effective porosity decreases, meaning the filter becomes more effective as it blocks up. The end result is that any nominal filter will be less “efficient”, in the sense of particulate retention capability, when it is new.
In terms of wine depth filters, these are typically in sheet or lenticular format. In almost all cases these filters are nominal, which is why performance across brands is not as simple as comparing sheet materials of the same stated porosity. Porosity typically spans a range (e.g. 1 – 2μm) for any given grade of filter sheet, to encompass the spectrum of retention capacity of that nominal grade (Figure 2). Further, some grades of sheet are described as “sterile”, but this again is a nominal term, since only an integrity-tested membrane can truly be considered to be a sterile filter.
Figure 2: Excerpt from a Becopad porosity chart, indicating both nominal porosity (μm, on the ordinate axis) and nominal flow rates (Lm-2min-1, on the abscissa). Porosity and flow rate are usually proportional.
Construction Material & Filter Type
Filter construction varies according to application. Situations involving harsh chemical treatments require highly resistant production materials, such as polypropylene. If no harsh chemicals are being used (caustic soda being a problem in some applications in this regard), a glass microfiber medium can be extremely effective. End caps on filters can be reinforced with glass fibres to provide increased strength is repeated sanitisation is required, such as for wine membranes. We have discussed in detail the differences between nylon and polyethersulfone (PES) wine membranes in a previous publication (Bowyer, Edwards and Eyre, 2013). Aside from chemical differences between these two polymers, there are significant physical differences that impact on the way the filters function, in terms of colour stripping, flow rates and even integrity testing.
If the application requires a high dirt-holding capacity, typically a depth filter will be used, such as the lenticular filters commonly used for cellar filtration or as wine membrane pre-filters. These have some thickness to the filtration medium, designed to entrap and retain particulates. The very nature of this physical entrapment mechanism makes the regeneration of depth filters quite difficult, unless the retained particulates can be dissolved and pushed through the medium, in which case a forward flush is equally valid.
If the application requires a higher flow rate, or of the filtration stream is not overly burdened with particulates or microorganisms, a pleated filter is more suitable, as it has less depth capacity but presents a much higher surface area to the flow stream, which allows a higher flow rate. Wine membranes represent the extreme of this philosophy, in that they have very high surface areas, facilitated by a thin layer and excessive pleating, but very little depth.
When visualising a membrane cross section it helps to think of a very thin sponge. In the case of nylon membranes, they are typically cast in a symmetrical manner in terms of cross section, so when they encounter load and block, they block on the surface and minimal “depth” in the membrane is used. PES membranes are typically cast asymmetrically, with coarser outer regions and a progressively tighter core, which allows them to not only flow faster but also to exhibit some “depth” capacity, a characteristic not usually attributed to membranes (Figure 3). This in turn can result in a longer service life.
Figure 3: A sectional comparison of nylon (left) and PES (right) membranes, illustrating the typical differences in symmetry between membrane types. Note the tighter, inner section of the PES membrane.
Filtration porosity is typically expressed in terms of microns (μm). Unless two particulate filters being compared are absolute-rated with the same β-ratio, porosity comparisons are relatively meaningless. Filters come in a range of porosities, and so can be optimised for performance according to the task at hand. In terms of wine this pertains to the specific loading that the wine will present to the filter. A question that is often asked is “what grade of filter should I use to filter my wine at X NTU to get it down to Y NTU?”. This is impossible to answer accurately without an understanding of the particle size distribution and colloidal status of the wine. Usually a filtration grade estimate is made based on a combination of experience and historical data.
Considering only the stated porosity of an absolute-rated filter, there should be little difference between a 0.5μm particulate filter and a 0.45μm membrane, since there is only a difference of 0.05μm, yes? Not so! Particulate filters and membranes are evaluated for efficiency in two different ways, and they are not transposable. Particulate filters use β-ratios, but membranes use log reduction values (LRV)
Log Reduction Value (LRV)
A log reduction value is another way of expressing filter efficiency. It is typically used to describe the efficiencies of membranes at removing micro-organisms, as the numbers become too high for β-ratios to be conveniently used. For example, the Parker Domnick Hunter Bevpor PH 0.45μm wine membrane is fully retentive of Saccharomyces cerevisiae, and also has a stated LRV of 9.1 for the organism Pseudomonas aeruginosa, meaning that the β-ratio for the latter organism is 109.1, or 1,300,000,000:1, or 99.9999999% (Figure 3). LRV’s are often expressed for several micro-organisms on a technical data sheet, but the organisms tested are not common to all filter manufacturers, and a direct comparison requires commonality of test organisms and conditions. LRV’s are also more meaningful for organism retention since organisms are able to deform, whereas particulates typically are not, and so a measured organism challenge yielding an LRV will provide more meaningful data than a simple particulate retention test.
Figure 4: Excerpt from a technical data sheet for the Parker Domnick Hunter Bevpor PH indicating LRV’s for several organisms across 3 levels of porosity.
Integrity testing is a method whereby a membrane can be confirmed as being integral, with no holes or leaks in or around the filter. The filter is wet out completely, then the upstream side of the membrane is sealed off and pressurised, and the pressure drop over time measured, often with the use of a specialised pressure measuring device called an integrity tester. The gas (usually N2), will slowly diffuse through the wet membrane to the open downstream side of the filter at a defined rate. All test parameters are specific to the housing and upstream pipework volume, filter type, size and porosity, and the test is strongly affected by temperature. Provided the pressure loss over the test period is below the calculated allowable value, the filter is declared as integral and fit for purpose. Since this process cannot be applied to cross-flow filters, they should be considered as non-sterile filters only.
Comparisons of different filters is not a straightforward process, and several factors should be taken into consideration. Ultimately, testing and/or process trials must be undertaken to evaluate true cost-effectiveness of any filter. A filter that is cheaper to buy initially may ultimately lead to higher ongoing costs in terms of filtration performance and associated staff time allocation for change-outs.
Paul K. Bowyer