Wine naturally precipitates solids during fermentation and wine storage, and these traditionally settle out over time in a natural clarification process. Heat and cold stabilisation are often used to speed up clarification. Producers may also choose to filter wines to achieve a greater level of clarity for a particular wine style or to remove microorganisms from the wine to ensure microbiological stability after packaging. Some contract bottling facilities may also require wine to be provided at a certain level of clarity if filtration is requested. Bowyer (2018) reports that around 80% of all wine, by volume, is now sterile filtered, compared to just 20% ten years ago.
Particulates in wine can be crystalline (commonly calcium or potassium bitartrate), amorphous/colloidal (protein, tannin, polysaccharides or metal complexes), microbiological (yeast, bacteria or mould) and other materials including silt, residual grape matter and fining or clarifying agents. The human eye can detect large particulate matter >100 µm but not yeast and bacteria that are much smaller in size, around 0.5-5 µm in diameter.
Turbidity and different stages of filtration
The AWRI generally recommends using turbidity as an initial indication of clarity. Wines just after fermentation contain both large and small particulates and colloidal material and are often well above 200 turbidity units (NTU). Wines at this stage are usually settled, fined or centrifuged to improve clarity. Wines between 100 and 200 NTU may undergo macro-filtration steps from coarse metal screens to earth filtration, with a finer earth filtration conducted between 60 and 100 NTU. Once a wine is below 60 NTU, particulate matter is much smaller in size and pad filters (often in conjunction with earth filtration) or cross-flow filtration can be used, with final micro-filtration steps such as membrane filtration conducted to remove yeast or bacteria.
Wines are considered ‘cellar bright’, or visually clear to the eye at below approximately 30 NTU. Some bottlers will require tighter turbidity clarification levels for filtration at bottling and some export countries have maximum turbidity specifications. A commonly used threshold for bottling is <1.0 NTU however an NTU reading of <1.0 is not necessarily a guarantee that the filter medium will not block up and thus often a filterability test is used in conjunction with turbidity measurement. Correspondence between turbidity measurements (NTU) and appearance is summarised below from Ribereau-Gayon et al. 2006.
Wine type
Brilliant
Turbid
White wine
<1.1
>4.0
Rosé wine
<1.4
>5.8
Red wine
<2.0
>8.0
Filterability index
The filterability index (FI) of a wine is an indication of the time needed to block a specific filter medium during filtration. A typical manual set-up is shown below, although semi-automated systems are becoming more common. The systems examine the flow-rate of wine through the filter at a constant pressure, and at the temperature of the wine to be filtered in the cellar.
One litre of wine is passed through a 25 mm 0.45 μm pore size membrane filter at a constant pressure (usually 2 bar/200 kPa) and collected in a graduated beaker or measuring cylinder, generally over a time period of 5-10 minutes. A stopwatch is used to time the filtration process and the volume of wine filtered in each 30 second period is recorded.
The filterability index (FI) is calculated as the ratio of the volume obtained between 30 and 90 seconds and the volume obtained between 120 and 180 seconds, as follows:
Interpretation of FI results
FI Ratio
Interpretation
1.0-1.5
Good
1.5-2.0
Moderate
>2.0
Poor – difficult to filter/may require prefiltration
If the wine being filtered blocks the filter during this time, the volume of wine that has been filtered should be noted. If <200 mL has been filtered, this also indicates poor filterability.
Alternative method for calculating filterability index
An alternative method (Laurenty 1972; detailed in Descout et al. 1976) is to measure the time taken to filter two different volumes or masses of wine, say 200 mL and 400 mL. The FI equation then becomes:
FI= (Time (seconds) to filter 400 mL) – 2*(Time (seconds) to filter 200 mL)
The time to filter 400 mL is usually more than double that of 200 mL. The wine is suitable for filtration if the FI is less than 20.
It is best if the same filter medium (membrane) can be used for the FI reading and the commercial filtration. Many laboratories, however, will use a standard membrane filter in a laboratory setting despite the cellar filtration using a different filter. In these cases, the FI is still useful to give an indication of the general ability to filter the wine, or any potential issues that might be encountered with the wine during filtration (e.g. polysaccharides) but may not replicate the exact conditions or flow rates experienced during bottling.
Note that results from the two methods for calculating FI cannot be directly compared to one another – they are calculated and interpreted differently.
Types of filtration
A number of different types of filters are used in winemaking. Coarse depth filters have traditionally been used as the first stage of filtration to remove suspended solids. Traditional depth filters have used an earth or cellulose matrix applied to a rotary drum vacuum (RDV) filtration unit or plate-and-frame filter housing. Depth filters retain particulate matter but generally do not remove microorganisms from wine. These filters have gradually been replaced by centrifuge and cross-flow filtration alternatives, mainly to avoid the use of diatomaceous earth. Following the first stage of filtration, wines are often filtered through sequentially tighter pad depth filters. Lenticular filters are pad filters reformatted into a filter housing. Surface filters such as membrane filters are used as a final filter to remove smaller microorganisms.
Absolute versus nominal filter ratings
Nominal filters have a wide range of pore sizes. The nominal micron rating indicates the ability of a filter to retain a nominal amount, or remove a minimum percentage (usually 50-90%) of material, of the rated pore size.
Absolute filters are made to have a more regular pore size. The absolute micron rating indicates the maximum pore or opening size of the filter medium. Anything larger than this pore size will not pass through the filter and will be retained on the filter surface.
What is the difference between depth and surface filtration?
Adapted from Baldwin 1992. Pictures reproduced with permission from Winetitles.
Depth filters are not absolute filters because they do not have a precisely defined pore size or structure. They remove particles that are larger than the aperture or pore size of the filter by mechanical retention of the particles, by adsorption, or by an electric charge that can attract and trap smaller particles, generally within or inside the filtration medium. They are used for high solids removal and high throughput, with diatomaceous earth, perlite and cellulose pads falling into this category. There are varying grades of diatomaceous earth, with particle sizes ranging from 2-40 µm. The rating is also affected by flow rate, temperature, pressure and juice or wine viscosity. Today many winemakers are bypassing depth filtration and using clarification techniques such as centrifugation, flotation or cross-flow filtration.
Surface filters are generally absolute filters with precisely defined maximum pore sizes and these include 0.65 µm and 0.45 µm or ‘sterile’ filtration. They remove particles like a sieve, with larger particles retained on the surface of the filter, which gradually blocks the flow. Absolute filtration is useful for sterile filtration at the bottling line and to guarantee removal of microorganisms. Prior filtration with a depth filter and other surface sacrificial filters at a higher pore size is usually required to prevent clogging the surface of a membrane filter.
Note that filters generally refer to remove of particulate matter rather than microorganisms. Many membrane filter technical data specifications will now also indicate ‘log reduction value’ (LRV) for specific microorganisms, when the filters are challenged with a standard 107 colony forming units per cm2. Fully retentive (FR) denotes the filter has restricted all cells passing through the membrane.
Cross-flow filtration
Cross-flow filtration is a surface filter that allows high throughput due to the tangential nature of the filter that continually cleans the filter surface. Most cross-flow membranes are only nominally rated at 0.2 µm which is theoretically tight enough for sterile filtration and removal of microorganisms but as it is only a nominal rating, and can’t be pressure tested for integrity before use, this means that wineries will still need to use a membrane filter if they want to achieve a sterile filtration.
What level of filtration do I need?
The following table collates typical recommendations from filtration media suppliers, including NTU levels against their filter and nominal micron rating, and is intended as a guide only. If the NTU of the wine is known, then it should be straightforward to determine which filter pore size to use. Contact the manufacturer for more detail.
Filtration purpose
NTU
Cuno sheets1
Seitz sheets2
Ekwip grade3
Coarse filtration
First racking
35-100
5H (5.0 µm)
10H (2.0 µm)
K900 (8 µm)
Z3 (2.0 µm)
Z5 (1.0 µm)
Polishing filtration
Second racking
12-35
30H (1.0 µm)
50H (0.8 µm)
K250 (4 µm)
Z5 (1.0 µm)
Z6 (0.8 µm)
Final filtration for dry wines
1.0-3.0
50H (0.8 µm)
60H (0.5 µm)
K200 (3.0 µm)
K150 (2.0 µm)
Z6 (0.8 µm)
Z7 (0.6 µm)
Filtration for sweet wines, higher risk wines or final polish pre-membrane filtration
It is important to test the integrity of a filter membrane; that is, to confirm that there are no holes or leaks in or around the filter (Bowyer 2015). Generally non-destructive techniques are used, such as the bubble point test or the pressure-hold test.
The ‘bubble point’ is the gas pressure at which the surface tension of water in the pores of a saturated filter is overcome and gas is allowed to pass through the pores. It is directly dependent on pore diameter (Zoecklein and Fugelsang 2018). The bubble point test can be conducted at the start of a bottling run before any wine is run through the filter and it shows the integrity of the filter being used and if any leaks are present.
The bubble point test will determine the pressure at which a continuous stream of bubbles is initially seen downstream of a wetted filter under gas pressure – normally nitrogen. Nitrogen is applied to one side of the wetted filter, with the tubing downstream of the filter submerged in a bucket of water. The filter must be wetted uniformly such that water fills all the voids within the filter media. When gas pressure is applied to one side of the membrane, the test gas will dissolve into the water. Downstream of the filter, the pressure is lower. Therefore, the gas in the water on the downstream side is driven out of solution. At some point, the pressure becomes great enough to expel the water from one or more passageways – establishing a path for the bulk flow of air. As a result, a steady stream of bubbles should be seen exiting the submerged tubing. The pressure at which this is a steady stream is noticed is referred to as the bubble point.
Note that the filter must be thoroughly and uniformly wet and the pressure recorded should be the pressure at which there is a steady stream of bubbles. If failure occurs, open the housing and check that the filter is installed correctly. Re-test and if it still fails, then replace.
Bubble points for typical wine filters (using air as the test gas)
65 µm membrane water 20.3 psig = 140 kPa
45 µm membrane water 26.1 psig = 180 kPa
If the recorded pressure is greater than or equal to the minimum expected bubble point listed above, the filter passes the integrity test. If the recorded pressure is lower than the minimum bubble point listed above, the filter has failed the integrity test.
Pressure hold testing is conducted with a highly accurate gauge to monitor upstream pressure changes caused by gas diffusion through the filter. Because there is no need to measure gas flow downstream of the filter, any risk to downstream sterility is eliminated.
Use a 0.05 kPa decrease in pressure at 140 kPa as a test parameter. That is, if the system loses more than 0.05 kPa, then the filter fails the integrity test.
Filter composition and flavour/colour stripping
Depth filters are typically made from diatomaceous earth, cellulose or a combination. Both materials are polar and bind to polar colour and phenolic components in wine and thus some colour loss can be expected.
It is a long-held belief in the wine industry that tight surface filtration such as membrane or sterile filtration below 0.45 μm will strip the aroma and colour from ‘big’ red wines. For this reason, many winemakers avoid the use of sterile filters in wine production, which can cause uncertainty about the microbial stability of the finished wine. Most modern day surface filters are made of nylon, polyethersulfone (PES) and polyvinylidene fluoride (PVDF). These are non-polar and thus cause minimal adsorption of colour from wines being filtered (Bowyer et al. 2013)
Surprisingly there have been few studies investigating the impact of different levels of filtration on wine colour and sensory characteristics. The question of whether the colour would drop out anyway over time is also raised. A study at UC Davis is currently investigating if filtration strips out colour and has any impact on flavour over time (Bohanan et al. 2011). Current results after six months show no sensory or chemical differences between control and filtered red wines.
The AWRI has also investigated wine texture and the role of macromolecules, and if these are removed by filtration. Samples of four commercial wines (Cabernet Sauvignon and Shiraz from 2013 and 2014 vintages) were collected from two commercial bottling facilities before and after cross-flow filtration and lenticular filtration; after 0.65 µm membrane; and after 0.45 µm membrane filtration. The 2014 Cabernet Sauvignon wines were filtered through both polyether sulfone (PES) and nylon 0.45 µm membranes. The average size of particles in all wines decreased significantly with cross-flow filtration and the concentration of polysaccharides decreased with 0.45 µm filtration, while tannin and colour remained unchanged. After 18 months of bottle ageing, the average particle sizes of filtered and unfiltered 2013 wines were similar, while the filtered 2014 wines contained smaller particles than the unfiltered wines. Sensory analysis showed no consistent filtration-related trends in textural attributes across all wines, although there were some significantly different aroma or flavour attributes for samples of different filtration grade within each wine. These results suggest that commonly applied commercial filtration practices have no impact on wine colour and minimal impact on sensory profiles of red wines.
The effect on each wine of different filtration grade: pre-cross-flow (PreX), post-cross-flow (postX), pre-lenticulars (PreL), post-lenticulars (PostL), post-0.65 µm membrane and post-0.45 µm membrane of either nylon (N) or PES (P). a) polysaccharide concentration; b) average particle size; c) total particle count. Results are shown as the average of triplicate analyses ± one standard deviation. (Reproduced with permission from McRae et al. 2017.)
Filtration temperature
Filtration is easier at higher temperatures. Filtration at temperatures between 8 and 12°C doubles the difficulty compared with filtration at 20°C. It is recommended to aim for filtration at 15-21°C, or around 18°C.
Filtration blockages
Wines being sterile filtered should be immediately bottled or packaged. Filtration of wines to a packaging tank, and then bottling several days later, allows time for microbiological growth to occur. Sweetening the wine after filtration to the packaging tank can also increase the risk of any available yeast from the environment to begin to referment the wine. Such wines should be bottled within a week of sweetening and should be filtered at this stage in line to the bottling line.
Wines can be cellar bright, have turbidity <1.0 NTU and then suddenly cause problems with blockages during membrane filtration. This is most often due to macromolecular polysaccharide complexes in the wine. They can originate from the grape but are more commonly associated with Botrytis infection. Polysaccharides are released by yeast during fermentation and lees contact and are also released by some bacteria. The polysaccharides can form colloidal hazes that cause the filters to block. Since polysaccharides form gelatinous aggregates when mixed with alcohol solutions, a simple test may be used to determine if polysaccharides are present in a haze, with a milky/cottonwool-like mixture forming if excessive levels of polysaccharide are present.
Affected wines can be easily treated with a range of glucanases or other enzymes to break down the polysaccharides, such that the wine can then be filtered with ease.
