The notes and records in this document are designed to show how specific equipment and materials can be analysed to reveal the characteristics that prevent predictable behaviour and so spoil efforts to produce ‘fine prints’.

The equipment and materials considered are specific to my own methods, but the underlying principles are transferable and the detail acts as a comprehensive ‘worked example’. The example is presented in the second half of this article.

The first half of this article provides the tools for a predictability analysis method, which are:

• A structured view of predictability
• A means to use such a view in order to elicit predictability issues and responses to those issues
• A format for a Predictability Matrix to capture analysis results

Predictability Analysis Method

Predictability View

The key to ‘fine art’ printing is repeatability – the ability to consistently reproduce results. Once that is possible there is a baseline from which the impact of planned variations can be determined. To achieve this baseline it is necessary to analyse the predictability of each artefact (camera, meter, film, etc…) employed in creating a given result. There are 5 reasons why an artefact may behave unpredictably:

• It is operating near or beyond its optimal range.
• It has no optimal range and will behave quite differently at different settings, even though a predictable linear or geometric response is expected. It will however behave consistently at the same setting.
• The controls are imprecise and so may not be set accurately which in effect means the behaviour is not necessarily consistent at the same setting.
• Some significant environmental factor changes.
• Ad-hoc behaviour, no predictability at all. This won’t be considered, since it can not be predicted.

The impact of the unpredictable can be removed from the process by one of 5 predictability controls:

• Measurement – if a suitable measure can be taken then the deviance from the prediction is known and compensation is possible, e.g. Development temperature drops by 2^o over the 4 minute development cycle, at 2^o colder development should have been 5 minutes. If the developer temperature was constantly monitored then an additional half minute of development would be given.
• Establishing Norms – normal modes of operation, e.g. always shooting a test at the same shutter speed.
• Comparative Standards – Effectively these work like measures, they give an indication of deviance from prediction. Examples are the 18% Grey card, a calibrated step wedge.
• Alternative controls – considering if different controls can be used to create the same effect, e.g. if an aperture can not be set accurately at 1/3^rd stop increments then perhaps the shutter sped can be. Or the film speed is set in order to effect an automatic exposure system to a 1/3^rd stop accuracy. Or else 0.1ND filter can be used instead of changing any (electro) mechanical settings.
• Acceptance – Ignoring the fact that some piece of unpredictable behaviour may impact the result. If the rest of the process is predictable and it is unlikely that the worst impact would be felt anyway, then the uncertainty can be accepted. This is possible in the case of a calibration. It could also be the case that the area of uncertainty has empathic relation to the process at hand. In which case, it is entirely through the vehicle of the uncertainty that the process will be fully explored. This case is possible in a pictorial sense although unlikely in a calibration.

In discussing adjustments arising from measures (above) it was suggested that an additional 30s development could make up for the stated drop in temperature during processing. This assumes that the cooling profile was linear and that the developer’s action was in linear proportion to the development time. The chances of having two conveniently linear responses are of course, very slight. So the reality will be more complex. However, the inaccuracy of an adjustment can be gross if the quantity being adjusted is large. In the development scenario we are adding 30 seconds to 4 minutes to give us 270s. If we get the adjustment wrong by 20% (24 to 36s) then the final time will be anything between 264s and 276s – an error on 270s of only +/-2.2%.

The linear assumptions are then, in this case, acceptable. If results belie this ‘acceptability’ then further tests must be devised to map the non-linear behaviour and an accurate adjustment made in compensation for the measurements. The need for this is not anticipated. However some errors are additive and so a 2.2% potential error in such development adjustment may add to other errors to take the process out of tolerance on the whole. Most film test systems work at the 1/3^rd f/stop resolution, which is about 5% of range and represents the typical overall tolerance required.

The unpredictability of a process can be wholly understood by considering the physical artefacts of that process. The equipment and the materials are containers of the process’s unpredictability. A predictability analysis considers each item of equipment and each piece of material in terms of:

• The four (considered) causes of unpredictable behaviour, and

• The five predictability controls

Analysis Method

The view on predictability above can be modified since:

• the number and meanings of the ‘causes of unpredictable behaviour’ can be re-visioned to match any working environment, and
• the number and meanings of the ‘predictability controls’ can be re-visioned to match any working equipment and materials

However, once a specific cause/control view is established the predictability analysis process is:

1. The purpose of the analysis is to define handling instructions for the equipment and materials of the image’s workflow process. Therefore for each such equipment or material item (artefact) consider:
2. All of the aspects of how the artefact is used – in fact consider ‘what are the physical controls of the artefact?’ (e.g. aperture, distance).
3. For each such physical control consider:
4. Each of the identified causes of unpredictable behaviour in turn and how such a cause may manifest itself with regards the nature of the physical control. E.g. a lens has ‘range’ behaviour in its aperture control which may manifest as a drop in contrast at wide apertures or a loss of definition at small apertures.
5. For each unpredictability manifest so discerned consider:
6. Each of the predictability controls of the analysis method and how such may ameliorate the unpredictability manifest (e.g. we can decide that it will be the ‘norm’ to avoid the largest and smallest apertures of any given lens). There may be multiple means to control single behaviour.
7. For each predictability control applied to each unpredictable behaviour of each physical control for each artefact of the workflow process we have therefore identified an effective atomic handling-instruction. The example at step 6 (above) shows how two such atomic handling instructions can be used to derive a higher-level handling instruction.

Predictability Matrix

Because the predictability analysis method provides a means to record results an approach of continuous improvement can be applied. I.e. no individual analysis need at any given time be exhaustive (or exhausting). Results can be refined over time. This ensures that the analysis is not avoided due to the investment of effort required at any given time. Also it means the analysis is revisited and inevitably so in the light of further experience.

Given the value of recording and refining the predictability analysis method there is a need for a coherent transcription scheme – a predictability matrix:

One such matrix is maintained for each artefact – camera-body, lens, film, developer etc…

The top-left of the matrix gives the artefact’s designation and if possible this will include the manufacturer’s logo, as this simplifies finding a given matrix. It is also here that the physical controls are identified.

Because an individual matrix will contain only the ‘causes’ and ‘predictability controls’ pertinent to the artefact a reminder of all of these causes and controls is given in the top right of the matrix.

The middle section of the matrix then presents the predictability causes and controls identified for the artefact’s physical controls. Only pertinent entries are included in order to keep the matrix size manageable. I.e. not all physical controls will be susceptible to all of the kinds of unpredictable behaviour defined by the method. However for each specific unpredictability cause identified there should be at least (and usually just) one predictability control defined.  

The predictability matrix does not include a possible ‘unpredictability’ for every physical control in every category of unpredictability causes. The analysis is intended as a structuring framework for a brainstorming approach to understanding the behaviour of the equipment. It isn’t necessary to ‘invent problems’, only uncover those that may have a true impact. The quality of the analysis will be shown up in practice – and will be extended if needed. A further ben efit of the analysis approach is that consideration of what is, or is not, important has been recorded – so that any such assumptions can be re-questioned later, if required. The analysis will mature in practice.  

The lowest section of the matrix then collates all of the foregoing in order to derive specific handling instructions for each physical control. There will always be one entry here for each physical control identified in the top left part of the matrix. There may not be any specific handling instruction arising for a physical control. By insisting that all controls are listed, even where no specific handling is noted, there is less chance of inadvertently forgetting to consider the control.  

The matrix has been colour-coded in order to show how the different parts relate to each other. Once the method is appreciated the colour-coding need not be replicated. The colours remind the user that:

• The (orange) physical controls should each be considered versus all of the (blue) causes of unpredictable behaviour, with an entry arising in the matrix where a concern is apparent.
• The (green) predictability control methods will each be considered to find one (or possibly more) ways of controlling each recorded concern, and at least one (yellow) response to each recorded concern is derived.
• Groups of (yellow) responses are collated in order to define the overall handling instruction for each (orange) physical control.

When the individual handling-instructions are collated it is expected that the detail will be expanded than that typically presented in the central, brain-stormed, portion of the matrix.

Worked Examples

Nikon FM3A Aperture, Shutter, EV Comp., Film Speed		Causes: Range, Linearity, Imprecision, Environment
		Controls: Measure, Norm, Standardise, Alternative, Ignore
Unpredictability Causes		Predictability Controls
Range	Lens flares at wide apertures reducing contrast	Norm	Do not work a widest 2 apertures
	Minor absolute shutter speed errors have major impact at very short exposures	Norm	Operate at <1/500ths
Linearity	Shutter speeds tend to be imprecise	Norm	All tests to be carried out at same speed
Imprecision	f/stops and shutter speeds indented at whole stops only	Alternative	With auto exposure use film speed dial to measure f/stop change
	Aperture differs when steeping up than down	Norm	Always set aperture in same direction
Environment	Mechanical behaviour impacted by temperature and humidity extremes	Ignore	Not considered worth worrying about in deriving appropriate EI and Development times
Handling Instructions
Aperture	For test works operate between f/5.6 and f/11; Open lens fully before stepping down to desired f/stop; To set a 1/3rd or 2/3rd intermediate aperture:
	Start with the camera set at a whole aperture
	Adjust film-speed so that auto-exposure system is in agreement with actual aperture/shutter speed settings
	Adjust EV Compensation dial by 1/3rd (or 2/3rd) f/stop increments (this is a discrete setting so will not be ‘guesswork)
	Adjust aperture so that auto-exposure indication and manual exposure indications are in agreement
Shutter	For test works operate at a fixed speed somewhere between 1/15th and 1/250th second
EV Comp.	Use in setting intermediate apertures as above.
Film Speed	No specific issues have been raised above, however since testing requires film speed to be set manually (i.e. not DX-coded) then the manual setting should be the normal mode of operation. Although, because film speed is a discrete (not analogue) setting it should not have a significant impact how the setting is effected. Similarly with regards use of a discrete EV compensation dial.