@braxus, a few notes in the hope of helping with your original question, and some specifics on why modulating development can be a legitimate tool even when scanning. Developing longer it changes the image in ways that are qualitatively different from a digital correction. Not impossible to approximate digitally, but requiring a combination of techniques and precision that goes beyond dragging a slider. Three mechanisms are at work:
1. Signal-to-noise at the scanner. Your scanner's sensor has a fixed noise floor, mostly thermal noise in the electronics and statistical noise from photon counting. These are always there at the same magnitude regardless of what negative you put in. Now: what does this have to do with contrast? A low-contrast negative has smaller density differences between adjacent tones. Smaller density differences mean smaller differences in the amount of light reaching the sensor, which means smaller electrical signal differences between those tones. But the noise hasn't changed, so the ratio of "useful tonal difference" to "random noise" gets worse. When you then stretch that compressed histogram digitally to restore contrast, you amplify signal and noise by the same factor. The noise can become visible.
Developing to higher contrast does something different: it increases the density spacing between tones on the film, before it ever reaches the scanner. Larger density differences mean larger signal differences at the sensor. You don't need to stretch afterward, or you need to stretch less, so you don't amplify noise, or you amplify it less.
Could you fix this digitally? Noise reduction software exists, but it works by discarding information (smoothing), trading detail for cleanliness. Chemical contrast gives you the separation without that trade-off. Also noise reduction software often costs money and time and requires you to waste more time after, in front of a computer.
2. Development reshapes the curve (instead of simply stretching it it). The H/D characteristic curve doesn't respond uniformly to development time. The toe is nearly development-invariant, being governed primarily by exposure and not by how long the film sits in the tank. The straight-line section (midtones) instead, steepens in slope with increased development, and the shoulder (highlights) extends. So N+1 development redistributes tonal information non-uniformly: you gain midtone and highlight separation while shadows stay largely where they were. This can have important creative/visual impact on your image if you know what you're doing and it achieves the desired visual effect.
Could you replicate this redistribution with a carefully shaped non-linear curve applied digitally? In principle yes, but you'd need precise knowledge of the film's characteristic curves at both N and N+1 to calculate the correct transform, and you'd be applying it to data that already has worse signal-to-noise (back to point 1). It's not something you eyeball with a curves tool, but a calculated transform applied to inferior data.
3. Adjacency effects. At boundaries between differently exposed areas, developer exhausts faster on the heavily-exposed side while reaction byproducts (bromide) accumulate there. Fresh developer migrates laterally from the less-exposed side. The result is local density enhancement at edges: micro-contrast that depends on the spatial neighbourhood (as opposed to depending just on the tone of a given point). More development time means more time for this lateral diffusion to operate, so the effect strengthens with push development. The magnitude scales with the exposure difference across the boundary. High-contrast edges get stronger enhancement than subtle gradations. This is entirely spatial and scene-dependent. A curves adjustment doesn't touch it at all. Unsharp masking approximates the idea, but it operates uniformly on all edges regardless of their photographic origin, so it's a blunter instrument.
In short: could a skilled digital operator approximate the aggregate result, noise-managed data, non-uniform tonal redistribution, exposure-dependent edge enhancement, using luminosity masks, zone-targeted curves, noise reduction, and tuned local sharpening? Probably. But that's a stack of corrective techniques applied after the fact to worse source data. One decision in the tank gives you a coherent result that addresses all three simultaneously.
And more to the point:
why would you? Many of us shoot film precisely because the process between exposure and image is rich, physical, and manipulable. You have this fascinating chemical variable sitting right there: development time, temperature, agitation, dilution, each one shaping the image in coherent, predictable ways. Trading that for an hour of tedious computer work is a strange bargain IMHO
But to go back on topic, and for your practical situation
@braxus. Others in the thread have given you good starting points on times and dilutions. The one thing I'd stress is: before you start tweaking development for contrast, make sure your temperature control is tight. Development rate roughly doubles every 10°C, so even a 2°C drift during processing is a ~10–15% variation in effective development. Enough to mask the difference between N and N+0.5. Get a reliable thermometer, measure immediately before pouring, and temper the tank if your room is unstable.
Once your baseline is repeatable, then the contrast adjustments will show up cleanly in your scans and you'll be able to dial them in.
.
