Coffee | One Roast at a Time

The last six roasts have made one thing much clearer.

The repaired Gene Café CBR-301 does not appear to be a simple version of the old machine that is now “working again.” It appears to behave differently.

That matters because a lot of my recent roasting logic has been built around repeatability. Same machine, same broad setup, same target temperature, same total roast time framework, then validate with weight loss and cup results.

The problem now is that the machine underneath that logic has changed.

So I compared three back-to-back cold/warm pairs, roasts 70 to 75, against the mature pre-repair 250 g / ~250°C baseline to see whether the repair had simply shifted the old behaviour left or right.

At the moment, the cleanest reading is this:

it has not shifted in one direction
It is slower in the front half, but faster and less clipped in the back half.

That means the heater fingerprint has changed.

What I compared

I looked at:

  • average rise through the roast
  • time to 200°C
  • time to 230°C
  • time to 245°C
  • total roast time
  • weight loss
  • first fan 2→3 event
  • how long fan 3 stayed active
  • whether it recovered to fan 2
  • what happened to temperature while fan 3 was running
  • what phase of the roast those changes happened in

Cold and warm averages on the repaired machine

Across roasts 70 to 75, the repaired machine currently averages:

Cold starts

  • Start temp: about 22°C
  • 200°C: 4:45
  • 230°C: 7:39
  • 245°C: 9:32
  • TRT: 13:19
  • WL: 14.19%

Warm starts

  • Start temp: about 68°C
  • 200°C: 3:28
  • 230°C: 5:57
  • 245°C: 7:44
  • TRT: 12:48
  • WL: 15.31%

So the warm start effect is still very real.

In fact, across the repaired-machine set, warm starts were on average:

  • 1:17 faster to 200°C
  • 1:42 faster to 230°C
  • 1:48 faster to 245°C
  • 0:32 shorter in TRT
  • about +1.1% higher in weight loss

So the machine has not become “less sensitive” to start state. If anything, the warm-start advantage is still compounding as the roast progresses.

Where the repaired machine is slower

Compared with the mature pre-repair baseline, the repaired machine is clearly slower in the front half.

On average:

  • cold starts are about 19 seconds slower to 200°C
  • warm starts are about 18 seconds slower to 200°C
  • cold starts are materially slower again by 230°C
  • warm starts are still later by 230°C, though less dramatically

That means the repaired machine is not simply “hitting everything earlier.” It is arriving at early milestones later.

That is the first important correction to the idea of a simple left shift.

Where the repaired machine is faster

Once the repaired machine gets deeper into the roast, the picture changes.

Compared with the mature pre-repair baseline:

  • cold starts move through 230→245°C faster
  • warm starts move through 230→245°C much faster
  • both cold and warm also move through 245→250°C faster

So the repair seems to have produced a machine that is:

  • slower early
  • later to engage meaningful protection
  • but stronger or less interrupted later

That is not how the old machine behaved.

The biggest difference, fan behaviour

The clearest fingerprint change is not actually total roast time. It is the fan/protection pattern.

Before the repair, the first meaningful fan 2→3 event usually arrived in the 200–230°C band, often around the 4-minute mark on a cold roast, and then generally stayed high through the rest of the roast.

That old behaviour looked like:

  • early intervention
  • earlier clipping of momentum
  • very little recovery back to fan 2

After the repair, meaningful fan 2→3 events have moved much later.

Across roasts 70 to 75:

  • they no longer arrive around the low-200s
  • they often arrive in the 230–245+ band
  • some arrive very late, near the upper end of the roast
  • recovery back from fan 3 to fan 2 is now common
  • some roasts show late oscillation rather than one early permanent step-up

That is a major mechanical difference in behaviour.

It means the repaired machine is spending much more of the roast below protection, then managing heat later rather than earlier.

Why that matters

This changes the shape of the roast.

With the old machine, the roast got pushed forward early, but was also clipped earlier.

With the repaired machine, the roast is slower to get going, but once it gets deeper into the roast it seems to retain more momentum because protection arrives later.

That is why the machine does not feel like a simple left or right shift.

It is a different shape.

Two cold Honduras roasts show why TRT alone is no longer enough

One of the clearest examples came from the cold Honduras pair.

Two roasts landed at almost the same TRT, but arrived at 245°C at very different times and produced materially different weight loss.

That matters because it shows that, on this repaired machine, same TRT does not necessarily mean same thermal work.

The pace to the checkpoints matters.

In practical terms, that means:

  • TRT is still the control I set
  • but time to 200, 230, and especially 245 now carry much more explanatory value than before

That does not replace TRT.

It means TRT now needs more context.

Current working interpretation

At the moment, my working position is this:

The repaired heater element has given the machine a different fingerprint.

Not just “more power” or “less power.”

More specifically:

  • slower front-half climb
  • later protection engagement
  • more late-stage momentum retention
  • more recovery from fan 3 back to fan 2
  • greater need to read checkpoint pace, not just final TRT

So the old TRT-to-weight-loss expectations cannot simply be carried over.

They remain useful as history, but not as current truth.

What this means going forward

The machine is usable.

The repair appears successful.

But it needs to be treated as a recalibration problem, not a continuity problem.

The next step is not to assume the old system still applies.

The next step is to build fresh anchor points:

  • cold and warm
  • multiple coffees
  • repeated target TRTs
  • observed pace to 200 / 230 / 245
  • final weight loss and cup outcome

That is how confidence gets rebuilt.

Not by assuming the machine is back to normal.

By learning what normal now is.