chasing RoR

[Steve]

Ok so I went back to 7.3.12.1064 just to be sure and this was the result.

DamianPNG.JPG (82.71 KiB) Viewed 6810 times

My settings are ROR multiplier 1 / smoothing 10 sec
Dropped at the very last snaps of first crack as best as I could hear because with this higher speed fan profile it is much louder, latter roast suggests ear muffs help hear the cracking better.
I had a peek inside with light and the agitation definitely looks more like a column fluid bed roaster now.

I cupped blind this morning with the following I came up with on a whim yesterday aiming to slow at yellow briefly then ramp evenly to first crack which did not really happen. This was using the stock Classic fan profile and the same green coffee.

ExpPNG.JPG (89.47 KiB) Viewed 6810 times

The other 2 in the cupping were of my latest 500g drum roasts, one of which was the same coffee as Damians / my profile, a washed PNG peaberry bourbon / a Project Origin sourced green. Buying notes: Beautifully balanced & clean with flavours of lime, starfruit and caramel and a tea-like finish.
I record MET manually and have added it for reference.

PNG.png (127.4 KiB) Viewed 6810 times

Cupping.jpg (2.7 MiB) Viewed 6810 times

From left to right A: My PNG, B: Damian's PNG, C: Pacamara espresso drum, D: PNG Filter drum.
Cups are numbered underneath.
This line up got me and I did not get any right, which is actually a nice surprise as it usually mean I will be learning something new.

I drank all I could of B and most of D, they were similar but B had a slight edge on clarity of flavour / delicate fruit notes and the fruit aspect was more jammy on D. I did not need to drink much of A as it was undrinkable flat and baked with some malty underdevelopment and persistent bitterness, the dry ground aroma smelt bad and the wet aroma break also said "dont drink me".

In conclusion Damian's profile displayed the coffee quite accurately in line with the green buying notes, which is a welcome surprise. I will keep cupping it over the next week or so to see how it develops.

Cheers

[kaffelogic]

I was hoping to hold off sharing the settings until I was confident with the result. But anyway, see what you think.

Here is the log of that roast
log0088.klog

Unfortunately it looks like there is something wrong with the way your roaster is recording ROR in the logs. Here is actual vs profile ROR from your log0088. The actual line should more or less overlay the profile line.

log0088.png (17.67 KiB) Viewed 6802 times

This does not look like a hardware error and the other lines all look ok so it would appear not to be affecting the roast. Could I suggest you reset your roaster (hold start down while you turn it on) or update the firmware (which also resets it)?

[kaffelogic]

Ok so I went back to 7.3.12.1064 just to be sure and this was the result.

DamianPNG.JPG

My settings are ROR multiplier 1 / smoothing 10 sec

Just a few technical notes from looking at your log Steve:
1. The wiggles in the temperature and ROR curves towards the start of the roast look like PID tuning issues. I would say the extra air has meant PID re-tuning may be needed. However, wiggles this early in the roast are perhaps not coming out in the cup?
2. The last 30 secs of the roast the power curve has flatlined. This has caused ROR to drop. Can be seen in the temperature curve too. The roaster has maxed out on power. Damian gets away with it because he has unusually good voltage where he plugs in. You can make things better by not using an extension cord or multibox, and plugging in as close to your supply mainboard as you can. This profile is flying the wee machine right up to its stall point! You are getting the last little bit of performance out, but without the headroom that makes for a plug-and-play consumer profile. Plus also, if you wanted to drop the roast any later the effect would get worse. Damian's log0088 peaked at 1.353 kW so you should probably set roast required power to 1350 or 1375. This would give you a warning when you turn the roaster on if the roast is at risk of stalling.
It's good to see the machine being taken to its limits like this.
Cheers

[Steve]

Ok so I went back to 7.3.12.1064 just to be sure and this was the result.

DamianPNG.JPG

My settings are ROR multiplier 1 / smoothing 10 sec

Just a few technical notes from looking at your log Steve:
1. The wiggles in the temperature and ROR curves towards the start of the roast look like PID tuning issues. I would say the extra air has meant PID re-tuning may be needed. However, wiggles this early in the roast are perhaps not coming out in the cup?
2. The last 30 secs of the roast the power curve has flatlined. This has caused ROR to drop. Can be seen in the temperature curve too. The roaster has maxed out on power. Damian gets away with it because he has unusually good voltage where he plugs in. You can make things better by not using an extension cord or multibox, and plugging in as close to your supply mainboard as you can. This profile is flying the wee machine right up to its stall point! You are getting the last little bit of performance out, but without the headroom that makes for a plug-and-play consumer profile. Plus also, if you wanted to drop the roast any later the effect would get worse. Damian's log0088 peaked at 1.353 kW so you should probably set roast required power to 1350 or 1375. This would give you a warning when you turn the roaster on if the roast is at risk of stalling.
It's good to see the machine being taken to its limits like this.
Cheers

Thanks for the notes Chris, I noticed the slight clap out at the end and have set my Roast Required Power to 1400, yep right on the limit for 236V, my voltage fluctuates up to 242 and I am on a dedicated power point not too far away from meter in garage.

If you have any suggestions for different PID parameters I am all ears.
The extra wiggly start does not seem to have any negative effects at this early stage?
I think the most important observation is that the power profile is in a steady inclining trajectory all the way through the roast, bumpy and not perfectly linear but with this very high airflow it seems not to matter.

I think the key with Damian's approach is that it is forcing massive amounts of convective heat through the beans right at the start. The agitation level at the start is akin to how the beans behave closer to first crack on the Classic profile. My thinking is that this very aggressive use of high airflow at the start is giving much better development to the core of the beans. What I have observed over many 1000s of roasts is when the heat is insufficient at the start, (lingering in some inbetween land) the beans always end up with this outer layer like shell where the early heat penetrates to a certain point, then the core gets "stuck" or that excess moisture is trapped and after the roast has completed the beans will have poor expansion and if cut open the insides will look darker or burnt compared to the outsides, the assumption is that excess moisture boils under pressure and causes damage. Obviously once into dark roast territory this becomes harder to observe or a non issue.

I did a few more roasts using an Ethiopian natural 100g and have attached a photo of the cut beans. The roast is very uniform all the way through with excellent expansion. I have already cupped this, its displaying pleasing aromatics and tasted great, classic light roast Eth natural that is well developed.

Further experimentation may reveal that Damian's fan profile can be gradually scaled back to lower fan Rpm keeping the same shape to give more of a buffer in terms of available power issues until the roast / cupping starts to suffer.

Trying not to jump the gun too much, but this profile has got me excited about the roaster again and I think this approach warrants further investigation.

EthCut.jpg (2.82 MiB) Viewed 6797 times

EthNat.JPG (84.6 KiB) Viewed 6797 times

[kaffelogic]

Thanks for the notes Chris, I noticed the slight clap out at the end and have set my Roast Required Power to 1400, yep right on the limit for 236V, my voltage fluctuates up to 242 and I am on a dedicated power point not too far away from meter in garage.
...
I think the key with Damian's approach is that it is forcing massive amounts of convective heat through the beans right at the start. The agitation level at the start is akin to how the beans behave closer to first crack on the Classic profile. My thinking is that this very aggressive use of high airflow at the start is giving much better development to the core of the beans.

On the basis of this my advice would be to try dropping the fan speed later in the roast. Damian, your log0088 reaches 1.2 kW at 5:32. You could start steadily reducing the fan speed from there to keep the power curve more or less flat for the rest of the roast. This won't actually cause a reduction in air speed, but will undo the increase in air speed which takes place as the beans dry out and expand. I estimate you could drop from 15720 RPM at 5 mins to 14640 RPM at 10 mins without actually causing the air speed to reducing during the roast.

My understanding of the need for heat early in the roast is based on the fact that the beans have higher thermal conductivity at the start. As they dry out they also become less conductive to heat. The way to benefit from this period of higher conductivity is to, as you say, get as much heat in as early as possible. In the past I have focused on doing this by increasing the temperature of the inlet air, and it has never occurred to me to use increased air to achieve this.

[kaffelogic]

If you have any suggestions for different PID parameters I am all ears.

The settings that are relevant are four specific heat adjustment settings.

Looking at the log it appears that the PID de-tuned oscillations (the nice smooth regular ones) start big, and fade away. This is over the part of the roast where the specific heat multipliers start at 2.1 and 4. That is, up to 80 deg C the proportional gain is multiplied by 2.1 and the derivative gain by 4. This taper offs until the regular PID settings are in use at 180 deg C. So the regular PID settings do not appear to be at fault, but in the part of the roast where they are modified, they appear to be over cooked.

It looks like if you reduced the multipliers you might achieve better tuning. You could try 1.5 and 2.25. This is a guess for the Kp multiplier, and using the principle of squaring the Kp multiplier to get the Kd multiplier.

The whole thing is a combination of guesswork and theory combined with trial and error.

[Steve]

Thanks for the suggestions Chris I will experiment further in coming weeks / months....

Another round of blind cupping, this time with some high quality commercial roasts.
The same PNG as previous post using Damians profile
My drum roast using the same PNG
Mountain Air Roasting U.S.A - Black Balsam Colombia - Custom built fluid bed roaster
Tim Wendelboe Oslo - latest subscription Colom 2019 - Loring Smart Roast 35kg
Stitch Coffee Syd - Brazil P.N - Loring Smart Roast - 70kg

Damian's new profile is holding up very well, great aromatics, sweet clean taste, overall well developed light - med roast.

My drum roast tasted like it experienced slightly higher temps than Damians KL profile, a bit more syrupy mouthfeel, the fruit is still there but more caramelized and jammy. I am guessing I will prefer this for espresso or a milk drink. Both roasts have the same black tea aftertaste which seems to be inherent in the green coffee.

Mtn Air and Damians PNG while being quite different coffees have very similar overall roast characteristics which is excellent because I have been looking to replicate this on the KL!

TW, is very very light but fully developed, sweet and clean, ages well.

Stitch - pretty sure this roast was baked, thin, straw / hay like, lots of astringency / bitter unpleasant finish.

Well done Damian!

[kaffelogic]

I was hoping to hold off sharing the settings until I was confident with the result. But anyway, see what you think.

Here is the log of that roast
log0088.klog

Unfortunately it looks like there is something wrong with the way your roaster is recording ROR in the logs. Here is actual vs profile ROR from your log0088. The actual line should more or less overlay the profile line.
log0088.png
This does not look like a hardware error and the other lines all look ok so it would appear not to be affecting the roast. Could I suggest you reset your roaster (hold start down while you turn it on) or update the firmware (which also resets it)?

Hi Damian
The explanation for the strange behaviour has been found. It is the 90-min bug. Whenever the roaster has been on continuously for over 90 minutes it starts to loose its ability to calculate ROR correctly. Eventually it reaches the point where it cannot follow a profile at all . The explanation is a bit obscure (catastrophic cancellation of floating point numbers), but there is now a fix available. The fix has been applied to the official release, and also to the two different beta releases currently available. If you update your firmware to the appropriate version the problem will go away in future roasts. See https://kaffelogic.com/community/viewto ... t=144#p902
Cheers
Chris

[Damian]

Thanks Chris

[Beanz]

Fascinating and generous input Damian.

I'm just starting the roasting journey and have an endless amount to learn. (4 roasts in!).

Can you advise, was it really FC at 7"30' ?

With Guatemala Huehuetenango and K-Classic at L3.5 FC has been at 5'03''. That being 5 minutes from pressing the go button on the roaster and sound recorder. (from memory, when reviewing the graph, the software seems to start the time count later??)

I get that different varieties behave differently, but, with a circa 2 minute difference to FC, are you aiming for a richer roast without burnt overtones?