sandc wrote:
@blondica73:

Have you already gone through the pressure profiling setup tutorial in the leva! manual? I'm asking because a screenshot in #4 shows that 7.9bar were achieved while 9bar were programmed. That is not normal. The error should be a fraction of a bar, like here: http://www.vimeo.com/301475180. Since the yellow plot of the pump power was not turned on, I can not tell if that happened because the grind was so mismatched that even full power couldn't get the pressure to 9bar (one has to grind finer after introducing preinfusion) or if the pressure profiling needs tuning. Factory defaults will not work, except by luck.

@sandc:
I have not gone through the pressure profiling setup/tunning in leva! manual, I guess I got lucky. I'll try going through it sometime this week. I'm still reading the manual and getting familiar with the setup.

I cannot run a pressure test on the mini Vivaldi as control module turns the singal to SNS off due to inactivity between power steps thinking that the coffee is ground too fine as it expects more flow. I wonder how the fellow from Germany ran the pressure test.
Here is a full graph

> cannot run a pressure test on the mini Vivaldi as control module turns the singal to SNS off due to inactivity

A work-around would be a temporary connection between clamps SNS and L of ito, so that the firmware can continue to sense the AC waveform during the test. You'll only have to do this once (pressure test data survives firmware upgrades, unless the EEPROM is explicitly erased), so a short piece of thin wire - for an insignificant current - would be sufficient. If you have configured SNS in the contacts menu to start the pump, that must be temporarily changed back to do nothing before that wire is installed.

PS: I assume you currently have two flow meters, one for each controller? Flow meters contain a nozzle which leads to a drop in flow rate. Having two might not be an issue but if it is: That fellow from Germany shares the original flow meter between the controllers. I'm not familiar with the details though (probably not a direct connection; while I don't know to what level the output of the Vivaldi's flow meter is pulled up to, I seem to remember that the Vivaldi supplies its flow meter with 12V or more; the absolute maximum rating for ito's digital inputs is 5V + 0.5V).

> Here is a full graph

Unfortunately a very small screen - hard to read. But it looks as if the yellow phase angle line is at 0° (in the middle of the screen) a.k.a. full pump power. That would mean the grind is too coarse: The machine can not reach the target pressure even with full pump power.

I was able to run the pressure test. I needed to push the shot button first, then start the pressure test. This way the SNS input has a signal to run the test.
In regards to the flow meters, I have the original MV and the one I installed right before the pump.
I'll be reading the other blog more careful to see how he shared the MV flow meter between the two controllers.

sandc wrote:That fellow from Germany shares the original flow meter between the controllers. I'm not familiar with the details though (not a direct connection from flow meter to both controllers).

I read the blog and it looks like he is using opto-isolators to isolate the ITO from MV. Some of the pictures on pg. 6 of that blog are no longer there. I wish I could see the final connections.

Performed the pressure test and changed the pump characteristics based on the output graph. Played with the PI values and ran a couple of shots. I did have to grind a lot finer than before. Here's the latest shot, used 20s of 1 bar preinfusion.

a - In your plot, minimal deviations from the target - e.g. a fraction of a bar around 43s - lead to disproportionately large swings in the yellow pump power curve. This does not look correct yet. The value of the proportional constant(s) might be too aggressive or you might have entered very different values for Kc<SP and Kc>SP (causing a highly asymmetric response to error). I'm used to smaller power adjustments with 230V EP5 pumps. Also, there might have been a problem in this specific shot, possibly channeling, that lead to an abrupt change of conditions at 43s.

b - At full pump power (0° from 25s-27.5s), the Vivaldi seems to have a noticeably lower pressure rise per second than my Rancilio Silvia. About half. If that is not just an effect of the long 25s@1bar preinfusion*:

1. Partly (the part after 27.5s), that could be a result of the configuration. For example, bias might need more tuning.
2. If the rise rate remains comparatively slow after more tuning, it could be due to the pump type in the Vivaldi.
3. The flow restriction by the two flow meters might have an effect.
4. The Vivaldi might have a slower rise rate by design, for example due to a larger air pocket in the brew boiler.

leva! can not recreate factory pressure profiles containing segments with higher rise rates than a specific machine can achieve at full pump power. Detemine its abilities with a shot without pressure profiling (full power).

c - did have to grind a lot finer than before

That is one of the reasons to have a PI: To be able to grind finer, in order to increase extraction. If that is not what you want (if you are overdosing and prefer underextraction), avoid a PI. Interestingly, the effect of a PI on flow reverts at very high pressures. At 13.2bar, for example, a PI can decrease the subsequent flow.

The other reason for a PI is to reduce the risk of channeling. This can be achieved by a short PI. Essentially just a slow-down in pressure build-up. S1 Vivaldi II users can buy a preinfusion chamber for that purpose. This video shows its effect. A similar pressure curve is stored in the latest leva! firmware unter Setup->Pressure->Profiles->Your profile's name->Presets PI->CHAMBER.

* Be warned that that such a long PI can have a huge effect on the temperature in the coffee puck; see these measurements.

I'll play some more with tuning. However, I'm impressed with the improvement in how coffee tastes. I did have different values for Kc<SP and Kc>SP, 25 and 90, very mismatched. I like the long preinfusion, I typically let the preinfusion go until I see the first drops (around 20s) of coffee at the bottom of the portafilter.
The MV pump is fairly lazy and I'm not surprised it takes a little time to come up to 9 bar. I also noticed that there's a difference between the pressure measured by the pressure sensor and what my MV shows.
All in all, I'm very happy with the upgrade.

More tunning, dropped the Kc to 20.

I think the blip around 40 sec is some software register overflowing and restting, but that is a guess.

Looks good but you still seem to have a small anomaly in the middle of the shot.

> the blip around 40 sec is some software register overflowing and restting, but that is a guess

There is no such thing as 'software register overflowing' in this context. Some guesses:

- Could be something that happens in the coffee. If you have a bottomless portafilter, you will be familiar with the jets of water that sometimes suddenly appear. What happens in the coffee is not necessarily a continuous state. There can be abrupt changes.

- Could be the effect of a heating element going on/off. That causes a voltage drop in your home's power lines over their resistance (to the amount of I * R). This affects the pump: It suddenly gets a bit weaker or stronger. Normally, leva! controls the heating, so it knows to the millisecond when a heating element goes on and can compensate before it affects the pressure (power factor correction in menu). But your machine has an independent controller which might turn 10A on or off at any time.

- Very fine grinds can have a tendency to suddenly stall. As the coffee gets older and the grind finer, this effect increases. Fresh coffee shouldn't have that problem.

- Could be a measurement error due to noise. Probably not in your case, because the pressure line looks fine, but to be sure: Install blind filter, create pressure, stop pump. Watch the pressure and the pressure band for a while. You shouldn't see spikes in the pressure and the noise band should not be excessive. Otherwise shield analog lines.

- Could be a control loop configuration issue. If the proportional or integral reaction is mismatched (usually too strong), a minimal disturbance can turn into oscillations. Oscillations can get bigger quickly. Plots show the state every 0,5s but the control loop has run 30 times in the meantime.

It might have been possible to say more if the plots would be a bit different. They are not fine enough (screenshots from a phone?) and the time base is too long (please use 30s) and the plots most relevant to pressure profiling are off (please enable all phase plots, pressure and pressure band).