Volumetrics, how accurate are they? - Page 2

Need help with equipment usage or want to share your latest discovery?

Postby rimblas » Jan 08, 2019, 2:38 pm

DE owner here. It turns out that, as Assaf well described, flow meters are inconsistent and they lie. The DE team spent much time (I heard 8 several months) finetuning the firmware to come up with a formula that uses the flow meter and the information about what the pumps are doing to come up with a good flow measurement. The result is that the DE is very good at dispensing the flow you want. "Very good" needs to be defined and it's all relative. I don't have scientific measurements, but if I ask for 1 ml/s for 30 sec the result will be within 5-10% and it's fairly consistent.
The DE can also work with a Bluetooth scale and measure the output and stop the shot at the requested weight. The most recent update finetuned the calculation formula, and we can now get within 0.1-0.2g of the target. I think this is pretty impressive. The formula had to take into consideration the flow. If the shot is flowing at 3ml/s or above, or under 1ml/s the "stop at weight" algorithm needs to be aware of that in order to stop the shot on time. Some pretty smart and talented members assisted with the calculations (since all of the tablet software is available).


Postby ShelbiRyan » Jan 08, 2019, 10:09 pm

AssafL wrote:Do you flush before the first pull? It may be that there is some extra volume in the pipe that fill up?

If it sometimes happens in reverse (first being 45 and the second being 39) then obviously this isn't it. But if it always goes up - it may be that.

BTW. This is a >10% error. It should be within a gram (even less). So there is something wrong.

You can connect to the flow meter with a scope or an Arduino and count the pulses. See if during the pull pulses are missing and at what point in the pull.

You can also pull the cap off the flowmeter (if it is a gicar) and see if the is debris or something clogging the nozzle. If you do don't bend the tabs - if they break it is an expensive part.

So this morning it actually happened the opposite. I woke up to a warm machine, did not flush. First pull was 40g. Second pull was 45g.....very annoying. But yes to answer your question, I have tried different flush cycles to try narrow things down.

One thing I have noticed is that if I flush inbetween shots (to rinse coffee from the shower screen) at a low pressure, my next volumetric shot can be out as much as 15g under weight. It almost seems like air is passing the flow meter and giving it inaccurate readings. I can tell when a shot will do this instantly. Depending on my dose and basket used, it usually takes about 24ml to fill my portafilter. So essentially my flow meter on the screen will quickly fill to 24ml, after a few seconds I'll get some coffee dripping and slowly the ml will climb up to my desired volume and my shot will finish pretty close to my desired weight. If I flush with a low pressure and repeat my profiled shot, the basket with fill with upwards of 50ml of water and my resulting shot will be 10-20 under my desired weight. To alleviate this I've been making sure to flush with a higher pressure. Doing that gets me around +/- 5g of my desired weight.

My last response from Rocket was that this is the same flow meter set up used in several different brands of espresso machines and that my error is consistency is "industry standard" and not isolated to the R Nine One. Can any GS3 AV owners chime in on this?

I have replied to Rocket and they are currently looking into things...

User avatar

Postby AssafL » Jan 09, 2019, 9:20 am

If you set the pump to a certain flow (steady flow rate - say 90ml/min) - and run it 10 times without a puck (10 pulls stopped by the flowmeter) what are the weights/volumes? Are they consistent?
Scraping away (slowly) at the tyranny of biases and dogma.


Postby ShelbiRyan » replying to AssafL » Jan 09, 2019, 9:55 pm

If I remove my portafilter and set a profile to a steady 8.5bar with a desired output to 52g and repeat a volumetric measurement 10 times the results are as follows:


These results are very good, and I'd be happy with this variation if this is how the machine acted in cup. Now here is the kicker, during that 10 shot interval I threw a slow a 10 second flush (manually with paddle) at 1bar expelling approx 30ml of water for the 11th shot. Then immediately pulled another volumetric shot that should of yielded around the 52g mark. It ended up yielding 42g.

So what I've concluded was that if a shot finishes at a relatively high pressure (6-9bar) or a flush of the same pressure before a shot is performed, then I can get my outputs relatively close. I had notified Rocket of this a couple of weeks ago.

User avatar

Postby AssafL » Jan 10, 2019, 10:38 am

Perhaps it doesn't reset the flowmeter dosing counter after the slow flush?

There may be a switch further down in the paddle that needs to be hit for it to reset the counter?

Maybe way to test this is to try doing the flush a bit longer or a bit shorter and see if the output changes appropriately. Or do a long slow flush and then whack the paddle and see if that does it (but then any physical effect may also be primed).
Scraping away (slowly) at the tyranny of biases and dogma.

User avatar

Postby kolu » Jan 10, 2019, 3:57 pm

Internal boiler pressure has great influence on this. Also, the way they route the drain hose from 3 way valve can cause not all the water to drain out from the brew path completely thus increasing your effective brew water by amount that the flowmeter can't register. Try pulling few shots where the flush amount/time and wait time before pull is constant. That sort of sorted out my similar problems with dosing accuracy at NS Aurelia II...

I had the R9-1 open on bench yesterday and honestly I don't like it, but that is probably just my personal problem or something.


Postby ShelbiRyan » Jan 15, 2019, 2:47 am

So I've received word back from Rocket. The only help they can offer is that they use the same high end flow meters as La Marzocco :roll: So essentially, no help at all. Very disappointing.

I haven't had a huge amount of time lately to play around with this, but I have found a way to keep my shot output very consistent. Using a more traditional shot profile, so basically a slow ramp to 8.5bar and hold. I can keep my shots within 1-1.5g by doing a high pressure flush (20-30ml above 6bar) before each shot. Not sure exactly why this is happening, and I was hoping Rocket could advise on this. But as of right now, no answers have been provided....

User avatar

Postby hankbates » Jan 19, 2019, 10:16 am

kolu wrote:Internal boiler pressure has great influence on this. Also, the way they route the drain hose from 3 way valve can cause not all the water to drain out from the brew path completely thus increasing your effective brew water by amount that the flowmeter can't register.

The case with the Livia 90 was that the recycle from the opv was routed to the drain. From what I can see in the parts schematics from both Bezzera and Pasquini (no doubt prepared by Bezzera for them) is that Pasquini never changed to a configuration which recycles excess flow back to the pump suction as was done by Bezzera in their own branded machines. You have to be awfully good with grinding and tamping to make use of flow control in the auto version of these machines.
My Livia is a 90s with an on-off brew switch, and I have closed off the opv recycle drain connection and rerouted the water to the top of the tank.

User avatar

Postby kolu » Jan 19, 2019, 1:32 pm

hankbates wrote:The case with the Livia 90

oh no, I thought the OP talks about Rocket R9-1. there the OPV is before the Flowmeter.


Postby ShelbiRyan » Jan 27, 2019, 1:13 am

Just a quick update to the volumetric problems I'm having with the R Nine One. Rocket has responded saying that my machine is working normally. It seems as though they will be making no efforts to address my concerns.

So as it sits, in order for me to have semi accurate volumetrics (say +/- 5grams) when using a profile with long low pressure pre infusion, a high pressure pre flush needs to be performed. The annoyance with this is that a second profile needs to be made to do this. So before every shot, I have to cycle to the flush profile then cycle all the way back to my desired shot profile. If this isn't done, the machine will dose anywhere from 15-20g less then the saved volume.

I will continue to try and figure this problem out when time allows, but as for now I will continue with my current work around :roll: