Tamper Base: Flat or Convex best?

In august last year I posted a blog about the difference, if any, between flat based tampers an convex based tampers:

http://kostverlorenvaart.blogspot.nl/2014/08/tamper-base-flat-or-convex.html

My conclusion then was that I could not detect a significant difference.

I do like my little effective Londinium Button tamper but I mostly use it to tamp a puck absolutely flat in order to measure the roast color with the Tonino device. For extractions I mostly use the Intelligentsia Black Cat convex. I have ordered the latest Londinium button tamper which allows 'nutating' of the coffee puck so I can't wait to test that.

This morning I noticed a tweet by R. Justin Sheperd reporting a dramatic "mind blowing' difference between flat and convex tampers:

I had not thought to test the difference with a refractometer and I have the VSTlabs TDS meter at hand so I set out to see if I can replicate the Sheperd results.

Shepard did not specify his method, what refractometer / app used, what baskets if any different, et cetera.

In my tests, using the LONDINIUM I machine, three different origins were used:  Sumatra Mandheling, Peru and Colombia, roasted Tonino # 107, # 116 and # 103 respectively, so all in the "light roast" spectrum, roasted 23 and 26 February.

Each time 18g of beans were ground on the motorized HG One with 83mm burrs, in the same grind setting for both tampers and into the IMS filter basket with size codes B68 2Th16.5 E.

The same naked portafilter used every time.

Preinfusion: 7 seconds every time.

I monitored the extraction time, flow and weight with the Acaia scale.

Mandheling and 27mm spanner

For every shot I used a freshly unwrapped syringe and I used three different filters for the measurements, frequently comparing the same extraction through another type syringe filter to make sure the filter was not causing a significant difference. No difference was seen, just the flow from the original VST filters is much better and they do not easily break under pressure like the cheaper filters do.

Distilled water was used to calibrate the refractometer between each set of  measurements.

Test setup

The Sumatra Mandheling had a slow flow, 35g out in 28s with 10.7 TDS yielding a 21.6% extraction for the flat base, 34g out in 48s and 10.7 TDS -> 20.9 % EXT, which is practically identical.

Peru, flat: 35g in 31s, TDS 9.5 -> 19.1% EXT

Peru, convex: 35g in 24s, TDS 9.0 -> 18.1% EXT

Colombia, flat: 36g in 37s, TDS 8.7 -> 18% EXT 

Colombia, convex: 37g in 36s, TDS 9.3 -> 19.8% EXT

In the 'flat Colombia' extraction, my distribution was probably a little sloppy as some early 'thin' drops appeared in one spot on the bottom of the filter basket but luckily during the pre-infusion this channel closed. Still a little lower % EXT which I think was caused ore by the early dripping than by the tamper.

My conclusion is that once more I cannot see any significant difference in the results between the two tamper bases.

I sent my findings to Andy Schecter who owns a vast array of high end machines, one of these a Londinium I like I have. His view is that my results have too much variance to catch the (small) difference between tamper bases. In his memory, when he tested it over a series of extractions, the difference was just 0.5% in extraction, nowhere near the 6+% that Sheperd saw using his setup.

Shepard told me he will write a more specific blog about his methodology and results so then we will know more. He suspects the difference in his case is his espresso machine which is not capable of the 7 seconds pre-infusion like my Londinium I can facilitate.

To narrow down my findings, I did some more tests this morning. As Andy Schecter specified, to get near-scientific I would need to to a large series of tests with just one bag of beans and discard any outliers in the results. 

Another friend on a Dutch coffee forum suggested that I should not only weigh the beans going into the grinder but also the weight of the portafilter before and after adding the coffee grinds. 

Kfir Shabo, another L1 aficionado in Israel, advised to dose 17g instead of 18g in the basket.

I took 4 measurements with beans from Tanzania, 2 roasted by a friend nearby and 2 roasted by myself. 

Tanzania roasted by John:

Flat: 17g in, 36g out, TDS 9, EXT 19.8 %

Convex: 16.7 in, 36g out, TDS 8.7 EXT 19.4%

Tanzania roasted by me:

Flat: 17.0g in. 36g out, TDS 10, EXT 21.9%

Convex: 17.0g in, 35g out, TDS 10.4, EXT 22.2%

I agree that I would need to do a lot more detailed testing to find a consistent difference between both tamper bases but I sure do not see a dramatic difference between both extractions.

PS March 6, 2015: A fellow on a Dutch coffee forum has also tested the difference between flat and convex tampers. Over a series of extractions he sees a very small advantage in the use of a flat tamper, but just like Andy Schecter reported earlier, more in the proportion of 0.5-0.1 % than the 10-12 times higher difference of "mind blown" Sheperd quoted at the beginning of this blog:

De koffie: El Salvador La Ilusion van Single Estate

Tamper vlak: Reg Barber 58,4 mm
Tamper convex: Espresso Gear Primo Convex 58mm
VST 18 grams filterbakje
18,5 gram gemalen koffie (exact afgewogen in filterbakje)

shot 1 : tamper vlak, 35,2 gr espresso, 30 sec, tds 10.1, EXT 20,01
shot 2 : tamper convex, 35,3 gr espresso, 26 sec, tds 9,9, EXT 19,69
shot 3 : tamper vlak, 34,9 gr espresso, 29 sec, tds 10,3, EXT 20,23
shot 4 : tamper convex, 35,1 gr espresso, 25 sec, tds 9,8, EXT 19,39
shot 5 : tamper vlak, 35,1 gr espresso, 30 sec, tds 9,9, EXT 19,56
shot 6 : tamper convex, 34,6 gr espresso, 23 sec, tds 9,4, EXT 18,33

After taking the measurements, as he was cleaning up, he realized he'd started off by first (before logging his measurements) tweaking the grinder for the day, testing for the best extraction using the flat tamper. The results might have been different had he first and foremost selected the best grind setting for the convex one but by then he was ready to start with the rest of the day and not repeat it all again. I'm grateful for his input!