The Air Pump Driven Espresso Machine

Need help with equipment usage or want to share your latest discovery?
Abe Carmeli
Team HB

#1: Post by Abe Carmeli »

This post was split from Time to retire vibe and rotary pumps?. I am posting it as I received it from Kevin Wu

The Air Pump Driven Espresso Machine

Every coffee geek is dreaming of building the perfect espresso machine. The following is my story.

There are many facets in the espresso brewing that will affect the result: Pre-infusion time, the amount of water for pre-infusion ..... etc.

But most people agree that stable water temperature and pressure during espresso brewing are the most important goals that we want to achieve.

Fortunately and coincidentally, the air-driven espresso achieve the above two goals in one package that no other machine currently on the market can do.

It all started with the observation of my home espresso machine, the Racilio Audrey, many years (1997) ago. I found that no matter how hard I try, the density and texture of the crema from my Audrey can't compete with the crema from commercial machines. For commercial machine, the viscosity is like honey and the crema looks glossy with a silky texture. On the other hand, from my Audrey, the viscosity is like water and the crema looks dull and rough.

My first idea about it was that the small piston pump of the Audrey can not produce the desired 9 bar pressure as the big rotary pump of the commercial machine does. I soon found out that this is not the case. The piston pump is capable of generating 9 bar pressure, only with a lot of jitter. The pressure will vibrate violently between 7 to 11 bar.
With the above observation, I have the following assumption :

The smaller the ripple of the pressure during espresso extraction, the better the resulting crema.

My first idea about improving the pressure curve of my Audrey is to add a dimmer on the piston pump. Via controlling the dimmer, I can totally control the following parameters:

1. The amount of water for pre-infusion. I do it by slowly turn on the dimmer to half power and the slowing turn off the dimmer.

2. The time duration for pre-infusion. It's easy. Just keep the dimmer off as long as you want.

3. The rising speed of the pressure curve. I do it by gradually turn on the dimmer to a higher power, for instance, 70%.

4. When the 9 bar pressure is established, I will slowly turn the dimmer to a lower power, for instance, 50%. Since the power of the pump is greatly reduced, the ripple of the pressure is also reduced proportionally.

The resulting crema is much better than the crema from the original no-dimmer machine. The result of the above experiment supports our previous assumption. So, if we push the condition of the assumption to the limit, it will become:
If there is no ripple for the water pressure during espresso extraction, the resulting crema will be the best one can get.

In other words, the absolute value of the water pressure (9 bar, 8.5 bar etc) is not so important. The most important thing should be eliminating the ripple of water pressure during espresso extraction.

Our assumption can explain why there are still so many aficionados for manual lever espresso machine. The crema from lever machine is really much better than the more convenient electric pump machine. That is because human palm is soft. And the soft touch of a barista pulling the lever machine will produce no vibration for the water pressure.

From my observation:

Lever machine > rotary pump >> vibration pump.

Although the result of lever machine is great, its manual nature is unacceptable. Modern espresso bar is so busy that volume control becomes a necessity.

So, my goal is to design a ripple-less espresso machine with the same functionality of conventional semi-automatic espresso machine currently used in most espresso bar.

I am a user of Versa Dynamic 2.0 record player which has a oil-less pump and a air tank to generate the necessary air pressure for its air bearing to work. Since Mr. Bicht is in the US and I am in Taiwan, I have to maintain and tweak the air system myself. So, I am quite familiar with the air pump, regulator, and air tank etc. Then, the idea of using compressed air as the pressure source for espresso extraction comes to my mind.

So far, the result is very great. There are some concerns on the Forum about the feasibility of using a compressed-air system on the espresso machine. The followings is my answer:

1. The oil-less air pump is expensive.
Ans : An oil-less air pump is about $350 which is very expensive. However, our target market is the high end market as the La Marzocco and Synesso. The price for a two group machine is at the $9000 range. Therefore, it's justified to afford a air pump in the machine.

2. The size of the pump and tank is too big and the pump generating too much noise that can not be fit into an espresso machine.
Ans : A stand alone box can solve the problem. This is the way Versa Dynamic record player do. We only need a 4X6 mm hose to connect between the espresso machine and the pump box. The stand alone box will comprise of air pump, air tank and regulator. The air pump will pump the air tank to 10 bar, and then regulated to 9 bar via the regulator and feed to the espresso machine. During espresso extraction, the air pump will keep rest so that no vibration from the air pump will affect the resulting espresso. The air pump is only activated when the machine is not pulling a shot.

The initial purpose of the air-driven espresso machine is to achieve a stable non-vibrating water pressure. Fortunately, by doing so, we also achieve another important goal that no other espresso machine can do : a constant water temperature during espresso extraction.

Modern two-boiler espresso machines use PID controllers to keep the water temperature in the coffee boiler to be a constant. But during espresso extraction, cold water is pumped into the coffee boiler and the water temperature will drop significantly. In the state-of-the-art machine, the Synesso Cyncra, cold water is pre-heated before entering the coffee boiler. However, there is still temperature difference between the pre-heated water and the water in the coffee boiler. On the other hand, during the espresso extraction phase of our air-driven machine, there is no water, cold or pre-heated, being pumped into the coffee boiler. Instead, compressed air is pumped into the coffee boiler. Since the mass of the air is very small, the temperature in the coffee boiler will not change at all, which is exactly what we desire. Only when the espresso extraction phase is over, our machine will open the valve to relieve the air pressure. Then fresh pre-heat water will re-fill the coffee boiler, and the PID controller will bring the water temperature in the coffee boiler back to our pre-set value very soon. And then, we are ready to brew another shot in perfect condition.

Since our machine has a much better thermal stability than the others, this enable us to use a smaller boiler (says 1 liter) than the 1.9 liter Synesso Cyncra while still enjoy a performance.

In the SCAA conference Seattle this April, our prototype Reneka Techno did not relieve the pressure after each brewing. We only relieved the pressure and refill the boiler when the water level is lower than the 1/3 level-meter. This will result a very slow recovery time for the temperature to recover to the preset value. The other drawback is since the boiler is still pressurized at 9 bar after brewing, we can not control the rising speed of the pressure for the next shot. In our next commercial model, we will relieve the pressure for every shot and then refill with preheat water. The resulting recovery time will be very short. And we can control the rising speed of the water pressure for every shot.

In conclusions, compared with other high-end espresso machine, our air-driven machine has the following features :

1. Although aim at the highest performance on the market, we still keep it user friendly. The operation of our machine is just the same as ordinary commercial machine on the market. We will have volume control. With just one touch of the button, a cup of perfect espresso will be made. On the other hand, machines like Versalab M3, Synesso Cyncra, spring lever machine all need human operation and monitoring during espresso extraction. They will take too much time from the baristas, therefore not suitable for busy commercial environment.

2. We achieve the two most important goal : ripple-free water pressure and ripple-free water temperature in one tidy package.

3. The amount of pre-infusion water and the duration for pre-infusion can be controlled via our on board process control chip. The rising speed of the air/water pressure can be controlled via a needle valve. We also devise a patent pending device that will make the temperature of the group head exactly the same as the coffee boiler. With this feature, the water that pass through the coffee cake will have a much stable temperature than all other machines.

It will be very appreciated if you have any questions and concerns about our air-driven machine. Your input will give us directions to make our machine even better.

Kevin Wu

User avatar

#2: Post by Jepy »

Abe, do you know if these guys will be in Italy for the coffee trade show next month?

Abe Carmeli (original poster)
Team HB

#3: Post by Abe Carmeli (original poster) replying to Jepy »

Perhaps an update is in order. To quickly answer your question: No. But I can check it out for you: what's the name of that show and where is it? I met with Kevin Wu, the owner on Saturday in New York. He is working on the machine, a new design for a commercial grinder, and a 1lb air popper home roaster. The guy is a busy bee. If all goes according to plan, I should be getting the espresso machine and the grinder for a review before April 2006, at which time they will be in Charlotte for SCAA 2006.
Abe Carmeli


#4: Post by Budha »

So what happened to the development of the espresso machine? Sounded very interesting.

Abe Carmeli (original poster)
Team HB

#5: Post by Abe Carmeli (original poster) replying to Budha »

Apparently not much. I haven't heard from the guy since our meeting last year.
Abe Carmeli


#6: Post by 'Q' »

wow - Funny I stumbled on this. I have been thinking along a lot of the same lines the past week or two for a machine. It's partly what spawned this idea.

User avatar

#7: Post by boar_d_laze »

Of possible interest to anyone still working on the problem of controlling "ripples" in a pressurized air or water supply:

If I understood Wu correctly, he placed a "surge tank" between the air pump and his (a) air-bearing tone-arm; and (b) brew water supply.

My experience with, surge tanks is that most of the perturbations in pressure supply are removed at the entry of the "tank," where the cross-sectional diameter of the supply increases, and pressure decreases. Not news to anyone who has any appreciation of fluid dynamics.


Elementary fluid dynamics and complexity theory suggest significant contribution to removing "ripples" can be made by replacing the "tank" with a "tuned manifold." Such tuning to create a delaminar flow by manifold geometry and baffling which create eddies within the directional flow. The eddies are sufficiently small, sufficiently evenly distributed, and of sufficient frequency, that although creation is chaotic across size, distribution, and frequency, velocity, etc., the flow would appear homogeneous to the point of complete laminarity.

Put more simply, although each individual snow flake may be individual, each makes the same contribution to the mass of stuff you have to get out of your driveway.

I was able to significantly improve pressure uniformity over unit time in the air supply to a variety of air bearings by using a tuned-pipe (3" dia, more or less "S" shaped, legs cut roughly to "pi" a k a "golden" proportionality, total volume < 2 gal.), baffled with aquarium floss, as compared to simple 5 and 10 gallon tanks. Improvements were noted with both reciprocating and rotary pumps. Indeed, flow from reciprocating pumps was smoothed to the point that it was better than air supplied by systems employing rotary pumps and traditionally shaped (i.e., bottle or cylindrically shaped) surge tanks, and essentially indistinguishable from rotary pump systems using a tuned manifold.



#8: Post by ntwkgestapo »

To smooth out ripples from my inexpensive, small, air compressor (cheap one from Wal-Mart) that I use for my airbrushing, I did a similar trick. I used 3 4in PVC pipes approx. 2.5 ft long, cross connected at top and bottom with 4in couplings. I filled the entire volume of the "tank" with polyfiberfill (the stuffing for most teddy bears, cabbage patch kids, etc), put an input port about 3in up from the "bottom", a manifold outlet right off the top and a drain valve at the very bottom of the system (it's intentionally "twisted" about 5 degrees to insure the drain valve is @ the bottom and water/moisture can accumulate there). My air pressure went to a straight, no pulsing (can't see any pulses in the paint flow, even @ 5psi output. The pressure in the tank fluctuates from 75psi to 100psi as the pump cycles) air flow. I mounted a triple manifold with 3 independent regulators (with moisture traps, pressure gauges and fine adjustment) and provide 4 connection points (for hoses) out from each regulator. Each connection point has a ball valve for independent "on/off" control. This allows me to have "fine art" (HA!) airbrushes regulated @ 5-20psi, "T-Shirt" airbrushes running @ 40-60psi and another port where I can do what ever I wish (usually a second EITHER "fine art" setup @ a slightly different pressure OR a "T-Shirt" setting at a different pressure).

This does a wonderful job of insulating my airbrushes from the pulsing of the small piston compressor pump. Before doing this the pulsing was easily seen in the paint flow (even @ 50 to 60 psi). Now, no visible pulsing! IF only I knew how to use the airbrushes! :twisted:
Steve C.
I'm having an out of coffee experience!
LMWDP # 164


#9: Post by Paul »

i find it interesting how ideas come around in circles. Here is a pic of a gaggia prototype machine that used a steam piston to effect sufficient pressure:

I have no idea as to how well this worked. I guess not very well.

LMWDP #084

User avatar

#10: Post by boar_d_laze »

I hadn't realized Rube Goldberg was Spanish. Ole Vey!