Ben 78, who is with John Henry Hydrogen has offered an HHO unit to test. Their site is straightforward, and they do not claim that their unit will work only if you wave magnets around, pour weasel urine into the tank, or use an "efie," a device which alters (illegally, of course) the O2 sensor(s) readings.
My tentative plan is to install the unit on my 2004 Honda Accord, which has a manual transmission, helping to eliminate variability in torque converter lockup vs non lockup, etc. The Accord will be equipped with a Scan Gauge 2 fuel economy computer, the operation of which is proprietary -- but it is reasonable to assume that is uses readings from either (or both) the mass airflow sensor or pulse width on the injectors to calculate injected fuel amounts continuously. It uses the vehicle speed sensor to read vehicle speed, and then does the calculation to compute instantaneous fuel consumption (as well as trip, day, etc.).
People who have used the Scan Gauge 2 have reported very high accuracy, as would be expected, because it is using the same data that the ecu uses to precisely inject the correct amount of fuel.
With a Scan Gauge 2, and without the complicating factor of any devices which tamper with the emission system (and which could potentially "trick" the Scan Gauge 2 in the same way that the ECU can be "tricked,") even on-road tests of a device like an HHO unit (which can easily be turned on and off) can be accomplished with pretty good repeatability. However, there is no substitute for a dyno, which allows consistency of load and speed that is very hard to achieve safely (if at all) on the road, so we will use a dyno at least to calibrate and gain confidence in the Scan Gauge 2.
We'd measure HHO output flow rate periodically throughout testing to keep the flow rate constant. If it appears that there is significant gain or loss from the HHO unit, then we will vary flow rates to see if the relationship between flow rate and gain or loss changes. We will also test the claims by Onecraftydude that the typical flow rates (1-2 lpm) are actually too high (even though the injected mass is incredibly small) and cause detonation -- an effect not reported by other HHO promoters, nor by the numerous university researchers who have worked with hydrogen injection.
We will also monitor O2 sensor readings with the unit switched on and switched off.
We will do a series of A/B A/B A/B tests one immediately after the other to keep temperature, humidity, tire pressure, engine condition, etc. all constant. This will eliminate many sources of variability and sources for criticism of of invalid tests. We will allow enough time between on/off cycles to allow performance to stabilize. If we find a load at which the unit is particularly effective, we will alter flow to see if performance can be optimized, recording changes and results.
The tests will be designed so that any garage inventor can replicate them at very low cost. This will help small promoters of the units to avoid feeling that they must make outlandish performance claims before having any real data.
Our working hypothesis will be that performance differences attributable to the unit may be very hard to measure, because the energy value of the injected hydrogen is such a small portion of the total fuel consumption (on the order of 1/1000 of the total). Thus, we will be careful to see if effects can be observed at very low loads, such as at idle, as well as at more normal road loads.
If this preliminary testing indicates significant efficiency improvements, then we will attempt to set up testing with a governmental agency using standard procedures and equipment. Also, if this preliminary testing indicates a significant performance change, we will measure the criteria emissions (CO, HC, NOx) to see how they are effected.
If you have suggestions, or questions you would like to have answered via this preliminary testing, please comment. I'd like this test to be sufficient for an engineer, scientist, or technically engaged person (and ideally the average person) to be able to say either: "Yes, it looks very likely that these devices work as advertised", or "No, it seems unlikely that these devices work as advertised."
Perhaps if the HHO unit shows significant performance changes, then we can come up with a plan for testing additions, such as Efie, magnets, etc.
The HHO Test Plan 
Re: The HHO Test Plan
Good Answers:
"Almost" Good Answers: