Well, it [the regression] gives the wrong answer. Aside from that, it is pretty good.
Depends on what wrong means. If being able to predict the dependent variable for the data set to within 5 decimal places, it works perfectly as described by the standard deviation of its error.
That a linear function can approximate an exponential that was hidden from view is not surprising. I will endeavor to verify this and report back.
Display MoreThe equation that calculates airspeed from blower power (3) is given by Mizuno in the patent that Ahlfors helpfully posted.
I do not understand the equations for power (1) and (2), however (3) seems straightforward.
I wonder how well (3) corresponds to the data we actually have? The exact equation is given here.
EDIT: this looks like it must be a different system from that described in the 2018 paper: 5W in the paper delivers ~5m/s air speed whereas from this equation 5W delivers 3.9m/s air speed.
The patent provides information that was not in the papers. Good find.
Note: the mathematical formula using exp(-Wb/w) _may_ be approximated with the linear function that I got from regression over a relatively small move (the 49 points range) in power -- I will have to work on that to prove it. My point is the exponential function is not _necessarily_ inconsistent.
With regard to Jed bringing in more digits -- the linear regression will still solve for something that will be VERY similar. I doubt the extra dights are significant and because regression is a robust numerical technique, the conclusion will not change over the relatively small range provided in the 49 points.
If Jed gives us the entire spreadsheet (not here) we can then fit the exponential using a regression of the log, or similar numerical technique. No big deal here -- nothing to see. This is all normal and our guess was right as proven by the patent.
Display MoreI never talked about random numbers added to any calculation. There is no random number. The whole set of the Blower Power (BP) and Air Speed (AS) data published by graph (1) and numbers (2) can be explained by two calculations and by the effect of multiple rounding off on primary (measured) and intermediate values:
- Blower Power was obtained mathematically by multiplying the Voltage and Current to the blower. Presumably, only the voltage was directly measured by an instrument. The current was probably obtained multiplying the voltage drop through a resistor by the inverse of its resistance;
- Air Speed was obtained mathematically by using a formula which fits the curve shown in Figure 4 of (3). This formula is probably similar to the exponential relationship shown in the formula (5) described at page 14 of a 2017 paper (4). However, due to the very narrow range of the 49 data sample provided by JR, the exponential relationship between BP and AS is undistinguishable from the linear one that you have proposed.
As you correctly noticed, the numbers provided by JR (2) include only 14 different values in a series of 49 pairs of data. This is true for both the BP and the AS series and, more importantly, the two values are always coupled in the same way up to the fifth significant figure. This only fact is by large more than sufficient to affirm with absolute certainty that the values of the two series are mathematically tied each other!
This is the list of all the 14 possible pairs of values:
1) I guess someone else decided you said it was a "random" number. That person was confused as to what you meant.
2) Trying to guess the quanti-sized voltage and current from this data series is a total waste of our time, i.e. like solving a meaningless puzzle or playing solitaire. Either Mizuno should give us this information, or we should just move on. Hopefully he will provide the data. There is real work to do here. The replicators data will come in soon. Rest up because this will heat up real fast.
Display MoreIf this was true, it should have been measured in the calibration. However, we found that the calibration got less heat to the RTD at higher temperatures. This makes sense to me because at higher temperatures the surface of the box (the calorimeter) is hotter so it transfers more heat directly to the outside than the the output airflow RTD.
I do not believe that this would be effected by the difference in surface emissivity of the reactor tube compared to the control tube.
Regardless, this effect can be eliminated with a future calibration run with a reactor of identical emissivity -- in my preferred design the experimenters would calibrate with the active reactor BEFORE it is loaded and is therefore identical in emissivity (to itself). Future replicators will doubtlessly do this if they get the same R20 results or even the better R19 results.
So - I'm not here arguing specifically about the R19 results (all done above) but about best methodology for replicators.
Mizuno uses a setup with two different reactors mounted simultaneously in an enclosure. He can then calibrate, and do active runs, without opening things up.
That is beneficial because it controls change in RTDs or blower, air temperature, etc, providing a run is done ABAB where A = cal and b = control, and data is kept to show the effect of reactor thermal inertia and room thermal inertia and possible external heat changes. If there is no excess heat the room temperature affect of the experiment itself should be stable over the entire run, which helps.
It suffers from artifacts due to differences between reactors: either design (e.g. different resistance heaters) geometry, color, or position within enclosure. Just small changes in distance between reactor and adjacent wall can dramatically alter airflow and hence temperature of reactor for same power. That can then alter efficiency and other more subtle issues like radiative heating of RTD (if this is not designed out).
Having removable insulation is an issue here - because it may not position exactly the same every time it is removed.
So good practice for this two reactor setup is:
Before experiment: photo two reactors, and position in enclosure showing symmetry. Measure heater characteristics (which should be nearly the same).
Keep geometry symmetric (between the two reactors) and constant - with large airgaps everywhere so that small changes in position of components do not alter airflow.
- keep reactor design identical
- calibrate twice, swapping reactor positions so the calibration reactor is checked in both positions
- do experiments with active reactor twice, in both positions. (if doing ABAB cal with active test then calibration and active tests are interleaved as one run, so separate calibration not needed. In that case do two test runs to show what happens when reactor positions are swapped).
- document input power measurement, measurement and PSU equipment, carefully. Document power type (AC or DC, if AC what sort of AC, etc)
A simpler one reactor setup is OK, but requires more care:
- Record room temperature in different positions during experiment
- Run with calibration (no Pd, or no Ni mesh at all)
- Run with active system
- Run with calibration again (by removing mesh)
- Record times of runs, correlate to room temperature measurements
- Note and if needed document the protocol to use that on assembly and disassembly calorimeter geometry stays identical - again ensuring air gaps are large is helpful.
- Input power measurement as above.
All this detail will help extraordinary results to be accepted as real, as will exact, timestamped, recording of all raw measurements.
In general strong positive results against a control will be more convincing than absolute positive results, because less work is needed to establish that result is positive. But a lot of care is needed to ensure control and active tests have identical conditions.
THH
Agreed more or less. Regardless, once multiple groups capture the "signal" of significant 6x COP 250 watt excess heat, the experiment will get done enough times with enough variation that it is generally accepted. Once two or three additional groups say they see the excess heat, the mainstream university and government labs will get involved and mainstream can then move to the next phases of LENR research -- the underlying physics and the engineering.
Display MoreQuoting me:
Therefore, the hypothesis that the Blower Power in the chart provided is a linear function of Airspeed is proven. It is clear that for _this_ data sample in the chart that the Blower Power shown is not that measured using volts x amps going into the motor, but instead is a linear function as shown of the airspeed measured. There may be other evidence of the relationship between blower power and airspeed, but the chart provided with the 49 data points is not it.
Quoting you:
Jed makes the point that there are still errors. Thus setting airflow = V * I * C and solving for each of the measurements you get slightly different values of C for each sample. The difference here is +/- 0.25%. How is that possible if, as seems certain from Ascoli's figure, somehow the airspeed is calculated from the measure (and quantised on measurement) V and I results
First -- it is a linear function with no error or random numbers as I have shown. I have changed my mind and think it likely that airspeed has been derived from blower power, and not the other way around. Why? 1) the hot wire anemometer would be really noisy -- much more than the 0.07% shown in the tabular data; and 2) an airspeed instrument manufacturer would not provide 5 significant digits of readout on an instrument that is has an 0.015 m/s accuracy and a thus has a limited resolution to two digits to the right of the decimal point.
Second "airflow = V * I * C " is wrong.
airflow = V * I * C + K is correct.
The equation is in slope/intercept form (i.e. y=mx+b) and the solution has an intercept constant of around 2 as I showed above. Forgetting the intercept constant will give us a huge inaccuracy. (Regression standard error with intercept .00005, without 0.0026, or about 50 times higher.) Why would we not include the intercept constant?
With regard to the quantization error, sure all instruments have such rounding. Mizuno will show us in his final paper exactly what the measurement method was and will tell us how inaccuracy is due to this. Here is a high accuracy lab unit with a usb interface:
Note that this hotwire unit has a relative humidity compensator because the specific heat of the air is different with the water vapor in it. Highest accuracy airflow calorimetry would need a relative humidity channel. The effect is quite small because the air at 25C is only around 1% water at 50% relative humidity and when heated in the experiment no water is added to the air. Therefore, this is a secondary effect which would not change the R20 results or the high excess power R19 results.
I think using power into the blower is relatively robust and I am satisfied with that as long as Mizuno calibrated the unit with the same blower power method.
Regarding the THH hypothesis that the higher power runs get more heat and thus temperature directly to the output airflow RTD:
If this was true, it should have been measured in the calibration. However, we found that the calibration got less heat to the RTD at higher temperatures. This makes sense to me because at higher temperatures the surface of the box (the calorimeter) is hotter so it transfers more heat directly to the outside than the the output airflow RTD.
I do not believe that this would be effected by the difference in surface emissivity of the reactor tube compared to the control tube.
Regardless, this effect can be eliminated with a future calibration run with a reactor of identical emissivity -- in my preferred design the experimenters would calibrate with the active reactor BEFORE it is loaded and is therefore identical in emissivity (to itself). Future replicators will doubtlessly do this if they get the same R20 results or even the better R19 results.
In summary, from the provided experimental design and data as communicated by Jed, there is nothing so far that I can see to invalidate the excess heat conclusions. The excess heat may not be as large due to the emissivity factor, but it is excess heat none-the-less. While there are details that we would like to see to get conclusive proof, they are hopefully coming from either Jed/Mizuno or from replicators.
THH or SevenOfTwenty or anyone else on the forum -- why is this wrong?
Put another way -- if you purchased a scaled up duct heater for your forced air heated home and it raised the forced air temperature by that amount for that power (50 watts makes 300 watts), and if you could capture the additional radiation by putting this furnace in your basement, would it not save you money over the winter compared to a purely electric forced air heater? This unit gets hotter for less power. If the data has been presented accurately, how can we deny that?
Revised fit with Jed's tabular data:
BlowerPwr = 1.713784 * Airspeed - 3.44502 + ResidualError
R-Squared = 0.9999
or if you prefer
Airspeed = 0.583436 * BlowerPwr + 2.010436 + ResidualError
R-Squared=0.9999
The standard deviation of the residual error for airspeed is 3e-5 or for those who don't use a computer 3x10^-5 = 0.00003
This is within rounding error of Jed's airspeed numbers which are good to 0.0001/4.17 = 2.4e-5, and Jed's power numbers 0.0001/3.7 = 2.7e-5.
(Note: the difference between my preliminary fit and the fit to the tabular numbers is because of the horizontal lines of the scale of the graph are not clearly defined, i.e. there are two 4.18's on the left scale and two 3.71's on the right scale. I assumed from reading the graph scales that the bottom line was 4.165 m/s and 3.690 watts, and the top line was 4.190 m/s and 3.735 watts. Those minimum and maximum numbers on the scale were imprecise.)
There is NO RANDOM number added to the Mizuno numbers. (If this was Ascoli's hypothesis this is wrong.) It is the straight formula that I show:
Airspeed = 0.583436 * BlowerPwr + 2.010436
I believe that Mizuno developed the above formula using calibration data from an anemometer and input power from the fan.
Also, the voltages and currents produce powers in Ascoli's table which don't match the tabular powers that Jed just provided us. They are from an old 2016 experiment. The new table shows powers at 14 levels between 3.7010 and 3.7237 watts for the 49 data points, with a mean of 3.71205. Ascoli shows powers in the 4.5 to 4.7 watt range. Assuming that this hypothesis is true, it needs to be modified for the new data. I tried to make this fit using a constant current of 420 mA and it looks not unreasonable (but still unproven) assuming voltage is measured to the 1 mV level. I don't think it matters as long as we know that Mizuno collected the blower power from the voltage and current going into the blower. If the current is measured to only 10 mA accuracy, the blower power is probably +/- 1 to 2 percent. This is not significant for the R20 and most of the R19 data. Perhaps the current is more stable than that. I did not look into the the blower motor specs so I don't know why the current should be stable.
It is kind of ridiculous, although I wondered if that was the case when I first made that graph. It does seem remarkably close. As I reported here above, I checked for that by the direct method: dividing blower power by (V*A). If this were a simple multiplication, it would give the same answer for each data point. Right? It's a spreadsheet. It doesn't give the same answer. Okay, so maybe Ascoli should say: ""The values of the Blower Power curve have been obtained by multiplying the values of the blower's Voltage and Current, multiplied by a very small random number." The question arises: why bother doing that? Does he think Mizuno is trying to fool people? To what end? What would be the point?
If the Mizuno (or I) felt that blower power power was a better measure of the air speed than the anemometer, we would use it. We would make this the primary measurement, and use the anemometer as verification, or as a method of calibrating the blower power. Why not?
Ascoli could have it correct, i.e. that the air speeds are derived by the formula I have shown from blower power. Or it could be reverse -- that the blower power is derived from the air speeds. Either way it is a linear relationship of the form Y = mX + b, or the inverse, X = (1/m) Y + Bprime. This is almost certainly the case. It is a linear relationship by formula in this graph. So what. Doesn't matter.
None of this changes any conclusions on the R19 or R20. We are arguing about insignificant airflow measurement estimates that are most certainly calibrate-able. In fact I believe that Mizuno DID calibrate the airfow to the blower power using the anemometer probe at different input powers (and deriving the linear relationship contained in the datapoibnts), and that is what is utilized in the analysis. While he _could_ have made a significant mistake in the calibration of the airflow mass and hence the temperature removal, he calibrates separately power in vs power out using this airflow estimate. All of this will be verified later by Mizuno, and MOST importantly, validated by the third party replicators.
I am sorry we wasted two days on this. It doesn't matter. I showed you all how to obtain the linear relationship of airflow vs blower power used in the chart, and that is really the end of this part of the minutiae. Let's move on to something important.
I did the whole data set that graph came from. 14,259 points. As I said =LINEST(O13:O14272,T13:T14272) = 1.7244.
Hi Jed,
Sorry for my delay in getting back to you. You are on the right track. As I said, from my data I got:
The fit is BlowerPwr = 1.722298 * Airspeed - 3.484167
Or Y = 1.722298 X - 3.484167
The R-Squared is 0.9994
The single variable linear regression finds m and b from Y = mX + b (slope intercept form) that is the best fit for the data.
You need to use the ARRAY formula entry method on Excel LINEST to get the b and the R-squared. You only returned the m, and you can see that your m = 1.7244 is the same within the precision of what I extracted from your graph of mine at 1.7223.
I am working now on a laptop so I cannot copy and paste the procedure for entering an array formula result in Excel. Essentially when you enter the =linest() function you have to highlight a rectangle of about 6x4 cells where it will place the result and hit (IIRC) control-shift-enter so that it places the linest() result in multiple cells instead of one cell. See https://www.vertex42.com/blog/…ray-formula-examples.html.
You will almost certainly find that you have the same result.
Hi Jed,
I took the points and made a table of airspeed vs blower power by interpolating the pixels vs the scale provided.
AirSpeed BlowerPwr
1 4.1818 3.7182
2 4.1851 3.7238
3 4.1821 3.7186
4 4.1803 3.7157
5 4.1835 3.7210
6 4.1833 3.7208
7 4.1789 3.7132
8 4.1789 3.7132
9 4.1792 3.7136
10 4.1833 3.7207
11 4.1803 3.7157
12 4.1835 3.7215
13 4.1880 3.7287
14 4.1821 3.7187
15 4.1836 3.7210
16 4.1805 3.7159
17 4.1791 3.7136
18 4.1791 3.7136
19 4.1821 3.7186
20 4.1836 3.7211
21 4.1806 3.7160
22 4.1821 3.7187
23 4.1791 3.7135
24 4.1797 3.7147
25 4.1763 3.7085
26 4.1776 3.7108
27 4.1821 3.7188
28 4.1792 3.7136
29 4.1835 3.7211
30 4.1835 3.7211
31 4.1835 3.7211
32 4.1821 3.7187
33 4.1763 3.7086
34 4.1791 3.7136
35 4.1791 3.7136
36 4.1835 3.7211
37 4.1835 3.7211
38 4.1806 3.7160
39 4.1835 3.7211
40 4.1835 3.7211
41 4.1819 3.7182
42 4.1789 3.7132
43 4.1789 3.7132
44 4.1763 3.7085
45 4.1776 3.7108
46 4.1851 3.7236
47 4.1848 3.7231
48 4.1784 3.7122
49 4.1748 3.7060
The fit is BlowerPwr = 1.722298 * Airspeed - 3.484167
The R-Squared is 0.9994
Therefore, the hypothesis that the Blower Power in the chart provided is a linear function of Airspeed is proven. It is clear that for _this_ data sample in the chart that the Blower Power shown is not that measured using volts x amps going into the motor, but instead is a linear function as shown of the airspeed measured. There may be other evidence of the relationship between blower power and airspeed, but the chart provided with the 49 data points is not it.
I invite anyone else on this forum to verify with the above data and show the same R-squared and regression fit.
Note: if you plot the points BlowerPower Vs. AirSpeed, you will see the linearity of the fit as a straight line with no outliers.
Jed, I don't want to waste your time with this, but this is the chart with the data points and I counted one every 5 seconds for 4 minutes, or 49 data points.
The LINEST function when it is applied for the blower power column against the air speed for these 49 points give the R-squared that I said.
To get that, you need to paste the results into an array so you can get the regression statistics:
Google LINEST site:https://support.office.com
It is possible I am confused about this. I intend to ask Mizuno when I get a chance. But as I said, I searched for a function that converts watts into air speed, and I cannot find one. There are small, random differences between them.
y=mx+b linear regression, i.e. LINEST if you are using Excel.
There are only 50 data points on the graph so if you can copy and past them here, we can try this ourselves. But -- I think we will find the r2 of nearly 1 will tell you that one is derived from the other.
I am curious as to why the blower power fluctuates at all, i.e. are we not using a voltage regulator that should keep the fan voltage constant, and does not the gas pressure impedance of the calorimeter over a 4 minute period stay constant (because the parts are fixed in place on the inside and the temperature barely changes). The power changes by 0.01/3.71 watts =~ 0.27% over a 5 second sample. If this was derived from input power, this is almost certainly volt meter measurement noise or amp meter measurement noise (both random), i.e. I would think that we cannot measure the blower input power to this level of precision.
Therefore, I believe that the "blower power" was derived from the air flow speed using the hot wire anemometer. It is likely subject to more noise, i.e. air burbles around the probe. Here is a typical "professional unit": https://www.instrumart.com/ass…x-6035-6036-datasheet.pdf
+/- 3% of reading or 0.15 m/s whichever is greater.
Here the airflow speed is varies 0.01/4.17 =~ 0.24%.
What is clear is that the fluctuations in velocity are almost certainly smaller than the accuracy of the anemometer, and that the anemometer is less accurate than the volt or amp meter, and thus the signal more likely to come from the common source of the anemometer. We are looking at measurement noise here and for the noise to have a r2 of almost 1 means that one signal is derived from the other.
If we want to confirm anything here, we need to actually create a signal. To do this, step the blower power in 5% increments from say 4 watts to 7 watts, and then measure the air speed so that we can actually say the air speed is proportional to the blower power, and not the noise. Using tiny 0.25% fluctuations at the extreme limit of the volt and amp meters and beyond the accuracy of the anemometer wouldn't help.
From engineering rules of thumb blower power is proportional to air mass flow. We can calibrate it and move on. If Mizuno will not, a replicator can using cheap instruments. This is a diversion here from the problem at hand -- measuring heat on R20. If the calorimeter+reactor is identical and if the internal heater in an identical R20 tube without reactants is used to calibrate up to 300 watts, all this talk about radiation to the RTDs is a theoretical modeling exercise waste of time.
Relicators: use the R20 tube to calibrate before you load it with reactant mesh. You have a heater in it already. Run the calibration with different powers (suggest varying by 25 watts from 0 to 300) at 1) vacuum < 10^-4 Torr, 2) helium at 2 Pa, 3) helium at 2000 Pa. The later two test conditions will prove that it doesn't matter what the conductivity is of the gas inside the reactor to the airflow calorimeter. All of this will prove that the RTD radiation effects, if any, have already been included in your calorimeter calibration. Thus, after the replicator has done this, he/she doesn't need to go thru the tedious thermodynamic modeling of the calorimetry.
Why does the blower fan motor need to reside physically touching the calorimeter box? Cannot the blower still function and be on the other side of insulation? Why are we needing to focus on this?
Generally speaking repeated load/deload cycles increase the permeability and reactivity of gas-loaded metals. Measuring the actual amount absorbed on each of the cycles though requires some very precise measurement of pressure and temperature.
I think we can do that with an extra empty "measurement tank" that we pump to vacuum, bake out, and then load with a known pressure and temperature of deuterium while keeping the measurement tank isolated from the reactor. With the reactor evacuated the first time, or only with a previously known amount of D2 gas pressure in it from previous loadings, we can then introduce the known amount of D2 at the known temperature and pressure from the measurement tank via a valve to the reactor input line. Does this seem to make sense to others here on the forum (including Alan)?
Jed and THH,
I hate to watch two of our best minds argue over something that is hopefully immaterial. This is like arguing over your high school algebra grade after you graduated from university. Does it really matter that you could have gotten an A instead of a B 10 years later? THH has made his point. Jed has made his rebuttal. To go back and forth here over the same thing is only frustrating.
If after the replication results come in in a few months this is material and has not been rectified in the revised experimental designs, we can revisit it then.
You have overlooked the reactor temperature. The thermocouple on it shows a temperature hundreds of degrees higher when there is anomalous heat compared to the calibration. Perhaps you think the thermocouple is malfunctioning. That is ruled out. It works correctly during calibration at low power and high power. When there is anomalous heat, it tracks the RTDs and thermometers, and agrees with them (based on the calibration). That cannot be a coincidence. It cannot be malfunctioning only when the other problems you postulate occur, exactly to the same extent.
Jed it right. The reactor gets really hot with only 50 watts. Even if the air flow calorimetry is somewhat mis-calibrated from direct radiation to the output RTDs, it still gets significantly hotter. This would be the important takeaway from replication: that R20 with only 50 watts in + Ni/Pd rubbed mesh + D2 gets much hotter than R20 or an otherwise identical control reactor with no Ni/Pd mesh and no D2. That is anomalous and that is what validators will show.
Jed and THH -- I suggest that arguing over calorimetry details with each other is unnecessary here. THH's critique is noted and the validators can take this into account to produce a more accurate calibration.
None of this changes what is most important to validate, that the Mizuno active reactor design (like R20) gets much hotter than a control. That is really important and will get all of our attention when proven. This proof will encourage many more people to focus on LENR and will put an end to the blind "Nature" articles calling this "unproven". We need this as a field so we can move to the next level.
If the unit is heating a room as was demonstrated in figure 1 and the COP is 10, what would happen if one built a brick oven structure around the unit making it more difficult for the heat to escape?
Would it reach a higher temperature and a higher COP?
Will it stop working at some higher temperature caused by this procedure?
If it keeps pumping out the same amount of excess heat and you insulated it with a brick oven structure, it will get hotter. If the reactor produces additional excess heat at higher temperatures, it could thermally run away by getting hotter, inducing the reactor to create more excess heat, which could make it get hotter yet again. If it thermally runs away, it could keep heating until it melts the nickel, melts the stainless steel container (dangerous), or until some other effect limits the output. The radiation goes as T^4 so that would ultimately also act to stabilize the unit if it is allowed to radiate. This excess heat vs temperature area is not fully explored in the R20 unit so caution is warranted. I would suggest replicating as Mizuno has done first before doing a different experiment. Once you have confirmed Mizuno's excess heat you can do other experiments using the Mizuno excess heat as the new "control" using suitable safety precautions.
Display MoreIt is connected as I said. That is clear from the paper and Table 1, which shows the pressure, and shows changes in the pressure. If it were disconnected you could not read the pressure, or pump out the gas, or add new gas.
If you "expect" some artifact then you are not a skeptic. You are a true believer. You have not found any reason to think there is artifact, and neither has anyone else. The results have a high signal to noise ratio. As shown in Fig. 5, the outlet air is 10 deg C hotter with excess heat than during a 50 W calibration. That is confirmed with other thermometers and thermocouples. It cannot be a mistake. It is not possible the fan is running that much slower; it would stop running completely, and burn. The flow calorimetry result is backed up by the fact that the reactor temperature during a 50 W calibration is 27 deg C, whereas with excess heat it is 380 deg C. So your "belief" is entirely based on faith, without a scrap of evidence to support it, and it is contrary to many physical laws. This is the opposite of skepticism.
If you were to say "I suppose there an artifact" I guess that would be skeptical. I would call it cynical or unreasonably pessimistic. But to "expect" an artifact where there no evidence for one, in a conventional instrument, using techniques that are employed by hundreds of thousands of HVAC engineers every day, is to "expect" a miracle.
I had the impression from the earlier answer you gave to me (about a week ago in the other thread) that the D2 valve and vacuum valve (separate valves) are shut "OFF" (closed) during a run, but that the pipes are still connected so that Mizuno can do the next run at a different pressure in the R19 run sequence by opening one valve or the other at the completing of a day's run. Further, the vacuum gauge remains connected to the conflat (the stainless steel cylinder) so that the pressure rise can be measured as negligible during a day's run (before the next pressure is set for the next day's run by adding D2 or pumping out the conflat with the vacuum pump). If you think this is correct, it would be helpful to the replication teams if you would confirm this to eliminate their confusion. Jed has indicated that the vacuum system has a good seal and does not leak so that the pressure stays relatively stable during a run.
For those replication teams who are extreme detail oriented, the heat lost thru the connected pipe can be measured with a thermocouple at the conflat end and at the valve or tank end of the inlet pipe, so that one can see the temperature drop between the two and then with a little bit of conductive metal and gas assumptions, get the heat transport out the connection pipe. I have not done this calculation, but I believe that it is minimal compared to the convective air (mass) heat transport out the calorimeter and the radiation out the conflat walls at the conflat measured temperature to the air and the calorimeter walls. I do not believe that this will throw off the calibration even if there is a vacuum in the pipe, or D2 at full conductive pressure (about >10^-1 Torr), i.e. I believe this small loss is immaterial to the result.
Good luck replication teams.
Here is an example of how closely the blower power correlates with the air speed measured by the anemometer:
Air temperature and other factors have no visible impact on the air speed. No doubt they do have an effect, but the fluctuations in electric power are so large they swamp these effects. You would need much more sensitive instruments to detect them.
I checked the engineering rules of thumb for ducted fans Jed and found that thermal mass transport (and hence power reported by Mizuno's calorimeter convective heat transport per degree C) is directly proportional to blower power, and that air temperature and pressure have null effect. The only thing that would have an effect is if the humidity was different as the specific heat is different for water vapor than air. Thus, 6.5 watts of blower power is the same amount of mass being transported per second past the thermometers, which if they have the same specific heat, would transport the same amount of heat for the same temperature difference. This is a nice feature of the Mizuno calorimeter.
It soon became clear that pressures of the order of 10-6 Torr were being reached. But the gauge’s design was attacked repeatedly. Morris Travers, who used it with Ramsay, complained that it was very sensitive to contamination by moisture.
The above quote concerns the Sprengel Pump. I suspect it suffices for a low cost experiment if and only if inlet and outlet cleaning continuously prevents unwanted intruders from entering the pump and the reactor.
One half liter can be evacuated in 20 minutes. Or should I say, it is claimed such a volume etc in twenty.
The Pirani type gauge according to my recollection uses a heated wire that will outgas any contamination from the wire by itself, i.e. the wire bakes itself off. It is essentially a fancy conduction gauge and is limited to around 10^-4 Torr in measurement. There are better gages (ion gauges) that can go to 10^-8 Torr if you or Mizuno feel that level of clean vacuum is important, but they can't measure higher pressures. Thus if you want to go from atmospheric down to super high vacuums, you need two gauges. I think 10^-4 Torr is enough.
Last comment is that you have to correct the Pirani gauge depending on the thermal conductivity of the gas, i.e. the vacuum is different at the same measurement for H2 than D2 or He or air. The gauge is really measuring thermal conductivity and then mapped into a vacuum depending on the thermal conductivity vs pressure of the gas.
My thoughts are that the Mizuno experiment needs simply a clean environment to load with no air or water vapor left in the baked out instrument. I believe that the Pirani gauge will tell you enough at the 10^-4 Torr level or equivalent thermal conductivity for D2 gas. At those levels the number of moles left of O2 is so small that you can rule out chemical once pumped out and it should also not interfere with the D2 loading and diffusion thru the palladium/nickel metal lattices.
Someone else can comment on your Sprengel pump. However, my recollection is that the full Mizuno R20 reactor is close to 6 liters so your pump down time will be a bit longer.
