Page 8 of 24 FirstFirst ... 34567891011121318 ... LastLast
Results 106 to 120 of 347
  1. #106
    Join Date
    Dec 2013
    Location
    San Antonio, Texas, USA
    Posts
    3,070

    Default

    Quote Originally Posted by RayG View Post
    Hi Rob,

    I'm not seeing any of the attachments, something about invalid attachment?

    "Invalid Attachment specified. If you followed a valid link, please notify the administrator"


    On the subject of who's saw used the soft steel, I'd argue that it's better just to be upfront and say who's saw it was. You are certain to get criticized either way, so this is the least painful way forward, also, I'd be totally open and upfront about the fact that you are considering starting a saw making business. People will be suspicious of your motives if you aren't totally open and honest.

    I believe and understand that you just want to make a better saw, I suspect that there will likely be others won't see it quite the same way as I do.

    Ray

    PS.. On the subject of saw smithing, I have a range of different sized dogs head hammers that are used for saw doctoring work, also a nice big proper saw anvil, and a few other saw doctoring hammers, I am missing the twist pein hammer... but I know a guy who has a couple.. one day I'll make him an offer he can't refuse..

    The dogs head is what you need for the hammering you are doing.
    I'm working on it Ray, how is it now?
    Innovations are those useful things that, by dint of chance, manage to survive the stupidity and destructive tendencies inherent in human nature.

  2. # ADS
    Google Adsense Advertisement
    Join Date
    Always
    Location
    Advertising world
    Posts
    Many





     
  3. #107
    Join Date
    Nov 2011
    Location
    Melbourne
    Posts
    6,973

    Default

    Dale
    Fantastic detective work

  4. #108
    Join Date
    Feb 2006
    Location
    Perth
    Posts
    27,756

    Default

    Quote Originally Posted by D.W. View Post
    Paul, when you say the data is precise, I think you're getting resolution and precision twisted around.

    Resolution implies showing a lot of detail, precise implies correct. There is a lot of resolution here, but the precision is not there for some of the measurements. The trouble is resolution implies or at least is perceived as precision by a lot of people.
    Nope - Paul has it right.
    "Precision" is the same as "resolution" and 'reproducibility"
    "Correctness" is the same as "accuracy"
    See http://en.wikipedia.org/wiki/Accuracy_and_precision

    FWIW I don't think sufficient "precision" has yet been demonstrated to adequately differentiate between saw types.
    Using precision for differentiation purposes between samples requires the measurement of standards or reference materials that bracket or are very close to the properties of the samples being investigated. This is required to take into account any non-linearities in the measurement system.
    And "accuracy" in this case is irrelevant as it not being claimed.

  5. #109
    Join Date
    Aug 2009
    Location
    Armadale Perth WA
    Age
    54
    Posts
    4,524

    Default


  6. #110
    Join Date
    Jun 2008
    Location
    Victoria, Australia
    Age
    74
    Posts
    6,132

    Default

    Quote Originally Posted by rob streeper View Post
    I'm working on it Ray, how is it now?
    Hi Rob,

    That's good, I can see the attachments and all good. I'm still wondering what it means.

    The Rockwell N45 (45kg) test obviously doesn't penetrate as deeply as the Rockwell C 150kg test, and I'm wondering if the results you are seeing are more a question of measuring the surface hardness.
    Is it possible the hammering produces a hardened surface, something like a hard skin on the outside of the plate while remaining softer inside? I've no idea, we need a metallurgist to help.

    Ray

  7. #111
    Join Date
    Dec 2013
    Location
    San Antonio, Texas, USA
    Posts
    3,070

    Default

    Quote Originally Posted by RayG View Post
    Hi Rob,

    That's good, I can see the attachments and all good. I'm still wondering what it means.

    The Rockwell N45 (45kg) test obviously doesn't penetrate as deeply as the Rockwell C 150kg test, and I'm wondering if the results you are seeing are more a question of measuring the surface hardness.
    Is it possible the hammering produces a hardened surface, something like a hard skin on the outside of the plate while remaining softer inside? I've no idea, we need a metallurgist to help.

    Ray
    Ray,

    I'll try to respond without violating anyone's presumption of confidentiality. I got a PM earlier today from an expert, from the sound of his post a real, honest-to-God-expert, who wrote that I was and am using the correct scale for thin materials. I encouraged my correspondent to post his opinion. I haven't received a reply as of yet but if we're fortunate we may get somebody who has unquestionable credentials to weigh in.

    Cheers,
    Rob
    Innovations are those useful things that, by dint of chance, manage to survive the stupidity and destructive tendencies inherent in human nature.

  8. #112
    Join Date
    Jun 2009
    Location
    Holland
    Posts
    72

    Default

    These look like reasonable numbers. 50-51 for 1095. One point less when sanded and 2 to 3 points more when work hardened. I wonder if this work hardening is only at the surface and won't change the ability to file very much?

    Here is some old data I found when googling arounf on the old tools list.

    http://swingleydev.com/archive/get.p...ubmit_thread=1

    The old saws all register lower values then your 1095 sheet. He also uses one of the N-tests. You can read about that lower down in that thread.

  9. #113
    Join Date
    Apr 2012
    Location
    Sydney
    Posts
    1,501

    Default

    Quote Originally Posted by rob streeper View Post
    Okay, levity and insults aside I've done the experiment I discussed with BobL.
    Nicely detailed post but I don't recall any insults. Certainly not here.
    You may not get 100% agreement with your findings but that doesn't make an insult!

    The testing makes interesting reading but the high and low values seem extreme.
    Often these would be excluded as outliers.
    With the saw that tested 60, can you file it?

    With thy the testing of an old Disston, rather than wrecking a good saw I was thinking of an old junker. If you sanded equally a bunch of spots around the saw plate to the same level, you could at least determine where an old saw was hard or soft in relation to the work done on that saw. Even if the those numbers are not comparative exactly to other saws, it may help explain the value of tensioning beyond just straightening.

  10. #114
    Join Date
    Dec 2013
    Location
    San Antonio, Texas, USA
    Posts
    3,070

    Default

    Quote Originally Posted by hiroller View Post
    Nicely detailed post but I don't recall any insults. Certainly not here.
    You may not get 100% agreement with your findings but that doesn't make an insult!

    The testing makes interesting reading but the high and low values seem extreme.
    Often these would be excluded as outliers.
    With the saw that tested 60, can you file it?

    With thy the testing of an old Disston, rather than wrecking a good saw I was thinking of an old junker. If you sanded equally a bunch of spots around the saw plate to the same level, you could at least determine where an old saw was hard or soft in relation to the work done on that saw. Even if the those numbers are not comparative exactly to other saws, it may help explain the value of tensioning beyond just straightening.
    I'm referring to the posts above by DW or David Weaver, a poster from Sawmill Creek with whom I engaged in a debate which was at times heated.
    Innovations are those useful things that, by dint of chance, manage to survive the stupidity and destructive tendencies inherent in human nature.

  11. #115
    Join Date
    Mar 2010
    Location
    US
    Posts
    3,075

    Default

    Quote Originally Posted by rob streeper View Post
    Ray,

    I'll try to respond without violating anyone's presumption of confidentiality. I got a PM earlier today from an expert, from the sound of his post a real, honest-to-God-expert, who wrote that I was and am using the correct scale for thin materials. I encouraged my correspondent to post his opinion. I haven't received a reply as of yet but if we're fortunate we may get somebody who has unquestionable credentials to weigh in.

    Cheers,
    Rob
    Rob, the reason the whole debacle on the other forum occurred is less because of the process you're using and more because the results are not reasonable or meaningful for at least some of the saws.

    Above again is clinging to the process but completely disregarding the results of the disston and atkins saws that are not correct:
    1) you have not struck at the tooth line where it really counts (as you described they are struck under the handle)
    2) you mentioned that you could file those saws without issue

    I would presume that a great deal of the assumption that older saws were harder has to do with those results, but professional toolmakers (you are not going to meet a more "honest-to-God-expert" than George Wilson) have suggested that's not the case.

    So regardless of whether someone who has knowledge about a portable tester vouches for your method, at least some of the results still need correction to even be meaningful and you need to do something about that.

    I have no interest in the process, I have interest in the results, and most people will have interest in the results.

    If you are going to compare old and new saws, you need to strike the old saws somewhere near the tooth line, and several times in different places to make sure you're not in an area that someone has hand hammered the saw near the tooth line.

    Do you follow what I'm saying about having results that are incorrect or have no utility? Regardless of what your process is or how finely you have it tuned, when you have results that are incorrect or that imply something incorrect, people will question the remaining results (I don't question the remaining results, the hang up is correcting the accuracy and finding out what other people already know, that disston and the other makers were not regularly making woodworking saws harder than 52 hardness because it presents a problem filing a saw).

    You need to step back from the process and take a skeptical look at the results, or at least caveat that they can't imply anything at the toothline of the saws.

  12. #116
    Join Date
    Mar 2010
    Location
    US
    Posts
    3,075

    Default

    Quote Originally Posted by rob streeper View Post
    debate which was at times heated.
    Well, I hope at some time, you come to view it as brisk and results focused and not personal, because it is brisk (which I favor) and results and utility focused (which I also favor). Fascination with process is OK, but a skeptical view of the results is always a good thing. If you can create a data set that stands up to skepticism and passes muster with other rule of thumb or practical tests, then you will be in good shape.

  13. #117
    Join Date
    Dec 2013
    Location
    San Antonio, Texas, USA
    Posts
    3,070

    Default An aside to discuss the saw of Maker 3

    Since the question has been raised as to whether or not enough data points have been acquired I have undertaken two more power analyses.
    Read about power analysis here: http://en.wikipedia.org/wiki/Statistical_power
    And using this program for performing power analysis: http://biostat.mc.vanderbilt.edu/wik...owerSampleSize
    Read about Student's t test statistic here: http://en.wikipedia.org/wiki/Student%27s_t-test

    I'll re-post this here because we'll be discussing it in some detail.


    Refer to the Wikipedia article on Power Analysis to understand 'power' as used in the program. I set the power value to 0.80, or 80 percent, which is a generally accepted power used in the design of studies for the testing of drugs and Class III medical devices in the United states.

    In short and with a fair degree of arguable word choices, a power analysis, as I used it for my most recent study posted yesterday, uses data from two historical sets of data points to determine how many sampling events should be conducted in a future experiment.
    Power analysis can also be applied in the analysis of historical data sets to determine whether or not they contain enough data points to have a certain degree of statistical confidence that the two data sets are different.

    'Paired' data sets contain equal number of measurements. In the figure reproduced below the data sets for each test article contain the results of 5 measurements, thus they are 'paired'.

    alpha as used in the program is this: http://statistics.about.com/od/Infer...d-P-Values.htm
    In these analyses I have been using an alpha of 0.05.

    The 'p' value is, again with a lot of interpretation to make it understandable here, a measure of the degree of statistical confidence that two sets of data points are truly different. P values less than alpha are generally considered to be 'statistically significant'. The smaller the p value, the more significant the difference. Small p values make physicians happy. For instance, a p value of 0.001 is considered to be more significant than is a p value of 0.04.

    Delta (δ) is defined as: A difference in population means. In this case it is the difference between the means of the measurements between Maker 1 and Maker 3, or δ = 52 - 30.8 = 21.2.

    Sigma (σ) is the standard deviation. I have chosen to use the standard deviation for the measurements for Maker 3's saw because it is largest and therefore will make for a more stringent power analysis. Thus σ is 4.15.

    In the program select the second tab 't-test' and you will see the input page that I used. In the first pull-down box from the top I selected 'sample size' because that is what I want to know. In the second pull down box from the top I selected 'paired' because my data sets have 5 measurements each and are thus 'paired'.

    In the next pane on the left I entered 0.05 for alpha. Below that I entered 0.80 for 'power'. On the right side of that pane I entered the value for delta at 21.2, and below that I entered the value for sigma as 4.15.

    Here is the output:

    Type of study: T-test
    Requested output: Sample size
    Design: Paired
    alpha=0.05 power=0.8 DIFF=21.2 SIGMA=4.15 M=0
    Sample size=2
    t-test confidence interval width=36.04322

    We are planning a study of a continuous response variable from matched pairs of study subjects. Prior data indicate that the difference in the response of matched pairs is normally distributed with standard deviation 4.15. If the true difference in the mean response of matched pairs is 21.2, we will need to study 2 pairs of subjects to be able to reject the null hypothesis that this response difference is zero with probability (power) 0.8. The Type I error probability associated with this test of this null hypothesis is 0.05.


    Okay, now for some discussion of that this means. The output of 'Sample size = 2' means that, given the difference in the averages measured hardness values between Maker 1 and Maker 3 with the standard deviation of the Maker 3 readings representing the expected data scatter, I only needed to take two hardness measurements on each saw to detect a statistically significant difference between the Maker 1 saw and the Maker 3 saw and I took 5 measurements, more than enough to support my assertion that the Maker 3 saw is softer than the Maker 1 saw.

    Now for the comparison of Maker 2 and Maker 3. Same same as above.

    Type of study: T-test
    Requested output: Sample size
    Design: Paired
    alpha=0.05 power=0.8 DIFF=18.4 SIGMA=4.15 M=0
    Sample size=3

    t-test confidence interval width=30.80268


    Again, I only needed to take 3 measurements on each saw to detect a statistically significant difference between the saws of Maker 2 and Maker 3 and I took 5 measurements, more than enough to decide that is very likely that the Maker 3 saw is softer than the Maker 2 saw.
    Innovations are those useful things that, by dint of chance, manage to survive the stupidity and destructive tendencies inherent in human nature.

  14. #118
    Join Date
    Mar 2010
    Location
    US
    Posts
    3,075

    Default

    Quote Originally Posted by Corneel View Post
    These look like reasonable numbers. 50-51 for 1095. One point less when sanded and 2 to 3 points more when work hardened. I wonder if this work hardening is only at the surface and won't change the ability to file very much?

    Here is some old data I found when googling arounf on the old tools list.

    http://swingleydev.com/archive/get.p...ubmit_thread=1

    The old saws all register lower values then your 1095 sheet. He also uses one of the N-tests. You can read about that lower down in that thread.
    Thanks for that, Kees. The results of those strikes shown at the top look like what we would expect to see for vintage saws.

  15. #119
    Join Date
    Mar 2010
    Location
    US
    Posts
    3,075

    Default

    Quote Originally Posted by BobL View Post
    Nope - Paul has it right.
    "Precision" is the same as "resolution" and 'reproducibility"
    "Correctness" is the same as "accuracy"
    See http://en.wikipedia.org/wiki/Accuracy_and_precision

    FWIW I don't think sufficient "precision" has yet been demonstrated to adequately differentiate between saw types.
    Using precision for differentiation purposes between samples requires the measurement of standards or reference materials that bracket or are very close to the properties of the samples being investigated. This is required to take into account any non-linearities in the measurement system.
    And "accuracy" in this case is irrelevant as it not being claimed.
    It appears there are several definitions of precision. I work in a technical field, and despite the technical designation on this definition, we use the former definition and not the latter for precision (as in, in professional use it implies accuracy), and use resolution to mean the fidelity of the data that is provided.


    pre·ci·sion
    prəˈsiZHən/


    • the quality, condition, or fact of being exact and accurate.
      "the deal was planned and executed with military precision"
      • marked by or adapted for accuracy and exactness.
        modifier noun: precision
        a precision instrument
      • technical
        refinement in a measurement, calculation, or specification, especially as represented by the number of digits given.
        plural noun: precision
        "this has brought an unprecedented degree of precision to the business of dating rocks"




  16. #120
    Join Date
    Mar 2010
    Location
    US
    Posts
    3,075

    Default

    Quote Originally Posted by rob streeper View Post
    Since the question has been raised as to whether or not enough data points have been acquired I have undertaken another power analysis.
    Read about power analysis here: http://en.wikipedia.org/wiki/Statistical_power
    And using this program for performing power analysis: http://biostat.mc.vanderbilt.edu/wik...owerSampleSize
    Read about Student's t test statistic here: http://en.wikipedia.org/wiki/Student%27s_t-test

    Refer to the Wikipedia article on Power Analysis understand 'power' as used in the program. I set the power value to 0.80, or 80 percent,which is a generally accepted power used in the design of studies for the testing of drugs and Class III medical devices in the United states.

    In short and with a fair degree of arguable word choices, a power analysis, as I used it for my most recent study posted yesterday, uses data from two historical sets of data points to determine how many sampling events should be conducted in a future experiment.
    Power analysis can also be applied in the analysis of historical data sets to determine whether or not they contain enough data points to have a certain degree of statistical confidence that the two data sets are different.

    'Paired' data sets contain equal number of measurements. In the figure reproduced below the data sets for each test article contain the results of 5 measurements, thus they are 'paired'.

    alpha as used in the program is this: http://statistics.about.com/od/Infer...d-P-Values.htm
    In these analyses I have been using an alpha of 0.05.

    The 'p' value is, again with a lot of interpretation to make it understandable here, a measure of the degree of statistical confidence that two sets of data points are truly different. P values less than alpha are generally considered to be 'statistically significant'. The smaller the p value, the more significant the difference. Small p values make physicians happy. For instance, a p value of 0.001 is considered to be more significant than a p value of 0.05.

    Delta is defined as: A difference in population means. In the case of the data above it is the difference between the means of the measurements between Maker 1 and Maker 3 or

    In the program select the second tab 't-test' and you will see the input page that I used. In the first pull-down box from the top I selected 'sample size'. In the second pull down box from the top I selected 'paired' because my data sets have 5 measurements each and are thus 'paired'.

    In the next pane on the left I entered 0.05 for alpha. Below that I entered 0.80 for 'power'. On the right side of that pane I entered the value for delta and below that I entered the value for sigma.
    There is no shortage in the number of data points, the issue is the data to begin with. I am a professional statistician of sorts, and cringe a little bit when I see basic statistical analysis described in such verbose terms as above. With a couple of small empirical data sets like you have provided, we can see that the variance in the strikes is fairly small and we could draw a strong conclusion from them as is. As a group of woodworkers, we don't need to know if the level of confidence is 0.87 or 0.94, it doesn't matter too much as long as its not 1/2 or something.

    But the confidence in the results doesn't mean they have utility.

    Let's presume that you struck several vintage saws under the handles because it wouldn't leave a visible mark. Let's also presume that all of the strikes are exactly correct. We see then that there is little enough variation in the strikes at the locations you took them to be concerned that resampling will provide a substantially different result.

    But again, it's focus on the process that's causing the problem, because no critical analysis of whether or not the under-the-handle strike is relevant has been done. There are two things of importance to a user of a saw (aside from the basic geometry of the saw), 1) how hard are the teeth at the tooth line (and any relevant area that could be expected to be teeth), and how does the saw feel in use in terms of tension.

    It could be that a saw plate that's harder up from the teeth is useful still because it would imply more tension (but still, the saw could not be harder than spring temper, or it would break and it has been demonstrated many times that you can take a vintage saw and touch the tip to the handle by bending it laterally and it will return to almost exactly the same position as you started).

    We don't know how much users care about hardness as it implies usefulness in a well-tensioned saw, but we do know that people using the saws, especially beginners, will be fascinated with the hardness at the tooth line because it implies the saw will stay sharp much longer.

    So the critical issue is before the data is even measured (and after given the problem with the values), and that's whether or not a handle strike tells us something relevant about the hardness of the saw in a way that has utility for a user. We don't know whether or not it does unless you do another test and strike a saw near the tooth line and the handle area and find out if there is a fairly common differential between the two (perhaps if you established that the difference was 10 on the C scale for a substantial sample, then you could go on implying that the hardness under the handle suggests fitness for use.

    What you're doing right now is similar to judging the usefulness of a plane iron or chisel by striking it at the tang.

    I don't think I'm saying anything that a lot of other people aren't thinking, but they are unwilling to engage in a long verbal debate. I have interest in you getting correct and relevant data because we may learn something from it once the data at least has some credibility across the board.

    Now, you have two data sets that don't agree with yours. One that was supervised by a professional that shows disston saws 44-48, and that of George (who had saws at his disposal to strike and ruin if he liked) that showed vintage saws between 40 and 45 on the C scale. It should tell you that you have some revisions to do or caveats to add to the results.

Page 8 of 24 FirstFirst ... 34567891011121318 ... LastLast

Similar Threads

  1. Hardening & Tempering
    By Dovetail in forum HAND TOOLS - UNPOWERED
    Replies: 6
    Last Post: 8th February 2014, 11:07 AM
  2. Case hardening
    By Pete F in forum METALWORK FORUM
    Replies: 11
    Last Post: 18th November 2011, 10:05 PM
  3. Induction hardening
    By morrisman in forum METALWORK FORUM
    Replies: 5
    Last Post: 2nd October 2011, 09:59 AM
  4. brass hardening
    By Eldanos of KDM in forum HAND TOOLS - UNPOWERED
    Replies: 14
    Last Post: 8th July 2010, 12:56 PM
  5. Timber hardening
    By boris in forum FINISHING
    Replies: 1
    Last Post: 31st January 2004, 11:01 AM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •