Marcus Bowman | 04/09/2017 08:03:23 |
196 forum posts 2 photos | Posted by John Lluch on 03/09/2017 21:16:28:
Posted by HOWARDT on 03/09/2017 19:37:07:
If it is the intention to keep the creations as CAD models and drawings, what format will be used. I worked as a designer all my life, and went to CAD in 1989. With changes to the core of the CAD systems old files become unreadable unless you convert immediately after the change. Also later minor changes make for incompatibility. The only sure way is to create jpg scans or similar, even PDF may not be an option. Just a thought, but been there and got the tee shirt on reading old files! About CAD, I am not aware of any format that would open (in an editable form) in all software platforms. However, for the purposes of dynamic display only, there are several file types that will open in most CAD systems. This does not solve of course the problem of maintaining an editable version of the 3D plans, which will be required to keep future compatibility. I am used to SolidEdge from Siemens because that's what I learned, and that's what I would chose, but I suppose SolidWorks can be considered a better choice because I understand it is more widespread. Other CAD systems have options to import SolidWorks files, though generally in a non fully editable way. Finally, if the possible legal issues about ownership can be solved, a good thing to do would be to open source the 3D plans files to let everybody propose improvements or correct errors by pushing updates based on experience while building the actual locomotives.
Interesting points, raising different issues. Firstly. although I like the idea of open source plans available in an editable format, the CAD platform would have to be open source, to allow anyone to edit or alter the plans. There are several possibilities, but there is a somewhat larger question. The nature of open source suggests that a user could make changes and re-post the plans. That might be useful for allowing the correction of universally accepted errors or making necessary modifications. But that leads to two questions: who is to say those errors and/or modifications are necessary (because the plans would then deviate from the original); and who then controls availability of the new open source plans? One solution is to use a platform such as GIT, where changes are tracked, and it is clear which versions are which, and, especially, which is the current version. Controlling the addition of modifications and 'enhancements' becomes important as plans are changed and morph into something else, over time. Take 'Pansy' for example. Doug Hewson's version is more accurate and up-to-date, incorporating much more modern approaches, so if I was to make a Pansy, I would choose Doug's drawings. But would it then be fair to call that LBSC's PANSY? It's not a straightforward question. I would not want to use any of the LBSC boilers, because they really need to be brought up to date to comply with current legislation and safety requirements. But does changing the boiler make those designs non-LBSC? We need some person or group to control this aspect. I might draw the distinction between the developmental course of Mach3 and LinuxCNC, for example. Mach3 was originally updated and amended by its creator. When it was sold on, modifications and bug corrections stopped, as the new owners focussed on Mach4 development. They control development, so that was their choice. LinuxCNC, on the other hand, is open source, and undergoes active development, with new versions available from a central repository. The development is driven by user suggestion and by the willingness of users to contribute to that development. Older versions remain available. A mechanism of some sort is required to manage any changes. It could either be the original draughtsman, or a user-driven group. What would not be good would be the model where development was frozen once the initial drawings had been released. As to the format of the plans; this could either be an agreed open source CAD package, or simply the DXF files. The difference is that having the CAD package available to everyone would allow anyone to edit the plans and perhaps incorporate their own mods, for their person use. The DXF files would be the most useful for the person who simply wanted to take and use the plans. DXF may allow some modifications within some packages, but would certainly allow the CAM stage to take place, to prepare G code files for manufacture by CNC. PDF files would allow non-CNC manufacture, just like existing paper plans. The benefit of PDF files is that printing could take place at any size or scale. Larger format plans can be printed by may local print/copy shops. JPEG files are not quite as useful, if the print size is to be made larger, as definition suffers when the file is enlarged beyond the native size. PDF, on the other hand, scales well. Marcus
|
Neil Wyatt | 04/09/2017 08:08:31 |
![]() 19226 forum posts 749 photos 86 articles | Posted by John Lluch on 03/09/2017 23:53:23:
Posted by Neil Wyatt on 03/09/2017 22:08:15:
Posted by John Lluch on 03/09/2017 20:01:39: Decimal numbers is what CAD/CAM and CNC will use, so when converting from inches to mm it all becomes a mess of dimensions with infinite (to be exact) decimal places. That causes trouble due to accumulation of errors and rounded display of dimensions. Not strictly true, as imperial fractions are all powers of two, they are rather well suited for representation by computers. Even the worst case of converting 1/128ths to millimeters it's only seven decimal places, exact.
Hi Neil, I am not totally sure of that because I'm not too used to it, but I recall having seen fractions in LBSC plans that were divisible by 3. This creates a periodic number with an infinite number of decimals. I hope my comments are taken in the spirit of light hearted musing on the subject, not an argument for arguments sake... It's the denominator that matters, you won't have seen an LBSC fraction whose denominator wasn't a power of two. So 1/128 (for example) Having written a BASIC interpreter I know about floating point representation. You've actually handed me a wonderful proof that imperial fractions are better than metric decimals... As you say figure such as 0.1 can't be stored precisely on a computer using usual floating point formats, but let's try a fraction, such as 1/128, the smallest you are likely to find (I have to input the fraction as the equivalent decimal): Oh look! The error is... nothing! Forget multiples of three, let's use a prime number as the numerator, how about 7/128: Wow! Perfect representation again. Clearly Imperial fractions are completely superior.
Duncan makes fair comment that, at 1:12, for example some dimensions become inconvenient. A 1" dimension becomes 1/12" which is rather awkwardly between 1/16" and 3/32" stock sizes. That's why I said many full size dimensions some out as convenient sizes not all. Due to a quirk of fate 1mm ~= 1/24" so fortunately 2mm is close enough to use without anyone noticing! This makes 1:12 a very convenient scale to work to if you want to use as many readily available stock sizes as possible. Metric buffs can work to 1:10 which is fine for most things but raises problems on popular rail track gauges. Neil |
Neil Wyatt | 04/09/2017 08:09:31 |
![]() 19226 forum posts 749 photos 86 articles | For the sceptical: www.h-schmidt.net/FloatConverter/IEEE754.html
|
JasonB | 04/09/2017 08:31:09 |
![]() 25215 forum posts 3105 photos 1 articles | Posted by Neil Wyatt on 04/09/2017 08:08:31:
but let's try a fraction, such as 1/128, the smallest you are likely to find (I have to input the fraction as the equivalent decimal):
Ah but what about a part dimensioned as "full" or "shy" you can't easily express that as a fraction Easy enough to do in decimals by increasing or decrasing by say 0.01mm or 0.0005" |
John Lluch | 04/09/2017 08:51:06 |
![]() 31 forum posts | Posted by Neil Wyatt on 04/09/2017 08:09:31:
For the sceptical: www.h-schmidt.net/FloatConverter/IEEE754.html
Hi Neil, You are an interesting man lol. Actually, while I was posting my 0.1 decimal to floating point binary conversion example, I wondered if you (or somebody) would refute that by showing fractions that are powers of two. I honestly didn't think this would be the case on a model engineering forums. But I'm absolutely delighted about your reply!. Having written a BASIC interpreter puts you in a strong position for that. My background is computer software development, and I too have written a couple of interpreters of common (old) computer languages, and a compiler for a specialised computer language aimed at the knitting industry. I am also the main developer of the HMI Draw and HMI View apps in the App Store, which also contain a real time pseudo-compiler to RPN for fast execution of Ruby expressions. So yes, you are right. Indeed, fractions commonly used in imperial dimensions have a better representation in computers.
|
IanT | 04/09/2017 09:00:45 |
2147 forum posts 222 photos | An interesting point Neil - but since we've drifted well away from LBSC - onto the use of floating point with it's various issues - another quick aside. I play with embedded devices occasionally and use a Forth factoring technique to simply avoid the use of floats completely. Simply put - decide the very smallest value to be worked to - say 1 thou - and that becomes your base integer value. An inch is then represented 'internally' as 1000 - so all arithmetic is integer based - conversion to/from the base factor occurring only on input and output. This also seems to be my own 'embedded' computers way of working too - generally thous seem easier for me to work with if anything is less than an inch (no built-in FPU I guess) Regards, IanT |
John Lluch | 04/09/2017 09:53:13 |
![]() 31 forum posts | Marcus. I would expect that as younger generations enter the hobby, the development of open source communities supporting and updating model engineering plans in CAD format will eventually become a reality. You are making very valid points on this subject. IanT. I also want to add that Integer numbers up to a certain figure are EXACTLY represented in floating point formats. Also floating point arithmetic of integer numbers is EXACT. I have no doubt that CAD software uses floating point numbers to store internal dimensions, but I sometimes suspect that the actual stored values are scaled up to a particular factor to prevent accumulation errors. I do not know about other CAD packages, but I never, and I mean NEVER, have seen or suffered an accumulation error on SolidEdge. It behaves for example as if 1 was internally stored and computed as 10000, or 0.1 as 1000 and so on. So unless you get to specify a really small number or a dimension with a ridiculous amount of decimal places, the chances of accumulation errors are nil. This, or the software uses a really clever approach at rounding floating point arithmetic errors.
Edited By John Lluch on 04/09/2017 10:17:25 |
Michael Gilligan | 04/09/2017 11:05:22 |
![]() 23121 forum posts 1360 photos | John, The fractions debate has, I think, been concluded ... but let me just mention one practical instance, from personal experience. The interface dimensions for the 5" x 4" 'Graflex' sheet film holder [which became a de-facto standard for the camera industry] were very obviously designed in 1/128ths. Years ago, Polaroid very kindly provided me with a copy of their drawings for the 545 film holder **LINK** http://camerapedia.wikia.com/wiki/Polaroid_545 and when working through these, it became immediately apparent that the clumsy-looking decimal dimensions were in fact 'rounded' conversions from exact fractions. I don't know how common the practice of designning in 'woodwork' units was; but I think it was fairly widespread in the USA, in the late 19th & early 20th centuries. MichaelG. |
John Lluch | 04/09/2017 12:23:21 |
![]() 31 forum posts | I miss a 'like' button in this forums. |
SillyOldDuffer | 04/09/2017 13:41:58 |
10668 forum posts 2415 photos | Posted by Michael Gilligan on 04/09/2017 11:05:22: ... I don't know how common the practice of designning in 'woodwork' units was; but I think it was fairly widespread in the USA, in the late 19th & early 20th centuries. MichaelG. I too am conscious that we risk deviating too far from Dean's original question, but it may be helpful to understand the mindset of early British designers. Before about 1900 decimals were not universally taught in British Schools and I expect the same was true in the US. Instead, pupils focussed on fractional representations because back then fractions were the basis of most practical systems of measurement: an ounce is 1/16 of an avoirdupois pound; a shilling is 1/20 of a pound sterling; a foot is 1/3 of a yard; an hour is 1/24 of a day etc. etc. In that context, fractions can simplify mental arithmetic. Despite the utility of fractional systems in simple cases, they bite in complex situations. In contrast the decimal system is smoothly suited to the more difficult sums required in finance, science, mathematics, statistics and advanced engineering. In these fields it is understood that a number can be expressed to any desired accuracy simply by calculating more decimal places. Decimal worked very well back in the day with Slide Rules and Log Tables etc, and they remain solid in modern times on computers. Fractions are still taught, and are still useful, but the demise of old measure has undermined their previous importance. Today most of us work in decimal. However, in the 19th and early 20th century, most practical Brits doing design would have had a natural bias towards fractions. It made design easier, though probably not optimal. In a 19th century workshop, fractions would have been well understood, decimals less so. But technology was being driven forward by numerate science, maths and engineering, in which most calculation was inexorably decimal. Electrical Engineering and Aerospace do not feature fractions. Even so I think decimal took about a century to replace fractional methods in British Design. I have Engineering Drawing textbooks that seem to bear this out. Parkinson (1933) is almost entirely fractional with decimals only appearing in the gear section. Abbott (1949), is mainly fractional but it's not unusual to find examples of mixed fractional and decimal dimensions in the same drawing, and several examples are entirely decimal. In my 1972 textbook, fractions have gone completely - Mott is decimal throughout. There's a change of approach in the 1972 book as well. The drawing examples in the earlier books are of Steam Age objects like Flywheels, Con-rods, Couplings and Wall Brackets. Mott has none of those: his examples are generalised with ideas that can be applied to any engineering object. For example, in Mott, the design of a lathe half-centre is about developing curves, not about drawing a centre of a particular size. Mott is about design effectiveness rather than keeping the arithmetic simple. Presumably the shift was partly made possible by the appearance of calculators and other mechanical aids. However it's done, design is hard work! It may take a long time but my feeling is that LBSC's designs are heading for oblivion in their original form. They are already difficult to implement in most of the world and signs of stress are apparent in the UK. Some items (like plates rolled to a gauge thickness) are no longer available and others are unpleasantly costly. My feeling is that the best way forward would be for Dean to capture the designs with their original dimensions. Once that's done, a second stage would be to pragmatically convert dimensions so that the engines can be easily built using metric materials. Metrication could be done by someone else. Conversion would also flush out LBSC's mistakes and, if necessary, provide an opportunity to to modernise the design. I suggest LBSC's designs are a product of their time. If LBSC were 25 today, I think he'd work in metric. After all his objective was to make live steam practical for the average builder, not to make life hard by insisting on obsolete ways and means. Dave |
Michael Gilligan | 04/09/2017 14:16:49 |
![]() 23121 forum posts 1360 photos | Excellent post, Dave ... Very well-reasoned. MichaelG. |
Bob Youldon | 04/09/2017 14:35:51 |
183 forum posts 20 photos | Hi all, Going back to the original posting regarding the LBSC designs, it must be remembered most date from the early twenties through to the mid sixties and the imperial measurement was in full swing; not one of his designs was intended to be produced by todays CAD, CNC, rapid prototyping etc, each were intended to be a hand built job using the limited facilities that were available then and all fitting was by hand by the builder, it doesn't matter a jot if it's measured in whatever unit the builder choses. I would suggest some of the correspondents read his Maisie instructions, or come to that Martin Evan's Springbok, both designs have been built in their hundreds, probably more examples than their full size counterparts! Steam locomotives are not complicated. Take a look at the drawings for Stephenson's Rocket, 1/64"s and 1/128"s abound, no metric there then! and that's where it all started. As I approach my mid seventies I will and can easily work in either imperial or metric, I have to; it doesn't faz me but don't decry one format as one being better that the other. When it comes to the actual process of constructing your locomotive, I'd like to see the first all CNC built boiler! Regards, Bob |
Neil Wyatt | 04/09/2017 15:40:17 |
![]() 19226 forum posts 749 photos 86 articles | Posted by IanT on 04/09/2017 09:00:45:
An interesting point Neil - but since we've drifted well away from LBSC - onto the use of floating point with it's various issues - another quick aside. I play with embedded devices occasionally and use a Forth factoring technique to simply avoid the use of floats completely. Simply put - decide the very smallest value to be worked to - say 1 thou - and that becomes your base integer value. An inch is then represented 'internally' as 1000 - so all arithmetic is integer based - conversion to/from the base factor occurring only on input and output. It is great fun devising ways to generate 'adequate' precision numbers using an 8 bit microcontroller. I used one to generate a table of sines*256, the calculation looked up the sine of an angle (in 0.5 degree steps) then multiplied the hypotenuse length by it, then dropped the low bit to give the length of the adjacent side. The table used one byte per step, but had to trap values of 87 degrees or more as they needed to be 0x100, in which case they simply returned the same as the input. Fine for graphics on a 64x128 screen! Neil |
duncan webster | 04/09/2017 19:44:08 |
5307 forum posts 83 photos | It is great fun devising ways to generate 'adequate' precision numbers using an 8 bit microcontroller Neil The expression 'should get out more' springs to mind. I've just spent at least an hour sorting out why some code gave silly unpredictable results. Turned out to be trying to handle numbers too large to be integers. I've had this trouble before, so it didn't take the days it took last time, but I shoud have spotted it straight away |
RRMBK | 04/09/2017 19:49:09 |
159 forum posts 18 photos | " Marcus. I would expect that as younger generations enter the hobby, the development of open source communities supporting and updating model engineering plans in CAD format will eventually become a reality. You are making very valid points on this subject. " I fully support the above comments and this is probably the most practical method by which we are going to introduce newcomers to a hobby which is clearly dwindling. You only have to look at the reduction in ME castings suppliers in recent years. Whilst the LBSC 31/2 & 5 " designs still sell, Reeves for example, no longer sell the 2.5 gauge designs they used to and there are very few G1 suppliers for the LBSC designs . Whilst I appreciate the technical details that Marcus throws up, If these 3D draftings are acknowledged as re drawn and any omissions or errors are the responsibility of the user to check prior to use, either for hand manufacture or CNC then I don't see what the problem is. I think there is a real opportunity here to revive interest in our King of Hobbies among a whole new generation, lets not pass it up. Kind regards Brian. Edited By Neil Wyatt on 05/09/2017 09:47:26 |
Dean da Silva | 04/09/2017 21:17:42 |
![]() 221 forum posts | If it comes down to it I will buy collections of these magazines if I have to. |
Jeff Dayman | 05/09/2017 13:46:56 |
2356 forum posts 47 photos | All the recent high-end numbers theory and metric/imperial debate to me is pretty much the opposite of what LBSC was trying to do - to enable Joe Blow to make a good running small locomotive with basic equipment (and maybe his only initial measuring tool being a 6" fractional scale rule). Could we get back to that theme please, as Mr da Silva started out to do, and dispense with the extensive theory, at least in this thread? |
John Lluch | 05/09/2017 14:52:54 |
![]() 31 forum posts | Posted by Jeff Dayman on 05/09/2017 13:46:56:
All the recent high-end numbers theory and metric/imperial debate to me is pretty much the opposite of what LBSC was trying to do - to enable Joe Blow to make a good running small locomotive with basic equipment (and maybe his only initial measuring tool being a 6" fractional scale rule). Could we get back to that theme please, as Mr da Silva started out to do, and dispense with the extensive theory, at least in this thread? Indeed, the thread is about making the hobby more accessible and appealing for the next generations. He also was concerned that the LBSC designs would eventually fall into oblivion. Dean da Silva proposal was creating updated 3D plans of existing LBSC designs, and possibly adapt them to what's available or more common today. One of the questions he raised was whether he should keep the plans in imperial units, or switch to metric. He said he would rather choose metric. I feel some responsibility for your discomfort because I first ignited the debate about the practical difficulties of moving existing plans from imperial to metric. My view was that I would rather stick to imperial for the existing LBSC designs, but only move to metric for any new developments. Others seemed to agree on that, or otherwise they highlighted the feasibility of moving to metric. However, instead of reviving the LBSC designs, my preferred proposal would be to work on a set of totally new locomotive designs, with several levels of difficulty, created from scratch in 3D CAD, and starting with a very simple entry level locomotive. The entry level locomotive should be easy to make for the average person, and require only simple tools to be completed. For example, it could be designed in a way that it would use a majority of laser cut parts, and avoid or minimise the need for castings and complicate machining. Essentially, a revival of the old LBSC spirit, but not by using his old locomotive designs, but by creating totally new ones.
Edited By John Lluch on 05/09/2017 14:53:21 Edited By John Lluch on 05/09/2017 14:57:49 Edited By John Lluch on 05/09/2017 15:01:26 |
Marcus Bowman | 05/09/2017 15:41:24 |
196 forum posts 2 photos | LBSC's magazine articles did that, of course. His subsequent books did the same thing for individual designs. In modern times, the equivalent is the ARM1G design, the book for which is shown here: http://www.g1mra.com/shop/ and its predecessors (Project and Dee), although ARM1G is designed to accomplish just what you describe, including lots of parts which can be bought from trade sources, such as frames, complete cylinder assemblies etc, so that the entry level can be tailored to an individual's pocket, capabilities and resources. There is a case for 3D modelling of LBSC's designs, as a project in its own right. There is another case for updating those designs, by converting sympathetically to metric dimensions (which would get my vote) to make them accessible for younger or newer builders (say under 70), and/or by updating the designs comprehensively to modern standards and practices, including modifications in the light of previous builders' experiences. There is also a case for preserving the 'words and music' of text and drawings from the magazine articles. Those preserve an important aspect of his success, which is the 'can do' spirit. I agree with earlier comments about LBSC's intentions, and his ability to describe how to do the job with limited resources. It would not, however, be my approach now, as I think that although it may (and I do say 'may' Drawings re-drawn by the magazine staff would be M.E. copyright, of course, as would the page layouts, but the words and original illustrations and drawings belong to LBSC's estate. Which takes us full circle, I think... Marcus
|
Niels Abildgaard | 05/09/2017 16:21:27 |
470 forum posts 177 photos | If we can have a society named Deans L&BSCR with a anual fee of say 25£ I will join imidiately. If it comes to voting I want metric and Fusion 360 or Onshape as CAD system.
|
Please login to post a reply.
Want the latest issue of Model Engineer or Model Engineers' Workshop? Use our magazine locator links to find your nearest stockist!
Sign up to our newsletter and get a free digital issue.
You can unsubscribe at anytime. View our privacy policy at www.mortons.co.uk/privacy
You can contact us by phone, mail or email about the magazines including becoming a contributor, submitting reader's letters or making queries about articles. You can also get in touch about this website, advertising or other general issues.
Click THIS LINK for full contact details.
For subscription issues please see THIS LINK.