4 decimal places in inch dimensions

I know it’s come up before but I want to explain why it is important to me. I’m the one that matters, after all.

I work in a high end decorative metal shop. It makes sense to draw things in SketchUp for clients’ aesthetic needs and architects’, (and my boss’s), technical needs. Great so far. Fractional dimensions are fine.

So when everybody’s happy and we know how things will look and fit together, it is time to send approved drawings to the shop. They’re accurate to a 64th I say. That’s as good as 6 decimal places I point out! I know these guys can’t find a 64th on their tape measures!

The trouble is that they use calculators in the shop. When you are calculating placement of scores of pickets on a long stair rail on a hot day, it seems like a special kind of hell to memorize the decimal equivalents of even 16ths, much less 64ths. And you can’t just figure the spacing then add one picket at a time. You need to place the first and last, and then the center pickets, then continue dividing the open spaces. You don’t want to find yourself welding the hundredth picket and realize you’ve been off by .01 for 99 pickets in a row.

They demand 4 decimal places. These are guys that toss steel around all day and ride motorcycles on weekends. So it is off to AutoCad I go, (which is easier than the scaled then changed dimension trick.)

Woe is me.

The values you site are display precision, not actual internal cal. precision. You have the option to change display to decimal if that helps. Of course you still have to understand the difference between seg length and actual based on how Su approximates curves. Since the shop guys use the computer it should be easy matter to make a recursion formula to give them spacing so to have an integer answer. That should save them time should would reduce trial and error approach=> I would not want to eliminate since measure twice cut once is stll and good idea unless all is machine controlled.
The current display precision is based on the accuracy one gets with a 32 bit machine which is 7 digits irrespective what you think with the fraction. I worked on some builds where we used long invar tapes to calibrate the measurement setup to reduce temp related errors since invar has close to zero CTE.

1 Like

Measuring to the nearest 1/100" of an inch is more precise than the nearest 1/64" … however, it is only accurate to 2 decimal places (i.e., 0.01", 0.02", 0.03" …). Similarly, 1/64" may equal 0.015625" which is definitely 7 decimals, but the precision is less than 2 decimals. Personally, I use decimal inches for almost everything and set the precision to 0.000000".

1 Like

You need to do the math and not rely on the display. I was addressing what the capability of a 32 bit machine is and it can only have a 7 digit accuracy or really a little more if if you want to count tenths of a digit. You can display many more if you want but accuracy is still only 7+ digits. I agree if you don’t what to get fooled never use the fraction format for accurate display. If you are talking accuracy and see a number such as .02 there is a question how many zeros follows if accuracy is the question, you do not know unless there is some pre understanding and at times that is done by over strike on the number zeros. If search is done you will find that issue discussed.
One needs to be careful and not confuse accuracy with precision. I can have MIC which measures with a .00000 precision but the accuracy is only 1 inch lets say. This number 50000.00 has same accuracy as .5000000

I guess that I should not have assumed that people would understand, given the thread, that I’m talking about Layout here, not SketchUp proper. Drawing is not the problem. Display is the problem – sending drawings to the shop, on paper using parallel projections and standard views.

The shop uses calculators not computers. They can’t type in 12’ 5-11/16". They can’t even type in 12’ 5.6875". They want to type in 149.6875.

I have to eat lunch with these guys. I give them what they want.

I’m curious about this since I’ve understood for a long time, (Lord knows where my notion comes from), that AutoCad has 15 decimal places of accuracy as long as you stay near the origin. How could they do this on 32 bit machines?

double precision integers – they use two bytes to store the number instead of 1 byte

Actually, the number of bytes depends on the programming language and hardware. In the most common cases today, a char is 1 byte, a short int is 2 bytes, and a long int is 4 bytes. Some languages have even longer integers available. But more relevant to 3D modeling are floating point types: a float is usually 4 bytes and a double is usually 8 bytes, both of which include the sign, exponent, and mantissa.

1 Like