When you get a key copied, the locksmith usually asks if they key is an original or a copy. This is because the process of copying a key is lossy; each time you make a copy from a copy, detail is lost. After a few generations the key will not open the lock, and after many generations the copied key will be blank.

image Recently on the [IRC RepRap channel](http://reprap.org/wiki/IRC), someone piped up and asked a similar question about self-replicating 3D printers: how many generations can a RepRap reproduce before there are serious issues with print quality? A couple of people chimed in with answers, which basically boiled down to the fact that with the right calibration you can always make a better printer from your existing printer. This certainly parallels my own experience - the RepRap parts I am able to print today on my printer are vastly better than the first prints I got from my [friend's printer](http://michaelcandy.tumblr.com/). It was a bootstrapping process: those initial poor quality parts got assembled into my printer, which I used to create higher quality parts to upgrade both our printers. So it looks like the process of copying a RepRap printer from itself is not lossy at all, but rather the opposite. I don't even know what that is called; [_enhancy_](http://www.youtube.com/watch?v=KiqkclCJsZs)? Case closed. Almost. I have always had one tiny problem with my RepRap that no amount of calibration could resolve: ellipses. Circles would always come out a bit elliptical on my printer. I never noticed till I was printing some large disks for a [replica hoverboard](http://www.geek.com/wp- content/uploads/2012/02/PDVD_000.jpg) a friend was building.
The problem came down to the printed pulleys I was using. They were not circular, and despite how much calibration or tweaking I did, I could never print a perfectly circular set. I had been using [CNC couplers](http://rover.ebay.com/rover/1/711-53200-19255-0/1?icep_ff3=9&pub=5575111803&toolid=10001&campid=5337638891&customid=&icep_uq=5mm+8mm+aluminum+flexible+coupler&icep_sellerId=&icep_ex_kw=&icep_sortBy=12&icep_catId=&icep_minPrice=&icep_maxPrice=&ipn=psmain&icep_vectorid=229466&kwid=902099&mtid=824&kw=lg) on my Z-axis for a [while](http://reprage.com/post/24304653117/aluminium-couplings), and decided it was time to get some decent [machined pulleys](http://rover.ebay.com/rover/1/711-53200-19255-0/1?icep_ff3=9&pub=5575111803&toolid=10001&campid=5337638891&customid=&icep_uq=timing+pulley&icep_sellerId=&icep_ex_kw=&icep_sortBy=12&icep_catId=&icep_minPrice=&icep_maxPrice=&ipn=psmain&icep_vectorid=229466&kwid=902099&mtid=824&kw=lg) to drive the X and Y axes. The machined pulleys do their job really well - print quality is up and circles now get printed, well, circular. RepRaps are 'enhancy' only if you are using high quality, machined parts for power transmission. Then you will be able to reproduce many, many generations of printers without any serious issues with print quality. However, RepRaps are 'lossy' if you are using printed pulleys for power transmission. In my experience, any defects will result in elliptical pulleys that no amount of calibration can correct. These pulleys get worse and worse with each new generation, introducing alignment issues to the detriment of print quality. So while it is true that RepRaps are self-replicating, there is a caveat - if you don't want any 'genetic defects', you're going to need some machined parts.