Air Noise
------
Regarding the noise, any significant noise source will bleed into the platter. It won't be louder than the signal coming from the groove (it should be pretty quiet by the time it got the cartridge), but you aren't making this just because it will play a record, you want the ultimate, right? If you wanted good enough, I doubt you would be looking at air bearings.
Current Encoder
------------------
Your encoder situation is tough. If your 1000 line encoder was perfect, you could come up with a scheme to do a phase locked loop with the pulse edges and get pretty good velocity info. The problem is that the average optical encoder is not very consistent on where the edges on the pulses are. They typically spec +/-45 degrees (electrically) or so. That's fine for reading position, but not for phase locked loops. The encoder is fine at 20 krpm, but the signal to noise, limited by quantization noise is pretty ugly in the 33 rpm range. Think 4000 counts (1000 lines x 4 for quadrature) / rev and 33 rpm (0.55 rev / sec) gives 2222 counts / sec. One count of error (smallest increment of error) and you get .05% speed error at 1 Hz bandwidth or 5% at 100 Hz. That's pretty bad by anyone's closed loop measure. Let's say that your encoder is good to +/- 18 degrees on edge phase accuracy (optimistic), You could see a 10x improvement by going to a phase locked interface, but that still isn't too great, and you can't really get a commercial product that would do that too easy.
Ideal Encoders
-----------------
I would shoot for 1,000,000 count/rev encoder, which would give you a 250x increase in velocity resolution, which should give you a fighting chance. Those were very exotic until the last decade or so and can be found on eBay from time to time. I've had good luck with Micro E encoders for a relatively low cost but super performance encoder.
http://www.microesys.com/m1encoders/data.html In an ideal world, I would suggest a big ring encoder like this, but it's likely out of the budget
http://www.renishaw.com/en/6453.aspx
Also, another option is something like a 1000 line sin/cos encoder. They have analog outputs and if you (your drive or controller) can interpolate the signal, 1000's X interpolation is not uncommon. The Micro E stuff really just does that and hides the details from you.
Controller
-----------
A lot of smart drives have more than enough functionality to do the velocity ramp up / ramp down on a push button, like you need, but they will be industrial PWM drives. They are miracles of modern manufacturing and control, but from your cartridge, wiring, and preamp's perspective, they are like radio stations. No grounding scheme will protect you from the 16 kHz (varies, drive to drive) carrier frequency, and you will likely hear noise from the motor, which isn't good in a turntable, no matter now low the level. What you need is a linear amp, but that pretty much rules out smart drives. I would probably look at galil (
www.galilmc.com). You might also look at the microcontroller based motion control market. I'm not super familiar with it, but they seem to be getting better all the time and you can get <<$100 controllers there. In any case, if you use a linear drive, you will likely need to deal with commutation in the controller because linear drives will need two phase inputs that control commutation.
Drives
-------
I don't think you will be able to get away with a PWM drive and get super low noise, audibly and electrically. I have used these linear drives with great success and the company has really great people. Remember that you need next to no power, so the smallest drive will do the job.
http://trustautomation.com/cm/Standard_Products/Drives.html
Realism
---------
If the linear amps and controllers and fancy encoders put this over the top, take a look at this
http://www.elmomc.com/products/solo-elmo-integrated-servo.htm. A buddy at work is using some of their stuff. As long as you don't get visions that it will be the most precise control loop in the world and realize that it is pwm, you get a very small, reasonably priced, stand alone product from a company that makes good stuff. A turntable project that gets done slightly less ambitiously than the ideal is much better than the ideal one that never gets done, and this might make the difference.
My Old Turntable
-------------------
I'll see if I can dig up any pictures of my old turntable. It didn't survive some re-evaluation of my priorities. It was two granite surface plates, mounted flat surface to flat surface. The top one was cut to be round. When some air was pumped inbetween the two surfaces, they made a great air bearing that would float << 1 micron height. After much experimenting, I found that a weed sprayer made a great compressor and would last more than a record on one pump. It solved the noisy pump problem. I used an old Gerrard turntable motor and a belt made of a bungee cord fiber. I was in the process of converting it to direct drive with a home brew brushless motor and an industrial controller system when I scrapped it. It was a lot of fun and shockingly good at sounding relaxed.
Let us know what path you choose, it sounds like you have a great project.
Erik
