If your AP has a max TCP throughput of 40Mbps when using a 20MHz channel, this would put the AP at 100% of its airtime utilisation when the AP was passing 40Mbps.
To ensure that you dont overload your AP's, it is always best to work to capping the limits within your design to be using 75% of the AP's airtime utilisation.
This being the case your max TCP throughput value you should be aiming at the AP consuming is 30Mbps.
The simple math now is that if you divide 30 by 5 you get 6 clients.
Note: In this example this would mean that each client would need to be drawing a constant value of 5Mbps.
Now all you need to do is find out what the max TCP value is of a r500 when uising a 20MHz channel using a single stream - you can do this by conducting a FTP or iperf test.
Note: you will need to use more than 1 client to consume all of the available bandwidth and graph the network interface of the FTP/iperf server (I would use a linux machine for this as its more accurate)
...ignore the 5Ghz max theoretical bandwidth (since you're specifying 20Mhz channelization) and substitute the 2.4Ghz figure of 300Mbps, divide by the 5mbps you specified, and arrive at a figure of 60 clients.
I would never expect to get close to "theoretical and optimal" performance, but since I'd configure my APs to support 40Mhz channelization (or more w/ ac support) and I'd expect some number of clients to support higher channelization and netflix should provide enough caching to smooth over momentary congestion issues... Well... You're still not going to get close to that theoretical number.
BTW: WiFi might not be your only bottleneck. Do you have 300Mbps of internet connectivity?
Here are independent throughput results found on the Ruckus website: http://www.ruckuswireless.com/carnet-performance-testing If I'm reading it right, the effective throughput of an R500 (with 60 clients pulling traffic) drops to 80Mbps, leaving something better than 1Mbps per client. If we use 80Mbps as our bandwidth cap, you get 16 netflix clients. (at 5Mbps each) ...but the "real world" nature of those results were disputed by other vendors.
So, compare the advertized performance numbers of a 7982 to those of the R500, scale the results appropriately and... you might get 10 or 12 simultaneous video clients per band. I'm guessing, 'cause I didn't do the math there. Plus, we totally lost track of your 20Mhz channelization specification.
The tests that Ruckus is highlighting seem to indicate that other brands would do worse.
..but feel free to search for your own throughput stress tests or conduct your own.
Quite honestly, I wouldn't prematurely go 20MHz, unless your deployment is really so dense that you've got one AP on each channel from 36 to 161 all within earshot of each other and still need more capacity.
If you go from 20 to 40, you can push a higher bandwidth to each client, which may make them happier in terms of being able to achieve a higher bitrate/throughput at a greater distance, and that would hopefully get them off the channel faster.
I've found that in moderate density, 40MHz is still quite appropriate and even leads to more satisfied users, since they see bigger numbers when running a speed test, and you still get good air efficiency because clients get off the channel faster.