187-207  8550 -R21 Identical to -021 except it has a 50-key function keyboard intended for use primarily in the teller workstations of financial institutions.  This keyboard is commonly used for basic teller applications which usually deal with limited, mostly numeric, input and output.  It is not intended for use with general purpose applications.

The 50-021 / -031 lacks the ESDI part of the BIOS.

Peter Wendt scrawled:
   If you pick the 50Z BIOS chips (2 x 512) and split-burn them down into 4 x 256 (as the 50-0xx uses them) you can convert a 50-0xx into an "almost 50Z" without needing the "50-to-50Z upgrade controller", which is a rare item. It only adds the ESDI code to the BIOS and needs a different @DF9F.ADF, which allowes to set an adress for the BIOS ROM on the controller.
   Did that a decade ago and it seemed to work fine with a 50-021 and a 120MB HD from a Mod. 70.

>Couldn't you also replace bits of the BIOS in software?  Get the BIOS copied into RAM as is sometimes done for performance, then overwrite bits of it.  I dunno if anyone has actually done this, but it sounds plausible.

   The 286s don't copy the ROM into RAM after POST like the 386s do - and even if: you should do that *before* POST to have the ESDI controller available for the boot process .... but I don't think that was what you *meant*.
   [Rem: The "upgrade controller" in fact *does* something similar: it copies some routines into an appropriate area in the "adapter ROM area" (C000 - DFFF) and sets the bootloader entry point to the start of that routine. Usual "check adapter segments for bootable controller ROMs" stuff.]
   Using "Split Copy" on an Eprommer is much better. You pick the "even" of the two 27C512 (64K x 8 bit) Eproms, read and store 0000-7FFF in the first and 8000-FFFF in the second process. 
   Then you pick the "odd" Eprom and repeat the step. The four files nicely burn into 4 x 27C256 (32K x 8 bit) Eproms and you have Odd-Low, Odd-High, Even-Low, Even-High.
   It's *a little* complicated - but not much. (Even I managed ...)  ;-) I think I still have the codes in my data cemetry somewhere.  In the 50-021-format.

The advanced Do-It-Yourselfers *may* pass this obstacle with this method:

- copy the 50Z-BIOS from 2 x 27C512 into 4 x 27C256 (split the "ODD" and "EVEN" eprom into ODD-Low and ODD-high, EVEN-low and EVEN-high. 0000h - 3FFFh = low, 4000h - 7FFFh = high at each set) This requires an Eprommer !
- install the new BIOS chipset on the 50-021 planar
- use the bare "Hardfile Riser card" 90X9441 from the 50Z
- use any of the IBM ESDI drives (30-160MB)
- reconfigure the machine. "1 ESDI harddisks" must be displayed in the
device-list on advanced diagnostic (unmodified 50-021 show 0 ESDI drives here on that attempt ...)

The 50-021 uses a proprietary MFM / ST-506 controller along with a CMOS version of a MFM 20MB harddisk - and that's the only drive supported on that controller.