Operator ASCII art

V, 78, 72, 71, 68, 65, 63, 62, 61 bytes

Ç=ü-/Àé X
ç^Ó/é Àä$
ç+/ÀÄM|ÀR+ 
ç=/Ä
ç¯/lòhYpX
çx/rxòl3Äjxlrx

Try it online!

As always, the neck-and-neck battle with 05AB1E is really fun!

Since this contains non-ASCII characters, here is a hexdump:

0000000: c73d fc2d 2fc0 e920 580a e75e d32f e920  .=.-/.. X..^./. 
0000010: c0e4 240a e72b 2fc0 c44d 7cc0 522b 200a  ..$..+/..M|.R+ .
0000020: e73d 2fc4 0ae7 af2f 6cf2 6859 7058 0ae7  .=/..../l.hYpX..
0000030: 782f 7278 f26c 33c4 6a78 6c72 78         x/rx.l3.jxlrx

This does create leading spaces in the output for = and -, but this seems to be allowed. If this is not allowed, feel free to comment and I'll roll it back.

Explanation

The "global command" (e.g. ç) applies a certain set of commands to every line that matches a certain regex. The syntax is

ç<compressed regex>/<commands>

This is the easiest way of simulating a conditional/switch statement. In my original answer, I simply created the entire ASCII-art on the right-hand side for each different character we need to search for. However, a lot of these outputs require similar commands. So I combined them. The first command ('Ç') is actually the inverse of the global command, it applies the command to every line not matching the regex. So the first command is:

Ç=ü-        " On every line not containing an '=' or an '-' (e.g. inputs '/', '+', and 'x'):
    /Àé     "   Insert *n* spaces
        X   "   Delete one of them

The following command is for inputs '=' and '-'. These two are conveniently easy and similar. After this command, we need no more processing for -.

ç^Ó         " On every line that starts with non-whitespace (e.g. every line not affected by our previous command):
   /é       "   Insert one space
            "   Move back a character
      À     "   Make *n* copies
       ä$   "   Of every character on this line

From here we just do some extra commands for each individual possible input. For +:

ç+/         " On every line containing a '+':
   ÀÄ       "   Make *n* copies of this line
     M|     "   Move to the first character of the middle line
       À    "   *n* times:
        R+  "   Replace the next two characters with '+ '

The command for equals is very straightforward. We just duplicate it with Ä. For /:

ç¯          " On every line containing a '/':
  /l        "   Move one character to the right
    ò       "   Recursively:
     h      "     Move one character to the left
      Yp    "     Make a copy of this line
        X   "     Delete one character
         ò  "   End loop (implicit)

The last one is the most complicated. It is basically a port of this answer.

çx              " On every line containing a 'x':
  /rx           "   Replace the first character with an 'x'
     ò          "   Recursively:
      l         "     Move one char to the right
       3Ä       "     Make 3 copies of this line
         j      "     Move down one line
          x     "     Delete one char
           l    "     Move one char to the right
            rx  "     Replace this char with an 'x'

05AB1E, 81 76 74 73 70 69 68 65 64 62 60 59 57 56 bytes

Currently in war with the V answer. I'm coming for you Dr. McMoylex :p.

Also in war with the Pip answer. I'll be watching you Mr. DLosc.

---

Code:

Ç6&"¹s<ú.s.Bívy¹'xQiÂðñ}, ¹×S)»¹'=Qƒ= ;ƒ¹})D¦»»Rû.c"#è.V

Or in a more readable form:

  Ç6&
"¹s<ú.s.Bívy¹'xQiÂðñ},
 ¹×S)»¹'=Qƒ=
 ;ƒ¹})D¦»»Rû.c"
#è.V

Uses the CP-1252 encoding. Try it online!

Python 3, 304 283 278 bytes

Simple enough, just makes a matrix of chars and applies the different operations based on which one it is. The = and - have trailing spaces if that's not too bad.

EDIT: Thanks to @Shebang and @Rod for their suggestions which ended up saving 21 bytes!

EDIT2: Thanks to @Artyer for saving 5 bytes!

t,s=input().split()
s=int(s)
r=range(s)
b=[[' ']*s for x in r]
exec(['for x in r:b[s//2][x]=b[x][s//2]=t','b=[t*s]'+'*2'*(t=='='),'for x in r:b[x][s-x-1]='+'b[x][x]='*(t=='x')+'t'][(t>'+')+(t in'x/')])
if t in'-=+':b=[[x+' 'for x in l]for l in b]
print(*map(''.join,b),sep='\n')