Welding glass is designed to reduce UV, IR and visible light by significant amounts and is available very cheaply in a range of different shades. 4”x3” sections of the standard types are available for around £2-3 (from welding suppliers & e-bay...), these should fit in a Cokin P holder. I've only seen shades 8 and above, but I've never looked for the lower numbers.

Optical quality is NOT controlled but is usually reasonable, if you need a guarantee of optical quality '10 stop' filters can cost ~£100, but cheaper vaieties are becoming available.

Visiblly these filters provide a slightly greater cut in red and blue, leaving a significant green tinge (conveniently the blockband of most FL-D filters corrects for this reasonably well).
‘Gold’ variants of some shades exist that are more neutral in colour but more expensive ~£20. Even with these colour correction will be required.

In use, the glass can be just held close in front of the lens, or fitted to a cokin mount, but a screw mount is preferable (especially for long exposures!) Taping/gluing the right size Cokin adapter to the welding glass, is easiest. Note ANY light leaks will be very visible, especially if it reflects off the back of the filter!

Nominally, the shade number is defined as ‘one plus 7/3 the absorbance’, and absorbance is the negative log10 of fractional transmission, enabling the light reducing power to be calculated.

This gives:
Shade:_____EV/ stops_____exposure multiplier
2:_____________1.4_____________2.6
3:_____________2.8_____________7
4:_____________4.3____________20
5:_____________5.7____________52
6:_____________7.1___________140
7:_____________8.5___________360
8*:___________10___________1 024
9*:___________11.4__________2 700
10*:__________12.8__________7 100
11:___________14.2_________19 000
12:___________15.7_________53 000
13**:__________17.1________140 000
14**:__________18.5________370 000
*- I have a standard one of these ** - I have a ‘Gold’ one of these.

This is complicated by the fact that the light source used to test welding shades is spectrally different from daylight, and the filters make no pretension of neutrality.

For day-lit scenes, I've found that the shade number to overrate the filters, (others tell me it’s by about a stop). Apparently they are pretty accurate for tungsten lighting, but I've not found them so useful in controlable dim light.

I've sucessfully used a shade 13 for photographing sunspots (unlike polarisers & standard ND filters it cuts the UV/IR to safe levels), though the image isn't even near the standards of NASAs SOHO. I've also used my shade 8 for more normal '10 stop' type shots. In a recent workshop session they've been used for photographing the filament of a spotlight.