I recently purchased a Tabor 8025 waveform generator (~US$120 from eBay), designed circa 2000. It is in quite good condition (only issue is failing batteries), and the outputs are all within spec.
This model has a <=100 MS/s clock, generates output sine waves at up to 50 MHz, has 1 M-word of RAM, switchable 25 and 50 MHz filters, segmented ARBs, GPIB control, 10 MHz reference input, and a 4X20 character LCD.
The only disadvantage I've discovered of this generator is that its internal waveform cannot be used as an AM source. So, a second generator would be required to have an AM modulated output. On the plus side, the AM input has a bandwidth of 500 kHz (quite high).
I believe that this model was also sold as a LeCroy LW110. While mine is the base model, a 2 channel version was also sold (Tabor 8026/LeCroy LW120), and a 4M-word RAM version. It may also be related to the Tabor 3361 (which is also the Fluke/WaveTek/Datron 301 or 302).
This generator has a couple unique features that are not available on other generators that I've used. The most interesting is that the sample clock has a variable rate. This allows pulses of nearly any length to be created (at any length >= 10ns). Other instruments I've used have some large step size for their pulse width resolution. The output rise/fall time is about 4.5ns (approx 75 MHz output bandwidth) as measured with a 200 MHz oscilloscope. This unit also supports segmented ARBs, where multiple pattern segments can be programmed, and then played in some order. This allows repetitive signals to be created which are much longer than the 1 M-word memory would normally allow.
There is an auxiliary sampleclock output at the back of the unit, which is a fairly clean sine wave (I don't have equipment to analyze it), at up to 100 MHz (the maximum sample clock rate) but its amplitude cannot be adjusted. The standard output can generate a 50 MHz sine wave, though by default it is very distorted. The distortion is because the unit disables its filters when outputting between 40 and 50 MHz. A much cleaner signal can be created by outputting a square wave, and enabling the 50 MHz filter, although the output amplitude will be reduced.
The unit has two PCBs (front panel + main board), and a NEMIC-Lambda power supply module. There is a small case fan, which is somewhat loud. The main board contains a processor, 10 MHz reference, 1.5625 MHz clock (CPU clock???), a pair of NiCd batteries (which power the SRAM during shutdown), six Altera MAX CPLDs (memory controllers, etc), an Analog Devices AD9767 dual-channel TxDAC, and filters.
It's tempting to try to upgrade the unit to be dual-channel.There is a jumper on the board to select between 1 and 2 channel, and also between the 1 and 4 M-word RAMs. But, without full schematics, it'd be very difficult to determine the component values for all of the passive elements of the filters (L and C values). Also, I don't know if my firmware even supports two channels. The ROM chip has a label marking it as a 8025 unit, so it may not be able to be easily modified to become an 8026.
Upgrading the RAM would be too expensive for its value (for me). It would involve populating 24 SRAM ICs, at about $3 each.