Perhaps show us some scope screenshots and describe how you measured it.
Sure (see attachment). This was measured with an ordinary 10x scope probe between positive and negative outputs. Note that the average voltage is exactly correct (4.0V), open circuit or under a modest load (50 ohms). But if the current limit is tripped, the oscillation goes away and the output is perfect.
Oh and does the fast/slow switch on the back do anything? Its supposed to be used to keep it stable with long cables on slow.
Funny you should mention that! On all six of my units, there is no switch, only a hole in the chassis covered with a silver sticker, and wire jumpers to fix the unit in the "normal" (slow) position. The stickers are colored and sized perfectly; it appears to be a professional job, or perhaps manufactured that way? In any case, per the user manual p. 27 ("Fast/Normal Operation"), this setting should have no effect for open circuit or resistive loads. The only issue would be if capacitance were added in the "fast" mode.
I checked the voltages on J307 (connector between the analog and digital boards), and thought I'd found a smoking gun when OV_SCR, the crowbar control, was 0.64V instead of the stipulated 5.0V. A check of the schematic, however, reveals that OV_SCR can never be higher than a diode drop above ground. I keep finding these small mistakes in the HP documentation, which unfortunately has the effect of making me doubt all of my earlier findings
FWIW, here's a photo of my analog PCB. There are definitely some variances with the service-manual schematic. For instance, U313 (in the current monitor amp) is an AD712 instead of an LM358 as shown; its gain is 2 instead of 1 as shown, due to R427 being a 10k resistor instead of a jumper; and the connection to R427 is made with a bodge wire. And these observations are just from one small area, since the wire caught my attention. Perhaps I have an early revision of the PCB?
Btw.: this tantalum cap error is present in the 6632B only, afaik.
Actually, on my 66332A the error is still there. I haven't changed anything yet because the bias voltages are all within spec, but if I did, what type of capacitor would you recommend? The SiBaTit ones you used are no longer available.