A cheap 3GHz Agilent 8133A pulse generator for my home lab

I recently told you about my home lab, as well as the home labs of Barrie Gilbert, Darryl Philips, Alan Martin, and Jim Williams. Well, seeing all that lab equipment worked me up into a frenzy. I got back on eBay and checked for some of the equipment I still wanted. In that blog post about my lab I mentioned that I had an HP 8131A pulse generator, but I wanted the faster, PRBS (pseudo-random binary sequence) capable 3.5GHz Agilent 8133A. Well lo and behold, there was a used one for sale and it was right here in Silicon Valley. I think I got a great deal. The seller was honest; he said one of the channels was dead. He was also very nice, Zack let me come over and pick it up the same day and didn’t even charge me for handling. Check out the Cogwell test equipment eBay store. So for 737 bucks, I got a dual-channel pulse generator that works just as well as the single channel models you see for 5 and 10 thousand dollars. Maybe the other eBay bidders thought that this was a single channel differential model with one of the differential outputs broken. No, the differential pair in the middle works just fine, all the way to 3.5GHz. It is the differential pair on the right of the unit that is wacky. Best of all, that channel is not completely dead; it just has a negative offset. This tells me someone hammered the protection diodes which are now leaking, and that is why there is a fixed negative offset that does not respond to any offset command. With one working channel I should be able to probe the signal chains and find the bad component. Below is a photo essay of my firing up my Agilent 8133A. Oh yeah, that is another cool thing; this is new enough to be Agilent, so that is extra cool. Thanks Zack, thanks Cogwell. Click on any picture to enlarge.

.

Here it is, hooked up with some SMA to BNC adapters. The 8133A is a gorgeous piece of equipment. Jim Williams loved to build his own pulse generators, but that makes it pretty hard for someone to duplicate his setups. Now I can do the same tests that the application engineers at high-dollar analog companies do. And my results should match theirs.

.

First thing was to just plumb all 4 channels into my lunchbox Tek scope. It does not have 50-ohm inputs so I use 50-ohm pass-through BNC terminators on the inputs. The pulse generator is down around 40Mhz. You can see the top differential pair working great, the bottom are offset, note I had the lowest trace on “invert” so it looks like it is hovering above ground, but it really is offset just like trace 3. See how the bottom differential pair is at twice frequency? That is because the scope is aliasing, its not real, stay tuned.

.

This is why I love my 400MHz Tek 2465B. No aliasing on an analog scope. But look how slow and rounded the signals are. That is because……

.

Yeah, I had the BW/LIMIT bandwidth limit button on. With it off, the signals look as snappy as you would expect from a generator with a 60ps rise time.

.

So now its time to get bold and hook the 8133A pulse generator to the Tek 11801B sampling oscilloscope. The mainframe has a 50GHz capability, and the SD24 sampling head modules I have in this baby are good to 20GHz. Nice. If you don’t know what a sampling scope is, read the Tek app note that Jim Williams lent me and I scanned and sent to Tek to put on their website. This scope has a 150kHz sample rate. To understand why it is a 50Ghz scope, read that note. Notice the crap display. It took me almost an hour to remember how to use the most evil user interface in the test and measurement world. You are looking at some funky ghost images due to the trigger being a little off. That is one pain- you have to feed this scope a separate trigger, but the 8133A has just such a trigger output.

.

A touch of the trigger level, and the waveform looks right. This is why Paul Grohe over at Texas Instruments tells me he keeps a Tek 2465B on top of his 11801C. He says with an analog scope you can use beam-find and get the traces in a few seconds. Then you have horizontal and vertical values you can set in the 11801B so you can find the traces.

.

I run the good differential channel on the 8133A up to 3.5GHz.

.

The trusty 11801B is painting the repetitive waveform at 200ps/div. The waveform is choppy, but I used short lengths of semi-ridged to go to my remote test head, so that is probably pretty close to what is really coming out of the generator. Next step would be to use no cable at all, just a coupler to an output

.

Here is more ghost aliasing bad triggering crap showing on the 11801B. None of this is real. Never believe a digital scope. The autoset on this scope never seems to work. Maybe it is the trigger amplitude I am feeding it. Be careful you don’t overload the trigger or the inputs themselves.

.

A touch on the trigger and the square wave shows up.

.

Oh yeah, here is how I plumbed the generator to the 11801B. I think I will look for more of these extender cables for the test heads. This is the way I verified that the right hand differential pair was bad. I kept one input on the middle channel and put the same test head on the suspect channel.

.

Sure enough, the right hand differential pair output is bad, it has a negative offset I can’t take out. What is great is that it still looks fast, so this is probably a protection diode leaking from getting hammered by an over voltage. Grohe tells me 2465B scopes do the same thing- if you really over-voltage an input it makes a permanent offset.

.

Here I run the generator up to 2GHz.

.

The red channel is still broken, but it is fast, so that tells me I should be able to fix it pretty cheaply.

.

Here is the fantastic SD-24 sampling head on the 1-meter extender cable. Grohe taught me it is critical to keep the caps on the inputs. The sampling bridge diodes on the input can get zapped by static very easily. Now its time to get a nice set of SMA cables from Digi-Key. You can see I had to use a coupler in the background to make a cable long enough to get the trigger output up to the ‘scope.