HP 3325A repair progress/questions. Is it a good idea to socket high-speed ICs?

[mrnuke]

TL;DR: Is it okay to socket an IC which amplifies signals up in the 20-30 MHz range?

I'm working on the repair of a 3325A function generator, which I obtained as "does not power on".

Imagine my surprise when I was probing around, and all of a sudden the thing comes to life. It turns out that several mounting screws were missing on the PSU assembly.  One of the screws is needed  to complete the ground path through the chassis. Another issue with the power supply was the leads on one of the pass transistors had broken off. That's fixed now.

After that was said and done, I noticed the unit did not produce and output signal. Not only that, it would happily display the following errors:

• FAIL 1
• FAIL 2
• FAIL 3
• ACAL FAIL

At this point, I'm freaking out. There's got to be a lot of stuff broken, and I'll probably never be able to get this fixed. So I follow through the service manual, get to the function circuits (section J) troubleshooting, and there's no input signal into the circuit. At that point I'm stuck, and I start following any troubleshooting procedure that didn't look like a complete wild goose chase. Every procedure says the respective circuit is good.

I finally get to the frequency mixer, and after a week of probing around there, I track the problem to the output amplifier IC: signal goes in, signal doesn't come out. It's one of those dreaded CA3000 transistor arrays, doped with Unobtainium. I pull the IC off the board, and much to my chagrin, the BC junction of one of the transistors is open.

TL:DR: When I do these sort of repairs, I like to put a socket instead of soldering the new IC directly. It saves the PCB from more thermal shocks if the component ever needs to be replaced again. In this case though, the signal has harmonics in the 30MHz range, and I'm not sure how much those signals like IC sockets. I'd really like to put in a socket, but I'm afraid that it might affect the performance of the instrument. Do you think it would be safe to use a socket in this case?

Schematic of the section attached below. Failing IC is U15. It's a CA3127E.

[edavid]

TL;DR: Is it okay to socket and IC which amplifies signals up in the 20-30 MHz range?

Yes, it's fine.

[mrnuke]

I was able to find some replacement ICs at a local surplus store. I'm still shocked that I was able to find exactly a CA3127E array. I decided to socket the IC for good measure.

[mrnuke]

And of course, the thing works! And it's within spec. I was afraid the thing might need re-calibration, but the amplifier I replaced is part of a much larger feedback loop.

One more question though. is 82W consumption at the outlet normal for the 3325A? On a 110V network here.

[Vgkid]

Your power draw seems normal. Specced as 90w with all options.

[David Hess]

I have used sockets for replaced ICs up to 100 MHz without problems however better than using a socket is to insert separate machined collet socket pins into the printed circuit board to make a low profile socket.  This also allows access to the area that the body of a socket would cover.

[krivx]

I'm sure there are cases where it's important but generally I don't worry too much about lead inductance for ICs that are already in DIP.

[David Hess]

I'm sure there are cases where it's important but generally I don't worry too much about lead inductance for ICs that are already in DIP.

It can be a problem with ECL and some high frequency analog ICs.  Old Tektronix products which used a mixture of TTL and ECL logic usually or always placed the ECL DIPs in Berg Miniserts which are the ultimate in low profile sockets.  Tektronix had their own through-hole IC package with leads on 4 sides which was good up to 500 MHz using the same low profile socket arrangement.

I like using collet socket pins for repairs because they look so clean.

[mrnuke]

I like using collet socket pins for repairs because they look so clean.

Damn! That looks amazing! Thanks for the advice!

[David Hess]

That photograph is of the power supply section of an ex-NASA Tektronix DC505 225 MHz universal counter/timer.  The black plastic IC sockets shown there are the infamously unreliable edge wipe sockets made by Texas Instruments.  The two I replaced had hot running ceramic DIP 723s which caused the sockets to actually get loose barely holding the ICs in place.  The 723 regulator is slow so the low parasitics do not matter but the extra effort of cutting the socket pins out and soldering them in place is not a big deal and as I said, I like the look.

The following photograph shows where I added a socket where there was not one previously.  The IC shown is a 100 MHz (actually 200/250 MHz) analog channel switch which *is* prone to oscillation.  I used the entire collet socket instead of cutting the pins out and it worked fine but if I did this repair today, I would cut the pins out and install them individually for better performance.

The cheapest way to get these machined collet pins is to buy SIPs and cut them out as needed with diagonal cutters.  The outside diameter of the collet will usually fit perfectly inside standard plated through holes if cleaned of excess solder.