pretty stange board

It has five busses that DUT pins and instruments can be connected to, which should worn. I have minions check for compatibility of the test board and various products, to make sure I haven't missed anything.

Most of our products have multiple channels, from 4 to 64, and we can test one channel at a time, so the 5-bus thing will usually work.

A "module" of four DPDT relays makes an 8:1 mux, which would allow 7 busses and open, but I don't think I need that.

A proper full crossbar might need thousands of relays.

What beautiful irony!

A question posed by a complete dimwit whose IQ is pegged at zero.

Right. I was imagining a bed of nails tester with a ton of pins.

NT

On Thu, 01 Oct 2015 09:55:03 +0100, Pomegranate Bastard Gave us:

Leave it to a dumbfuck like you to not know what was asked about.

clock. That is easy to generate by upping the frequency and picking your edge.

On Thu, 1 Oct 2015 09:17:49 -0400, rickman Gave us:

With the shit you have been spewing, you have no edge.

ADI makes a chip that uses digital logic to make the phase shift, just two flipflops, which is easy and divides the LO by 2. But this

seems to do the phase shift internally without dividing down, from 30 MHz to 1.4 GHz.

I haven't studied the data sheet extensively (maybe some RF jock can do that) but that phase shifter sounds impressive.

and they record the operating cycles.....though these can be rolled back in SW. It's curious to see reports of adjacent relays with an order of magnitude (or two) difference in operating use.

I've had two 40ch that bricked - enough to make the 3497x do a flash dance, when dead 40ch was inserted. It looked like complete loss of firmware here, based on prel examination. Maintenace on these plug-in modules is a reconditioned unit swap, vs repair at Agilent, with no failure analysis provided. Not completely comfortable with this.

RL

I don't see anything in the data sheet that describes it other than, "a

up, but a digital divide would not be good for an analog mixer.

If the relays are unreliable, they should plug into sockets.

All relays have a predicted life cycle, and these are rated based on the contact breaking and mating power level - to be accounted for by the end user.

The fact that the relays tested basically functional, as did all the other simple parts on the pluggins - and that it was the control IC seeming to have 'lost' its firmware - that aggravated me. I would like to have been able to fix this, figure out how to make the module more bullet-proof, or to get at least some idea of how to avoid it in the future.

As is, we just keep on banging away, hoping it doesn't happen again, or that later revs of the combined hardware and firmware don't show the same behaviour. Busted hardware is my nickel, and should not be part of Agilent's revenue stream.

RL

some reed relay are rated for 50e6/100e6 cycles at load/no load that's a few years at 1Hz

-Lasse

It applies a phase shift to the LO, so there's an unmodulated reference input; possibly a simple phase-shift oscillator is locked to the reference. If the quadrature LO internal signals were AGC controlled, that'd do it. Disturbingly, there are four(?) external trims applied...

I never use reeds, because they are *not* reliable.

Den torsdag den 1. oktober 2015 kl. 21.57.28 UTC+2 skrev John Larkin:

got any number for that?

-Lasse

No, but I've had a lot of reeds fail when they shouldn't.

And they are big, expensive, electrically noisy (twang in the mag field) power hogs and and have horrible thermal offsets. You can't put them too close together or the mag fields can make trouble.

I don't like reed relays.

The tiny Fujitsu DPDT surface-mount relay that we like is very reliable and costs 84 cents.

...and sockets are even less reliable.

Machined-pin sockets seem to be absolutely reliable. I've never seen one fail.

My clothes dryer needed a relay replaced, and the only part I could find, that would actually fit, was an aerospace-rated sealed unit which was a tenth the size of the original, had over twice the current rating (and more contacts, to boot). And, its gold-plated pins fitted into machined-pin sockets.

I suspect a spare PCB is cheaper than machine-pin (if you can find them) relays and sockets. Heck, ten replacement PCBs might be more economical.