TSP  

Could you please tell me what material you used for the moving contacts? I think it's some kind of berylium bronze.

@@user-jekish the naterial was beryllium, suitably heat treated. This was plated with gold. Hard gold on the fixed contacts and soft gold on the moving ones. With similar gold on both, contact became unreliable after a few weeks.

Thank you for your reply, Pure beryllium?

@@user-jekish This is a long time ago, but I think I remember it being beryllium-copper.

it's common to leave a DC block and 10dB attenuator permanently attached to the analyzer until you actually need the extra performance removing them brings. But I think this is only feasible at low frequencies, not for 50+GHz?

You can do this up to 110GHz - although working with 1mm connectors is always a pain.

Good idea, but unfortunately it would be useless. without precise trimming of the individual switching elements and without calibration, is this thing worthless😢

All those gates in the FPGA are the emergency supply of magic smoke; too bad it didn't get deployed in time to save that one pad.

The RPGA controls more than just six relays. For example, the correction values, the number of switching cycles of the individual stages and the serial number are saved.

@@DD1PW I mean it still sounds like something a 50 cent micro controller could do

It depends ​@@zyeborm, if this devices is used a lot inside other products, then the price won't be that high. Standardisation may be another reason to use certain components everywhere

There is also a chance the high power drivers for the relays are some propitery BS, forcing you to go FPGA to not emulate.

This is a ~20 year old design. Small FPGAs are relatively commonly used for simple switching logic, like blinking LEDs or actuating solenoids. More commonly so 20 years ago than now.

@@tommihommi1 Sounds more like a CPLD then.

@@vincei4252 It probably is. Could be a PLA or PLD as well. Xilinx did a lot of programmable logic chips and still do. If we were to know the part number the guessing game would be over. But I suspect that it isn't remotely as impressive as people's first reaction to hearing "FPGA".

They do this because for the BOM of this part the CPLD cost is irrelevant and they likely use this for many designs. I would not do this simply because each solenoid still needs a dedicated driver and therefore a high-current output shift register would require fewer parts and less things to fail. The simplicity of the interface being asynchronous may be an advantage for them across various products. Difficult to know without knowing how they have built the firmware of the FSU.

Daisy chain a few shift registers would be cheaper, for sure. And more repairable, better to debug. That's just how companies get lazy and lame as time goes on. And the reason why instruments from 1960s-1990swere mostly better engineered that what you get nowadays, with few exceptions.

it still sounds like overkill, what's the cheapest FPGA in that package? What's a mcu that'll do this job cost? 50 cents? a whole dollar?

@@zyeborm Yes, from a pure part cost perspective it might be expensive. From a BOM cost perspective it is perhaps one less item to keep on hand. Even if there is "a million" other options on the market for driving 6 solenoids. Personally would have gone with a 74128 address decoder and some NPN/N-fet transistors for power handling. Or just a suitably high output power capable shift register. (driving a bank of solenoids is not a niche application after all, so of course there is dedicated (almost jelly bean) chips for this field.) But perhaps this 100 pin QFN packaged "FPGA" allows them to parallel a bunch of the IO ports to handle the power. (if the solenoid is of suitable low voltage, since I don't expect the chip to handle more than 5V, if even that much... But it might be a high voltage IO capable one.) Back EMF is also a "non issue" for a totem-pole CMOS/fet output. (since it is low impedance and able to source the desired current when actively driven low.) So on paper they could save a lot of time on the pick and place machine, freeing it for other tasks. And this is quite valuable. But I also wouldn't be the slightest surprised if they actually did something more traditional, as in only use the logic chip for basic decoding, and handle driving of the solenoids with regular discrete transistors and back EMF diodes. In the end, I don't know. But the decision for an "FPGA"/programmable logic isn't that wild as one might first expect. (remember, Xilinx did make PGAs and PLAs too, as well as CPLDs.) Lastly, the concept of a 50 cent microcontroller is rather modern. Even in volume orders. Programmable gate array chips have existed for a long time and often been cheaper than microcontrollers. (especially 20-30 years ago.)

@@todayonthebench and this board was manufactured in 2004

@@tommihommi1 Yup, back when dirt cheap micros were less of a thing, programmable logic devices were common as mud and also often quite cheap. I am not surprised to see a Xilinx chip here.

Stop commenting with this AI crap please, it's such a waste of time and resources.

First time I saw this. Interesting

If thi makes you angry, you really need some damn perspective karen