I am progressing on a medium speed oscilloscope project (targeting ability to show 10~20 MHz signals). On analog front end there is a lot of signal control to cater for
- Different Signal Voltage (compensated step attenuator)
- AD/DC coupling
- Op Amp Gain Control
- Bias Control
The research I have done to date has pointed me to using reed relays to control where the signal goes given the settings input. While there is quite a bit of positive to using reed relays (low cost, no real forward voltage, can easily handle signals to 100 MHz, easy to find components to 100V and above), it looks to my eye a bit clunky with about 20 relays needed for only 2 channels.

I could reduce the count a bit by using an e-pot on some amplifiers but the tolerances look frightening and many seem to be designed for audio and higher frequency signals is an issue for many.

I believe there are GaS options but they seem to be quite pricy where compared to the very affordable reed relays. Generally analog multiplexers have a significant amount of forward voltage that would be a problem in getting the reading correct. VGAs could help some but that would still leave a lot of reed relays.

My question is if there is another good solution that I am overlooking?
Thanks

 

Welcome to AAC.

I think your attempts at avoiding relays is misguided. Relays are standard practice for this application because anything else is going to create artifacts.

This TI reference design might be a little helpful for perspective. The relays used in it are just what would be expected Teledyne MCR sealed reed relays. Anyone who has dug into scopes and other instruments of high quality will find the silver cans and logo very familiar—for a reason.

​

So aside from an unease with what seems “a bit clunky” do you have some other reason for avoiding them? Do you have any examples in a similar application where they are not used? You very articulately describe the advantages of reed relays but your objections lack a similar clarity.

It isn’t my intention to be harsh in this response, and I apologize if I seem so. It’s just that you answered your own question and the confirmation that your answer is correct can be seen everywhere you look in the test instrument domain, so I think you should take your own advice.

One more thing: think about all of the compensatory circuitry and special case accommodation a silicon solution would entail. Have you ever used an instrument that didn’t click at range changes? I can’t recall one.

 

Thanks Ya’akov
No harshness taken. Valid question. Relays do take up a bit of footprint, though that is hardly a deal breaker on this application. If there was a magic AMUX that had no forward voltage, nicely linear and could handle the bandwidth, that might have been interesting, but it does seem to not be the case. The low capacitance and no forward voltage make them the simple choice. Mainly checking to see if I was missing something. I actually had already used them on the schematic but when you see about 10 per channel it made me wonder if there was a better way. My thinking was better to check before starting routing.
Thanks for confirming that reed relays are the likely optimal choice. I had read where I believe it was Tek had actually a cam and leaf arrangement directly on the board to minimize capacitance.

 

In non-instrument applications there are many mux options of course, but when signal fidelity is critical, a piece of wire (with a controllable gap) beats a piece of silicon that acts like a resistor any day.

 

Some scopes just use a multipole rotary switch to set the gain for each channel. At least it was done that way back when scopes used tubes for the circuits. So that is another option. But it can be bulky as well.

 

The book "The Art and Science of Analog Circuit Design" by Jim Williams has a chapter 7:
"Signal Conditioning in Oscilloscopes and the Spirit of Invention" by Steve Roach
which discusses the requirements of an oscilloscope input circuitry and a possible
solid state design. No relays... This is a "concept" and it's not clear how far it was taken
to implemetation. It's interesting... (Oh, target is DC to 500 Mhz)