EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: ezalys on April 16, 2022, 12:52:47 am
-
I have a piece of coax coming from a cryostat that carries a signal I want to capture, but I also want to avoid creating a ground loop. I ordinarily would break this ground loop using a transformer, but the signal needs to be coupled from 20 Hz up to 1 GHz. Is there such a thing as a 50 ohm matched instrumentation amplifier? The ideal electrical model for what I want is essentially a battery powered oscilloscope, but is there a way to provide high common mode rejection in an amplifier circuit?
-
You could use a battery powered amp, then go through some sort of isolation by splitting the low frequency and high frequency parts then recombining after the isolation barrier, phase and gain is very important near the crossover frequency. But more importantly what are the signal levels you are dealing with, and what level of either common mode voltage level or isolation are you looking for? And the amplifier is it only unity gain or do you need some amplification?
-
If you just Need It Done, you can get fiber optic oscilloscope probes. They expen$$$ve!
-
You might look around for an LMH6401 demo board if the common-mode voltage and other parameters aren't out of line.
nctnico sells a nice differential probe made with a similar chip (LMH5401) that handles DC-1 GHz easily.
-
Might it be easier to isolate the other connections going to the cryostat? Or use a scope that can be powered from an isolated supply.
-
AD8129 is the fastest I know, but why not just make something yourself from a differential pair made from fast transistors?
-
nctnico sells a nice differential probe made with a similar chip (LMH5401) that handles DC-1 GHz easily.
The common mode impedance to ground isn't exactly instrumentation amplifier level of course, it won't break a groundloop to the same extent.
You could make your own high frequency high input impedance instrumentation amplifier with two LMH6559 and a LMH3401/LMH5401.
-
I like the fast diff pair idea. I'm still a bit new to thinking about differential signal paths. I'd be using this to sense the shield and center conductor of a piece of coax. What's needed to convert a minimum of the common mode signal into the differential mode? I know these things are usually done using balanced twisted pair connections but I don't have that option. I have to sense the shield and center conductor.
-
Maybe you could use a differential (active) oscilloscope probe, measuring between the centre conductor & shield of your cable? You don't say if you are using an oscilloscope to capture your signal, but if you are it would be a quick & easy (if expensive) way to solve your ground loop problem.
What you need would be a wide bandwidth (GHz+) probe, which usually handle signals up to a few volts, not the lower bandwidth (<100MHz) high voltage probes that can handle mains.
-
I like the fast diff pair idea.
I was thinking something like BFP720FESD or BFP840FESD differential pair with couple hundred Ohm emitter resistors, 1k collector resistors and either a high Ohm resistor to a negative rail or active current sink, 100 Ohm base resistors. Still, in retrospect, GHz ... maybe optimistic.
The integrated developers can make much smaller differential pairs than you, two LMH6559 and a LMH3401/LMH5401 will likely have superior CMRR at a GHz even if the discrete solution works as an amplifier rather than an oscillator.
-
Why is the LMH6559 necessary? On first glance, it looks like I can use the LMH3401 to just sense the shield and center conductor, but I must be missing something.
-
Without the LMH6559 there's a relatively low impedance path from the coax shield to circuit ground.
Just because it's differential, doesn't mean it breaks the ground loop.
-
The LM3401/5401 datasheets are very coy about cmrr.
The LMH6881 may offer better performance with a quoted 20dB of CMRR@1GHz.[attachimg=1]
AD offer a similar product with more boastful specs. ADL5561 They offer an eval board for $100 and their giving the Gerbers for free.
[attachimg=2]
What witchcraft is going on here?
[attachimg=3]
-
LMH3401 has common mode input common mode output graph (though LMH5401 does not). Figure 42. I assume Sdc21 for the LMH6881 is the same thing.
That is specifically what you need to know, because common mode rejection just means the common mode is rejected from the differential output signal, not if you take a single ended output signal when using it as a differential to single ended converter.
-
I though a bit more and have questions…
1: do you need to pass signals all the way from 20 Hz - 1 GHz or are there ‘gaps’ in the required frequency range?
2: how much common mode voltage do you need to accept? At what frequencies? Do you need to provide electrical safety isolation?
3: what level of signal quality do you need at the far end? I know some experiments require full analog capture but others might be OK with pulse detection? Do you have some estimates of noise levels etc.?
4: have you looked at RF over fiber? Given you don’t need to pass DC this analog approach could be helpful. (Mind you, operation down to 20Hz could be tricky.
5: if you’re working in a team, a Principle Investigator might say “we need a really good measurement” but not give any specifics. Being able to ballpark some costs & timelines could help you get more concrete answers.
I heard a great expression recently: “desirements.” Requirements are things that will block you from finishing the job. Desirements are the things that turn out NOT to be requirements when someone learns they’ll up the cost by $XX,XXX.
-
I'm sure there's a reason for this vagueness, probaly because it doent much matter in their intended app.
Here's an app note from AD. This suggests that a couple of buffers driving a regular diff amp will work.
Your still at the mercy of intrinsic CMRR of the ouput opamp and resistor matching.
I'm sure strays will kill you stone dead unless you have the dark powers.
My uninformed guess is that you'd get better signal integrity using an ADL5561 or similar and send it differentially and a classic diff amp at the RX end.
This lets you use lower gain at the TX end and increase bandwith. The ADA4960 ds shows another way. 0.1pF!
-
No gaps. 1 volt of CM — no galvanic isolation. 10 nV/rtHz white noise, with 1/f knee between 10 and 100 KHz. Pulse detection might be interesting — I need full time traces for the low frequency component but the HF parts can be gated. 1 GHz is a “nice to have” but the hard requirement is 500 MHz