Measure fast differential signals (100MHz) using instrumentation amp

[levnu]

Hi
i'm trying to measure w/ Scope a fast differential signals (100MHz, 0.3V, Vcm = 1V) without loading them by the scope input - means
that when i connect to a scope and set it to 50ohm i loss 6dB and when i set to 1 MOhm input i get impedance mismatch and its bad

i thought about next solution do you think its possible under such conditions?


So now I'm looking for Instrumentation Amplifier w/ next specification (and it seems impossible to find):
- Can operate w/ signals up to 500MHz
- Small signal AC is up to 0.5V @ 100MHz
- the small signal is riding on a DC V common mode of about 1V
- The required Gain is about A=2 (linear)
- Differential input single ended output (as standard for In-amps)

- is it possible to find such In-amp? if not what can i do? how can i reject the V common-mode and extract the ac
w/o using any capacitors or things like that...\

- Maybe some transformer?

 

[Borber]

Two channel oscilloscopes usually can measure differentially by using two probes and one channel inverted.

 

[levnu]

i attached my setup schema does your answer solves that?


- - - Updated - - -

i need to measure differential signals from DC to 500MHz

 

[Borber]

Measure with two 10:1 probes one for each side of the component and do the math on oscilloscope subtracting channels like ch1-ch2 and this is high impedance differential measurement without any adapter.

 

[levnu]

Thanks
- I need to think about your advise before i can assure it suits me...

- On the other hand ...
Do you think it is possible that i'll build by my self a wide-band (DC-500 MHz) instrumentation amplifier by using 3 wide-band op amps?

 

[FvM]

It's definitely possible. A different question is if you are able to make it with acceptable performance, because it depends on your high speed analog design skills.

 

[crutschow]

You should be able to measure that directly with the scope, as suggested. What scope are you using?

 

[levnu]

You might be right i need to verify it ... don't you think that 10:1 probe will make me have issues w/ the matching to 50 output?
because today i'm measuring via SMA connector than coax cable and SMA input to the scope ... i'm not sure if 10:1 probe will have same results for freqs like 500MHz
my scope is Agilent DSO-X-3034A

 

[crutschow]

The 10:1 probe should have a high enough impedance that it won't significantly disturb the signal from a 50 ohm source. But you can't have any significant length (more than a couple inches) of unterminated coax connected to the source or you will get reflections.

 

[FvM]

The 10:1 probe should have a high enough impedance that it won't significantly disturb the signal from a 50 ohm source. But you can't have any significant length (more than a couple inches) of unterminated coax connected to the source or you will get reflections.

A standard 10:1 passive probe has an input capacitance of 10 to 15 pF, resulting in a 200 to 300 MHz low-pass corner with a pure resistive 50 ohm source. So "won't significantly disturb" isn't right if you want e.g. 500 MHz bandwidth. The real problem is however probe connection inductance. Any connection method where the probe tip isn't coaxially plugged to a 50 ohm connector involves an considerable amount of inductance and ruins the signal quality. Who ever tried to monitor a fast edge digital signal with a passive probe knows about the problem.

Common mode rejection is another thing. The suggested dual probe method works fairly for low and medium speed signals, but has no satisfying results in the 100 MHz range.

In so far I can well understand why levnu wants a differential probe amplifier. I'm am often using a single ended active probe for critical measurements and occasionally a differential one and don't want to miss it.

For low source impedance, you can also use a 10:1 or 20:1 resistive probe, a 450 respectively 950 ohm series resistor at the end of a 50 ohm cable. It can achieve several GHz bandwidth and costs about nothing when soldered in place.

 

[crutschow]

A standard 10:1 passive probe has an input capacitance of 10 to 15 pF, resulting in a 200 to 300 MHz low-pass corner with a pure resistive 50 ohm source. So "won't significantly disturb" isn't right if you want e.g. 500 MHz bandwidth. ...............................

Common mode rejection is another thing. The suggested dual probe method works fairly for low and medium speed signals, but has no satisfying results in the 100 MHz range.

.........................................
For low source impedance, you can also use a 10:1 or 20:1 resistive probe, a 450 respectively 950 ohm series resistor at the end of a 50 ohm cable. It can achieve several GHz bandwidth and costs about nothing when soldered in place.

Yes, I made an incorrect assumption about the probe impedance at 500MHz.

I can see common-mode rejection is important for a signal for with common-mode noise, but will that be a problem for the OP's signal with a constant 1V common-mode voltage?

I suppose it is obvious but it should be emphasized that when using the 50Ω cable and the 10X or 20X (-50Ω) series resistance, the cable must be terminated in 50Ω at the scope input. It should also be noted that this will result in a 10% (with 10X) or 5% (with 20X) amplitude loss from a 50Ω source.

 

[FvM]

I can see common-mode rejection is important for a signal for with common-mode noise, but will that be a problem for the OP's signal with a constant 1V common-mode voltage?

I suppose it is obvious but it should be emphasized that when using the 50Ω cable and the 10X or 20X (-50Ω) series resistance, the cable must be terminated in 50Ω at the scope input. It should also be noted that this will result in a 10% (with 10X) or 5% (with 20X) amplitude loss from a 50Ω source.

I presume, DC common mode won't require differential measurement at all, just a trace offset or AC mode. I was expecting high frequent CM signals, but the OP didn't tell exactly.

I suppose it is obvious but it should be emphasized that when using the 50Ω cable and the 10X or 20X (-50Ω) series resistance, the cable must be terminated in 50Ω at the scope input. It should also be noted that this will result in a 10% (with 10X) or 5% (with 20X) amplitude loss from a 50Ω source.

Resistive probes require a 50 ohm termination at the oscillsocope, it's a built-in standard option for oscilloscopes > 300 MHz anyway. The signal attenuation is of course the same as with a high impedance passive probe with respective divider factor.

 

[levnu]

i would use a standard differential prob but its too expansive and i need to have few such measurement points
Do you think i can design a simple Instrumentation amplifier by using 3 wide band op amps for e.g. from TI?

 

[FvM]

Do you think i can design a simple Instrumentation amplifier by using 3 wide band op amps for e.g. from TI?

As previousl mentioned, how should we know if you can design it? It's at least worth a try.

I would probably prefer buffers followed by a true differential amplifier stage because it involves less stability issues (no cross-coupled amplifiers). But the sketched instrumentation amplifier can work too.

If you are satisfied with a bit below 500 MHz, I suggest THS3001 which has excellent large signal bandwidth and time domain signal quality, and is relative sweet-tempered. You should become familiar with the specifics of current mode OPs to use it correctly. Voltage mode OPs have always more overshoot and less slew rate.

 

[levnu]

Thanks,
i'll try simulate that and than create a PCB board that will include THS3001 or similar one (i would prefer not to be a bit low than 500M but i'll start from this point you suggested)

 

[levnu]

Here is how i simulated it looks what as i desire it except that @ 200M the peak amp falls from 0.7 to 0.5 and i want to have a BW of at least 500M
do you think i can find such CFA? maybe even some overkill?

 

[FvM]

The Rf values are probably not yet set optimally, that's related to what I previously mentioned about becoming familliar with CFA circuit design. You should also take it as fact that the simulation models only represent part of the real part behavior.

You can use faster OPs, e.g. THS3201. If it works in a real circuit, depends on the layout etc.

 

[levnu]

Thanks - This exactly what i have tried see below


I think that the 500MHz results are almost ok i wish i could find a bit better solution but still its seems to be quite good isn't it?

The Rf values ... you mean all the resistors values around that need to be find tune?
Is there something i can do w/ the phase shift?

 

[FvM]

Rf is setting the loop gain of CFA. It should be set according to datasheet recommendations in a first order. Rf of the input stage is too high now, Rf of the second stage too low.

But I don't believe that you can finally design the circuit in a simulation. You should start with real hardware and gain experience.

 

[levnu]

Thanks that's what i'll do