Balanced Attenuator Confusion

[Guille]

Hi everyone!


I've been testing an HP 4436A 600Ω balanced attenuator but there's something about the balanced connection that I just can't get right.

The setup described in the manual is simple: A signal generator is configured to 600Ω output and hooked to the attenuator HIGH and LOW input, the GROUND input goes to cable shield. The HIGH and LOW outputs are then connected to a multimeter to check how much voltage is being dropped, lastly 2 * 300Ω shunts are placed to match the high impedance of the voltmeter to the 600Ω output impedance of the attenuator.



The problem is I reproduced this setup but my readings are off. The attenuation is adjustable and will work allrightish from 0dB down to around -80dB but from -80dB to -120dB the voltage doesn't drop at all. This is my implementation of the setup (I swapped the DMM for a lockin amplifier with floating differential inputs because of the very high attenuations).



Initialy I suspected a hardware error but the attenuator is fine, no resistor has drifted out of tolerance, all switches make clean connections, everything is up to spec. To drive the point home the manual (attached) specifies the attenuator can be set up with an unbalanced connection so I checked and sure enough when configured as the image below it works flawlessly down to -120dB



Hooking the generator output to attenuator LOW (instead of HIGH like in the picture) will also work equally well so both T attenuators are working perfeclty. This has me thinking it's got to be user error, there has to be something about the balanced setup that I'm misunderstanding  .


Thank you so much.

[BillyO]

In a shielded balanced circuit the shield should never be used to carry signal.

Do not attach the shield to the differential amplifier.  Attach the high to one input and low to the other.  Leave the shield unconnected at the input to the differential amplifier.

BTW, I'm not saying that's your problem, but it's not the way to do things.

Also, what frequency are we talking about here?

What is the input impedance of the lock in amplifier?

What is the frequency response of the lock in amplifier?

There are possibly more questions...

[Guille]

Hi BiilyO

Do not attach the shield to the differential amplifier.  Attach the high to one input and low to the other.  Leave the shield unconnected at the input to the differential amplifier.

Thank you, will try this and report back.

Also, what frequency are we talking about here?

What is the input impedance of the lock in amplifier?

What is the frequency response of the lock in amplifier?

The signal generator is putting out 1Vrms sine wave at 1 kHz, the attenuator is rated from DC to 1 MHz and the lockin is rated from 1 mHz to 102.4 kHz, it's input impedance is 10 MΩ. Oh btw I suspected the shunt resistors poor tolerance being the problem but I got them painstakingly matched to 0.0010% tolerance and the problem still persisted...

There are possibly more questions...

Feel free to ask

[BillyO]

Feel free to ask

What lock in amplifier is it?

[Guille]

It's a Stanford Research SR830 (https://www.thinksrs.com/downloads/PDFs/Manuals/SR830m.pdf), it's been used recently by other colleagues and I trust it to be in working order. It's a dual lockin so there are no phase problems (i.e. output cannot suffer from phase missalignments). I made sure to let the readings stabilize for 6 * time constant to reach a 99% accuracy at each measurement. I also repeated the measurements on different days and they all agreed.

[BillyO]

Well, it seems t be up to the task.  Not sure what is going awry.  I guessing you are using the unattenuated frequency as the reference, right?

Maybe someone with more experience with these measurements can chime in.

[Guille]

Yep it has stumped me and the electronic engineer at work for the past week haha.

Yes that's correct, I'm using the TTL signal out of the Sync port in the generator to reference the lockin. I don't think it's the lockin though, the readings look perfect with the attenuator unbalanced.

Thank you for your time

[Axisofed]

Brainstorming here...

Often I suspect leakage being a problem, especially when trying to measure very low level signals.

Is everything after the attenuator isolated from power ground as the instructions suggest?  Could you be getting 1 khz hopping on your AC?  I could see this changing when you have a signal referenced to ground vs. not. 

Have you disconnected the signal from attenuator to amplifier and verified the measurement drops to nothing?

Try changing the frequency and see what happens, just for another data point.

[Guille]

Hey, thanks for your answer.

Have you disconnected the signal from attenuator to amplifier and verified the measurement drops to nothing?

I tried this today but the signal level indeed dropped to zero when switching off the source.

Try changing the frequency and see what happens, just for another data point.

Good point, this I did too, changed from 1 KHz to 10 KHz the other day and there was some dependency with frequency which suggest some parasitic reactance somewhere.

Today I'm thinking perhaps there is something wrong with this lockin's differential measurement because I get different readings when I swapp it's inputs: Input A to High with Input B to Low will measure 45 uVrms while Input B to High and Input A to Low will measure 25 uVrms. This is weird to me, both inputs are floating and AC coupled... They should measure the same voltage no matter their orientation.

To check if the differential input was indeed the problem I changed the lockin for a low noise preamp + DMM setup, the preamp floats on batteries and feeds into the DMM which has floating inputs too, just like the manual suggests. I measured the balanced signal this way and sure enough now the measurement is perfect, all the way down to -120dB...

[Vovk_Z]

To check if the differential input was indeed the problem I changed the lockin for a low noise preamp + DMM setup, the preamp floats on batteries and feeds into the DMM which has floating inputs too, just like the manual suggests. I measured the balanced signal this way and sure enough now the measurement is perfect, all the way down to -120dB...

There is some ground line noise, like common mode noise, or common ground wiring problem. 

[Guille]

Yep could be but I don't really see how when the inputs are floating from ground. The setup works with the preamp + DMM combo but it doesn't work with the lockin. Both setups use differential AC coupled floating inputs, the same cables, same terminator resistors, same attenuator... The only thing that changes is the lockin.