A question on Purifi 1ET400A input impedance - Lars Risbo?

[pma]

Purifi do not publish much about electronic circuitry of their modules. Recently I needed to know input impedances without the buffer. The datasheet says:

however, according to my repeated measurements, it is evident that this information is wrong.

My test circuit and voltages measured, with additional resistors in series with input connector, looks like this and I also shoe measured ac voltages:

Resistor values R4, R7, R8, R9 I have calculated from measured voltages.

Simulation says this and is in conformance with measurements


For R10 and R11 shorted, I get this, both measurements and simulation shown:

So, my analysis says that input impedance of 1ET400A (without the buffer) is highly non-symmetrical and we can hardly speak about true "balanced" input even in case when signal source has moderate output impedance.

@Lars Risbo , willing to reveal more?

 

[Lars Risbo]

hi, i would also expect the input to be symmetric. I need to ask my colleagues from the ET team (I work on speakers and on vacation now ).

These node voltages are AC voltages - right? there is some common mode control.

cheers,

Lars

 

[pma]

Hi Lars, thank you. Yes, these node voltages are AC rms voltages. They are measured with high precision AC voltmeter. The Purifi model resistor network values may differ a bit from my calculation, but not too much. Input impedance at 1ET400a +input against common I measured as 3.738 kohm. Not the 2.2k as stated in the datasheet.

 

[BR52]

I'm interested in this as well, because my 1ET400A running in a no buffer mode as well.

 

[pma]

I think that somewhere was a recommendation by Purifi company for the signal source output impedance to be less or equal 50 ohm. The lower the better.

 

[BR52]

I think that somewhere was a recommendation by Purifi company for the signal source output impedance to be less or equal 50 ohm. The lower the better.

My driver (DAC) has an output impedance of ~100 Ohm, so if your observation is confirmed (seems for me is true) I'm thinking of a unity gain buffer.
With a symmetric load, the 100 Ohm to 2.2 k Ohm was more or less OK.

 

[Bruno Putzeys]

Use the below as the simulation model. Even though the input currents are unequal, the impedances are balanced. This is easy to understand, since adding equal source impedances will not affect cmrr (other than resistor matching precision obvs). This impedance balance is the only balance you should care about, not equality of input current.

Thus, one should never attempt to equalise input currents of a diff amp by scaling the inverting and noninverting legs of the resistor network. At that point, any additional source impedance would immediately unbalance the input and degrade cmrr massively. It wouldn't work anyhow since the currents would be if the input voltages are symmetrical with respect to ground, whereas the whole point of differential signalling is not to care where you are with respect to ground.

The line in the data sheet that says 2.2k to ground is only there to allow you to estimate max input current (and relates to the inverting input as you can see). Otherwise the input impedance must simply be treated as 4.4k, since using a source with 4.4k output impedance would result in exactly half the gain.

Please note that to be sure of a reply, it's best to contact us via our web site. Lars only happened to spot this thread.

 

[BR52]

Use the below as the simulation model. Even though the input currents are unequal, the impedances are balanced. This is easy to understand, since adding equal source impedances will not affect cmrr (other than resistor matching precision obvs). This impedance balance is the only balance you should care about, not equality of input current.

Thus, one should never attempt to equalise input currents of a diff amp by scaling the inverting and noninverting legs of the resistor network. At that point, any additional source impedance would immediately unbalance the input and degrade cmrr massively. It wouldn't work anyhow since the currents would be if the input voltages are symmetrical with respect to ground, whereas the whole point of differential signalling is not to care where you are with respect to ground.

View attachment 461564

The line in the data sheet that says 2.2k to ground is only there to allow you to estimate max input current (and relates to the inverting input as you can see). Otherwise the input impedance must simply be treated as 4.4k, since using a source with 4.4k output impedance would result in exactly half the gain.

Please note that to be sure of a reply, it's best to contact us via our web site. Lars only happened to spot this thread.

Thanks, Bruno, for Clarifying