Question on SSM2141

Hi Group,

I see some strange behavior in some SSM2141- wonder if some of you can shed light on this.

Link to 2141 data sheet: http://www.analog.com/media/en/technical-documentation/data-sheets/SSM2141.pdf

This circuit (see attached) is a minimalist balanced-to-balanced circuit, picks up a differential signal on input, outputs a servo-balanced replica, gain trimmed by the 2K "p17"

C61-62 are 2u2 polyester (they are here because the 2-3mV DC input offset of the IC annoys the previous stuff) and with the internal 25K resistors in the 2141 (from data sheet above) this suggests a lower Ft of ca 2.9Hz - and all is as expected at +Input, Pin3..

However, across C62, at the inverting input, Pin2, I get a -3dB point at ca. 15Hz, pointing to ca 4K8 Ohms load instead of the expected 25K..??

I have a feeling that the load/cutoff here relates to the absolute "symmetry" of the input differential - it seems to behave best at precisely-ground-symmetric signals, but this should surely not be a requirement for an input like this?

Have I mucked around with transformers for so long that I've finally lost my solid-state realism?

I've never used a 2141 but it looks like a pretty conventional non-inverting stage.

Is the + input 0V or receiving audio signal too?

With the + input at 0V the HPF pole should be the expected input R to virtual earth.

If the + input is getting signal too, the effective input impedance will be affected by the moving virtual earth - input.

The pole caused by the 100nF in series with + input Rs could cause a measureable different at 15Hz (but perhaps in the other direction).  A strong signal at the + input opposite polarity to the - input signal, will make the apparent input resistance seem lower as the virtual earth - input is swinging away from the - input signal.

I would not expect a 5:1 effective impedance change.  I do not see the servo you mentioned?


JR

I'm driving both + and - inputs, with a differential signal as expected, just not one that is nicely symmetric around ground/0V.

The 100nF's are only psu decouplings - only caps in audio are the two 2u2's at input.

(the servo I mention is at output stage, that part behaves well enough  )

Jakob E.

I guess the + input is augmenting the "virtual ground". If you ground pin 3 so that there's no + contribution then the op amp output will do what is necessary to make - also at ground. But if anti-phase signal is on pin 3 then the output will NOT do what is necessary to make - at ground. It will do what is necessary to make - match what is at +. Look at the - pin with your scope. Is it not always close to 0V? If not, then it's not really a virtual ground and thus the 25K resistor will look like it's less accordingly.

gyraf said:

I'm driving both + and - inputs, with a differential signal as expected, just not one that is nicely symmetric around ground/0V.

Click to expand...

DC symmetry won't affect what I am talking about but for the standard situation of a nominal sine wave signal at + input, and equal voltage opposite polarity signal at - input. The virtual earth - input is actually swinging  1/2 the voltage in the opposite direction as the minus input... this would make the 25k input R look like roughly 16k.

If the voltage at the + input is larger than the voltage at the - input the effective impedance drop will be larger.

Note that effective impedance drop is only for opposite polarity normal signals, CM signals moving in the same direction will see the  impedance at the - input increase, which is why the nominal 50k + input Z balances with the 25k - input (because of 1/2 CM signal swing) for good CMRR.

The 100nF's are only psu decouplings - only caps in audio are the two 2u2's at input.

Click to expand...

my bad,  on my display the node at 0V looks like a loop. 

JR

(the servo I mention is at output stage, that part behaves well enough  )

Jakob E.

Click to expand...

> Have I mucked around with transformers for so long that I've finally lost my solid-state realism?

Yes, you've lost it.

This type 4-R diff input has (as JR says) different impedances depending which input and single-ended or balanced or floating inputs.

The most usual way to make this negligible is to scale the impedances "very high" compared to sources. 25K is indeed >> 100r, our usual source or line effective impedance. If caps are needed, make them very-very large, 10uFd maybe.

Ah, you want film caps, and not soup-can size, so you will observe what you observe.

Or buffer it. A TL072 will do. Input Z can be 1meg, so 0.05ufd would be ample, and small. Maybe less total cost than the two 2uFd film caps. There's sexier JFET-in chips if you are ashamed to use '072.

> the IC annoys the previous stuff

If "previous stuff" is well defined, measure it all together and don't worry about oddities in connections that won't be used.

PRR said:

Or buffer it. A TL072 will do. Input Z can be 1meg, so 0.05ufd would be ample, and small. Maybe less total cost than the two 2uFd film caps. There's sexier JFET-in chips if you are ashamed to use '072.

> the IC annoys the previous stuff

Click to expand...

Or use a THAT 1206. So it's $6 instead of $4 but you get high Z and bootstrapped CMRR.

Thanks guys, things are clearer now.

Yes, you've lost it.

This type 4-R diff input has (as JR says) different impedances depending which input and single-ended or balanced or floating inputs.

Click to expand...

Thanks. This is exactly the point - somehow I painted myself into the corner of ASSuming the 0-Ohm virtual-ground-point at the inverting opamp inputwas behaving like a 0 Ohm. Where it in reality can get significantly into the negative impedances depending on where's the +input's at..

Or use a THAT 1206. So it's $6 instead of $4 but you get high Z and bootstrapped CMRR

Click to expand...

Hmm, yes. Only it's sadly not pin/function compatible with existing layout.

If "previous stuff" is well defined, measure it all together and don't worry about oddities in connections that won't be used.

Click to expand...

Point taken. That will be tried.

Thanks guys - really appreciated..

Jakob E.

gyraf said:

I'm driving both + and - inputs, with a differential signal as expected, just not one that is nicely symmetric around ground/0V.

The 100nF's are only psu decouplings - only caps in audio are the two 2u2's at input.

Click to expand...

What is the source impedance of each leg of the differential input signal?

That's the catch - source impedance vs. Frequency is rather undefined and different under different conditions. Transformer dosent care - but this mucks up in various ways under different conditions.

Jakob E.

I would second what PRR said... resistor impedances can be scaled up considerably for line level interface without a huge noise penalty.

Or use a transformer..... it's only money.

JR

JR> Or use a transformer.....

MODERATOR!! Someone has hacked JR's account/computer!

PRR said:

JR> Or use a transformer.....

MODERATOR!! Someone has hacked JR's account/computer!

Click to expand...

NAH...  people get angry when I say to use a digital microprocessor for everything ....

While I spent an entire career being prudent about design cost effectiveness (and linearity), different strokes for different folks who may not share my philosophy (a lot of that going around lately).  I don't much mind how other people spend their money, it concerns me more how they want to spend mine. 

JR

gyraf said:

That's the catch - source impedance vs. Frequency is rather undefined and different under different conditions. Transformer dosent care - but this mucks up in various ways under different conditions.

Click to expand...

I wonder if this weirdness is the reason the 2141 "fell out of favor" among gear designers.

FWIW, the THAT and TI INA parts do not seem to exhibit this behavior.