Constructing a telephone hybrid

[supersat]

I've built an Arduino shield based on the reference circuit in a TI app note for an MSP430-based modem (page 9/pdf page 13 of http://www.ti.com/lit/an/slaa037/slaa037.pdf). The plan is to allow people to implement experimental software modems on the AVR, so the shield is primarily an interface between the phone line and Arduino analog inputs and PWM outputs. (And before you ask, this shield is intended to be used on private PBXes only--not connected to the PSTN.)

The problem is that during initial testing, I'm getting no audio from the phone line. After taking another look at the TI reference design, I'm beginning to wonder if it's wrong. There are a couple of questions I have about it, and I'm hoping someone can help explain its theory of operation.

First, the two output op-amps have resistors from VREF to their non-inverting inputs. Is there a reason for this? VREF should be a low-impedance source, so I'm not sure what those resistors are doing.

Second, I'm seeing a signal from the phone line on one side of 500 ohm resistor, but not the other. This leads me to believe that the low-impedance op-amp outputs are overriding any incoming signal. I don't see how this hybrid circuit works. Perhaps the 500 ohm resistor is supposed to be on the other side of the 33k/pot voltage divider?

Any insight would be much appreciated.

[Circlotron]

First, the two output op-amps have resistors from VREF to their non-inverting inputs. Is there a reason for this?

Ideally, opamps should have equal resistance on both inputs (feedback resistor appears paralleled with input resistor) so that the voltage drop in these resistors caused by input bias current will be equal and opposite on each input and so cancel out and not cause an output offset.

[calexanian]

Remember that telco will only engage the line if it sees what it deems to be a proper impedance, additionally it does not like seeing ANY ground reference. Isolation transformers are a MUST!!! or 600 volt .1 uf caps on both lines ad then use a difference amp to sort it out to a ground referenced unbalanced signal.

[supersat]

Remember that telco will only engage the line if it sees what it deems to be a proper impedance, additionally it does not like seeing ANY ground reference. Isolation transformers are a MUST!!! or 600 volt .1 uf caps on both lines ad then use a difference amp to sort it out to a ground referenced unbalanced signal.

Yes, there is a 600:600 ohm isolation transformer in that circuit. Don't be fooled by the ground symbols in the schematic--similar grounds are just tied together, and the phone line side "ground" is not connected to anything else.

I can definitely see the dial tone signal on one side of 500 ohm resistor, so I know the line is being engaged. The signal just isn't making it to the input op-amp.

[TerraHertz]

I agree, that TI example schematic is wrong.
The input resistors on the op-amps are correct - as pointed out above op-amps should be provided with equal impedances on their inputs. This minimizes offset voltage, and variations over temperature range.
But the way the Rx amp's inputs are wired up doesn't make sense. Ideally the nodes of the two op-amp outputs are zero impedance opposing polarity versions of the Tx signal. Hence no portion of the received line signal will be detectable there.

At a guess I'd say the wire to pin 12 of the op amp, via the 33K resistor, should actually go to the RH end of the 500R resistor. Then the Rx op-amp will see Rx signal, but Tx signal can be nulled out via the pot.

A quick google found this: http://www.randolph-telecom.com/articles/AN-5,%20Transformer-based%20phone%20line%20interfaces%20(DAA,%20FXO),%20Rev1.pdf
Which makes more sense. For one thing, why waste an op-amp on symmetrically driving the local side of an isolation transformer?

[ludzinc]

I see nothing wrong with the 1.5V reference - it's to bias the output of the opamps.

I've been fascinated with telephone hybrids for ages - you can find my take on them here:

http://ludzinc.blogspot.com.au/2013/03/the-not-so-obvious-telephone.html

[David Hess]

First, the two output op-amps have resistors from VREF to their non-inverting inputs. Is there a reason for this?

Ideally, opamps should have equal resistance on both inputs (feedback resistor appears paralleled with input resistor) so that the voltage drop in these resistors caused by input bias current will be equal and opposite on each input and so cancel out and not cause an output offset.

This usually only applies to operational amplifiers which lack input bias current compensation and have high input bias current.  With input bias current compensation, the input bias currents are uncorrelated and with low input bias current, additional offset from mismatched source resistance is insignificant.

Second, I'm seeing a signal from the phone line on one side of 500 ohm resistor, but not the other. This leads me to believe that the low-impedance op-amp outputs are overriding any incoming signal. I don't see how this hybrid circuit works. Perhaps the 500 ohm resistor is supposed to be on the other side of the 33k/pot voltage divider?

I do not see how the design shown in the TI application note could possibly work as shown.  The idea is that the transmitted signal is subtracted from the received signal which is what the balance adjustment is for.  I think the connection to the 33k resistor going to the non-inverting input of the lower operational amplifier should be on the other side of the transformer primary.