THAT 1646 unbalanced to balanced line driver

Some time ago I started a thread here to design a power supply. I will order the pcb's somwhere next week and do some prototyping. However a power supply needs something to power. I made it primarily to be used with an unbalanced to balanced line driver board. I have also made a design of that board and I'd like to hear your thoughts on it. It isn't exactly headphone related as it will be used to send an unbalanced signal out of a mixing console to some powered dynacord monitors. The length of the cable is about 20m. Right now it is used with an unbalanced cable but there is quite a bit of hum, hiss and cellphone noise.
 
The circuit consists of an OPA227 at unity gain driving a THAT 1646. The OPA227 can be replaced with the OPA627, but only for small channel counts or by those with slightly to much money. the connector is an amphenol ACJS-MHD double trs. I couldn't find an eagle library for it so I made one myself. 

 
It's a bit hard to see but the trace to pin 12 on the connector isn't connected to the ground plane

Have you tried the THAT1510 opamp?

The THAT1510 is a preamplifier. It is generaly used as a mic preamp. It takes a balanced signal and amplifies it up to 60dB. So I think it isn't really usable in this situation.
edit:
And it's exactly the wrong way round. it has balanced inputs and unbalanced out. If that were the aim, I would use a THAT1246. Which is a balanced line reciever.

Sorry - for a moment I thought that the THAT1510 might be used as a normal opamp.

No problem, just check the datasheet and you'll see 

Is U1 there because you want to increase the input impedance?  THAT1646's input impedance is 5K ohms, which is low-ish, but still tolerable for most sources (certain tube gear excepted).
 
At any rate, you should add a high-ish value (~100K is good) resistor from U1's + input to ground.  Otherwise, if you disconnect the unbalanced source, the opamp input will float and lose its bias.  What would happen is most likely a large DC offset at the output.

U1 is there because it is stated in the data sheet that it should preferably be driven directly from an opamp's output to maintain specified performance (page six in bold). Several articles I have read indicate that they actually mean it and that when it isn't used in this manner it tends to be very noisy. Will add the resistor tonight, a large dc offset would be bad .

Quote:

steven2992 said:

U1 is there because it is stated in the data sheet that it should preferably be driven directly from an opamp's output to maintain specified performance (page six in bold). Several articles I have read indicate that they actually mean it and that when it isn't used in this manner it tends to be very noisy.

Click to expand...

What do you think is driving this circuit from your source?

An opamp, but with at least a meter of cable between them. I think the key word is directly ie not more than a few cm. 

Quote:

steven2992 said:

No problem, just check the datasheet and you'll see 

Click to expand...

 
I'd already done that much before I saw this thread. You're such a modest guy, are you.
 

Quote:

steven2992 said:

An opamp, but with at least a meter of cable between them. I think the key word is directly ie not more than a few cm. 

Click to expand...

Not to be a pain in the *ss, but the datasheet doesn't say anything about a few cm, only that the source should have low output impedance.  How much impedance do you think a meter of cable adds anyway?
 
Btw, in the March-April issue of multi media manufacturer magazine (from Audio Amateur Inc, which also publishes audioXpress and Voice Coil), on page 23 is an article by Gary Galo about the implementation of balanced line drivers and receivers using THAT1646 and THAT1200.  Here is the link to that issue for your reference:
http://audioamateurinc.com/digital/m3/issue/210/

You are never a pain in the *ss , and you are right(as always). I've done a bit more reading and U1 isn't neccesary and could be ommited. But I like the piece of mind it gives. You can connect any piece of gear no matter what output impedance it has. So I think I'm going to leave it in anyway. 
 

not only would I leave the op-amp in but I'd allow for gain, too!
 
(I need that right now in something I'm working on.  there are times that unity gain is good but there are other times you need gain.  since the op-amp is recommended to be close to the THAT chip (I've read that a bit, too) why not give it the option of gain and have it be switchable.  it can go down to near unity (being mostly a buffer) and can give gain if/when needed.

Adding a gain option is possible but it would add some resistors and jumpers/switches. The unbalanced to balanced conversion in the that1646 already gives you 6dB of gain and I don't need any more since it is just a line driver. I could add some test points/pads so that gain resistors can be added in. 
 
edit:
 
How does one put components, or at least silk screen, on the bottom of the board? Or just remove the silkscreen of a single component? I could put 2 resistors on the bottom of the board and you would leave them both out for unity gain or cut one trace and put them in for gain. 

I changed the board layout to include a 100k resistor from input to ground and 2 optional resistors to set gain. The gain resistors are on the bottom of the pcb.
 

 
If you don't want any gain the resistors can be left out and if you do want gain you need to cut one trace: