multiple output balanced buffer board

Hi All!,

I designed this board to have multiple balanced outputs from a single differental input signal.
It is based on an NE5532 l differential amplifier, (or integrated like INA134), and balanced inverting output buffers made with 6k8 resistors for gain set (unity), and 75 ohm output resistors. (also made with NE5532).

I have built a version with one input and one output, and worked fine. But the version with multiple outputs distorts the signal.. . So there must be a design error... if you can see the picture...

The output of the differential amplifier goes directly to the  inverting input of each output buffer... maybe the mistake is there??

Any help much appreciated
JAY X

all the 'B'site missing the input resistor

ijzerpro said:

all the 'B'site missing the input resistor

Click to expand...

Ditto

Also I would use .1% 10ks around U3.  Nice simple DA

Gary

Hi!

Well, i finally found the cause of the distortion... ¡excessive level of signal input!. I think i should place diodes at the input of the differential amplifier.

0.1% resistors would help reducing noise, yes, or i could use an integrated diff amp, like the INA134, THAT 1240, etc.

ijzerpro, the input and outputs are balanced, but the connections in between are unbalanced, that is normal.

Thanks a lot for your help!
JAY X

JAY X said:

Hi!

Well, i finally found the cause of the distortion...

Click to expand...

Are you sure? If your schemo is actual, there should be a 6k8 res between pin 1 & pin 6 of all the line-driver stages. How have you checked the operation? 

think i should place diodes at the input of the differential amplifier.

Click to expand...

I don't think they would do anything for your issue.

0.1% resistors would help reducing noise, yes,

Click to expand...

No. They wouldn't change significantly the operation.

or i could use an integrated diff amp, like the INA134, THAT 1240, etc.

Click to expand...

You could, but in the state of things, the problem with the output sections is much worse than any hypothetic problem with resistor tolerance.

In fact, out of the possible strategies of implementing transformerless balanced outputs, you have chosen the worst, particularly because it does not interface with unbalanced equipment. It can generate high levels of distortion that you would figure out only by listening individually to each output. This is known as the "Tascam issue".
There are other possibilities for EBOS (Electronically Balanced Output Stage) that are as simple to implement, such as impedance-balanced (which is basically a single active stage and a dummy load on the negative output leg), ground-sensing (which is also a single active stage with a sensing resistor on the negative leg).
In terms of performance and (almost) unconditional interfaceability, none of these will beat the THAT 1646 chip, but they are satisfactory for most studio applications.

Hi Abbey!

The output stage is a classic two inverting output. The first one inverts the input signal, and from its output, there is a derivation that goes to the second inverting stage...

I saw this type of output stage in old mixing consoles as balancing buffers before balanced buses.. but with tl072.

In any case i have build it in a single layer board, to run test signals at -12db from a soundcard,  into the 8 outputs, as i don't have a multitrack recording program like cubase, protools, etc.. It distorts with increasing signal level...bad..

I thought to build this board with line drivers.... but i took this approach... :

I will post a picture of the board asap!

JAY X

As several previous posts have pointed out, there is a fundamental design flaw in you schematic. No need to search any further, as shown this is simply not going to work properly.

Samuel

some pictures...

hi!

yes... i see... i missed to draw a 6k8 resistor... but i placed on the board.. be sure!! because i draw the schematic with a program but made the board with another different... ;D

jay x

So can you post the schematic you used for the PCB layout (I assume your layout software needs a schematic too?)? What is the used power supply and decoupling?

Samuel

JAY X said:

Hi Abbey!

The output stage is a classic two inverting output. The first one inverts the input signal, and from its output, there is a derivation that goes to the second inverting stage...

Click to expand...

I know it's classic, and it's known to have issues.
1) The least is that you loose 6dB if you use it unbalanced
2) In unbalanced mode, it does not cancel common-mode noise, which the other implementations proposed do.
3) If you connect it to an unbalanced input without taking proper precautions, one leg is shorted to ground, causing not only to lose 6dB, but also inducing nasty distortion, that can propagate into the other outputs via ground /rails.

Tascam have used it for years in some of their products (I believe even the DA series had them) and it's always been a problem integrating them in a studio environment where they would be on a patchbay.

I saw this type of output stage in old mixing consoles as balancing buffers before balanced buses.. but with tl072.

Click to expand...

In an application where the load is correct and reliable, it's ok, but for a generic output that may end up being connected to unknown or uncontrolled destinations, it's not.

In any case i have build it in a single layer board, to run test signals at -12db from a soundcard,  into the 8 outputs, as i don't have a multitrack recording program like cubase, protools, etc.. It distorts with increasing signal level...bad..

Click to expand...

Maybe your souncard is unbalanced...

Yes, while it can be a pain sometimes one benefit of using modern CAD tools is that there is good agreement between the schematic and the physical PCB.

JR

Let me pop in again to say (as always):
The XLR pin#1's to not connect to the circuit common (OK, OK call it ground). Any XLR pin#1 connects only to the chassis very near the jack.

On another note.  The 0.1% resistors don't need to be 0.1% resistors. They need to be matched resistors. So get a big box of 1% resistors, and match the values for an individual op-am.

Hi all!

well, ok, it is a bad design. So next time i will use line drivers like that 1646 or DRV 134. I just used the parts i had available at home...

The pin1 is tied to the groundplane via a jumper, so it can be disconected in case the board ends in a metal box...

My soundcard has unbalanced outputs with +4dbu output level...

I made the schematic with PROTEUS ISIS, and the pcb with abacom Sprint layout 6.0. lately i do it this way, no autorouting, no autoplacing, just draw tracks and conections, placing parts according to a schematic done in another program... for me it is easier, till i find a better schematic software... ;D

JAY X said:

Hi all!

well, ok, it is a bad design. So next time i will use line drivers like that 1646 or DRV 134. I just used the parts i had available at home...

Click to expand...

Any design that works is a good design, while pretty much every design ever done could be better. I know I'd like to revisit a lot of mine.  :-[

The pin1 is tied to the groundplane via a jumper, so it can be disconected in case the board ends in a metal box...

Click to expand...

If you revisit this I would suggest considering an alternate topology even using the same parts to address external grounds and internal 0V (not the same thing).

I would suggest using a differential topology (ideally balanced but differential works), and using the differential on the input to reference the two legged input signal to the one internal (local) 0V. Then configure the outputs as differential between the local 0V and output chassis ground. This may require more resistors but will handle ground errors with math, not brute force (like lifting or shunts between sundry grounds). 

My soundcard has unbalanced outputs with +4dbu output level...

Click to expand...

Still benefits from balanced wiring and differential inputs.

I made the schematic with PROTEUS ISIS, and the pcb with abacom Sprint layout 6.0. lately i do it this way, no autorouting, no autoplacing, just draw tracks and conections, placing parts according to a schematic done in another program... for me it is easier, till i find a better schematic software... ;D

Click to expand...

For simple designs you can get away with that but without that 1:1 correlation between schematic and PCB you can make extra work for yourself finding design/layout errors. FWIW this strict 1:1 relationship between schematics and PCB is a relatively recent (only decades) development,  so many classic schematics in circulation contain errors. This is useful to know when studying old schematics. Sometimes if the design does not make sense as drawn it is because the schematic is drawn wrong.  8)

JR