Bridge/Parallel with balanced input

[sangram]

Hi

Have no yet managed to get an authoritative reply on this one yet, and my searches came close but none of them answered it directly.

Basically I have the following in mind. Four LM4780 dual power amp chips, two each per channel. And each chip running off its own power supply. The two outputs of each chip are paralleled. Then the two chips are connected in bridge mode. As the input I have a pure balanced connection off my Delta 66, with floatable earth. Also a CDP converted to balanced through the use of a DRV134 line driver, and switched passively after volume control.

Paralleling apparently works quite happily with two inputs and outputs tied together, sharing a single feedback loop and input resistor, but needing an output current sharing resistor.

Here's my problem, when it comes to bridging the amps, specially now I have four power supplies and a stereo balanced input (2 XLR/TRS).

Where does each amp draw its input ground reference from, and where does the summing happen? I have not found an easy answer.

I have a few possible configurations in mind. Let's forget the complexity of parallel for a minute and assume one chip is one amp, which for this discussion, it is - so two amps/channel.

1. + and - of the balanced connection to one each of the amps' + input, and the reverse (+ or -) connection to the signal ground connection of the same amp. Fully floating and independent supplies for each of the amps, so the supply ground does not short the input, and the load summing up both the signals.

I am unsure if this will even work, as there's no clear ground reference for the signals at the input side, except the other end of the balanced connection, and since it's going to a dirty ground (comparatively) it may inject noise into the input. Also since that injected signal may not be equal on both the amp grounds (since there are two independent supplies/grounds) the noise may be too much.

2. Separated supplies for the two amps, and each amp referenced to the cable shield and one pole of the balanced signal. One driven by (+) and the other driven by (-) Will grounding will be a mess - because we need to tie the amps together at power ground and input ground, then star grounding for all the amplifiers... Or can we leave the supplies floating and the bridge will still work?

3. The two amps fed from + and - sides to the + and - inputs on the chips, and reversed on the other amp, (as opposed to using the other side as the connection for signal ground) to get fully balanced opposite polarity signals. What I can't get my head around is the signal ground input. Can it be left open or does it need to be tied to the cable shield? I am assuming the two amps can be left with floating supplies in this case.

TIA

Sangram

[sangram]

Hate to bump, but gotta...

Is the question too long?

[theChris]

"Paralleling apparently works quite happily with two inputs and outputs tied together, sharing a single feedback loop and input resistor, but needing an output current sharing resistor."

-- ok, if this is true, then use it. I'd probably still use the output resistors.



each amp is similar to any opamp. basically the power amp can output a voltage between +Vcc and -Vee. the referecne is defined by the resistor from +in to ground and from -in to ground.


1.) nope. these are HIGH GAIN AMPS. this means hooking + to + and grounding a terminal will give you OPEN LOOP GAIN. this is in excess of 10,000 at DC, and has a bandwith under 100hz. what you want is to use the circuits with NEGATIVE FEEDBACK to limit gain to 10-20 and allowing a bandwith of 50khz+. anytime you see amps, think feedback. even amps without "feedback" have it anyways... if you know the voltage gain, you probably have feedback.

2.) if the source is not AC coupled, it may connect the grounds anyways.

3.) again, no. this gives open loop gains. you need lots of feedback to move that "very high gain" to "confortably high gain".



i suggest all amps's have the ground star grounded to the chassis. further, the sheild of all balenced connections connects to the chassis. earth ground probably will connect to the chassis as well.

+Vcc and -Vee for each IC can be done now. this is apparently 1 transformer per IC, as per your design.

+(-) In connects to the + input of 2 of the ICs. a inline capacitor of 1uF (mylar or poly) is used to Ac couple this signal. a 10k resistor connects to ground. a 12k resistos connects from output to - in of each amp. a 1k resistor connects to ground from each amp.

this can be seen in the datasheet. the datasheet also shows variaotions on this to reduce DC offset and other things.

[sangram]

Thanks Chris

It's actually become 1 transformer per channel, but with separated rectifiers, but I can see we'll have to have one common ground and a phase inverter anyway, as my balanced source is not fully balanced, the return lead is referenced very slightly above ground and not the mirror image of the hot lead (like a DRV 134 does). So out goes that plan.

Thanks for the advice.

[theChris]

AC coupling? this should remove any small DC offset.

look into classic Instrumentation amplifers or such. you may be able to use ina2134 or some other inamp with good results.

[tiltedhalo]

The DRV134 can have all DC offset practically removed by inserting a 10uf film bi-polar cap between the out pins and sense with zero degradation of the signal.
It is very easy to tell if you have input imbalances with a bridge configuration useing the DRV's you will notice Hisssssss, this signifies signal in-eqaulities, the input should be fed to the DRV134 via a 1K 1% input resistor to set sensitivity and a 22.1K from signal in to ground for referance point. You may have to adjust your gain because the DRV's offer 4db of internal gain, so if your amps are straight forward 25 to 27db you will get close to 32db overall, any noise will become brillantly apparent.
If you have zero DC offset at your pre-amp I would skip the input coupling caps.
Only be concerned with each bridged/ paralleled pairs ground points, if you are useing dual bridge/caps for each chip tie those specific grounds together AS CLOSE AS POSSIBLE. you will probably experiance hum anyway, then it will become a process of elimination.

[sangram]

Thanks for the replies guys.

1. theChris: Yes, the design calls for AC coupling due to the offset of DRV134. I was aware that the 30-35 mV of offset would be too much for my poor 4780s/speakers. The other idea was a +- buffer built with a dual opamp, with a coupling cap at the common input, to give low to 0 offset at the output.

2. tiltedhalo: Thanks for the tip on the gain of DRV134. Maybe an input-buffer is a better way to do things in that case. I see what you're saying about the sense pins, but does that help the output offset or input offset?

3. Yes, I am using separate dual bridge/caps for each paralleled amp, and plan to have thick ground bars connected from the CHG terminal (see Peter's board) across the boards.

4. Why should I have hum if my speaker is not referenced to ground at all, in a bridged configuration?

Regards

Sangram