Bridge or Parallel?

[Roushon]

GainClone Bridge or Parallel? Which one sounds good? The following are the combinations possible. I am confused which one to choose.

1. Bridge connection
2. Parallel connection of two inverting
3. Parallel connection of two non-inverting

I have been reading a lot about GainClone LM3886 here. I gather from other's opinion that Inverting always sounds good. The confusion is that if that is true then which one will sound good 1 or 2?

Now I have all the parts to build my dream project, the reason why I became member of this forum. I plan to have both buffer and DC-servo in the circuit. I am reading the application notes AN-1192 also. But it will be helpful to listen to experienced people.

Looing forward to your suggestion.....

[analog_sa]

1. Double the volts but if your speakers impedance dips - current starvation. Bridge + parallel is a better plan.

2/3 Differences between invering/non-inverting are a bit exaggerated. Parallel is very likely better with 90% of 'modern' speakers. Precise matching of resistors and offset trimming are a must.

Buffered is probably fine if you can build a good buffer. Servo is simply not needed with these chips. If you have input offset rather use capacitor/transformer than a servo.

[Roushon]

Thanks analog_sa for your reply. At the moment I am just reading and reading in different threads. Here is a rough plan of my stereo Gain Clone design for one chanel.

1. Parallel of two inverting LM3886TF design with 0.1% resistors.
2. DC-servo using AD8620 or LF412. Though people say that DC-servo is not necessary, but all the good designs have them.
3. Separate buffer for each inverting stage using OPA627. This is somthing new I am planning which I have not seen before.

Any suggestions or opinion are wellcome...

Once I am finished with the circuit I will post that in the forum for opinion and corrections.

[Franz G]

... I recommend you a t-network for feedback!

Franz

[TwoSpoons]

I suggest you look at the app notes in the data sheet for the LM4780. In particular the THD+N graphs comparing bridge mode with parallel mode. You will notice the parallel arrangment has worse THD+N in almost every case. I have no idea why - maybe the inherent symmetry of the balanced topology helps eliminate crossover distortion ?

At any rate, thats why I've chosen to build using bridge mode.

[Roushon]

Originally posted by Franz G
... I recommend you a t-network for feedback!

Franz

Originally posted by TwoSpoons
I suggest you look at the app notes in the data sheet for the LM4780. In particular the THD+N graphs comparing bridge mode with parallel mode. You will notice the parallel arrangment has worse THD+N in almost every case. I have no idea why - maybe the inherent symmetry of the balanced topology helps eliminate crossover distortion ?

At any rate, thats why I've chosen to build using bridge mode.

Thanks for the suggestions. I will look at t-network and the app notes of LM4780 and get back soon. The application notes AN-1192 suggests that the best design is first parallel and then bridge. But I do not want to build so big a amplifier at the moment. And then I got confused about choosing from bridge and parallel.

[Franz G]

I will look at t-network

I am sure, it will be an advantage with the t-net even with bridged and/or paralleled amps, as the chips have the same parasitic capacitances.

Edit: I just had a look into AP1192: the t-net seems specially interesting for the bridged version: with the t-net you could use the same input-Z for the inverting amp like the noninverting amp!

This should boost the quality of an bridged amp!
End of Edit...

You will not be disapointed about this detail!

Franz

[Roushon]

Originally posted by TwoSpoons
I suggest you look at the app notes in the data sheet for the LM4780. In particular the THD+N graphs comparing bridge mode with parallel mode. You will notice the parallel arrangment has worse THD+N in almost every case. I have no idea why - maybe the inherent symmetry of the balanced topology helps eliminate crossover distortion ?

At any rate, thats why I've chosen to build using bridge mode.

I have looked at the datasheet of LM4780. I think the following two procedures will give different performance:

1. parallel connection of two separate inverting LM3886.

2. parallel connection of two inverting amplifiers (equivalent of LM3886) in LM4780.

The reason I feel is the length of some paths will be more in 1 than 2.

[Roushon]

Originally posted by Franz G


I am sure, it will be an advantage with the t-net even with bridged and/or paralleled amps, as the chips have the same parasitic capacitances.

Edit: I just had a look into AP1192: the t-net seems specially interesting for the bridged version: with the t-net you could use the same input-Z for the inverting amp like the noninverting amp!

This should boost the quality of an bridged amp!
End of Edit...

You will not be disapointed about this detail!

Franz

I just noticed a thread on t-network. Have not yet explored it fully. Many many people seem to be interested in t-network. I will also try it in my circuit and post here the result.

Thanks for telling me about it.

[Roushon]

A curiosity: feedback resistor should be as close to the chip as possible. and the track to it be as short possible. even using SMD is suggested. I am just wandering that using a t-network will make both the track and the total path length more. will a t-network also solve this problem of closeness; I mean with a t-network one need not bother to make the path length least possible. I am now thinking of a t-network with SMD.