Comparison between TDA7293 and LM1875

[peufeu]

I'm building an 8 channel gainclone for active speakers. Crossover will be done digitally by a computer.

I needed a mute function for the protection circuit which will disconnect the drivers if there is LF on the tweeter output in case of wrong wiring, a software bug in the crossover, crashing computers, or me using the wrong impulse responses (these things happen !). The protection circuit will also detect DC at output and oscillations, and use an Atmel micro for its brains.

I was hesitating between :

- LM3886
- TDA7293
- LM1875 + mute relays.

I have the chassis with heatsinks and 8 spots to put my chips, so I tried the three of them.

LM3886 was eliminated right from the start as the mute input seemed to act like another signal input ! Feeding it with a current source just made the chip oscillate, which is highly suspect, and feeding it with a resistor isn't optimal as the current will vary with the voltage on the V- rail, hence changing the internal bias of the input stage...

Just after that, one of the chips commited suicide so I cant' comment on the sound. Good bye LM3886 !

So there was TDA7293 and LM1875 left.

TDA7293 Pros :
- Almost impossible to destroy
- Very hard to make it oscillate
- High power

TDA7293 Cons :
- Too many pins
- Complicated wiring
- Expensive if you need 8 !

LM1875 Pros :
- Small and easy to wire
- Can use the SMD resistors for feedback a la 47 Labs
- Dirt cheap

LM1875 Cons :
- Very picky about layout
- Likes to oscillate (and I mean, it likes it a lot)

Sound tests doing a blind test with my girlfriend switching the amps and me trying to guess, on Portishead's Roseland NYC Live.

- When there is crowd ambience in this live recording, I can pick out the two easily, as the LM has very good "air" and detail, whereas the TDA squashes the details.

On other music, it really seems the LM is more detailed, airy, and the TDA erases the details and sounds darker.

The LM sounded a bit too bright though.

Thus I think I'll use the LM chips and pay careful attention to wiring and feedback resistors to tame the brightness.

Comments ?

[GregGC]

Sound like you have it all figured out.
Pay good attention to the shielding of the amp. And the wiring of course, as you said.

[Peter Daniel]

That's pretty much in line with my observations regarding TDA7294 and LM3875. The first one sounded dark, uninvolving and lacking air.

I still have 5 of TDA7294. I could sell them at good price

[peufeu]

I had some TDA7294, and as the 7293 was supposedly an "improved" version, I ordered a pair... but they sound exactly the same.

I'll keep them for the day I build low-cost active speakers for friends, maybe. Or for my computer... Even TDA amps are better than a "consumer quality" amp and speakers...

[peufeu]

Now an interesting question : how do I ground this mess ?

From what I understand, I think I know how to do it properly, but there are some unavoidable ground loops in any shared power supply amp...

First I have the power supply. Standard 200VA toroid, high speed diodes (does not ring on switchoff, I checked), 30.000 uF per rail. Rails run about +/- 27V.



I have taken care that no capacitor charging pulses travel through the ground (see schematic)

From the power supply ground I make a star point going to all the chips decoupling caps (power gnd). Then I add a star point going to all the speaker outlets (out gnd). Then I take a single wire going to a little board with all the input RCA's, and this is the point the chassis is connected to ground (see schematic).

Now all the shielded cables from the RCA go to the power opamp's input stages.

From V+, V- and power gnd, I draw wires to each amp, going through the local decoupling caps, and to the amp's V+ and V-.

Speaker wire goes from the opamp's otuput to the socket, return ground follows the same path to the speaker ground star point.

Note that there are three ground near each opamp's : the node between the decoulping caps, the input ground, and a speaker ground wire passing by.

To minimize current loops, I'm keeping together the forward and return paths of all currents. That's why I have 4 wires going from the power opamp to the supply : V+, V-, and two grounds. I don't want the pulses from the decoupling caps going through the same wire as the speaker signal.

Now what do you think of this ? Is it good ? Is it the only way to put 8 unbalanced amps in the same box using the same transformer ?

Thanks !

[Pedja]

Pierre,

First, your feedback network certainly needs a shorter path to the ground.

Second, if your value of the chips decoupling caps is not less than a few hundreds of uF then I’d also do some other things the different way. I altered a bit your drawing (sorry, it was the simplest way):





As you can see I’d rather go with local ground points (consisting of local caps central point, speaker ground, feedback ground and input signal ground) that will be connected to the main ground point (central point of the rectifiers or, if you must use them, those 30.000uF caps). Of course, main ground point and local ground points should be connected by low impedance (short and fat) paths.

Also, the thing could be done without any main ground point: you could route both negative end of the positive rectifier/main cap and positive end of the negative rectifier/main cap independently to each section where they will be jointed.

Pedja

[richwalters]

The LM1875 is a bit hairy, a typ mid 80's era amp; that is other configurations than given in Nat Semi applications note.... it's liable to take off. I'm not really convinced that Nat Semi refined their homework on this chip amp, in that THD is hard to optimise at higher frequencies, or rather not optimised in their appls circuit. THD is quite low, but:- note; only stated at 0,015%@ 1Khz. Some chip amps can sound quite awful at higher frequencies;.....once it was considered for use in the console industry as a bus/headphone amp (with excellent S/N ratio) but never made it' debut ....as there are better variants around.

The non-inverting arrangement is about it: I've never had any luck with other configurations.
For performance, the layout needs to be carefully optimised, with double sided PCB with via's that is the C's and LS return in the right place for min THD and no bird's wiring.


rich

[peufeu]

Originally posted by Pedja
... [/B]

Wouldn't this create a ground loop ?
Remember, the input grounds are joined at the input RCA's, and also in the signal source, so what you'd do would make some charging pulses go through the signal cable shielding !

[Pedja]

I can’t see if charging pulses run through the interconnect shields.

This way the input signal grounds shouldn’t be joined at the RCAs.

Pedja

[peufeu]

I can’t see if charging pulses run through the interconnect shields.

Imagine there are several amps in the box sharing the same power supply.

If we connect the signal cable shield to the power ground node near each amp, we get a ground loop : between the multiple power ground nodes (one for each amp, where the decoupling caps join) and the RCA grounds.

This way the input signal grounds shouldn’t be joined at the RCAs.

They are joined in the preamp anyway, which is even worse.