My small contribution

[syn08]

Here's my small contribution to the audio DIY world...

It's basically a combination of the LM4702C driver with the LT1166 power output stage biasing system, replacing the QMult in the original LM4702 application note. Arrownac sells the LT1166 sells for about $2, well worth replacing the QMult, as long as the impact on AC performance is kept to a minimum by bypassing the Gm stage.

At Vc=+/-44V the breadboarded version has a bandwidth of 28KHz (0.1dB), Distortions at f=20KHz and Po=87W (just before clipping), were unmeasurable using my rudimentary TI UAF42 based notch filter which, based on previous measurements, makes them better than 0.01%.

Clipping occurs at Vo(peak) = 39V and is clean and symmetrical. For a 4ohm load, clipping occurs at Po=150W and is because of the output Darlingtons base current, which is larger than the 5mA that the LM4702 can provide. Triplets are required to further push the output power (this was not experimented so far).

Schematic is attached, I'll follow up with some pictures.

Planning to design the PCB ASAP...

[CBS240]

At first look it apears this circuit should be triple darlington, otherwise what is the point of D1, D2?

[Tim__x]

D1 and D2 give the LT1166 some voltage headroom. Look up the chip on www.linear.com. They won't be absolutely necessary with a triple darlington.

[syn08]

Experimental setup.

[syn08]

Results.

Vc = +/- 44V, f=20KHz, load is 8ohm, output power 87W (just before clipping).

Upper trace, 20V/div, lower trace 1V/div. Horizontal, 20uS/div.

[syn08]

The minimum operating voltage for the LT1166 is ±2V, so
it is necessary to bias the part with adequate voltage from
the output stage. D1 and D2 add 0.7 V to each LT1166 supply branch.

[djk]

Try taking out R11, I bet it will still work just fine.

If it does, then Q1,2,3,4 and associated parts are not needed.

These parts are in the application note because they need to provide the voltage drop across the bias IC. The LM4702 sources the current from pin 2, and then sinks it into its own internal CCS on pin 3.

I would recommend you slap a cap from pin 2 to pin 3 of the LM4702, try 1µF~100µF.

I would also do a triple as the LM4702 has very low output current.

[syn08]

The LM4702 can provide only 5mA of output current. That's definitely not enough to both bias the LT1166 (min +/-4mA at +/-2V) and provide the base current for the output darlingtons. Therefore, some extra current is required to bias the LT1166 and that's the purpose of the current sources. They could be replaced by simple resistors to the +/- supplies, however this will negatively impact the PSRR. The LT1166 PSSR is only 19dB...

[syn08]

Originally posted by djk


I would recommend you slap a cap from pin 2 to pin 3 of the LM4702, try 1µF~100µF.

Could you elaborate on this? A couple of hundreds of pF to ground from each output could be a frequency compensation method (but then the LM4702 gain stage is already compensated). Experimentally, for 500pF the SR is seriously affected, but for 1uF-100uF...?!

[djk]

You're right, the LM4702 is only spec'd at 3mA minimum.