..About quasi class a bias for a headphone amp..

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am applying some current sources I found in document an103 see link


from fig 4 I am using a 2n4392 (see chart on last pg) to bias the neg rail to the output of a TPA6120A2 (same chip as the ths6012 I believe) for max current for the chip of ~12ma (nominal).

While observing the polairity of the source shown I adj the pot for supply amps but I find the current is split about 5.5 ma through the neg power pin and 6.5 ma through the curr source to the output.

If I watch only the power pin and go for 12ma I get 17ma total curr through the neg rail.

SHOULD I watch the total on the rail to keep the chip below max

OR is the curr going "in" the output through the curr sc. to be ignored as to curr used by the chip???

It does seem to sound best at the 12ma total.

AS to why I do this only my ears KNOW that.:D

This sounds much better than a resistor and a resistor while syurpie smooth is better than w/o bias.

I would be very interested to learn that you were able to make these distinctions in a proper double blind comparison among properly matched multiple amps with ABX switching, "comparing" on a single prototype with the soldering iron in your hand really doesn’t count as adding to audio knowledge

current source biasing has clearly measurable advantages with older IC processes that couldn't make good pnp transistors, but you're really unlikely to be do anything good (or even detectable) by loading this 100 MHz, complementary bipolar, 400 mA output op amp with sub 1 ppm distortion specs

my guess is that the tpa6120 already has its output running at > 5 mA bias and you're only upsetting the natural npn/pnp Vbe distortion cancellation you would expect from its intended operation in push-pull class A - external bias will turn off one transistor and give you the full nonlinear Vbe modulation around zero crossing


Joined 2003
Paid Member
> SHOULD I ...

What you should do depends on what you want to do.

I am afraid I don't understand what you want to do.

> ...watch the total on the rail to keep the chip below max

What max? You seem to have found the "Characteristics" numbers on page 3 of TI's datasheet... this gives a typical supply current plus a tolerance. If you found a TPA6120 that sucked 15.1mA you could return it as possibly defective (except when you are buying ones and twos, the factory won't listen to your complaint; if you bought a million, you could get credit for any that were substantially outside these specs).

> to bias the neg rail to the output .. for max current for the chip of ~12ma (nominal).

I do not understand the point of adding load to a chip to make it appear to eat the maximum supply current of a not-quite-reject chip. Even if a high-demand chip "sounded better" (this is true of many underbiased chips), sucking from the output is not the same as actually having a high bias current.

If the idea is to run Class A: that depends entirely on signal level and load. Making 1.4V into 10K, a 1mA bias is deep Class A. Making 3V into 32Ω, as in a headphone amplifier, 12mA bias is Class A only up to 2 milliWatts, an audible but hardly musically-loud level. OTOH, 12mA in some 300Ω phones will stay Class A to 21mW, which in some phones is loud.

As jcx says, it is mighty hard to explain what "only my ears KNOW" considering the nature of the TPA6120, and the fact that we don't have your ears or your exact test conditions.

If you like 12mA, do it. The chip won't be hurt by much-much larger external pull-down currents. Up to 100mA if you heatsink the chip (and FET!) very well.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.