LM4780 - problems with high idle current and no oscillations?

[gootee]

That's very thoughtful of you and is appreciated. Glad you got it working.

One thing that I still don't like the sound of is the "bypassing" of large electrolytic caps with small caps, especially if the small caps are low-loss types like film. Doing that could cause high-frequency ringing or maybe even oscillation.

And small-value caps can't do any good when they're working into and through the inductance of the power supply rails that are between them and the point of load, anyway. In a typical linear unregulated PSU, small-value caps are only needed and only useful at the point of load, to short-circuit ("bypass") high frequencies at the power pins that could otherwise form a sneak feedback path and could cause oscillation, and for "decoupling" at the power pins, i.e. to supply the fast transient current demands that can't come through the rails' inductances without causing unacceptable disturbances of the rail voltages, and which couldn't be supplied through the rail inductances with correct timing, anyway. (OK, sometimes they might be needed around the rectifiers, but then you'd probably want to use snubbers, i.e. with resistors in series.)

Even in snubbers, where small film caps can also legitimately be used, the resistor is the actual snubber and the cap is ONLY there so that the resistor won't have to deal with frequencies much lower than those that need snubbing, so the resistor won't have to dissipate as much power.

In most of the places in a chipamp where you might want a small-value cap, an X7R ceramic (i.e. not low-loss) is usually a much safer bet, and therefore better, than a film type.

[abraxalito]

Originally Posted by gootee

One thing that I still don't like the sound of is the "bypassing" of large electrolytic caps with small caps, especially if the small caps are low-loss types like film. Doing that could cause high-frequency ringing or maybe even oscillation.

Yep, agreed about this having the very real potential to make matters worse.

And small-value caps can't do any good when they're working into and through the inductance of the power supply rails that are between them and the point of load, anyway. In a typical linear unregulated PSU, small-value caps are only needed and only useful at the point of load, to short-circuit ("bypass") high frequencies at the power pins that could otherwise form a sneak feeback path and could cause oscillation,

Yes, in the case of chip amps the PSRR degrades pretty quickly as frequencies go up.

... and for "decoupling" at the power pins, i.e. to supply the fast transient current demands that can't come through the rails' inductances without causing unacceptable disturbances of the rail voltages, and which couldn't be supplied through the rail inductances with correct timing, anyway.

Here's where I find myself scratching my head though. Chipamps drive fairly long (over 1m typically) wires to speakers. So the interconnect is inductive and notionally 1uH as a minimum. At the far end, the drive unit also looks pretty inductive too, with 100's of uH. So how can it be that there's any 'fast transient current demands'? Inductors are immune to such - they oppose fast changes in their current.

(OK, sometimes they might be needed around the rectifiers, but then you'd probably want to use snubbers, i.e. with resistors in series.)

Yes - here you want to absorb the RF burst, not just recirculate it.

In most of the places in a chipamp where you might want a small-value cap, an X7R ceramic (i.e. not low-loss) is usually a much safer bet, and therefore better, than a film type.

Actually X7R is pretty low loss - they have usefully low ESRs. As an example, Kemet's 1206 1uF 50V is 12milliohm at its SRF of 3.4MHz. They're just not as low as NP0, and hence not as high Q. I don't recommend NP0 for decoupling duty at all as they're too expensive

[AndrewT]

Originally Posted by abraxalito

.....................
Here's where I find myself scratching my head though. Chipamps drive fairly long (over 1m typically) wires to speakers. So the interconnect is inductive and notionally 1uH as a minimum. At the far end, the drive unit also looks pretty inductive too, with 100's of uH. So how can it be that there's any 'fast transient current demands'? Inductors are immune to such - they oppose fast changes in their current.

each supply rail passes a halfwave copy of the output signal. The two halfwaves add up to the full cycle output AC.

A halfwave that starts from nothing and at the other end stops abruptly contains harmonics that theoretically extend out to infinity.

[abraxalito]

Originally Posted by AndrewT

each supply rail passes a halfwave copy of the output signal. The two halfwaves add up to the full cycle output AC.

Yep.

A halfwave that starts from nothing and at the other end stops abruptly contains harmonics that theoretically extend out to infinity.

Yes - but do the half-waves in a real amp actually start and stop in an infinitesmally small time? If so then the matching between the two halves would have to be phenomenally good not to get a big inductive kick back from the load. Does anyone know what the current slew rates are at zero crossing for a typical chip amp?

Its a very good point and may well be a reason why gootee has suggested a bandwidth around ten times the audio bandwidth. Its also one good reason for why I only design with bridged amps these days

[AndrewT]

I don't think 10times the audio bandwidth is adequate.
That only gets one to 200kHz.
I suspect the harmonics >10MHz have an influence on the way a ClassAB PushPull amplifier operates. These are the frequencies that the HF decoupling is supporting (supplying transient current).