But since we know P

and we are looking for

θ

θ

(4)

Again it must be noted that the value of

θ

upon the system designer’s amplifier application and its cor-

responding parameters as described previously. If the ambi-

ent temperature that the audio amplifier is to be working un-

der is higher than the normal 25˚C, then the thermal

resistance for the heat sink, given all other things are equal,

will need to be smaller.

Equations (1), (4) are the only equations needed in the de-

termination of the maximum heat sink thermal resistance.

This is of course given that the system designer knows the

required supply voltages to drive his rated load at a particular

power output level and the parameters provided by the semi-

conductor manufacturer. These parameters are the junction

to case thermal resistance,

θ

ommended Thermalloy Thermacote thermal compound re-

sistance,

θ

SIGNAL-TO-NOISE RATIO

In the measurement of the signal-to-noise ratio, misinterpre-

tations of the numbers actually measured are common. One

amplifier may sound much quieter than another, but due to

improper testing techniques, they appear equal in measure-

ments. This is often the case when comparing integrated cir-

cuit designs to discrete amplifier designs. Discrete transistor

amps often “run out of gain” at high frequencies and there-

fore have small bandwidths to noise as indicated below.

Integrated circuits have additional open loop gain allowing

additional feedback loop gain in order to lower harmonic dis-

tortion and improve frequency response. It is this additional

bandwidth that can lead to erroneous signal-to-noise mea-

surements if not considered during the measurement pro-

cess. In the typical example above, the difference in band-

width appears small on a log scale but the factor of 10 in

bandwidth, (200 kHz to 2 MHz) can result in a 10 dB theoreti-

cal difference in the signal-to-noise ratio (white noise is pro-

portional to the square root of the bandwidth in a system).

In comparing audio amplifiers it is necessary to measure the

magnitude of noise in the audible bandwidth by using a

“weighting” filter (Note 16). A “weighting” filter alters the fre-

quency response in order to compensate for the average hu-

man ear’s sensitivity to the frequency spectra. The weighting

filters at the same time provide the bandwidth limiting as dis-

cussed in the previous paragraph.

Note 16: CCIR/ARM:

A Practical Noise Measurement Method; by Ray

Dolby, David Robinson and Kenneth Gundry, AES Preprint No. 1353 (F-3).

In addition to noise filtering, differing meter types give differ-

ent noise readings. Meter responses include:

1.

RMS reading,

2.

average responding,

3.

peak reading, and

4.

quasi peak reading.

Although theoretical noise analysis is derived using true

RMS based calculations, most actual measurements are

taken with ARM (Average Responding Meter) test equip-

ment.

Typical signal-to-noise figures are listed for an A-weighted fil-

ter which is commonly used in the measurement of noise.

The shape of all weighting filters is similar, with the peak of

the curve usually occurring in the 3 kHz–7 kHz region as

shown below.

SUPPLY BYPASSING

The LM3876 has excellent power supply rejection and does

not require a regulated supply. However, to eliminate pos-

sible oscillations all op amps and power op amps should

have their supply leads bypassed with low-inductance ca-

pacitors having short leads and located close to the package

terminals. Inadequate power supply bypassing will manifest

itself by a low frequency oscillation known as “motorboating”

or by high frequency instabilities. These instabilities can be

eliminated through multiple bypassing utilizing a large tanta-

lum or electrolytic capacitor (10 µF or larger) which is used to

absorb low frequency variations and a small ceramic capaci-

tor (0.1 µF) to prevent any high frequency feedback through

the power supply lines.

If adequate bypassing is not provided the current in the sup-

ply leads which is a rectified component of the load current

may be fed back into internal circuitry. This signal causes low

distortion at high frequencies requiring that the supplies be

bypassed at the package terminals with an electrolytic ca-

pacitor of 470 µF or more.

LEAD INDUCTANCE

Power op amps are sensitive to inductance in the output

lead, particularly with heavy capacitive loading. Feedback to

the input should be taken directly from the output terminal,

minimizing common inductance with the load.

Lead inductance can also cause voltage surges on the sup-

plies. With long leads to the power supply, energy is stored in

the lead inductance when the output is shorted. This energy

can be dumped back into the supply bypass capacitors when

the short is removed. The magnitude of this transient is re-

duced by increasing the size of the bypass capacitor near

the IC. With at least a 20 µF local bypass, these voltage

surges are important only if the lead length exceeds a couple

and ground leads minimizes the effect.

DS011832-13

DS011832-14

www.national.com

13