LM324 and crossover distortion

The TL07x are all intended for low audio distortion, and the specs say so. The TL06x however, were intended for low quiescent current applications, and that data sheet conspicuously does NOT mention 'low harmonic distortion'. The TL08x should be pretty good, though; its output drive is very similar to TL07x.

More important, crossover distortion is only one source of distortion; at high frequency, there are slew rate limits, and at near-rail conditions some saturation will occur. These all cause harmonic distortion.

One of the scariest sources is thermal; there are lots of op amp circuits which have terrible distortion at ~1 Hz, and many audio power drivers are weak at the low frequencies, too.

Reply to
whit3rd
Loading thread data ...

I found your PDF on it here:

formatting link

I'm not sure how increasing the emitter areas of the output transistors implements the sliding class A.

Without any PNPs for current source loads it must have taken some intuition to arrange things so that everything canceled out and left the output at zero volts!

Reply to
Bitrex

Interesting. This gives me some insight into why there are so many varieties of op-amps available today.

Yes. I understand those limitations.

That one is new to me.

Thanks, John

Reply to
John - KD5YI

Those are just current multiplications. Work your way thru all the math to see how it works.

Yes ;-)

Actually my specialty over all these years has been tracking and matching over temperature, voltage and process corners.

At Motorola I was nicknamed "Vbe Thompson", because there wasn't anything I couldn't compensate.

See Tom Frederiksen's book for additional information on my talents ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

That's another well proven vanilla dual op amp that seems to have been around forever. Slightly better spec than a 741, iirc, but you would need to compare the data sheets to get all the details...

Regards,

Chris

Reply to
ChrisQ

This is why engineers can get pathologically detailed about qualifying what they say.

Looking at the schematic for the thing it looks like they've made a nice class AB output stage that's not going to have much crossover distortion at all -- but I'll bet that some of that 0.003% THD comes from crossover.

--
Tim Wescott
Wescott Design Services
 Click to see the full signature
Reply to
Tim Wescott

Okay. Thanks.

Reply to
John - KD5YI

I will do that. Looking at it again I believe it has to do with the interaction between Q14 and Q13, if most of the current from Q11 is considered to flow through Q14 that will create a Vbe drop that is larger than the Vbe of the transistor with a larger emitter area. At the quiescent point that will cause a current to flow through the output transistors, and small increases in current through Q11 won't change that much, keeping the output operating class A.

A large increase in current through Q11 will cause Q14's Vbe to rise, but this will also cause the current in Q13 to rise, and if I plug the equation for Q13's Vbe in terms of the current through it from Q11 into the equation for the current through Q13 there will be a certain current through Q11 that causes the current through Q13 to go to zero, marking the transition from class A/AB to class B.

Reply to
Bitrex

Indeed. Ignoring Maxim, it's hard to do worse.

Reply to
krw

Is there a simple explanation for poles and zeroes in feedback networks for us mere mortals?

Grant.

Reply to
Grant

LM324 is the quad version of the dual LM358. I use resistive pull down on LM324 output because the output spec shows sourcing is >2 times sinking ability. That was on 12V single supply operation. For running the chip at higher voltage I imagine you'd want to use a current sink in place of a simple resistor.

Hmm, I vaguely remember a spike at crossover when I first noticed it on LM324, and a pulldown resistor to 0V fixed that on 12V operation, about

1k. I don't use LM358/324 for audio work, too slow.

Grant.

Reply to
Grant

Here's how I understand it: In any feedback system the danger is having the loop gain be greater than unity at the frequency where the phase shift through the feedback loop shifts through 180 degrees. In op amps the main cause of phase shift is the capacitance inherent in the amplifying devices, like the drift capacitance between base and emitter and the diffusion capacitance between the base and the collector.

Each capacitive element constitutes a pole in the transfer function, adding up to a maximum of 90 degrees of phase shift, and each amplifying element generally adds at least one pole. So, if a transfer function has three or more poles there will be almost guaranteed feedback stability problems, unless something is done to prevent the phase of the feedback loop shifting through 180 degrees.

The way the LM324 handles this situation is it uses only 2 gain stages, and puts a (relatively) big honking capacitor around the second stage. This starts rolling of the gain at a low frequency, and also causes the other poles (phase shifts) in the amplifier frequency response to move to a higher frequency. The net result is that when the feedback loop is applied, by the time the frequency gets high enough for the high frequency poles to start causing phase shifts greater than 180 degrees, the amplifier's open loop gain is so low that the feedback loop gain will always be less than unity. The downside of this approach is that it reduces the amplifier's bandwidth and slew rate - the slew rate is mainly determined by how quickly current can be pumped in an out of the compensation capacitor.

Another approach, which is what I think Jim is talking about, is to add a zero in the open loop transfer function, or in the feedback loop. So if an amplifier with three poles is described by a transfer function like 1/[(s + a1)(s + a2)(s + a3)], you put a zero in the transfer function to cancel one of the poles, like (s + a1)/[(s + a1)(s + a2)(s + a3)] = 1/[(s + a2)(s +a3)]. In practice adding a zero will also add a pole, but that pole can be made at a much higher frequency where it is of no consequence. This allows you to add another gain stage, and reduce or eliminate the dominant pole capacitor, giving greater gain (better gain-bandwidth product) and faster slew rate.

I may have errors in my understanding, I'm sure the folks here will point them out for both of our benefits. :)

Reply to
Bitrex

[snip]

I don't know. I've been doing it for so long, I'm not sure where to begin.

Get out your schoolbooks and study up on Bode plots, as a starting point.

Poles cause increasing phase lag, zeroes cause leading phase. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Jim, do you mean zeroes cause increasing phase lead?

John

Reply to
John - KD5YI

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.