EMRFD Message Archive 1597

Message Date From Subject
1597 2008-04-25 21:18:30 Frank Op amps and crossover distortion with RC loads
I've built a set of cascaded op-amp based active audio filters.
Running from 8V and ground, I used one of the quad op amp sections to
create 4V for the "center" voltage.

They're Sallen-Key networks, effectively a 300Hz highpass followed by
a 3kHz lowpass. Thing is, with S-K networks, that means the output
pin of an op amp is feeding a capacitor, that in some cases, goes
right to the negative input, which is at virtual ground.

I'm using low cost op amps. LM324/224, MC33174. In all cases, I'm
getting an interesting current-based "crossover distortion".

It's like this....measure the voltage with respect to ground (4V) from
each of the two legs of a capacitor, and of course there's a phase
difference. Right at the crest of the voltage across the cap, the
current flow reverses. If this cap is being fed by the op amp (even
through like a 2.2k ohm resistor", this means the op amp has to
reverse its direction of output current from "sink" to "source" or
vice versa. This shows up as a glitch in the output voltage waveform:
the more gain I've got going on the worse the glitch. And...it's
audible. The location of the glitch varies with frequency, but not
amplitude - that is, it's at the same phase relative to the crossover.
Two glitches per cycle of the sine wave, for each time in the cycle
the capacitor current reverses. The only way I have been able to get
rid of this is by using *extremely* high series resistors in the
circuits - that is, the output of the op amp must have a minimum of
about 47k ohms before any capacitor. I am not intending to feed any
capacitor directly, but in fact, if I have a unity-gain buffer feeding
a Sallen-Key highpass, then it is feeding a capacitive input...and if
I stick that 47k resistor in to isolate, the gain gets uselessly low.

Look at EMRFD Fig 3.58. The center op amp of the top three, for
instance, has a 0.047uF cap from output to neg input. I've got
exactly that circuit in my rig-up, and am getting the crossover
distortion.

I've added a set of scope photos showing this "glitch" in the output
of the op amp....four of them, in the folder entitled "FM_Scope_photos".

So, the questions are thus:

1) Has anybody else seen this? Percentage-wise, the distortion
percentage is low, but it's quite audible and worse, during that
glitch period, the circuit is momentarily RF sensitive - not a good
thing to have around a transmitter.

2) Is there a better op amp to consider for this application?

I'm thinking that the theoretical explanation would fall into
something having to do with the natural pole in the op amp and perhaps
all I need is a higher speed op amp that can handle the
current-direction switch better.

However, EMRFD is loaded with example circuits that look like mine,
and it is written that one can use a 741 or 1458 - both of which I've
tried in some of my proto circuits, with exactly the results I've
shown in the photos.

Thanks for any help!

FM
1598 2008-04-25 22:15:50 Bob Larkin Re: Op amps and crossover distortion with RC loads
Hi Frank,

Yes, that quirk has been with the 324 and related op amps. One method to
move the cross-over region around is to put a bias current on the
output. This is wasteful of current, but you can watch the crossover point
move, with a 'scope, as you change the load resistor. You might start with
a 1K resistor from the output to the + supply.

I don't think I am up-to-date on the current op amps, and the best
alternatives. Perhaps others could comment on this. Certainly, changing
devices can solve the problem.

It is interesting that a long-time "standard" would have such an annoying
problem!

73, Bob W7PUA

At 04:18 AM 4/26/2008 +0000, Frank wrote:

>I've built a set of cascaded op-amp based active audio filters.
>Running from 8V and ground, I used one of the quad op amp sections to
>create 4V for the "center" voltage.
>
>They're Sallen-Key networks, effectively a 300Hz highpass followed by
>a 3kHz lowpass. Thing is, with S-K networks, that means the output
>pin of an op amp is feeding a capacitor, that in some cases, goes
>right to the negative input, which is at virtual ground.
>
>I'm using low cost op amps. LM324/224, MC33174. In all cases, I'm
>getting an interesting current-based "crossover distortion".
>
>It's like this....measure the voltage with respect to ground (4V) from
>each of the two legs of a capacitor, and of course there's a phase
>difference. Right at the crest of the voltage across the cap, the
>current flow reverses. If this cap is being fed by the op amp (even
>through like a 2.2k ohm resistor", this means the op amp has to
>reverse its direction of output current from "sink" to "source" or
>vice versa. This shows up as a glitch in the output voltage waveform:
>the more gain I've got going on the worse the glitch. And...it's
>audible. The location of the glitch varies with frequency, but not
>amplitude - that is, it's at the same phase relative to the crossover.
>Two glitches per cycle of the sine wave, for each time in the cycle
>the capacitor current reverses. The only way I have been able to get
>rid of this is by using *extremely* high series resistors in the
>circuits - that is, the output of the op amp must have a minimum of
>about 47k ohms before any capacitor. I am not intending to feed any
>capacitor directly, but in fact, if I have a unity-gain buffer feeding
>a Sallen-Key highpass, then it is feeding a capacitive input...and if
>I stick that 47k resistor in to isolate, the gain gets uselessly low.
>
>Look at EMRFD Fig 3.58. The center op amp of the top three, for
>instance, has a 0.047uF cap from output to neg input. I've got
>exactly that circuit in my rig-up, and am getting the crossover
>distortion.
>
>I've added a set of scope photos showing this "glitch" in the output
>of the op amp....four of them, in the folder entitled "FM_Scope_photos".
>
>So, the questions are thus:
>
>1) Has anybody else seen this? Percentage-wise, the distortion
>percentage is low, but it's quite audible and worse, during that
>glitch period, the circuit is momentarily RF sensitive - not a good
>thing to have around a transmitter.
>
>2) Is there a better op amp to consider for this application?
>
>I'm thinking that the theoretical explanation would fall into
>something having to do with the natural pole in the op amp and perhaps
>all I need is a higher speed op amp that can handle the
>current-direction switch better.
>
>However, EMRFD is loaded with example circuits that look like mine,
>and it is written that one can use a 741 or 1458 - both of which I've
>tried in some of my proto circuits, with exactly the results I've
>shown in the photos.
>
>Thanks for any help!
>
>FM
>
>
1599 2008-04-25 22:22:33 Russell Shaw Re: Op amps and crossover distortion with RC loads
Frank wrote:
> I've built a set of cascaded op-amp based active audio filters.
> Running from 8V and ground, I used one of the quad op amp sections to
> create 4V for the "center" voltage.
>
> They're Sallen-Key networks, effectively a 300Hz highpass followed by
> a 3kHz lowpass. Thing is, with S-K networks, that means the output
> pin of an op amp is feeding a capacitor, that in some cases, goes
> right to the negative input, which is at virtual ground.
>
> I'm using low cost op amps. LM324/224, MC33174. In all cases, I'm
> getting an interesting current-based "crossover distortion".
>
> It's like this....measure the voltage with respect to ground (4V) from
> each of the two legs of a capacitor, and of course there's a phase
> difference. Right at the crest of the voltage across the cap, the
> current flow reverses. If this cap is being fed by the op amp (even
> through like a 2.2k ohm resistor", this means the op amp has to
> reverse its direction of output current from "sink" to "source" or
> vice versa. This shows up as a glitch in the output voltage waveform:
> the more gain I've got going on the worse the glitch. And...it's
> audible. The location of the glitch varies with frequency, but not
> amplitude - that is, it's at the same phase relative to the crossover.
> Two glitches per cycle of the sine wave, for each time in the cycle
> the capacitor current reverses. The only way I have been able to get
> rid of this is by using *extremely* high series resistors in the
> circuits - that is, the output of the op amp must have a minimum of
> about 47k ohms before any capacitor. I am not intending to feed any
> capacitor directly, but in fact, if I have a unity-gain buffer feeding
> a Sallen-Key highpass, then it is feeding a capacitive input...and if
> I stick that 47k resistor in to isolate, the gain gets uselessly low.
>
> Look at EMRFD Fig 3.58. The center op amp of the top three, for
> instance, has a 0.047uF cap from output to neg input. I've got
> exactly that circuit in my rig-up, and am getting the crossover
> distortion.
>
> I've added a set of scope photos showing this "glitch" in the output
> of the op amp....four of them, in the folder entitled "FM_Scope_photos".
>
> So, the questions are thus:
>
> 1) Has anybody else seen this? Percentage-wise, the distortion
> percentage is low, but it's quite audible and worse, during that
> glitch period, the circuit is momentarily RF sensitive - not a good
> thing to have around a transmitter.

That's a very well known effect with the LM358-type output stage.

Most op-amps are properly biased class B where the top and bottom
output transistors are both slightly conducting at crossover, minimizing
crossover dead-band distortion.

The LM358 is a class-A npn darlington, with a single pnp pulldown.

The deadband is three diode drops or ~2V, so you *must* prevent the
current from reversing in AC applications.

Connect a resistor from the output to ground such that the DC current
is slightly larger than the highest peak AC current.

> 2) Is there a better op amp to consider for this application?

The LM358/324 is as good as any opamp when used properly. A class-B
opamp will save some power supply current in the absence of signal.

> I'm thinking that the theoretical explanation would fall into
> something having to do with the natural pole in the op amp and perhaps
> all I need is a higher speed op amp that can handle the
> current-direction switch better.
>
> However, EMRFD is loaded with example circuits that look like mine,
> and it is written that one can use a 741 or 1458 - both of which I've
> tried in some of my proto circuits, with exactly the results I've
> shown in the photos.
>
> Thanks for any help!
>
> FM
1600 2008-04-25 22:32:13 Russell Shaw Re: Op amps and crossover distortion with RC loads
Frank wrote:
> I've built a set of cascaded op-amp based active audio filters.
> Running from 8V and ground, I used one of the quad op amp sections to
> create 4V for the "center" voltage.
>
> They're Sallen-Key networks, effectively a 300Hz highpass followed by
> a 3kHz lowpass. Thing is, with S-K networks, that means the output
> pin of an op amp is feeding a capacitor, that in some cases, goes
> right to the negative input, which is at virtual ground.
>
> I'm using low cost op amps. LM324/224, MC33174. In all cases, I'm
> getting an interesting current-based "crossover distortion".
>
> It's like this....measure the voltage with respect to ground (4V) from
> each of the two legs of a capacitor, and of course there's a phase
> difference. Right at the crest of the voltage across the cap, the
> current flow reverses. If this cap is being fed by the op amp (even
> through like a 2.2k ohm resistor", this means the op amp has to
> reverse its direction of output current from "sink" to "source" or
> vice versa. This shows up as a glitch in the output voltage waveform:
> the more gain I've got going on the worse the glitch. And...it's
> audible. The location of the glitch varies with frequency, but not
> amplitude - that is, it's at the same phase relative to the crossover.
> Two glitches per cycle of the sine wave, for each time in the cycle
> the capacitor current reverses. The only way I have been able to get
> rid of this is by using *extremely* high series resistors in the
> circuits - that is, the output of the op amp must have a minimum of
> about 47k ohms before any capacitor. I am not intending to feed any
> capacitor directly, but in fact, if I have a unity-gain buffer feeding
> a Sallen-Key highpass, then it is feeding a capacitive input...and if
> I stick that 47k resistor in to isolate, the gain gets uselessly low.
>
> Look at EMRFD Fig 3.58. The center op amp of the top three, for
> instance, has a 0.047uF cap from output to neg input. I've got
> exactly that circuit in my rig-up, and am getting the crossover
> distortion.
>
> I've added a set of scope photos showing this "glitch" in the output
> of the op amp....four of them, in the folder entitled "FM_Scope_photos".
>
> So, the questions are thus:
>
> 1) Has anybody else seen this? Percentage-wise, the distortion
> percentage is low, but it's quite audible and worse, during that
> glitch period, the circuit is momentarily RF sensitive - not a good
> thing to have around a transmitter.

That's a very well known effect with the LM358-type output stage.

Most op-amps are properly biased class B where the top and bottom
output transistors are both slightly conducting at crossover, minimizing
crossover dead-band distortion.

The LM358 is a class-A npn darlington, with a single pnp pulldown.

The deadband is three diode drops or ~2V, so you *must* prevent the
current from reversing in AC applications.

Connect a resistor from the output to ground such that the DC current
is slightly larger than the highest peak AC current.

The output stage has a 50uA current sink, so if your AC load is less
than 50uApk, you don't need the extra resistor.

> 2) Is there a better op amp to consider for this application?

The LM358/324 is as good as any opamp when used properly. A class-B
opamp will save some power supply current in the absence of signal.

> I'm thinking that the theoretical explanation would fall into
> something having to do with the natural pole in the op amp and perhaps
> all I need is a higher speed op amp that can handle the
> current-direction switch better.
>
> However, EMRFD is loaded with example circuits that look like mine,
> and it is written that one can use a 741 or 1458 - both of which I've
> tried in some of my proto circuits, with exactly the results I've
> shown in the photos.
>
> Thanks for any help!
>
> FM
1601 2008-04-25 23:34:14 Frank Re: Op amps and crossover distortion with RC loads
1602 2008-04-26 05:37:35 Allison Parent Re: Op amps and crossover distortion with RC loads
1603 2008-04-26 13:15:18 Harold Smith Re: Op amps and crossover distortion with RC loads
The MC3403 quad opamp is the same as the LM324 except that it has a class-A
output stage which is free of crossover distortion. There is also a dual
equivalent to the LM358, but I don't remember what its part number is. A
3403 will drop into the 324 circuit with no changes and work the same, but
without crossover distortion. I don't know what the current drain will be,
but the increse will be less than that obtained with the 1k pulldown.

de KE6TI, Harold

On Sat, Apr 26, 2008 at 8:37 AM, Allis
1625 2008-04-29 18:23:03 jr_dakota Re: Op amps and crossover distortion with RC loads
Now you know why I bang my head against my workstation and mumble
incoherently whenever I see an audio frequency circuit with a LM324 or
LM358 .....

As others have mentioned there is a crossover distortion problem which
was unavoidable with the technology back then to make it run on such a
low (well for those days) 5V single supply. It wasn't designed for
audio, it was designed to interface analog circuits/sensors to TTL 5V
digital circuits. It's a good DC amp for transducers in 5V digital
circuits which is what it was mainly intended for ... Actually it was
groundbreaking when it came out and ushered in the age of low voltage
op amps

Analog Devices has a free PDF book by op amp guru Walter Jung that has
a chapter on the history of op amps including a bit