LM324 and crossover distortion
Mark
there was a brief discussion in the Wein bridge osc thread about
crossover distorion in the LM324.
Someone mentioned preloading by tieing a 10K betwennt the output and
ground.
I got to thinking abut this in some detail.
Yes a 10K preload will get rid of cross over distortion if the output
is lightly loaded.
But if the output is AC coupled and NOT lightly loaded, I'm thinking
that the preload doesn't really "get rid" of the crossover distortion,
it just shifts it to another voltage.
I woud think if the output is AC coupled and you want to guarantee no
corssover distortion anywhere in the waveform, the pre-load would have
pull at least as much current as the load, making it in effect a class
A stage.
Commnets..
Mark
Phil Hobbs
You're right, the whole point of the trick is to make the output class
A. Whether that's a good idea depends on what the load is. It helps
that the 324 is usually used in single-supply applications, where the
load is generally referred to the negative rail.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
John Larkin
Tim Wescott
Yes, the point is to make it a class A stage. I'm not sure about the
LM324, but on the LM358 you can use a pull-up resistor that can not only
make the stage class A, but will let the op-amp output go up to the
positive rail instead of 1.5-3V short of it.
Just about any op-amp is going to have crossover distortion. The
question isn't "is it there" but "how bad is it". Just run a sine wave
through an amplifier using an older op-amp and look at the output --
there'll be a flat spot in the middle, that gets more pronounced as the
frequency goes up. That's from the crossover distortion of the output
stage interacting with the thing's internal slew rate limit.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html
Wimpie
Hello Mark,
I agree on this. As you can guarantee that only the upper or lower
output transistor is conducting, there is no cross-over distortion
anymore.
Best regards,
Wim
PA3DJS
www.tetech.nl
cassiope
Even shifting crossover to a higher voltage can be an improvement
depending
on the kind of signals being processed. If the signals have a varying
amplitude
about 0V, crossover will bite you every "cycle". If the crossover
distortion is
shifted to some other voltage, preferably well towards the rail, it
will bite you much
less frequently for many signals. And in proportion to signal level,
the distortion
is reduced.
But I'm with John L, the 324 family sucks - get a better opamp. The
technique of shifting
the crossover can reduce distortion even with a better opamp, but
unless you're scrutinizing
every penny there are many better choices out there.
Jim Thompson
Cassiope,
Why don't you produce a list of "better" OpAmps and their cost per
OpAmp?
...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Remember: Once you go over the hill, you pick up speed
ChrisQ
>> Even shifting crossover to a higher voltage can be an improvement
>> depending
>> on the kind of signals being processed. If the signals have a varying
>> amplitude
>> about 0V, crossover will bite you every "cycle". If the crossover
>> distortion is
>> shifted to some other voltage, preferably well towards the rail, it
>> will bite you much
>> less frequently for many signals. And in proportion to signal level,
>> the distortion
>> is reduced.
>>
>> But I'm with John L, the 324 family sucks - get a better opamp. The
>> technique of shifting
>> the crossover can reduce distortion even with a better opamp, but
>> unless you're scrutinizing
>> every penny there are many better choices out there.
>
> Cassiope,
>
> Why don't you produce a list of "better" OpAmps and their cost per
> OpAmp?
>
> ...Jim Thompson
The '324 must have been one of the first quad op amps around, mid
seventies ?, We were quite excited by it's arrival at the time. Turned
out to be too noisy for audio work, but plenty good enough for a myriad
of other applications. The little bit later Raytheon 4136's and similar
were better, but the 324 was still a major milestone in the art, imo.
Regards,
Chris
ChrisQ
>
> The '324 must have been one of the first quad op amps around, mid
> seventies ?, We were quite excited by it's arrival at the time. Turned
> out to be too noisy for audio work, but plenty good enough for a myriad
> of other applications. The little bit later Raytheon 4136's and similar
> were better, but the 324 was still a major milestone in the art, imo.
>
> Regards,
>
> Chris
Not forgetting the ubiquitous LM339 quad comparator, probably around the
same time and still using in designs now. Not bad at all for a 35+ year
old design. Dirt cheap as well...
Regards,
Chris
Bitrex
In my experiments the problem with the LM324 for audio is also its slew
rate - with a slew rate of 0.3 V/usec the maximum amplitude of a sine
wave at 20kHz would be about 1.5 volts before running into slew rate
limitations, and at the same amplitude a square wave will stop looking
like a square wave at a much lower frequency.
Bitrex
I'm sorry, I did the math wrong - that should be 15 volts! But the
problem with square waves still occurs!
Bitrex
Sorry, I did the math wrong - that should be 15 volts P2P! But the
Jim Thompson
I have buckets of both in my parts bin :-)
Also loads of TL08x, which had become my favorite jelly bean OpAmp by
around 1980.
Jim Thompson
Neeerp !-) Keep trying...
S = Omega*VP
So VP = S/Omega
20kHz and 0.3V/us => VP = 2.387V, or Vp-p = 4.7746V
My very first OpAmp design, MC1530/31, circa 1963-64, had 6V/us
slew-rate :-) But it required external compensation.
And it's still being sold, 48 years later...
http://www.lansdale.com/part_search.php?search=MC1530&Go.x=0&Go.y=0
I've even had inquiries to re-design it on a modern process :-)
Bitrex
> On Thu, 31 Mar 2011 19:28:17 +0100, ChrisQ<me...@devnull.com> wrote:
>
>> ChrisQ wrote:
>>
>>>
>>> The '324 must have been one of the first quad op amps around, mid
>>> seventies ?, We were quite excited by it's arrival at the time. Turned
>>> out to be too noisy for audio work, but plenty good enough for a myriad
>>> of other applications. The little bit later Raytheon 4136's and similar
>>> were better, but the 324 was still a major milestone in the art, imo.
>>>
>>> Regards,
>>>
>>> Chris
>>
>> Not forgetting the ubiquitous LM339 quad comparator, probably around the
>> same time and still using in designs now. Not bad at all for a 35+ year
>> old design. Dirt cheap as well...
>>
>> Regards,
>>
>> Chris
>
> I have buckets of both in my parts bin :-)
>
> Also loads of TL08x, which had become my favorite jelly bean OpAmp by
> around 1980.
>
> ...Jim Thompson
What do you think of the 4558? It seems a lot of consumer audio
equipment uses that one.
Jim Thompson
Never had occasion to use it. I stopped doing any significant number
of discrete designs by ~1990 and instead specialized in custom
mixed-signal chip design.
Bitrex
Whoops! I forgot the factor of 2*pi, and that it should be the peak
amplitude, rather than the peak to peak amplitude. I got 6 hours of
sleep last night, that's my excuse.
>
> My very first OpAmp design, MC1530/31, circa 1963-64, had 6V/us
> slew-rate :-) But it required external compensation.
Is the LM324's slew rate so poor mainly because it's designed to be
low power?
Jim Thompson
<bit...@de.lete.earthlink.net> wrote:
So? That's about my norm... us old farts don't need much sleep ;-)
>
>>
>> My very first OpAmp design, MC1530/31, circa 1963-64, had 6V/us
>> slew-rate :-) But it required external compensation.
>
> Is the LM324's slew rate so poor mainly because it's designed to be
>low power?
Partially. It's a complex function of first stage gm, the use of
"pole-splitting" compensation (only two stages of gain), etc. My
gimmick in the MC1530/31 was the use of a ZERO (lead) in the
compensation.
>
>>
>> And it's still being sold, 48 years later...
>>
>> http://www.lansdale.com/part_search.php?search=MC1530&Go.x=0&Go.y=0
>>
>> I've even had inquiries to re-design it on a modern process :-)
>>
>> ...Jim Thompson
Jim Thompson
Forgot another bragging point for the MC1530/31... sliding-class-A
output stage... no cross-over distortion ;-)
John - KD5YI
> Just about any op-amp is going to have crossover distortion. The
> question isn't "is it there" but "how bad is it". Just run a sine wave
> through an amplifier using an older op-amp and look at the output --
> there'll be a flat spot in the middle, that gets more pronounced as the
> frequency goes up. That's from the crossover distortion of the output
> stage interacting with the thing's internal slew rate limit.
>
Tim -
Is that true for the TL072 whose specs say .003% THD?
John
whit3rd
> On 3/31/2011 11:34 AM, Tim Wescott wrote:
>
> > Just about any op-amp is going to have crossover distortion.
> Is that true for the TL072 whose specs say .003% THD?
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.
Bitrex
I found your PDF on it here:
http://www.analog-innovations.com/SED/MC1530-TeachingExercise.pdf
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!
John - KD5YI
> On Thursday, March 31, 2011 1:11:01 PM UTC-7, John - KD5YI wrote:
>> On 3/31/2011 11:34 AM, Tim Wescott wrote:
>>
>>> Just about any op-amp is going to have crossover distortion.
>
>> Is that true for the TL072 whose specs say .003% THD?
>
> 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.
Interesting. This gives me some insight into why there are so many
varieties of op-amps available today.
> 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.
Yes. I understand those limitations.
> One of the scariest sources is thermal; there are lots of
> op amp circuits which have terrible distortion at ~1 Hz,
That one is new to me.
Thanks,
John
Jim Thompson
<bit...@de.lete.earthlink.net> wrote:
Those are just current multiplications. Work your way thru all the
math to see how it works.
>
>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!
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
ChrisQ
>
> What do you think of the 4558? It seems a lot of consumer audio
> equipment uses that one.
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
Tim Wescott
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.
John - KD5YI
> On 03/31/2011 01:11 PM, John - KD5YI wrote:
>> On 3/31/2011 11:34 AM, Tim Wescott wrote:
>>
>>> Just about any op-amp is going to have crossover distortion. The
>>> question isn't "is it there" but "how bad is it". Just run a sine wave
>>> through an amplifier using an older op-amp and look at the output --
>>> there'll be a flat spot in the middle, that gets more pronounced as the
>>> frequency goes up. That's from the crossover distortion of the output
>>> stage interacting with the thing's internal slew rate limit.
>>>
>>
>> Tim -
>>
>> Is that true for the TL072 whose specs say .003% THD?
>
> 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.
>
Okay. Thanks.
Bitrex
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.
k...@att.bizzzzzzzzzzzz
<jjla...@highNOTlandTHIStechnologyPART.com> wrote:
>On Thu, 31 Mar 2011 07:08:17 -0700 (PDT), Mark <mako...@yahoo.com>
>wrote:
>
>>Hi,
>>there was a brief discussion in the Wein bridge osc thread about
>>crossover distorion in the LM324.
>>
>>Someone mentioned preloading by tieing a 10K betwennt the output and
>>ground.
>>
>>I got to thinking abut this in some detail.
>>
>>Yes a 10K preload will get rid of cross over distortion if the output
>>is lightly loaded.
>>
>>But if the output is AC coupled and NOT lightly loaded, I'm thinking
>>that the preload doesn't really "get rid" of the crossover distortion,
>>it just shifts it to another voltage.
>>
>>I woud think if the output is AC coupled and you want to guarantee no
>>corssover distortion anywhere in the waveform, the pre-load would have
>>pull at least as much current as the load, making it in effect a class
>>A stage.
>>
>> Commnets..
>>
>>Mark
>
>Get a better opamp!
Indeed. Ignoring Maxim, it's hard to do worse.
Grant
Is there a simple explanation for poles and zeroes in feedback networks
for us mere mortals?
Grant.
Grant
>On 03/31/2011 07:08 AM, Mark wrote:
>> Hi,
>> there was a brief discussion in the Wein bridge osc thread about
>> crossover distorion in the LM324.
>>
>> Someone mentioned preloading by tieing a 10K betwennt the output and
>> ground.
>>
>> I got to thinking abut this in some detail.
>>
>> Yes a 10K preload will get rid of cross over distortion if the output
>> is lightly loaded.
>>
>> But if the output is AC coupled and NOT lightly loaded, I'm thinking
>> that the preload doesn't really "get rid" of the crossover distortion,
>> it just shifts it to another voltage.
>>
>> I woud think if the output is AC coupled and you want to guarantee no
>> corssover distortion anywhere in the waveform, the pre-load would have
>> pull at least as much current as the load, making it in effect a class
>> A stage.
>
>Yes, the point is to make it a class A stage. I'm not sure about the
>LM324, but on the LM358 you can use a pull-up resistor that can not only
>make the stage class A, but will let the op-amp output go up to the
>positive rail instead of 1.5-3V short of it.
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.
>
>Just about any op-amp is going to have crossover distortion. The
>question isn't "is it there" but "how bad is it". Just run a sine wave
>through an amplifier using an older op-amp and look at the output --
>there'll be a flat spot in the middle, that gets more pronounced as the
>frequency goes up. That's from the crossover distortion of the output
>stage interacting with the thing's internal slew rate limit.
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.
Bitrex
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. :)
Jim Thompson
>On Thu, 31 Mar 2011 12:52:38 -0700, Jim Thompson <To-Email-Use-Th...@On-My-Web-Site.com> wrote:
>
>>On Thu, 31 Mar 2011 15:28:35 -0400, Bitrex
>><bit...@de.lete.earthlink.net> wrote:
[snip]
>>>
>>> Is the LM324's slew rate so poor mainly because it's designed to be
>>>low power?
>>
>>Partially. It's a complex function of first stage gm, the use of
>>"pole-splitting" compensation (only two stages of gain), etc. My
>>gimmick in the MC1530/31 was the use of a ZERO (lead) in the
>>compensation.
>
>Is there a simple explanation for poles and zeroes in feedback networks
>for us mere mortals?
>
>Grant.
>>>
[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.
John - KD5YI
> On Fri, 01 Apr 2011 10:19:09 +1100, Grant<o...@grrr.id.au> wrote:
>
>> On Thu, 31 Mar 2011 12:52:38 -0700, Jim Thompson<To-Email-Use-Th...@On-My-Web-Site.com> wrote:
>>
>>> On Thu, 31 Mar 2011 15:28:35 -0400, Bitrex
>>> <bit...@de.lete.earthlink.net> wrote:
> [snip]
>>>>
>>>> Is the LM324's slew rate so poor mainly because it's designed to be
>>>> low power?
>>>
>>> Partially. It's a complex function of first stage gm, the use of
>>> "pole-splitting" compensation (only two stages of gain), etc. My
>>> gimmick in the MC1530/31 was the use of a ZERO (lead) in the
>>> compensation.
>>
>> Is there a simple explanation for poles and zeroes in feedback networks
>> for us mere mortals?
>>
>> Grant.
>>>>
> [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
Jim, do you mean zeroes cause increasing phase lead?
John