Op amp bandwidth, and is there an affordable fast op amp ?

[michael_967]

Hi All, I have been tinkering with op-amp voltage followers, adding DC offset and the like, experimenting to eventually try making a function generator, just for fun and to learn.
The datasheets all have a "gain bandwidth" but my tests always show distortions way below the advertised bandwidth.

Here is how I test:
* Used my scope's function generator to output a sine wave.
* Configured the op-amp as a voltage follower (buffer): direct connection between the output and the inverted input, function generator signal connected to the non-inverting input.
* I used +12V/-12V for the op-amp supply
* compared on my scope the wave at the output with the original wave from the generator.
* Increased the frequency until I saw  distortions of the output signal.
* I also changed the amplitude of the original signal to see if it has any impact. It does, a lower amplitude signal, like less than 1V peak-to-peak, fares better at high frequencies, which makes sense.

I tried first with an LM741CN, and I started seeing distortions at about 50kHz
I then used an LF356N which has a "wide gain bandwidth" of supposedly 2MHz, but I could not get above 100KHz-150KHz before getting distortions.

Here are my questions:
-Why are the distortions starting well below the advertised "gain bandwidth" on the datasheet? Most likely I am not interpreting this data correctly. Is there any other parameter on the datasheet that are more meaningful ?
-Is there any affordable (couple bucks) op-amp which would give reasonably good results at up to 1MHz ? I mean to say that a 1MHz sine wave signal at the input would be output without any noticeable distortion.

Thank you very much for your insights!
Mike

[Kleinstein]

OP-amps get good linearity from extra loop gain that can correct linearity errors. So one gnerally needs high  GBW than just circuit gain times frequency.
A major source of INL in the open loop case is the cross over of the output stage. So it helps if there is no cross over and the OP only provides current of one polarity. This way one can reduce the distortion with little loop gain left.

Another point to whatch for the the slew rate limit. With a large amplitude the slow rate can be to high. This is especially true for BJT based OP. FET based OPs tend to have a higher slew rate at the same GBW.

For low cost, fast OPs, there are quite a few audio OPs, like NE5534 or the dual NE5532/LM833/MC33078.
There are few fast audio ones, but currently the availability can be tricky. So best look at your source and use the search function on the available parts there.

With much more then 10 MHz it starts to get tricky on a bread board and a prototype PCB starts to need some care.

[TimFox]

I think the distortion you see is a function of another variable, the "slew rate", which is the maximum derivative dV(t)/dt.

A 6 V pk-pk sine wave at 50 kHz has a maximum slew rate (passing through zero) of 2pi x f x V_pk = 0.9 V/us.
The 741 has a specified typical slew rate of only 0.5 V/us.

Note that in a typical op amp, the slew rate is determined by the maximum current of the input differential stage and the compensation capacitor across the next (current-to-voltage) stage.
Therefore, as you approach the slew rate (by increasing frequency or voltage swing), you approach the maximum output current of the first (non-linear) stage, and therefore that stage starts to distort.
When you hit the slew rate, the feedback in the circuit is "switched off", since the poor mistreated amplifier is already doing its best, and the distortion gets out of hand.

[Avelino Sampaio]

From what I understand, based on the formula presented by TimFox, the maximum frequency of the LF356 for a voltage of 10vp is: 191khz (12/(2pi*10)). This week I ordered the LM4562 from LCSC, which is geared towards audio. It has a sr=20v/us and BW=55MHz. To reach 1MHZ you will not be able to go beyond 3vp. (  20/(2pi*3vp) = 1.06MHZ  ).

[David Hess]

-Why are the distortions starting well below the advertised "gain bandwidth" on the datasheet? Most likely I am not interpreting this data correctly. Is there any other parameter on the datasheet that are more meaningful ?

You are seeing distortion from the slew rate limit.  Look for a specification like "full power bandwidth" to take this into account.

-Is there any affordable (couple bucks) op-amp which would give reasonably good results at up to 1MHz ? I mean to say that a 1MHz sine wave signal at the input would be output without any noticeable distortion.

Sure, there are lots of them, but what you primarily need is an operational amplifier with a higher slew rate, and these do exist.  125 volts per microsecond is going to be required for 1 MHz at 20 volts peak-to-peak.

The AD817 meets those requirements, however a slower and more expensive part like the LT1468 will likely have better distortion.  I limited my search to 8-pin DIP parts which can operate on +/-15 volt supplies.

[TimFox]

From what I understand, based on the formula presented by TimFox, the maximum frequency of the LF356 for a voltage of 10vp is: 191khz (12/(2pi*10)). This week I ordered the LM4562 from LCSC, which is geared towards audio. It has a sr=20v/us and BW=55MHz. To reach 1MHZ you will not be able to go beyond 3vp. (  20/(2pi*3vp) = 1.06MHZ  ).

A rule of thumb, from W Jung in various audio articles, is that the (output) slew rate (in V/us) should be at least (0.5 to 1.0) X (peak value of output voltage), for audio (<20 kHz) operation with low distortion. 
This, of course, is greater than the theoretical slew rate of a sine wave, for reasons including what I mentioned above about open-loop non-linearity of the input stage. 
For audio amplifiers (op amps or power amps), this implies SR > (5 to 10) V/us for a 10 V peak (7 V rms) output voltage, while the maximum mathematical slew rate of a 10 V peak sine wave at 20 kHz is only 1.26 V/us.

[michael_967]

Thank you all guys, I learned something new   

[David Hess]

Almost from the earliest days of the operational amplifier, a circuit like that shown below was used to boost slew rate.  Fairchild showed it several times in their application notes in the 1970s, and Tektronix used the same idea in their 7A29 vertical amplifier (1979?) for reasons which are not clear to me; maybe the engineer was having an attack of the clevers.  It allows slow 741 and 308 class amplifiers to be used in audio applications.  This particular variation is from Walter Jung.

[MrAl]

Hi All, I have been tinkering with op-amp voltage followers, adding DC offset and the like, experimenting to eventually try making a function generator, just for fun and to learn.
The datasheets all have a "gain bandwidth" but my tests always show distortions way below the advertised bandwidth.

Here is how I test:
* Used my scope's function generator to output a sine wave.
* Configured the op-amp as a voltage follower (buffer): direct connection between the output and the inverted input, function generator signal connected to the non-inverting input.
* I used +12V/-12V for the op-amp supply
* compared on my scope the wave at the output with the original wave from the generator.
* Increased the frequency until I saw  distortions of the output signal.
* I also changed the amplitude of the original signal to see if it has any impact. It does, a lower amplitude signal, like less than 1V peak-to-peak, fares better at high frequencies, which makes sense.

I tried first with an LM741CN, and I started seeing distortions at about 50kHz
I then used an LF356N which has a "wide gain bandwidth" of supposedly 2MHz, but I could not get above 100KHz-150KHz before getting distortions.

Here are my questions:
-Why are the distortions starting well below the advertised "gain bandwidth" on the datasheet? Most likely I am not interpreting this data correctly. Is there any other parameter on the datasheet that are more meaningful ?
-Is there any affordable (couple bucks) op-amp which would give reasonably good results at up to 1MHz ? I mean to say that a 1MHz sine wave signal at the input would be output without any noticeable distortion.

Thank you very much for your insights!
Mike

Hi there,

Op amps have two important specifications when looking for reasonable bandwidth. That is because the "power bandwidth" is dependent on the output level as well as the gain bandwidth factor.  This means that you can not say that a particular op amp does not work at some frequency like 50kHz because that's not enough information.  We also need to know the output level.  Assuming a sine wave for testing, that means we need to know the peak of the output.  For example, a 1MHz op amp may work fine at 20kHz with only 1 volt peak output sine wave, but at 10 volts peak output sine wave it may become very very distorted.  The reason is that the gain bandwidth factor is really like a small signal specification, and internally the op amp output stage has a limited response time usually referred to as the "slew rate".
The slew rate is the ramp time for the output to reach a certain level given a certain input condition that would cause the output to go higher than it was before the input was applied.  For example, the LM358 is often specified as 0.5 microseconds per volt.  Notice the units of volts are in that.  That means that if the input is such that it should force the output to go up by one volt, it would take 0.5 microseconds to reach that point.  Now you know a sine wave is curved, so if the slew rate isnt fast enough, the output can not follow the sine wave.  That causes a very ugly looking output even with a really clean sine wave input.  Obviously you dont want that.
The trick is to calculate the maximum output peak for a sine wave given a certain frequency and the gain bandwidth and the slew rate.  You probably already know that we use GBW to estimate like so:
fmax=GBW/Gain
That's easy enough, but we also need to look at the slew rate.
Since the slew rate is taken to be a ramp, we can compare a ramp to the maximum rate of change of the rise of the test sine wave.  It just so happens that the maximum rate of change is located at zero volts, so it's when the sine wave crosses zero.
To calculate that we can start with a general sine wave:
v=A*sin(w*t)

where A is the peak and w=2*pi*frequency and t is time and v is the voltage of the sine at any time t.
To find the slope which is the rate of change of voltage, we take the derivative of that and get:
dv/dt=A*w*cos(w*t)

Now since the max slope is when the sine wave goes through zero, we set t=0 and we get:
dv/dt=A*w

Writing this out, we have:
dv/dt=A*2*pi*f

and so on the right is the slew rate of a sine wave at the zero crossing.
Now we can solve for f:
f=(dv/dt)/(A*2*pi)

and with dv/dt being the slew rate we can replace that with the simple "sr" and get:
f=sr/(A*2*pi)

and this is with sr in volts per second.
Converting 0.5 volts per microsecond to volts per second, we multiply by 1e6 and get:
sr=0.5*1e6

and inserting that into the formula for f we get:
f=0.5e6/(A*2*pi)

and since A=3 as above we get:
f=0.5e6/(3*2*pi)=0.5e6/(6*pi)

which of course comes out to:
f=26525.82384864922 Hertz.

So we see that the maximum frequency of operation for that op amp is about 26kHz when the output is 3 volts peak.  Of course if we decrease the max output peak we get a higher frequency, and if we increase the max output we get a lower frequency.
That's how the slew rate fits into the calculation for the max frequency of an op amp and you can always solve for the slew rate 'sr' given the frequency you intend to use, then look for an op amp that has that slew rate or better.

[magic]

Sure, there are lots of them, but what you primarily need is an operational amplifier with a higher slew rate, and these do exist.  125 volts per microsecond is going to be required for 1 MHz at 20 volts peak-to-peak.

The AD817 meets those requirements, however a slower and more expensive part like the LT1468 will likely have better distortion.  I limited my search to 8-pin DIP parts which can operate on +/-15 volt supplies.

Another pricey option which meets your criteria is the LT1354~LT1365 series, designed specifically for fast slewing.

BTW, LT146x are slower than 125V/µs, about on par with the much cheaper LM4562.

[radiolistener]

* I used +12V/-12V for the op-amp supply

I tried first with an LM741CN, and I started seeing distortions at about 50kHz
I then used an LF356N which has a "wide gain bandwidth" of supposedly 2MHz, but I could not get above 100KHz-150KHz before getting distortions.

Here are my questions:
-Why are the distortions starting well below the advertised "gain bandwidth" on the datasheet?

It depends on Voltage swing. I think you're needs to see on "Slew rate" specification.

Here is relation between Slew rate and Bandwidth:

Bandwidth (Hz) = Slew rate (V/s) / (2 * pi * Vpk) = Slew rate (V/us) * 1000000 /  (2 * pi * Vpk)

In your case Vpk = 12 V, and:

LM741 Slew rate= 0.5 V/us, so Bandwidth = 0.5 * 1000000 / (2 * pi * Vpk) = 6631 Hz

LFLF356N Slew rate = 7.5 V/us, so Bandwidth = 7.5 * 1000000 / (2 * pi * Vpk) = 99471 Hz

Which well corresponds with your measurements.

If you want more wide bandwidth with output amplitude 12 Vpk, then you're needs to look for op-amp with higher Slew rate:

Slew rate = (2 * pi * Vpk * Bandwidth) [V/s] = (2 * pi * Vpk * Bandwidth)/1000000 [V/us]

For bandwidth 2 MHz at amplitude 12 Vpk it should be at least:

Slew rate = (2 * pi * 12 * 2000000)/1000000 = 151 V/us

or better...

[Avelino Sampaio]

I found an animal (LM7171) with sr=1400v/us and bw=200MHz. For less than $5, at LCSC.

https://www.ti.com/cn/lit/ds/symlink/lm7171.pdf?ts=1659363453842

[David Hess]

Sure, there are lots of them, but what you primarily need is an operational amplifier with a higher slew rate, and these do exist.  125 volts per microsecond is going to be required for 1 MHz at 20 volts peak-to-peak.

The AD817 meets those requirements, however a slower and more expensive part like the LT1468 will likely have better distortion.  I limited my search to 8-pin DIP parts which can operate on +/-15 volt supplies.

Another pricey option which meets your criteria is the LT1354~LT1365 series, designed specifically for fast slewing.

BTW, LT146x are slower than 125V/µs, about on par with the much cheaper LM4562.

A little bit of care is needed because faster parts are often two-stage designs which lack precision yielding more distortion if that is imporant.  The LT1468 is a high precision part but the faster AD817 only has an open loop gain in the 1000s.  The LT1354 is 10,000s.  The LM4562 is intended for low distortion audio applications and has the associated high open loop gain needed for low distortion.

Where gain is required and unity gain stability is not required, decompensation can raise the gain-bandwidth product and slew rate, but parts which support this are rare.  An alternative is to add a fast gain stage, often with local current feedback, within the feedback loop which will multiply the gain-bandwidth product and slew rate, which is what the circuit I showed earlier does in a clever way.

[Vovk_Z]

I found an animal (LM7171) with sr=1400v/us and bw=200MHz.

LM7171 is good, but it has much cheaper but possibly similar (almost as fast) brother LM6172. LM6172 is both fast and cheap. At least it was cheap some years ago when I used it last time. It is almost a jelly bean.

[Zero999]

Faster isn't always better. A fast op-amp needs more effort to stabilise: good decoupling and a careful layout are important.

[mag_therm]

-Is there any affordable (couple bucks) op-amp which would give reasonably good results at up to 1MHz ? I mean to say that a 1MHz sine wave signal at the input would be output without any noticeable distortion.

Mike

I am using AD8055 and MAX 4450/51 at 455 kHz as diff amp and  audio as Sallen Key. They work well.
I think the Max, which is not as fast slew, but single supply with inputs to +/- 0.3V outside rails, has superseded the AD.
These are basically intended for 50 Ohm RF, and the Ri and Rg have limitations.
The Max can be used over Gain of +/- 1 to +/- 25

[magic]

A little bit of care is needed because faster parts are often two-stage designs which lack precision yielding more distortion if that is imporant.  The LT1468 is a high precision part but the faster AD817 only has an open loop gain in the 1000s.  The LT1354 is 10,000s.  The LM4562 is intended for low distortion audio applications and has the associated high open loop gain needed for low distortion.

DC gain doesn't matter at 1MHz

Both LT parts are specified for distortion; 1468 distortion rises more steeply above the audio band and I wouldn't bet that it's still better at 1MHz. Besides, I suspect OP doesn't care about such level of THD if LM741 was good enough until it hit slew limiting.

[michael_967]

I have compiled a list of all the op-amp versions that were mentioned so far in this thread.
This is for my own benefit but I thought I would share it, maybe it can be useful to someone else.
Each op-amp designation has a link to its datasheet.

https://docs.google.com/spreadsheets/d/1Ok1HbdHBmdXHOwoE1VE_a0oBrc59Y_9-zwQK15-6EdA

[MrAl]

I have compiled a list of all the op-amp versions that were mentioned so far in this thread.
This is for my own benefit but I thought I would share it, maybe it can be useful to someone else.
Each op-amp designation has a link to its datasheet.

https://docs.google.com/spreadsheets/d/1Ok1HbdHBmdXHOwoE1VE_a0oBrc59Y_9-zwQK15-6EdA

Hi,

That's a nice list.  I think i was going to use that #6 op amp in a project a few years back but then it was $5 USD.  I needed faster response for a special kind of power supply regulation circuit.

The list is great, but maybe you could add a column for GBW too that would nice to see at a glance.

[Terry Bites]

Given the help this thread has given with determining the specs I'd go to a vendor site like digikey, plug in the numbers and sort buy price and in stock!

[Zero999]

From what I understand, based on the formula presented by TimFox, the maximum frequency of the LF356 for a voltage of 10vp is: 191khz (12/(2pi*10)). This week I ordered the LM4562 from LCSC, which is geared towards audio. It has a sr=20v/us and BW=55MHz. To reach 1MHZ you will not be able to go beyond 3vp. (  20/(2pi*3vp) = 1.06MHZ  ).

A rule of thumb, from W Jung in various audio articles, is that the (output) slew rate (in V/us) should be at least (0.5 to 1.0) X (peak value of output voltage), for audio (<20 kHz) operation with low distortion. 
This, of course, is greater than the theoretical slew rate of a sine wave, for reasons including what I mentioned above about open-loop non-linearity of the input stage. 
For audio amplifiers (op amps or power amps), this implies SR > (5 to 10) V/us for a 10 V peak (7 V rms) output voltage, while the maximum mathematical slew rate of a 10 V peak sine wave at 20 kHz is only 1.26 V/us.

The full power bandwidth of an audio amplifier only has to extend around 1kHz for slew rate to not cause distortion. Large voltage swings are produced by bass frequencies, not treble. An audio amplifier with a peak output voltage of 10V, will never have to output a 20kHz 10V peak sine wave, when playing music or speech. The lower frequency parts of the signal will clip, causing nasty sounding distortion before the high frequency elements approach the slew rate limit.

Almost from the earliest days of the operational amplifier, a circuit like that shown below was used to boost slew rate.  Fairchild showed it several times in their application notes in the 1970s, and Tektronix used the same idea in their 7A29 vertical amplifier (1979?) for reasons which are not clear to me; maybe the engineer was having an attack of the clevers.  It allows slow 741 and 308 class amplifiers to be used in audio applications.  This particular variation is from Walter Jung.

The 741 is fast enough to use in audio applications without any tricks to improve the slew rate. There are other valid reasons not to use it for audio, but speed isn't one of them.

[Terry Bites]

Check out DSL line drivers, they're fast and have massive current output per $.

[exmadscientist]

Check out DSL line drivers, they're fast and have massive current output per $.

DSL and powerline communication drivers are fantastic if you've got the power rails they like and the output headroom to make them happy (they're not RRO, not even close, don't go there). If not, well, you can join me in being sad that the cheap, available parts don't work out.

My go-to fast cheapies are TLV3541 and ADA4891. Nothing too special, but they're cheap for this kind of stuff (and oddly cheap for any ADI products), and they're 5.5V CMOS so they're fast stuff with low input bias currents, which is at least a little uncommon. Plus duals and quads and even a triple for the 4891 are available. Downsides are SMT only (I contend that's actually an upside in this speed range) and they're not actually that fast at 150-ish V/us. Oh, and they're 5.5V CMOS. If you really need more speed, well, there's always the scary current-feedback types or their partially-tamed VFB variants (both easier to use than they look as long as they're nowhere near a DIP or any kind of breadboard)....