Opamps power bandwidth

Hi!

Is power bandwidth of an opamp, an important design desicion?

Can an opamp start oscillating above this power bandwidth?

I use mc33078. The datasheet says It has 120khz power bandwidth@ vout: 27vpp/RL:2k
Thd: 1%.

Thanks for your advise!
Jay x

More is generally better but not really necessary. In normal music there is a significantly lower amount and level of top octave content. Historically FM and magnetic tape use equalization to boost and then restore HF content to original levels, reducing HF hiss at the same time.

Back to your question, full 20 kHz power bandwidth is not really necessary, and different designs behave differently. Just like circuits become less linear and distortion increases near clipping, distortion increases as we approach slew rate limiting too. I have never seen this specified in a data sheet and most modern high performance IC op amps deliver adequate slew rate for full power bandwidth. Back in the 80s I thought about a test to characterize this, but abandoned it as modern IC op amps got so much better than needed for relatively slow moving audio.

Another subtle point involves how amplifiers slew limit.  In conservative design (IMO) the amplifier topology nicely LPF the input before any input signal can possibly slew limit (analogous to clipping in the time domain). Trying to be all things to all people, some amp makers would defeat the input stage LPF to deliver a higher slew rate on the test bench.

  JR

Hi John,

Thanks for your comments.

I asked this question, because i am worried about the not so good frequency response of my last project.

With a sine sweep test Signal @-0.5dbu,  (aprox),i got:

20hz : 0.318dbu
50hz: -0.314dbu
100hz: -0.425dbu
1k: -0.502dbu
5k: -0.894dbu
10k: -1.784dbu
20k: -3.427dbu

TrueRTA  20hz-20khz filter on.

Other pieces of equipment i tested, showed better frequency response at 20khz.
(+/-0.5dbu).

This is why I think there is something wrong in my design...
  ???

Jay x

> With a sine sweep test Signal @-0.5dbu

That's not real "power". It is 1/10th of what the chip is rated, and at 1/5th the rated frequency. Power Bandwidth is not an important spec here.

What gain? If you try to get gain of 55dB out of that chip, the 16MHz gain-bandwidth product will give -3dB near 20KHz.

Otherwise: question everything else in your circuit and test rig.

Hi!

My test signal was around 775mv (0dbu). I explained myself badly, sorry.  :

Ok, i see the datasheet, so this is not the problem.

My problem is another.  I have a roll-off at 20khz. It is not normal. I expected -0.5dbu or -1dbu. not more.
The closed loop of my opamp stages is around 110khz. I may have a filter some where...

jay x

And the data of the frequency response..

And the schematic.

With the values of the schematic, in theory i should not have that roll-off at 20khz...¿right?

Or i could try to change the feedback capacitor to increase the bandwidth to around 300khz for each stage...but normally this is not necessary...

jay x

JAY X said:

With the values of the schematic, in theory i should not have that roll-off at 20khz...¿right?

Or i could try to change the feedback capacitor to increase the bandwidth to around 300khz for each stage...but normally this is not necessary...

jay x

Click to expand...

from some quick ciphering I see two feedback poles cascaded in series, both tuned to be half power (-3dB) at roughly 100kHz...

In series they will add to -6dB at 100kHz and -2dB at 50kHz.

This will probably sound OK but is lower than your 300kHz target.

JR

PS; Of course check my math

John's math is correct, also, the attenuation at 20KHz is -0.34dB, is that relevant to you?

Curve in #4 seems to show -3.5dB @ 20KHz, not -0.34dV.

I'm also troubled by +1dB at 20Hz; no musical effect but it shouldn't do that. System error? DC-cancellation in ADC off-kilter?  {EDIT} Or is this the effect of C31 C32 in the DRV134 NFB?

PRR said:

Curve in #4 seems to show -3.5dB @ 20KHz, not -0.34dV.

Click to expand...

Yes according to the curve the cut-off seems more abrupt, I was merely calculating the attenuation of a 2-pole cascaded low pass, which with a -6dB attenuation at 100KHZ produces -0.34dB at 20KHz. John Roberts seems to do the same thing.

To the OP : Is the curve taken from the same circuit you posted?

Anti-aliasing filter in AD used to measure? Or some other form of band limiting filter in measuring device?

Hi All!

I finally discovered the problem: !PHASE¡

Yesterday I went to visit some friends , they had  good scopes and signal generators. We discovered that each stage in my circuit is slightly out of phase in some degrees: 5-10º, and this adds up in each stage.
This is why I have this frequency response, as if the low end made the high frequencies disappear a bit..
I changed some feedback capacitor values for smaller ones, and that improved the phase response of the circuit.

The circuit i posted is not the full circuit, because I wrongly thought my problem was in the final stages... but no.. :

Lessons learned.: :

1. Get a proper scope.... : Pc soundcard scopes...not so useful...
2. To have a large but not too large input bandwidth.
3. Mantaining  in each circuit the same Resistor/capacitor Ratios.
4. It is also about experimenting with resistor/caps values...

I expect to fix all of this in the next  week or so, i will post a new frequency response graph ASAP.

Thank you very much for all your comments!

jay x

JAY X said:

Hi All!

I finally discovered the problem: !PHASE¡

Yesterday I went to visit some friends , they had  good scopes and signal generators. We discovered that each stage in my circuit is slightly out of phase in some degrees: 5-10º, and this adds up in each stage.
This is why I have this frequency response, as if the low end made the high frequencies disappear a bit..
I changed some feedback capacitor values for smaller ones, and that improved the phase response of the circuit.

Click to expand...

Your circuit is minimum-phase; as such, it's frequency response and phase response are bi-univocally related (in other words, one is a transform of the other). You cannot single out phase as the issue. Basically, you have a global frequency response issue, part of it being probably in the digital measurement chain.
The measurement screen shows you used a 44.1kHz sample rate, so your 20kHz measurement point is too close to the Nyquist frequency. You see the attenuation due to the anti-alias and reconstruction filters, resp. in the A/D and D/A converters. In other words, you're measuring your DAW's performance!
For proper evaluation at 20kHz, you need minimum 48kHz SR; to be safe, use Double Speed (88.2/96k).