I am wanting to retrofit LM4562s into my old Dolby 363 units, which currently use noisy, slow, and dirty 4558s. I will, of course, put 100nF caps between the power rails on each opamp. I was also going to put 100pF caps across the feedback resistors in order to curb any possible oscillations. Any downside to doing this?
Bandwidth limitations if the feedback resistors are large, but you probably already thought of that...
100pF is rather large for this, 10pF to 22pF ought to tame any tendancy to oscillate.
Are you sure LM4562 is an improvement for noise? Voltage noise isn't the only kind (I can't see a current noise spec for the 4558, but the LM4562 isn't great for current noise at 1.6pA, 4 times worse than the NE5534A, 2.3 times worse than the NE5532...
With circuitry using values around 10k the current noise and Johnson noise will dominate over opamp voltage noise, if the circuit uses impedances below 1k, then the opamp's voltage noise will tend to be the limiting factor. In between all three factors may be contributing.
And the elephant in the room is the issue with LM4562 popcorn noise in some chips...
Are you sure LM4562 is an improvement for noise? Voltage noise isn't the only kind (I can't see a current noise spec for the 4558, but the LM4562 isn't great for current noise at 1.6pA, 4 times worse than the NE5534A, 2.3 times worse than the NE5532...
With circuitry using values around 10k the current noise and Johnson noise will dominate over opamp voltage noise, if the circuit uses impedances below 1k, then the opamp's voltage noise will tend to be the limiting factor. In between all three factors may be contributing.
And the elephant in the room is the issue with LM4562 popcorn noise in some chips...
That elephant was already put to bed with numerous reports of new LM4562s being just fine. Besides, this is a line-level circuit (~ 1.23v rms) where any noise is likely ≥100db down and thus inaudible.
Well, that's interesting because Jim Williams, in his original 363 modification article, used 10 to 22 pf around the NE5532 and LM833 opamps he was suggesting....of course that was before the advent of the LM4562. The 4562 is also a 55Mhz chip (compared to the 5534's 10MHz), and I was thinking that perhaps that might necessitate more aggressive feedback---eh?
Well at higher frequencies smaller capacitances are needed for the same impedance so that doesn't square.
I think the main issue is ensuring that any stray capacitances that could lead to instability are overwhelmed by the direct feedback cap, ensuring there's enough phase margin (for instance capacitance to ground could erode phase margin). Stray capacitances are not frequency dependent, they are layout dependent. That suggests a pretty standard value would fit most situations.
Yes, I see your point. But let's say that you have an existing circuit where you can't change the layout and are unsure of its stray capacitance The higher-bandwidth chip is certainly more PRONE to oscillation than the slower one, so wouldn't the extra HF rolloff be safer? I don't see how 100pF would affect performance in the audio band.
Your assertion that the higher bandwidth chip is more prone to oscillation isn't necessarily true, it all depends on the phase margin at the unity gain intercept.
I upgraded to LME49720 in some key locations of the 5050 MKIII-2 and improved the supply decoupling. I did not make any other changes, I did check for oscillation and didn't see any.
In design validation and debug we usually change just 1 thing at a time in the event that the result is not what is anticipated it is much easier to determine what change caused the problem and that's what I would suggest here. Try a step by step approach on a single module (assuming this is a MKIII-8 or similar) and once you are happy with the performance you can clone everything on all of the others.
I upgraded to LME49720 in some key locations of the 5050 MKIII-2 and improved the supply decoupling. I did not make any other changes, I did check for oscillation and didn't see any.
In design validation and debug we usually change just 1 thing at a time in the event that the result is not what is anticipated it is much easier to determine what change caused the problem and that's what I would suggest here. Try a step by step approach on a single module (assuming this is a MKIII-8 or similar) and once you are happy with the performance you can clone everything on all of the others.
I put the bypass ceramic 33 pf capacitor on top of the feedback resistor when I changed 4558 to 33078. I just bent hooks in the leads of the new capacitor, hooked them over the old leads, add solder. Had it not stopped the oscillation, I could have sweated it off and put on a 68 pf or 100 pf. But 33 pf did stop it.
The local power supply cap, I drilled #46 holes through the board next to the +-12 rails to the op amp. I scraped the coating off the pcb land. Then I bent the ceramic .1 uf cap leads over, and soldered them to the lands. I only used 1 cap for 2 33078 op amp packages, that was enough.
4562 is not THAT much faster than 33078.
I installed both caps at one time. Taking the case off and on & flipping the board over is time consuming enough I didn't want to do half the job then check it to see if I'd gone far enough.
OK, you got me there---I don't understand that. I don't see any specification like that on the datasheet. Please explain.
Thanks---that's good advice. If you'll re-read my original post, you'll see that this is for my Dolby 363. And, just for the record, there aren't any resistors higher than 20KΩ in this signal path, so I don't think there will be any affect on audio frequency response here regardless of the feedback capacitor value.
I meant, of course, that the values between 10pF and 100pF as a feedback capacitor won't affect audio band frequency response.
Hmmmmm...I don't get it---it looks to me that the "intersect" point on that graph is at about 85°.......
It's about the same as the datasheet for the LM4562. So does that mean that neither will oscillate?
Well, the LM4562 IS the LME49720 (changeover of part code).
70-85° of phase margin gives you a healthy buffer for nonspecified effects (like parasitics, reactive loads) before you'll hit oscillations.
Please read this to give you some greater context: http://www.ti.com/lit/an/slyt087/slyt087.pdf
70-85° of phase margin gives you a healthy buffer for nonspecified effects (like parasitics, reactive loads) before you'll hit oscillations.
Please read this to give you some greater context: http://www.ti.com/lit/an/slyt087/slyt087.pdf
Well, I'm not so sure about that:
IC Slew Rate Gain Bandwidth Power Bandwidth
33078 7v/µSec 16 MHz 120KHz
LM4562 20v/µSec 55 MHz 10MHz
Anyway, the T.I. article on Op Amp Stability (thanks DPH!) says:
"The value of the feedback capacitor should be just large enough to achieve the desired overshoot response, because larger values cause a loss of high-frequency performance......Optimal performance is gained when the acceptable over-shoot is obtained at an acceptable frequency response."
In my application (Dolby 363 upgrade), the feedback resistor varies from ~6.5KΩ to 20KΩ. So, if replacing the ICs with LM4562s, what is the optimum value?
That's assuming that you want a large bandwidth. Do you care whether an audio amplifier has a cut-off frequency of 1 MHz or of 20 MHz?
Show me an amp with 0.001% THD at 20kHz which isn't high bandwidth though
(discounting passive filtering before or after the feedback loop).
(discounting passive filtering before or after the feedback loop).