It doesn't look like harmonics, unless it's only some quite-high-order mode(s). You might want to try increasing the scope's sweep speed (by a lot), to see if you can 'zoom in' on the higher frequency that is creating the 'thick' look. Knowing its frequency (or frequency components) might be a clue. Hopefully, you will be able to get the scope to trigger on some part of the high-frequency portion, after the sweep speed is set high-enough to allow examining it in more detail.
You might also want to see if you can probe around and find the high frequency anywhere else in the amp.
I guess you could also experiment with putting a small series resistance before the zobel and the load, to see if it goes away. And I suppose you might also get some clues by varying the Zobel's component values. You could even try putting a Zobel network on the opamp's output, just to see what effect it might have.
However, if you haven't yet tried increasing the feedback cap's values, I would at least try those, before changing or adding anything else.
You might also want to see if you can probe around and find the high frequency anywhere else in the amp.
I guess you could also experiment with putting a small series resistance before the zobel and the load, to see if it goes away. And I suppose you might also get some clues by varying the Zobel's component values. You could even try putting a Zobel network on the opamp's output, just to see what effect it might have.
However, if you haven't yet tried increasing the feedback cap's values, I would at least try those, before changing or adding anything else.
At first I thought it was the effect of the transducer, however it's not the case...
With both input and output disconnected, only scope probe on output...
Scope set to 0.2us and 0.5v/div... It's there and quite large, about 0.9 Vpp...
The frequency, if I calc'd right, (1/(0.2us/div x 3div / 2.5lambda)) is roughly 4.167MHz...
Pic included
Thanks!!
With both input and output disconnected, only scope probe on output...
Scope set to 0.2us and 0.5v/div... It's there and quite large, about 0.9 Vpp...
The frequency, if I calc'd right, (1/(0.2us/div x 3div / 2.5lambda)) is roughly 4.167MHz...
Pic included
Thanks!!
Attachments
Allrighty...
I think I understand why good GND layout is important now...
I'm seeing the same oscillation everywhere around the opamp...
It's on the input pin, with the input wire (shielded) disconnected...
It's also between the case ground and the input gnd... And between the gnd ternimal on the board and the input gnd...
For my purposes, a bit of interference isn't a big deal... However, I have read that oscillations can kill an amp, is this correct?
Thanks!!
Michael
I think I understand why good GND layout is important now...
I'm seeing the same oscillation everywhere around the opamp...
It's on the input pin, with the input wire (shielded) disconnected...
It's also between the case ground and the input gnd... And between the gnd ternimal on the board and the input gnd...
For my purposes, a bit of interference isn't a big deal... However, I have read that oscillations can kill an amp, is this correct?
Thanks!!
Michael
I swapped the lm833 for a tl072 and it really cut down the oscillation...
I also noticed when I adjust r6 towards class A, it reduces the oscillations/noise as well...
Now, it seems I'm jsut gettign a bit of noise... The noise decreases if I unplug the sheilded input wire from the board... about 50 mVpp max...
The noise almost completely disappears when I turn on the function generator and have the amplitude down...
Here is a scope pic with transducer attached..... As before, near clip the current phase is a bit messy, but at slightly lower amplitude, its a clean sine...
So, that fixed it.... not a great fix, but it will do for me...
I will certainly have to come up with a resonant driving system... As is, it's quite sensitive to tune and the function generator wanders a bit.. .
However, my next step should be the output transformer...
Thanks!!
I also noticed when I adjust r6 towards class A, it reduces the oscillations/noise as well...
Now, it seems I'm jsut gettign a bit of noise... The noise decreases if I unplug the sheilded input wire from the board... about 50 mVpp max...
The noise almost completely disappears when I turn on the function generator and have the amplitude down...
Here is a scope pic with transducer attached..... As before, near clip the current phase is a bit messy, but at slightly lower amplitude, its a clean sine...
So, that fixed it.... not a great fix, but it will do for me...
I will certainly have to come up with a resonant driving system... As is, it's quite sensitive to tune and the function generator wanders a bit.. .
However, my next step should be the output transformer...
Thanks!!
Attachments
On to the trafo....
I tried a few trafo calcs.... mostly for SMPS design...
They seem to be telling me that I should be using 21 AWG as max. size due to skin effect at 28kHz...
Which also indicates I should use 2 primaries in parallel to get the required current handling ability...
It also seems like a pretty small toroid is required...
I havent found a calc that will do ferrite E/I cores, will keep looking...
I tried a few trafo calcs.... mostly for SMPS design...
They seem to be telling me that I should be using 21 AWG as max. size due to skin effect at 28kHz...
Which also indicates I should use 2 primaries in parallel to get the required current handling ability...
It also seems like a pretty small toroid is required...
I havent found a calc that will do ferrite E/I cores, will keep looking...
Some random thoughts:
SMPS calculations are based on squarewaves. A 28Khz squarewave will have energy out to 800Khz, and strong current at 84Khz. If not running a squarewave, a plain old 18ga will be fine. No exotic core needed either.
Do you have the unused section of the opamp configured as a follower with the input grounded? If left open in will oscillate and modulate the supply rails.
I would try the input jack isolated from the chassis and use a 1:1 transformer.
SMPS calculations are based on squarewaves. A 28Khz squarewave will have energy out to 800Khz, and strong current at 84Khz. If not running a squarewave, a plain old 18ga will be fine. No exotic core needed either.
Do you have the unused section of the opamp configured as a follower with the input grounded? If left open in will oscillate and modulate the supply rails.
I would try the input jack isolated from the chassis and use a 1:1 transformer.
DJK, your random thoughts are great!
I don't have the unused opamp connected at all.. I will do that...
My input jack has a plastic case so it's not groounding on the case, I will pick up a small 1:1 audio tranny next time I get to the market...
Megajock, I bundled the wires to and from the caps, I was thinking there might be some inductance between them... I thinking about shielding them... or at least spreading them out a bit... Any suggestions on the value of the decoupling caps??
Thanks!!
I don't have the unused opamp connected at all.. I will do that...
My input jack has a plastic case so it's not groounding on the case, I will pick up a small 1:1 audio tranny next time I get to the market...
Megajock, I bundled the wires to and from the caps, I was thinking there might be some inductance between them... I thinking about shielding them... or at least spreading them out a bit... Any suggestions on the value of the decoupling caps??
Thanks!!
An electrolytic cap (value?) and a 0.1uF X7R ceramic in parallel, from each power rail to ground, where each power rail enters the board, would probably be good.
I would twist each power supply wire and its ground return wire tightly together, between the power supply and the amp board. That will minimize their 'enclosed loop area', minimizing both radiated and received effects of varying electromagnetic fields. (See: 'Faraday's Law' and/or 'Maxwell's Equations'.) The wires should be perpendicular to the amp board, for at least two inches or so, if possible.
You might also want to add a 10uF-or-so electrolytic in parallel with each of the 0.1 uF decoupling caps, at the opamp.
Good catch, djk! Unused opamps should always be put into a known state, usually with the + input to ground and output directly shorted to - input.
I would twist each power supply wire and its ground return wire tightly together, between the power supply and the amp board. That will minimize their 'enclosed loop area', minimizing both radiated and received effects of varying electromagnetic fields. (See: 'Faraday's Law' and/or 'Maxwell's Equations'.) The wires should be perpendicular to the amp board, for at least two inches or so, if possible.
You might also want to add a 10uF-or-so electrolytic in parallel with each of the 0.1 uF decoupling caps, at the opamp.
Good catch, djk! Unused opamps should always be put into a known state, usually with the + input to ground and output directly shorted to - input.
I put the fans in series, jsut to see if they were the cause... but no change...
However, it's much quieter now and I can hear the toroid humming... it's quite noticable... It is a low frequency hum...
I think I will try removing the power connector... It has some kind of noise filter built in...
Gootee, I will put some caps on there as soon as I find out what vlaue I should use..
Thanks!!
However, it's much quieter now and I can hear the toroid humming... it's quite noticable... It is a low frequency hum...
I think I will try removing the power connector... It has some kind of noise filter built in...
Gootee, I will put some caps on there as soon as I find out what vlaue I should use..
Thanks!!
If everything else is assumed OK, and your opamp is fast-enough, I guess the standard method is to try increasing the value of one or both feedback caps, i.e. C3 and C2 (I think) on the original schematic.
But, are the opamp's power rails oscillating? What type of decoupling caps go from each opamp power pin to gnd? I don't see any decent-sized (and high-ish ESR) output cap for the regulators, on your pcb diagram. If the opamp power rails are oscillating, it is probably originating from the regulators. The regulators should have output caps RIGHT AT their output pins. Usually a small electrolytic with high-enough ESR is *required*, there, to keep the regulators stable.
I would probably also solder a 0.1uF or 0.22uF X7R ceramic or film cap directly between the two opamp power pins.
Also, I don't remember how you said you did your input signal 'ground'. But it should connect ONLY to the ground reference point for the opamp input. The input's signal 'ground' should not be thought of as a ground. It's merely the reference voltage for the input signal.
I also don't know how you ran you ground returns. You would not want any large ground-return currents to share a conductor with any small-signal ground-return currents.
But, are the opamp's power rails oscillating? What type of decoupling caps go from each opamp power pin to gnd? I don't see any decent-sized (and high-ish ESR) output cap for the regulators, on your pcb diagram. If the opamp power rails are oscillating, it is probably originating from the regulators. The regulators should have output caps RIGHT AT their output pins. Usually a small electrolytic with high-enough ESR is *required*, there, to keep the regulators stable.
I would probably also solder a 0.1uF or 0.22uF X7R ceramic or film cap directly between the two opamp power pins.
Also, I don't remember how you said you did your input signal 'ground'. But it should connect ONLY to the ground reference point for the opamp input. The input's signal 'ground' should not be thought of as a ground. It's merely the reference voltage for the input signal.
I also don't know how you ran you ground returns. You would not want any large ground-return currents to share a conductor with any small-signal ground-return currents.
Gootee, thanks for the list...
For the opamp power, right on the regulator outputs I have 100uf electros...
I still have the 0.33uf tantalums on the opamp input... removing them made no difference... so I put them back on...
I'll try a cap between the two power pins of the opamp...
As for the ground... everything is connected together... I was thinking I would have to make a new board to separate the grounds... however, looking at it now, I think it won't be difficult to do that... ther is only a cap and resitor that I will have to lift(r1, c1)...
Otherwise its just a ground plane...
I will also try the feedback caps...
Thanks!!
For the opamp power, right on the regulator outputs I have 100uf electros...
I still have the 0.33uf tantalums on the opamp input... removing them made no difference... so I put them back on...
I'll try a cap between the two power pins of the opamp...
As for the ground... everything is connected together... I was thinking I would have to make a new board to separate the grounds... however, looking at it now, I think it won't be difficult to do that... ther is only a cap and resitor that I will have to lift(r1, c1)...
Otherwise its just a ground plane...
I will also try the feedback caps...
Thanks!!
I found something interesting...
I removed the resistor that was part of the zobel network...
The oscillation has about halfed in frequency and more than halfed in amplitude...
I broke one side of the resistor when I was disconnecting it, I will cut it in half and find out if it's carbon/coil... it was suposed to be non-inductive...
Perhaps it isnt...
I removed the resistor that was part of the zobel network...
The oscillation has about halfed in frequency and more than halfed in amplitude...
I broke one side of the resistor when I was disconnecting it, I will cut it in half and find out if it's carbon/coil... it was suposed to be non-inductive...
Perhaps it isnt...
krazatchu said:
In my simulation, I just found that with R6 set to its 500 Ohm max, even 10 nH of inductance between the opamp supplies and the opamp supply pins causes a significant increase in the amplitude of the ringing at the opamp output. And 10 nH corresponds to probably less than 1/2-inch of PCB trace.
With power supply (and a single main ground-return) leads simulated with 50 nH and .002 Ohms (i.e. about 2 inches in length), I see 327 mV p-p max, at over 1 Mhz riding on part of the waveform at the opamp output (with the more-sustained portion having a maximum amplitude of about 28mV p-p). But if 10 nH and .0005 Ohms (about 1/2 inch of pcb trace) is added between the +/-15v supplies (which were modeled as ideal voltage sources) and the opamp power pins, the ringing increases to about 591 mV p-p max (and to 50 mV p-p max for the more-sustained portion). That is all probably also heavily dependent on other parasitics, etc. But the trend might still be valid.
By the way, adding a 10 uF capacitor (with 1 Ohm ESR) in parallel with each 0.1uF bypass cap at the opamp pins virtually eliminated the sustained portion of the ringing, in both cases.
Adding a Zobel network (5 Ohms & 0.1 uF in this case) from the opamp's output to ground reduced the more-sustained portion of the ringing's amplitude by a factor of about 3 (but didn't seem to help it die out, much sooner).
Adding the on-board decoupling for the main power rails should also help.
It also appears that stray capacitances between the nodes in and around the potentiometer and ground would make things worse.
krazatchu said:
Oh.
So, maybe adjusting the Zobel component values could help. Or, maybe running a separate ground for the Zobel could help.
It sure seems like putting a small resistor (< 10 Ohms) with a small inductor in parallel (typically < 5 uH, I think; maybe even just some # of turns of magnet wire wrapped around a non-inductive resistor) in series with the output should help, maybe a whole lot.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.