krazatchu said:
Yeah. That's a fairly-common "wirewound" type. Those are inductive, although the lower-valued ones usually have increasingly-negligible inductance. You might want to try a film-type resistor. The standard cylindrical metal film ones are actually a little bit inductive, too, since they usually use a spiral groove. Some of the "flat" film types might be better (I don't know exactly how they're made, though). Caddock, for example, makes some nice 'non-inductive' power film resistors, with TO-220 and TO-126 or similar case styles, which are usually on the order of $5 each, plus or minus a few dollars. The bigger resistor manufacturers have somewhat-similar ones. They would need a heatsink to be rated at more than about 2 Watts, though.
krazatchu said:
I can't be sure, at this point. Feedback can sometimes be counter-intuitive, that way (for me, at least). I have had cases where an opamp with a very large bandwidth and high maximum slew rate was certainly very picky about being properly compensated and power-supply-decoupled, to prevent oscillation. That's fairly common. But I have also had circuits that oscillated because an opamp didn't have ENOUGH bandwidth.
Neither of those is really a particularly high-frequency opamp. Do they both have FET input stages? Does one of them have a lot more input capacitance than the other?
BY the way: At the negative opamp input, do you still have the original setup for the RF input filter, where the capacitor connects directly between the neg opamp input and ground? If so, that could definitely be a problem related to oscillation, and you should replace the original series resistor with two series resistors, each with half the original value, and take the cap to ground from between them (with double the cap value). OR, you could use two original-value resistors and the original cap value.
I couldn't find any t0-220 film resistors at the market here..
So I bought a bunch of 50ohm 2 watt cabon resistors...
And wired them parallel
The high frequency oscillation is now cut down by a factor of about 10 in frequency and about 2x in amplitude...
I ran out of time today, so I will play with it a bit more tomorrow...
Separate the input ground, etc etc...
Thanks!!
So I bought a bunch of 50ohm 2 watt cabon resistors...
And wired them parallel
The high frequency oscillation is now cut down by a factor of about 10 in frequency and about 2x in amplitude...
I ran out of time today, so I will play with it a bit more tomorrow...
Separate the input ground, etc etc...
Thanks!!
Why are you adopting such a high power capacity for your output Zobel?
Ensure it passes minimal current at your normal maximum operating frequency (select the cap appropriately) and roughly match the resistor to your intended load impedance.
In normal operation the Zobel should be cold. You should not need any more than 1W or 2W.
Ensure it passes minimal current at your normal maximum operating frequency (select the cap appropriately) and roughly match the resistor to your intended load impedance.
In normal operation the Zobel should be cold. You should not need any more than 1W or 2W.
Is there a practical minumum limit to the zobel cap value? It seems the smaller I go, the less power passes thru the resistor...
Simulation gives me these values when using a 5ohm resistor, and a 28kHz sine input:
100nF = 2.82W
47nF = 626mW
25nF = 177mW
10nF = 28.0mW
1nF = 283uW
100pF = 2.84uW
10pF = 28.4nW
1pF = 284pW
Simulation gives me these values when using a 5ohm resistor, and a 28kHz sine input:
100nF = 2.82W
47nF = 626mW
25nF = 177mW
10nF = 28.0mW
1nF = 283uW
100pF = 2.84uW
10pF = 28.4nW
1pF = 284pW
The wires from capacitors to board should be kept close to each other to minimize inductance and you should try putting something like a 100uF electrolytic and 100nF or so plastic or ceramic cap between each rail and speaker output, in parallell with the capacitors already there, but on the board.
Is ground of the circuit connected to the chassis? If it isn't there could be feedback as the output of the amplifier goes into the transformer which has some capacitance to chassis ground. The connection to chassis should be done with a short wire directly from the board.
Is ground of the circuit connected to the chassis? If it isn't there could be feedback as the output of the amplifier goes into the transformer which has some capacitance to chassis ground. The connection to chassis should be done with a short wire directly from the board.
It seems I burnt a few of the outputs... Not sure what I did but three of the output transistors are showing short between all pins...
I will pick up a few more and some 100uf electro's and 100nF ceramics for on-board decoupling...
While I had the heatsinks off, I separated the input ground, added a cap accross the opamp inputs and reconfigured the input as per Gootee's sugestion...
For clarity, Gootee, is this what you mean? see pic...
Thanks!!
I will pick up a few more and some 100uf electro's and 100nF ceramics for on-board decoupling...
While I had the heatsinks off, I separated the input ground, added a cap accross the opamp inputs and reconfigured the input as per Gootee's sugestion...
For clarity, Gootee, is this what you mean? see pic...
Thanks!!
Attachments
krazatchu said:
Yup!
krazatchu said:
Main plane, maybe. The bypass currents return to where? I think I would also solder a 0.1uF or so cap directly from one opamp power pin to the other opamp power pin (on other side of board), since the two opamp power supply bypass caps aren't grounded near the same spot.
VERY IMPORTANT: Don't forget to put a 10 uF or so aluminum electrolytic from each opamp power supply pin to main ground, too! (i.e. in parallel with the 0.1uF caps)
My simulation showed that the 10 uF caps, there, eliminated the HF portion of the ringing at the opamp output, and fixing the input filter eliminated the large transients that preceded the HF ringing.
Edit:
Just now, I found that doubling C2 (opamp feedback cap) and halving C3 (main amp feedback cap) almost-totally removed the last glitches that were showing up at the opamp output (after fixing input filter and also adding 10uF || 0.1uF from each opamp power supply pin to ground). Actually, I tried C2=100pF and C3=22pF (I had both at 56pF, before.). But "your mileage may vary.". [Note, too, that I am not modeling your regulators. Also note that I am using a model for an LT1115 audio opamp. I should probably change that to whatever you are using. Edit: OK, I quickly tried it with LT1122 and it still looks the same, at least.]
Actually, with the LT1122 (14 MHz Gain Bandwidth Product, max slew rate 60V/us, and JFET inputs), the input filter fix has a much larger good effect (eliminating ringing at opamp output) than adding the 10 uF bypass caps does. The LT1115 that I WAS using had a GBP of 40 MHz but a max slew rate of 10V/us, and, I assume, BJT inputs. Interesting.
At any rate, I think that you might still want or need to increase C2's value and/or decrease C3's value. The actual values will probably depend on your opamp, and lots of other factors. With the LT1122 instead of the LT1115, I had to futher-increase the value of C2, to eliminate some remaining HF ringing at the opamp output. If you don't mind a little distortion, and maybe just slightly less output amplitude, it might be safest to use something like 220pF and 22pF, or 330pF and 33pF, for C2 and C3. But, again, you'll probably need to experiment with them.
Of course, still add the 10uF (or even more) caps from opamp power pins to gnd, especially with those regulators in there. (It might actually be even better to solder the 10uF-to-47uF caps directly to the regulator's output pins, and then to gnd, instead of putting them over on the opamp's power pins.)
At any rate, I think that you might still want or need to increase C2's value and/or decrease C3's value. The actual values will probably depend on your opamp, and lots of other factors. With the LT1122 instead of the LT1115, I had to futher-increase the value of C2, to eliminate some remaining HF ringing at the opamp output. If you don't mind a little distortion, and maybe just slightly less output amplitude, it might be safest to use something like 220pF and 22pF, or 330pF and 33pF, for C2 and C3. But, again, you'll probably need to experiment with them.
Of course, still add the 10uF (or even more) caps from opamp power pins to gnd, especially with those regulators in there. (It might actually be even better to solder the 10uF-to-47uF caps directly to the regulator's output pins, and then to gnd, instead of putting them over on the opamp's power pins.)
It seems my mj15032 and 33 were also burnt.. so I replaced them as well as the mje340...
With the input un-gnded the output is floating high agian.. .
When I ground the output, its stays level but the dim bulb tester glows a little and doesnt go out like it did b4...
Something else must be burnt.. perhaps a cap ... but there is no visbile indicators..
The board is gettig quite messy, with the additions... I think I will make up a new layout this weekend... And hopefully I can correct the problems this time around...
Thanks!
With the input un-gnded the output is floating high agian.. .
When I ground the output, its stays level but the dim bulb tester glows a little and doesnt go out like it did b4...
Something else must be burnt.. perhaps a cap ... but there is no visbile indicators..
The board is gettig quite messy, with the additions... I think I will make up a new layout this weekend... And hopefully I can correct the problems this time around...
Thanks!
djk said:
Good point. Even just using a socket can make fast opamps misbehave.
If a new layout is made, it would also be a good idea to keep any copper ground areas a little farther away from the opamp's input and output pins' pads. Any stray capacitance there might be bad news. And it probably couldn't hurt to try moving the regulators closer to the opamp power supply pins, while leaving room (and pads etc) for both the large and small caps from each of the power pins to gnd.
Here is the hi side layout...
I like this layout more than the current one, as it's alot of trouble to modify the circuit... I have to unbolt the heatsinks and pull off the driving transistors every time I want to get at the copper side...
The low side would be a mirror ....
Any forseeable problems with this?
Thanks!!
I like this layout more than the current one, as it's alot of trouble to modify the circuit... I have to unbolt the heatsinks and pull off the driving transistors every time I want to get at the copper side...
The low side would be a mirror ....
Any forseeable problems with this?
Thanks!!
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.