I think that I can "swallow" using a set of "Parallel 4780" boards. I am searching for them now. This would be 4 chips and 4 Boards. P to P is DEFINATELY out the window for the 4780.
I think that in the parallel/bridged configuration and given the fact that I intend to keep the voltage down, that ease of using a PC board for this might make sense.
Using a quality heat sink along with a "heat spreader" should be OK.
I will probably get "fancy" with the heat spreader and try to design something that helps conduct heat to the heatsink, or use some sort of extrusion to make the spreader that helps with additional heat dissipation.
There are all kinds of CPU cooler heat sinks out there to play with. I also have a friend in the metal recycling business that I can look at all kinds of scrap copper parts and buy at scrap metal price.
I am looking at the board I posted earlier and see no problem modifying 4 boards to do the job.
I will sell off the "extra" PS boards on Ebay.
Using a quality heat sink along with a "heat spreader" should be OK.
I will probably get "fancy" with the heat spreader and try to design something that helps conduct heat to the heatsink, or use some sort of extrusion to make the spreader that helps with additional heat dissipation.
There are all kinds of CPU cooler heat sinks out there to play with. I also have a friend in the metal recycling business that I can look at all kinds of scrap copper parts and buy at scrap metal price.
I am looking at the board I posted earlier and see no problem modifying 4 boards to do the job.
I will sell off the "extra" PS boards on Ebay.
The spec sheet shows that the internal compensation results in 180 degree phase shift at a gain of about 5. Anything less and you are sure to have oscillation (180 degrees results in positive feedback). Any higher gain and you increase the phase margin and get improved stability for capacitive loads and you will also have improved step response. A good design point is where there is a phase margin of 60 degrees--this occurs at a gain of about 20, which is what National recommends. Obviously, some chips may have characterics different from the spec sheet that will allow them to operate at lower gains, but it isn't good practice to count on deviations from spec in your design.
See Phase margin - Wikipedia, the free encyclopedia for a short discussion on phase margin.
Hi Neil,
Yes, I understand the phase margin / stability issue and I've not had reason to go below a gain of 10 but the lads here are struggling with a design that requires lower gain and I thought I would quickly try it out. It satisfied my curiosity too.
I agree and took the time to point out that my results may not be repeatable.
It would have been nice if National had used a couple of the NC pins for external frequency compensation.
Yes, I understand the phase margin / stability issue and I've not had reason to go below a gain of 10 but the lads here are struggling with a design that requires lower gain and I thought I would quickly try it out. It satisfied my curiosity too.
I agree and took the time to point out that my results may not be repeatable.
It would have been nice if National had used a couple of the NC pins for external frequency compensation.
I think I was just trying to see if I could still remember how to read a Bode plot...it's been a long time
.
Non-inverting? Chances of that being truly stable are slim indeed. There appears to be an odd design flaw inside.
My guess is inverting mode, gain 17.
Can you check that on your scope?
P.S. I'm also thinking 50k input load and 330uf "signal quality" caps instead of the 220uF power caps shown.
Last edited:
This is interesting because the amplifier sounds remarkably different, depending on which input is used for source.
The + input, if connected to source, is prone to misbehave, exactly like an undercompensated op amp. That, of course, sounds "forwards" at best, if not screaming awful, and thus pointless. However, you can flip it around to use the - input and get a useful amplifier.
I cannot imagine why the scope doesn't show this, but my faith in measuring technology just hit an all-time low. Perhaps try some other measuring options?
The + input, if connected to source, is prone to misbehave, exactly like an undercompensated op amp. That, of course, sounds "forwards" at best, if not screaming awful, and thus pointless. However, you can flip it around to use the - input and get a useful amplifier.
I cannot imagine why the scope doesn't show this, but my faith in measuring technology just hit an all-time low. Perhaps try some other measuring options?
Last edited:
If there's nothing modified, in inverting you'll have 1k input impedance => sound changes. If the input impedance is set the same, the only difference is that the output signal is inverted, i.e. upside down. You get the same oscillation on rectangular signal. What do you mean by "undercompensated"? I want to make sure that we understand the term in the same way. In undercompensated opamp you'll have overshoots.
I was referring to what would happen if you unplugged the unity-stable op amp from your buffer and replaced it with the non-unity-stable version of that op amp. The consequences of that are quite similar (albeit difference in severity) to choosing either inverting or non-inverting for LM3886.
Indeed, perhaps there are "other factors" involved; however, its still true that LM3886 is less prone to screech when used inverting. Edit: "less prone to screech when used inverting" is all the data that I have so far; however, it seemed to be a nice clue.
Last edited:
Daniel, the same opamp behaves differently at different gains and/or configurations. That's why all those graphs are shown in the datasheets.
By compensating I understand to have the same gain in all the required bandwidth. And a good example of it is this: Probe Compensation (Calibration) - Developer Zone - National Instruments
By compensating I understand to have the same gain in all the required bandwidth. And a good example of it is this: Probe Compensation (Calibration) - Developer Zone - National Instruments
Okay. Which of those three photos represents disproportionately Louder and slightly blurry uppermost midrange output?
Over the past hour, it has remained true that "less prone to screech when used inverting" is all the data that I have.
EDIT: Exception: typical gainclone uses large caps at close proximity to the chip to straighten out the frequency response. . . but that is done with noise. Far smaller caps are workable when the chip is used inverting.
Last edited:
My web browser displayed these in a different order.
However, I have no choice other than to pick the tallest waveform (overcompensated), as the distortion point is fiercely louder than the rest of the audio spectrum.
For reference, this noise is present with the chipamp.com kit at approximately 2khz to 3.2khz
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.