I was going over my old lm4780 gainclone (2x parallel lm4780 - Old Chipamp.com kit) and I was changing some parts as well as trying to solve my ground loop issue.
One other thing that I wanted to change was the capacitor in the feedback circuit that limits DC gain. I was using a standard electrolytic and I thought that was not correct. In checking on Chipamp.com's literature, I see that they provide a standard electrolytic for this position. See their PDF instructions here: http://chipamp.com/docs/lm3886-manual.pdf
I realize the voltage seen at this cap will be small, but wouldn't a standard electrolytic capacitor behave poorly if the voltage it sees is ever in the wrong polarity?
One other thing that I wanted to change was the capacitor in the feedback circuit that limits DC gain. I was using a standard electrolytic and I thought that was not correct. In checking on Chipamp.com's literature, I see that they provide a standard electrolytic for this position. See their PDF instructions here: http://chipamp.com/docs/lm3886-manual.pdf
I realize the voltage seen at this cap will be small, but wouldn't a standard electrolytic capacitor behave poorly if the voltage it sees is ever in the wrong polarity?
The voltage seen by that cap is miniscule; however, the capacitance value in the manual and datasheets is usually FAR too small. A larger cap, plus a tiny bypass cap per highest treble needs, may work better than the wrong size cap.
Experiment:
Collect several identical 100uF caps, install one of them, and parallel on more of them until arriving at a value that passes lowest bass undistorted.
As for the ground loop, the PG+, PG- scheme seen in online kits works fine for monoblocs but is a bit unfortunate for ordinary stereo use.
For stereo (uses one transformer) it is ultimately simpler to bond PG+ to PG- at the output edge of the power supply board to form a real Power Star Ground and then run 1 (ONE!) ground cable from there to the amplifier board. Generally, star grounding uninstalls ground loops.
Other source of ground loop is when a grounded source, such as a computer is connected to a differently grounded amplifier, resulting in a loud buzz. There's a variety of solutions to that problem, such as as the standard ground loop breaker circuit, a professional double insulated amplifier build, or an input transformer.
Experiment:
Collect several identical 100uF caps, install one of them, and parallel on more of them until arriving at a value that passes lowest bass undistorted.
As for the ground loop, the PG+, PG- scheme seen in online kits works fine for monoblocs but is a bit unfortunate for ordinary stereo use.
For stereo (uses one transformer) it is ultimately simpler to bond PG+ to PG- at the output edge of the power supply board to form a real Power Star Ground and then run 1 (ONE!) ground cable from there to the amplifier board. Generally, star grounding uninstalls ground loops.
Other source of ground loop is when a grounded source, such as a computer is connected to a differently grounded amplifier, resulting in a loud buzz. There's a variety of solutions to that problem, such as as the standard ground loop breaker circuit, a professional double insulated amplifier build, or an input transformer.
If we want this capacitor to not intrude on the performance of the amplifier, our design goal must be to ensure that there will be no significant voltage across it. A good design will set the dominant bass pole with the input capacitor. The rule of thumb that I use as a starting point is to separate poles by a decade; however this is not always practical.
So let's say you set the input pole at 2 Hz, which would assure negligible impact at 20 Hz. Then we would set the feedback pole at 0.2 Hz. Once you start calculating, you will find that it is often the case that this isn't practical. Then you start juggling parts. You want to keep feedback resistors as low as practical, for noise purposes. But thermal drift can become significant with lower resistances ; often a larger (1 or 2 watt) resistor provides better performance. It is crucial to select low noise resistors that are linear in the audio range for this job, because any nonlinearities introduced by components in the feedback loop will be introduced into the amplifier. Vishay makes some power resistors that are appropriate for this task. Then we do as Daniel recommends to determine the smallest capacitor that will subjectively not intrude on the performance of the amplifier.
Another advantage of this approach is that you can use a film cap for the input and this will be the only cap that is directly interacting with the low bass signal; in other words its "sonic signature" (if you're into that ) will be overwhelmingly dominant.
Also as Daniel says, commercial designs often fall far short of the parameters I specify. Some of them sound terrible too, and with 20 minutes of design time (maybe less) and a couple dollars worth of parts they could have been so much better. I've seen commercial designs that used the datasheet circuit verbatim, and the datasheet circuit is almost never optimal.
So let's say you set the input pole at 2 Hz, which would assure negligible impact at 20 Hz. Then we would set the feedback pole at 0.2 Hz. Once you start calculating, you will find that it is often the case that this isn't practical. Then you start juggling parts. You want to keep feedback resistors as low as practical, for noise purposes. But thermal drift can become significant with lower resistances ; often a larger (1 or 2 watt) resistor provides better performance. It is crucial to select low noise resistors that are linear in the audio range for this job, because any nonlinearities introduced by components in the feedback loop will be introduced into the amplifier. Vishay makes some power resistors that are appropriate for this task. Then we do as Daniel recommends to determine the smallest capacitor that will subjectively not intrude on the performance of the amplifier.
Another advantage of this approach is that you can use a film cap for the input and this will be the only cap that is directly interacting with the low bass signal; in other words its "sonic signature" (if you're into that ) will be overwhelmingly dominant.
Also as Daniel says, commercial designs often fall far short of the parameters I specify. Some of them sound terrible too, and with 20 minutes of design time (maybe less) and a couple dollars worth of parts they could have been so much better. I've seen commercial designs that used the datasheet circuit verbatim, and the datasheet circuit is almost never optimal.
Last edited:
I wasn't using a DC blocking input cap. In my case, wouldn't a bipolar be a good idea then since the cap is having a bit more influence on the audio performance of the amp? I believe my cap would put my feedback pole around 1.6 hz. (100uF and 1k resistor)
My resistors are all the small (1/4w?) Vishays that came with the Chipamp.com kit except for the output and zobel resistors. (22k/680 ohm feedback, but I changed them to 22k/1k)
I am very limited to what I can fit on the chipamp.com boards while still being able to mount the chip to my heatsink. Even fitting the feedback capacitor was very hard. I feel like I am going to have to make some more serious changes since as Daniel said, this feedback capacitor should probably be bypassed with a small bypass cap but I have no absolutely no room to add one. When I hooked up my amp last night, I had some oscillation on both channels, so I am having a problem that is common to both boards, and now I have to check for any damage.
My resistors are all the small (1/4w?) Vishays that came with the Chipamp.com kit except for the output and zobel resistors. (22k/680 ohm feedback, but I changed them to 22k/1k)
I am very limited to what I can fit on the chipamp.com boards while still being able to mount the chip to my heatsink. Even fitting the feedback capacitor was very hard. I feel like I am going to have to make some more serious changes since as Daniel said, this feedback capacitor should probably be bypassed with a small bypass cap but I have no absolutely no room to add one. When I hooked up my amp last night, I had some oscillation on both channels, so I am having a problem that is common to both boards, and now I have to check for any damage.
If you design the circuit so that there is (virtually) no voltage across the capacitor, then there is no mechanism for the capacitor to introduce distortion into the circuit. This is why I recommend having the dominant pole on the input. Since you don't have an input cap, you have aced yourself out of this option.
I suppose that you could use a nonpolar there, but it is my opinion that a proper design wouldn't require it. Also your choice of values will possibly not be wide enough.
What was the reason for changing the values? Did you consider how doing so changes the performance of the circuit? What kind of resistors did you use?
You can tack small parts to the bottom of the board too. A small capacitor probably wouldn't get in the way.
If you browse Mouser, you will see that they have all different physical sizes of electrolytics. It's worth a look. There's tall thin capacitors and real short wide ones too.
Is this the first time you powered it up since making changes? Is there a low pass filter on the input?
Well I just looked at the circuit in your link. Are you using that circuit? It is suboptimal. And the capacitor is not optional.
Put the 680 ohm resistor back and replace the capacitor with a 100 uF value for a pole of 2.3 Hz. Configure a low pass input circuit.
Will a 6.3 by 11 mm capacitor fit on the board? Here you go. http://www.mouser.com/ProductDetail/Panasonic/EEU-FC1E101S/?qs=sGAEpiMZZMvwFf0viD3Y3c5%252bJs48NCph%2fwab8Ea1yao%3d
Put the 680 ohm resistor back and replace the capacitor with a 100 uF value for a pole of 2.3 Hz. Configure a low pass input circuit.
Will a 6.3 by 11 mm capacitor fit on the board? Here you go. http://www.mouser.com/ProductDetail/Panasonic/EEU-FC1E101S/?qs=sGAEpiMZZMvwFf0viD3Y3c5%252bJs48NCph%2fwab8Ea1yao%3d
I would have thought avoiding an input cap would be the better solution as long as my source has no DC.
I lowered the gain since the circuit has far more than I need. I didn't think I would hurt anything. I used Vishay 1/4W. Since I am supposedly still well withing the stable range of the chip, I didn't think there would be a problem. DC gain is minimized and my DC offset varies slightly with the input pot anyway. The 4780 is supposed to be stable down to a gain of 10. Am I missing anything else?
I have about 4 mm of clearance under there due to the shape of my heatsink.
Also, I'm sorry for the confusion but that PDF is actually the manual for the 3886. They no longer have the manual for the 4780 online. But if you read the instructions the capacitor actually is optional (at least according to the instructions). If not using the cap, you solder the feedback resistor using different holes.
The 4780 board is not like that 3886 one though. On my board there is no place to solder a feedback capacitor so I had to do a makeshift point-to-point attachment of the feedback capacitor in series with the resistor and there is simply no more room for me to put a bypass capacitor in parallel either unless I start cutting up my heatsink.
I feel the boards and heatsinks are really limiting me. I am probably going to have to get new heatsinks or have them modified so that I can fit the circuit mods that I need. Not sure what to do. I thought this would be an easy fix. This is only a backup amp.
Yes and no. You lose control of the low frequency poles. I hope you understand the benefits of controlling these; I tried to explain it above.
It depends on the layout. It might or might not be a problem.
You realize that a cap on the input would eliminate this, don't you?
Using the lowest gain for stability is usually not optimal. Among other issues, you're lucky if it really is stable.
That's terrible. Maybe you could mount the capacitor on the other side of the board?
I already pointed out that it's not optimal. In fact, my recommendation is not optimal either; but it's a little better.
I don't think that cap is optional.
Well, it's a learning experience.
This board doesn't put all the resistors right by the heatsink, does it? You know where I'm going.
Putting an extra cap in the signal path is better than having the feedback capacitor affecting the circuit? The DC offset is going to wander a bit anyway due to everything else so I thought the volume pot part of the problem was not a big issue. Maybe I am wrong but have seen others doing it this way. Are there any other reasons besides these?
Perhaps the best question to ask would be, what schematic would you recommend for a parallel 4780?
I have a gain of 23.. lowest recommended gain according to the datasheet is 10, I believe. Still a fair margin, but of course that is something I should look at as a possible source of my instability.
Just so you understand why my space is so limited, I am mounted in a channel heatsink which looks something like this:
An externally hosted image should be here but it was not working when we last tested it.
It absolutely is, and I already explained why. Distortion introduced in the feedback loop is going to be amplified. That is why you want the dominant pole on the input.
It's not about DC offset.
See above.
You can still do it, but it is suboptimal.
I missed the parallel part! But for parallel you must use precision resistors; 0.1% is recommended.
You also need an RF filter on the input. Only one input resistor and one RF filter per channel is required.
Output swamping resistors after the snubber are required too. 0.22 ohms should work.
In a perfect world.
What kind of bypass caps do you have on the power supply pins? You better have a high frequency cap there as well as an electrolytic.
RF filter and output snubber are very important too.
Indeed.
This is a marginal kit that does not lend itself to tweaking, in my opinion.
Move some of the parts to the bottom of the board. This will give you some wiggle room.
I have electrolytics and small bypass caps on the supply pins. I forget the values. There is a snubber on each output of the chip and .1 ohm resistors on the outputs. (This was the value they recommended in the kit as well as on the datasheet.. perhaps internally there is enough resistance for such a small resistor to be adequate?)
What do you use for an input RF filter?
What do you use for an input RF filter?
1K resistor in series with the input, and a 100 pF capacitor in parallel to ground, after the 1K resistor.
0.1 is fine, but for parallel 0.22 is probably better. Try to match them up as best you can.
Last edited:
Yes, installing the input coupling cap lets you roll off signal that would otherwise have "become jammed" at a too small feedback coupling capacitor. Possible fixes towards clearer sound include, making the feedback coupling capacitor Larger or making the input coupling capacitor Smaller. This is your H1/H2 bass harmonic balance control. Instead of a monotonous extreme, it needs to be set for a favorable harmonic balance.
- Status
- Not open for further replies.