Hi, I have just finished building a 3886-based Gainclone. However, I suspect that it is oscillating since it is getting warm, but not hot, even when there is no music playing. My old Gainclone, with a 3875, has always been very cool when it is not playing.
So, my question is: is it normal that a 3886 GC is getting a bit warm? Or should I start searching for oscillation? Would be nice to know to avoid going on a wild goose chase...
Some data: Voltage is +/- 27 volts. No zobel or LR (yet). Point2point. Star grounding with separate signal and power stars (the same set-up as my previous GC)
/Robert
So, my question is: is it normal that a 3886 GC is getting a bit warm? Or should I start searching for oscillation? Would be nice to know to avoid going on a wild goose chase...
Some data: Voltage is +/- 27 volts. No zobel or LR (yet). Point2point. Star grounding with separate signal and power stars (the same set-up as my previous GC)
/Robert
The conditions are almost identical, the only thing differing is the chip, the 10k mute resistor and the PS has two 2200uF caps/chip instead of one 10000uF/rail. But the wiring is of course a bit different, which could explain that it picks up something to make it oscillate.
I tried looking at the output with the oscilloscope, but could not see anything apart from a little noise, ~1 mV. I also tried building a simple RF detector with a diode, condenser and a resistor, but the voltage over the resistor was zero when applied to the output.
I will try adding the R+C network (there are already 100nF caps over the PS caps) and see if it helps, but it would have been fun being able to actually see the oscillations.
/Robert
I tried looking at the output with the oscilloscope, but could not see anything apart from a little noise, ~1 mV. I also tried building a simple RF detector with a diode, condenser and a resistor, but the voltage over the resistor was zero when applied to the output.
I will try adding the R+C network (there are already 100nF caps over the PS caps) and see if it helps, but it would have been fun being able to actually see the oscillations.
/Robert
I have now tried adding an R+C network (4R+100nF), but the chip is still getting warm. I measured the quiescent current to 52 mA, which apparently is within normal range.
I'm wondering, if it were oscillating, would it not draw more current? Is there any way of actually "seeing" the oscillations?
I'm wondering, if it were oscillating, would it not draw more current? Is there any way of actually "seeing" the oscillations?
Yep, I wouldn't say my lm3886s run cool, and they are on 10 x 10 x 4 cm sinks each.... Are both equaly hot...? If not, look at the bad channel's speaker with an ohmmeter...
Likely your scope isn't good enough to see HF either... so you cant rule out oscillation....
I think if it doesn't go into thermal protection during normal use, you are OK.
Likely your scope isn't good enough to see HF either... so you cant rule out oscillation....
I think if it doesn't go into thermal protection during normal use, you are OK.
ditto
Each of my 3886s are on an AMD heatsink. Warm if under half volume, hotter than the bad place if above. I re-attached the original PC fans as I use this amp in my garage and CRANK-IT-UP! I run it at near top volume as all the way will hurts your ears in just a few seconds - but there is no distortion!
Each of my 3886s are on an AMD heatsink. Warm if under half volume, hotter than the bad place if above. I re-attached the original PC fans as I use this amp in my garage and CRANK-IT-UP! I run it at near top volume as all the way will hurts your ears in just a few seconds - but there is no distortion!
A handy way to find HF oscillations is to use an ordinary transistor radio tuned between AM stations. Bring the antenna near the DUT and see if you hear anything. Don't short anything out! Maybe cover the antenna with some shrink tubing or similar. This can work to find RF coming from power supplies as well. It's not definitive, but another tool to add to the arsenal. IMO, you probably don't have oscillations, or you would have seen it on the scope. It's probably just the quiescent current of the device heating things up.
that's for both channels though.OK, I am feeling re-assured that there is no oscillation going on. It is only using 50mA (which seems about right from the data sheet) and I cannot see anything on the scope. I will try the AM radio thing though, but since other peoples 3886s also gets warm I think mine is normal.
I just found it odd that it gets warm both with and without any load, but the heat sink on my test rig was a lot smaller than jaycees. (In fact, it was a flat piece of aluminium 25X12 cm for both channels. I guess the fact that my 3875 never even gets warm made me over confident...)
I was looking at the data sheet and +-27V into 4 ohms will produce about 40W, which will need a heat sink with 1.9 C/W. So I will have to get a big heat sink, which will hopefully keep it cool.
Thanks everyone for your help!
/Robert
I just found it odd that it gets warm both with and without any load, but the heat sink on my test rig was a lot smaller than jaycees. (In fact, it was a flat piece of aluminium 25X12 cm for both channels. I guess the fact that my 3875 never even gets warm made me over confident...)
I was looking at the data sheet and +-27V into 4 ohms will produce about 40W, which will need a heat sink with 1.9 C/W. So I will have to get a big heat sink, which will hopefully keep it cool.
Thanks everyone for your help!
/Robert
you might want to check for good thermal coupling with the heatsink. In doubt, sand the heatsink with some fine sandpaper to make sure it's flat. If you are using isolated heatsinks (a good idea!) Mica will do but make sure it's quite thin, and you have to use paste thinly.
Personally, I used the grey "sil pad" things with mine (the TO3P ones fit just fine) and i dont have a heat transfer issue. Saves the mess of paste, too.
Personally, I used the grey "sil pad" things with mine (the TO3P ones fit just fine) and i dont have a heat transfer issue. Saves the mess of paste, too.
I finally got around to finishing the amp. As you can see in the image I built two heatsinks from scrap pieces of aluminium. They seem to be enough since they only get luke warm when playing at fairly high volume.
I used the approximation formula 50/sqrt(Area) = C/W to get an idea of how big they shold be.
It seems that the 3886 requires a lot more heatsinking than the 3875 in my previous GC.
How it sounds? Much better, I think, than the NAD 7240 it replaces. (I use it to drive the woofers in an active system. Another GC is driving the tweeters, which might explain why I prefer to match it with another GC)
I used the approximation formula 50/sqrt(Area) = C/W to get an idea of how big they shold be.
It seems that the 3886 requires a lot more heatsinking than the 3875 in my previous GC.
How it sounds? Much better, I think, than the NAD 7240 it replaces. (I use it to drive the woofers in an active system. Another GC is driving the tweeters, which might explain why I prefer to match it with another GC)
Attachments
LM3886 warmer than LM3875 with no signal
The LM3886 has a higher idle current than the LM3875 so it will will be warmer than LM3875 even with no signal as it has to dissipate more power.
The LM3886 has a higher idle current than the LM3875 so it will will be warmer than LM3875 even with no signal as it has to dissipate more power.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.