Sounds like a proper heatsink might be the most effective upgrade.
Check out:
http://business.search.ebay.com/_Manufacturing-Metalworking_W0QQsacatZ11804QQsassZbarrredboss
He'll cut the piece for you for free. One buy could be all the GC heatsink you ever need.
Check out:
http://business.search.ebay.com/_Manufacturing-Metalworking_W0QQsacatZ11804QQsassZbarrredboss
He'll cut the piece for you for free. One buy could be all the GC heatsink you ever need.
uncle_leon said:
A couple of possibilities.
1. Your music just sounds like that, and your other amp is just masking the top end.
2. You're bouncing off the protection circuitry.
3. Oscillation
For #1 there isn't much you could do other than get used to it, or apply some lowpass filtering to take the treble down a tad.
For #2 A better heatsink maybe? I don't know how loud you were listening, but mine has a case made of 3" by 1/8" aluminum angle. And stuck on there with heatsink paste it gets a little warm at my normal listening levels. Certainly not anywhere near what I would call hot.
For #3 You need some way to damp the oscillations. Maybe add in the "zobel" components if they came with your kit. I've never had that problem, so I can't say whether it will help, but it's certainly cheap to try.
When I first built mine, I noticed that good recordings sounded better than on my old amplifier, but bad ones sounded worse. I wouldn't be surprised if that's all it is. Now you can hear how your recordings suck.
-Nick
uncle_leon said:My amp sang for me today!
So, my initial impressions were:
- very expressive, life-like vocal (MJ, Kate Melua)
- well detailed treble
- very tight&controlled bass, extending deep into low frequencies
- sound is a bit on the bright side
- very good instrument separation
But after some more listening, I decided that there is something wrong with tonal balance. Bass seems to lack 'filling'. It does slam nicely, but it is nowhere near richness of my NAD 3225PE.
The other problem is upper-mids - they are screamy and overexposed. Add the fact, that there was way too much trebles, and you have one really shrilly&shouty amp.
And I don't like being shouted at... After hooking speakers back to NAD, the feeling was like if I was sat in an soft padded, comfy armchair. Relaxed, rich, but polite sound, enjoyable even in spite of the lack of that last detail. Gainclone seems to be much more involving, but for now, NAD remains my main amp.
So now I'm wondering WHY? Can anyone please give me a hint as to possible causes of such imbalanced sound?
I also have one less important question, as to optimal temperature for this chip. During normal, moderately loud listening it was getting more than warm, but wasn't burning. But after putting the volume really loud, it got to over 80deg.C. Is it still safe or should I install fan/larger heatsink?
That's how it sounded to me too, compared to good valve amps, and that is why I only built a couple of GCs. Good value, transparent, but stressful in the long term IMHO.
85C is very hot, I'd go with a larger heatsink.
On one of my GC's (parallel LM4780, see link in my sig line) I installed a thermoswitch that shorts at 45C and lights a red LED. Even with the fan turned off it can go for hours without reaching 45C, but when you play it loud it immediately gets hot and trips the switch.
Not sure what advice to give you on the sound, are you using a zobel network on the output? You might want to add/tweak it.
On one of my GC's (parallel LM4780, see link in my sig line) I installed a thermoswitch that shorts at 45C and lights a red LED. Even with the fan turned off it can go for hours without reaching 45C, but when you play it loud it immediately gets hot and trips the switch.
Not sure what advice to give you on the sound, are you using a zobel network on the output? You might want to add/tweak it.
What are you using for a heatsink anyways?
What's the impedance of your speakers?
Voltage of your rails?
with an 18v transformer, you're probably only around 30-34v at the rails and shouldn't be running all that hot unless either something is wrong, or your speakers are very low impedance, or you've got a very small heatsink.
-Nick
What's the impedance of your speakers?
Voltage of your rails?
with an 18v transformer, you're probably only around 30-34v at the rails and shouldn't be running all that hot unless either something is wrong, or your speakers are very low impedance, or you've got a very small heatsink.
-Nick
The temperature must have had some substantial influence, because when I auditioned the amp at low to moderate volume levels, the sound was much better balanced.
I found another problem though - the amount of bass is rather low. And when I tried "fixing" it with Winamp's equaliser (I simply put 60Hz at +12dB), strange thing happened... The bass did not became more dominant at all, but instead whenever there was a big "oomf" (or I should rather say "whenever there SHOULD be an oomf"), the rest of the sound was fading for a moment. I suspect that there must be some problem with power. Maybe those tiny-thinny wires, that I patched the rectifier bridge with, are a problem. I'll swap them against some thicker cable and see what happens.
I found another problem though - the amount of bass is rather low. And when I tried "fixing" it with Winamp's equaliser (I simply put 60Hz at +12dB), strange thing happened... The bass did not became more dominant at all, but instead whenever there was a big "oomf" (or I should rather say "whenever there SHOULD be an oomf"), the rest of the sound was fading for a moment. I suspect that there must be some problem with power. Maybe those tiny-thinny wires, that I patched the rectifier bridge with, are a problem. I'll swap them against some thicker cable and see what happens.
I replaced those tiny 0.1mm wires on the rectifier board with proper 0.75mm ones, but it made any difference, it was not audible. The amp still seems to be very current-limited.
I also checked out all caps and they look alright to me... My multimeter doesn't measure capacitance, so I can't make sure they are really ok.
Edit:
Oh, and by the way, I'm getting pretty high noise floor from that crappy laptop soundcard. I guess a pot on input would help supress it? The noise comes from the source, of course, the amp itself is perfectly silent.
E2:
In case it matters, the speakers are Mission 70 MKII, they are rated at 75W, 8Ohm. I think they should be really easy load on any amp. Even some utterly poor 5W amplifier I took off some cheap active PC speakers managed to drive them to reasonable volume levels without clipping.
I also checked out all caps and they look alright to me... My multimeter doesn't measure capacitance, so I can't make sure they are really ok.
Edit:
Oh, and by the way, I'm getting pretty high noise floor from that crappy laptop soundcard. I guess a pot on input would help supress it? The noise comes from the source, of course, the amp itself is perfectly silent.
E2:
In case it matters, the speakers are Mission 70 MKII, they are rated at 75W, 8Ohm. I think they should be really easy load on any amp. Even some utterly poor 5W amplifier I took off some cheap active PC speakers managed to drive them to reasonable volume levels without clipping.
There are only 4 caps:
- two 10uF next to each rectifier bridge
- two 1500uF next to LM3875 chip
PCB picture, list of the parts and schematics can be found here.
- two 10uF next to each rectifier bridge
- two 1500uF next to LM3875 chip
PCB picture, list of the parts and schematics can be found here.
Try reversing
I am far from the first to do this, but try reversing the order of the caps. Like 1500 ufd on the rectifiers and 10 ufd on the pins of the chip. Then maybe 220 ufd where you have the 1500 ufd.
Using 10 ufd caps right on the power pins give a lower esl than any 1500 ufd cap. I have installed 22 ufd 50V FM caps in a couple amps lately. Being away from the power supply they do not increase diode noise, like using a low esl cap at the rectifiers.
If your rails are low enough, the Silmic II 10 ufd 35 volt caps are perfect. Plus 7.5 cents each at Digikey. These will work great to about 30 volt (dc) rails.
What I really recommend for bass, mids and highs is 10,000 ufd Panasonics, bypassed with 0.047 ufd film and 0.4 ohm resistor in series. Then a 220 ufd ZL cap, and 10 - 22 ufd low esl soldered to the bottom of the power pins. And a 0.1 ufd film cap (100 volt) going rail to rail.
This combo is quiet and stores a lot of energy. But the real cat is the small cap on the power pins. It makes a huge difference in the dynamics and percieved bass.
George
I am far from the first to do this, but try reversing the order of the caps. Like 1500 ufd on the rectifiers and 10 ufd on the pins of the chip. Then maybe 220 ufd where you have the 1500 ufd.
Using 10 ufd caps right on the power pins give a lower esl than any 1500 ufd cap. I have installed 22 ufd 50V FM caps in a couple amps lately. Being away from the power supply they do not increase diode noise, like using a low esl cap at the rectifiers.
If your rails are low enough, the Silmic II 10 ufd 35 volt caps are perfect. Plus 7.5 cents each at Digikey. These will work great to about 30 volt (dc) rails.
What I really recommend for bass, mids and highs is 10,000 ufd Panasonics, bypassed with 0.047 ufd film and 0.4 ohm resistor in series. Then a 220 ufd ZL cap, and 10 - 22 ufd low esl soldered to the bottom of the power pins. And a 0.1 ufd film cap (100 volt) going rail to rail.
This combo is quiet and stores a lot of energy. But the real cat is the small cap on the power pins. It makes a huge difference in the dynamics and percieved bass.
George
1500uf simply is not enough capacitance per rail to provide decent smoothing. A quick test on the bench shows massive supply ripple even with just a 10Vpp 1Khz sine into 8ohms, at 20khz it is simply horrid. This is even with very good low ESR caps. This effect will have detrimental effects on PSRR (of course) and produce a lot of distortion and make the amp sound thin, certainly not great.
Such low capacitance on the rails will also result in a lot of mains induced ripple (100 or 120hz) being on your rails (very easy to see). This is also not a good thing at all. Even with very good diodes arranged as a dual bridge you should have at a minimum 10,000uF per rail at 30V to produce a decently smooth power supply. 20K being much better. 4700uf may be passable, but I would not personally go that low.
A pair of 1500uf may be simple, but it will not produce an amp that can deliver the goods. It is easy enough to prove both in simulation and with the most rudimentary of scopes.
There is simply no transformer with low enough impedance that you can get away with 1500uf rails. Especially at higher frequencies where the inductance of the transformer secondaries plays a large role.
Cheers!
Russ
Such low capacitance on the rails will also result in a lot of mains induced ripple (100 or 120hz) being on your rails (very easy to see). This is also not a good thing at all. Even with very good diodes arranged as a dual bridge you should have at a minimum 10,000uF per rail at 30V to produce a decently smooth power supply. 20K being much better. 4700uf may be passable, but I would not personally go that low.
A pair of 1500uf may be simple, but it will not produce an amp that can deliver the goods. It is easy enough to prove both in simulation and with the most rudimentary of scopes.
There is simply no transformer with low enough impedance that you can get away with 1500uf rails. Especially at higher frequencies where the inductance of the transformer secondaries plays a large role.
Cheers!
Russ
The last LM3875 amp I built easily betters my previous efforts. First I built the Peter Daniel basic kit (which sounds brilliant) then modified the psu by adding 15000uF per channel (which sounded dreadful - slow, detached bass and no excitement - just as others have reported).
My second build was a Peter Daniel premium LM3875 kit, but with 100uF black gate N on each amp pcb with 0.1uF BG Nx bypasses on the pins of the chips. I built a cap bank with panasonic FM caps to give me 18000uF per rail (or there abouts..). No snubbers were used - I hoped the paralleled 1000uF caps used in the 18000uF bank would not sound as slow as the single 15000uF caps I'd tried in my first amp. This amp was a big step up on my first effort.
My latest amp has 4.7uF BG N on the rectifier board, with a 39ohm 0.1uF snubber (optimised to reduce noise at 256KHz). I have a single pair of 2200uF 100V sikorel caps upstream of the amp boards, which have 940uF sikorel 125 (2x 470uF) per rail per channel. I have no low value bypasses on the chip pins, but a single 47uF BG standard between V+ and V-. All resisitors are 2W Kiwame except for the feedback, which is a Welwyn RC55. I managed to shorten the feedback loop considerably by mounting this to the rear of the chip, soldered directly to where the pins come out of the chip body. A Kiwame would have been too large for use here. The volume pot is a 1meg PEC with 39ohm Kiwame in fake-law.
There are far too many variables for me to claim why my new amp sounds so much better than my old efforts. I now have raw dynamics and bass weight without loosing the open mid and top of my first basic PD kit. My suspicion is that the sikorels behave like faster smaller caps (totally unqualified statement, i know) so don't seem to slow the sound as much.
I ended up buying loads of the 470uF 35V sikorels on Ebay (100 i think!) and I'd happily mail eight to anyone who wants to give them a try... I'd be interested for someone with more experience than I to try them and report back.
My second build was a Peter Daniel premium LM3875 kit, but with 100uF black gate N on each amp pcb with 0.1uF BG Nx bypasses on the pins of the chips. I built a cap bank with panasonic FM caps to give me 18000uF per rail (or there abouts..). No snubbers were used - I hoped the paralleled 1000uF caps used in the 18000uF bank would not sound as slow as the single 15000uF caps I'd tried in my first amp. This amp was a big step up on my first effort.
My latest amp has 4.7uF BG N on the rectifier board, with a 39ohm 0.1uF snubber (optimised to reduce noise at 256KHz). I have a single pair of 2200uF 100V sikorel caps upstream of the amp boards, which have 940uF sikorel 125 (2x 470uF) per rail per channel. I have no low value bypasses on the chip pins, but a single 47uF BG standard between V+ and V-. All resisitors are 2W Kiwame except for the feedback, which is a Welwyn RC55. I managed to shorten the feedback loop considerably by mounting this to the rear of the chip, soldered directly to where the pins come out of the chip body. A Kiwame would have been too large for use here. The volume pot is a 1meg PEC with 39ohm Kiwame in fake-law.
There are far too many variables for me to claim why my new amp sounds so much better than my old efforts. I now have raw dynamics and bass weight without loosing the open mid and top of my first basic PD kit. My suspicion is that the sikorels behave like faster smaller caps (totally unqualified statement, i know) so don't seem to slow the sound as much.
I ended up buying loads of the 470uF 35V sikorels on Ebay (100 i think!) and I'd happily mail eight to anyone who wants to give them a try... I'd be interested for someone with more experience than I to try them and report back.
Hartono said:
Hi Hartono, the heatsink is 6.5cm L x 6cm W x 4cm T, aluminum, with 10x8 fins. At normal listening levels I could not record any temperature above 42C (which simply HAS TO be low enough). I have massive experience with cooling computer processors (including water cooled / below 0C systems), so believe me, I did everything about this part right.
This "bass fade", occuring when over-equalizing low frequencies (as I descibed above), persists even at lower volume / low temperature. So I guess it wasn't the temperature that was the problem after all.
Is it possible that I damaged my caps somehow, while I was soldering them?...
Panelhead, sharpi31 - thanks for your ideas, I will certainly come back to them after my amp starts working properly. For now, it is not the matter of taste or quality, there's simply something wrong with the amp.
Russ White - I know that those 1500uF caps seem way too little, but everyone I asked told me that swapping them against some proper big caps only made their amp sound worse. I'm not an expert so I just rely on other's opinions.
uncle_leon said:
Oh I would keep the 1500uf caps, they are good for decoupling, but they should indeed be supplemented at the power supply side.
There is nothing peculiar about the way the LM3875 or any other chipamp "sees" the power supply. It the same as any other amp in that regard. While it does have decent PSRR it is not so great that you can ignore having stiff rails.
This is not to say you should throw any old cap you like in there. The thing should be suitable for the task.
There is nothing to argue here. This is a simple matter of math/engineering and very simple to test. 1500uf is simply not enough, not that it won't work, and not that some folks might say it sounds better (for who knows what reason) but I can tell you flat out it measures terribly, and all things tangible say 1500uf on the rails will never a great amp make.
In the end all that matters is that you are happy with what you built. Have fun!
Cheer!
Russ
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.