RudeWolf,
Riga is an absolutely gorgeous town! Latvia I assume.
Not to be rude, but I would implore you to build Tom's 286 design for a reference level 2ch/theater system and use Tom's LM3886 Done Right for the desktop speaker system. You certainly don't need 32V dual rails either way given that a desktop system is usually listened to nearfield. Even with 83dB efficient speakers (for example), power needs are still small and well within the LM3886DR. I would actually be happy with the SMPS-86 & LM3886DR combo (even with the +/- 24V rails) and be done.
Food for thought. It's your money after all, so I better shut up!
Best,
Anand.
Last edited:
How dare you bring sense into this!
I've built a Modulus-86 before, so I have an idea what the amp is capable of. Currently I'm just fantasising about a 1U compact desktop amp for a high resolution near field system. Maybe even do a fully active 4ch build to experiment with DSP.
And yeah, Riga sure is pretty on a good day!
I've built a Modulus-86 before, so I have an idea what the amp is capable of. Currently I'm just fantasising about a 1U compact desktop amp for a high resolution near field system. Maybe even do a fully active 4ch build to experiment with DSP.
And yeah, Riga sure is pretty on a good day!
Is your desktop speaker a highish impedance design? 10 ohms or greater perhaps? The HP-1 ultra high resolution headphone amp will pump out 3000 milliwatts into 20 ohms.
Since we are fantasizing and all
Best,
Anand.
The HP-1 isn't all that different from the 286, both are high performance IC composites. So, I'd rather have that extra current on tap, if needed. I'm just wondering, how much sinking I'd need to strap on to them chips at 26V rails.
That way I might be able to get away with something like this for a case - https://www.modushop.biz/site/index.php?route=product/product&path=67_193_108&product_id=517
Maybe even stuff some extra radiators in.
That way I might be able to get away with something like this for a case - https://www.modushop.biz/site/index.php?route=product/product&path=67_193_108&product_id=517
Maybe even stuff some extra radiators in.
I am running nearly ±33 V rails in that amp because the SMPS is a ±36 V version and I have the voltage dialled as far down as I could. For future builds, I'd use the ±30 V version.
I didn't measure the heat sink temperature at idle. I would expect it to maybe reach 25 ºC in a 20 ºC room. At max dissipation with a 4 Ω load, now that's another story. The heat sinks on the ModuShop chassis reached 65 ºC, which is a bit too hot for an external heat sink. This was with both channels running into a 4 Ω load at max dissipation.
I had the music cranked quite loud to test the amp. The heat sinks reached about 40 ºC (based on my super scientific hand-on-heat sink temperature sensing).
For a 4-channel amp intended for music reproduction, in particular in a multi-way system, this heat sink and power supply fit the bill but shouldn't be any smaller.
I'm thinking to do exactly that for my desktop amp. Do beware that you'll only get about 15 W into 4 Ω with both channels running as the SMPS reaches its max output current on the peaks around there. For 8 Ω load, you'll get just shy of 30 W (limited by the ±24 V rails). For a desktop amp, it's fine unless you're purposely trying to wreck your hearing.
With 1U you'll be pressed for the heat sink. Definitely do the math on that. The 2U Mini Dissipante would be about perfect for a desktop amp.
Tom
Last edited:
They're both high performance composite amps, but that's where the similarities end. It's very different delivering a ±18 V swing into a 32 Ω load than delivering that same ±18 V swing (or ±24 V if you use the SMPS-86) into a 4 Ω load. I show the math here: LM3886 chip power amplifier thermal design.
At ±24 V rails (assuming SMPS-86 here), I recommend 2.1 K/W per channel (so 1.05 K/W for the Modulus-286). This assumes 4 Ω load and music reproduction (14 dB crest factor). For 8 Ω load, you're looking at 3.8 K/W per channel (so 1.9 K/W for the Modulus-286).
I don't see any heat sinks on that enclosure. You'll need an actual heat sink. You could put it internal if you can get enough air flow. You could also make it fan cooled. There are all kinds of super quiet fans out there now for cooling computer CPUs. They'd work well.
Tom
I got a surprise during testing of a 3886 assembly for a desk top active monitor.
The two 3886 run from ±35Vdc and output was set to 1W @1kHz into 8r0 and 2W into 8r0 to mimic average levels of a loud audio signal.
The common heatsink was ~1C/W and ambient was ~20°C.
Within a few minutes the chipamps were too hot to touch. I'd guess >60°C and maybe >70°C
The two 3886 run from ±35Vdc and output was set to 1W @1kHz into 8r0 and 2W into 8r0 to mimic average levels of a loud audio signal.
The common heatsink was ~1C/W and ambient was ~20°C.
Within a few minutes the chipamps were too hot to touch. I'd guess >60°C and maybe >70°C
I something not right, ie oscillations ?.
Arguably the heatsinking of my Behringer B2031 speakers is not particularly good and the the transformer adds to the module heat load.
When running flat out for hours the amp modules get decently warm but nowhere as hot as that.
Dan.
No oscillation, plain simple 1W and 2W outputs into resistive dummy loads.
Look at the datasheet, Figure 36.. I can see that at low output powers the heat that needs to be dissipated is quite high.
I tried to read off from the plots what to expect for my 1W & 2W test, but resolution down at that end is very poor.
@ 5W output on the ±35Vdc line, I see ~23W of dissipation. @ 20W of dissipation it looks like National are showing an output power of ~3W into 8r0.
The temperatures were a big surprise even though I am using a heatsink about double what National recommend in Figure 34.
Fig 34 shows 3C/W for 1 channel @ 25°C with 8r0 and 69Vdc
@ Ta=20°C and 70Vdc, 8r0 should be around 4C/W
two channels becomes 2C/W and I used ~0.9 to 1C/W
If you have been following my posts you will see I consistently recommend doubling the National recommendation for heatsinking the 3886.
Look at the datasheet, Figure 36.. I can see that at low output powers the heat that needs to be dissipated is quite high.
I tried to read off from the plots what to expect for my 1W & 2W test, but resolution down at that end is very poor.
@ 5W output on the ±35Vdc line, I see ~23W of dissipation. @ 20W of dissipation it looks like National are showing an output power of ~3W into 8r0.
The temperatures were a big surprise even though I am using a heatsink about double what National recommend in Figure 34.
Fig 34 shows 3C/W for 1 channel @ 25°C with 8r0 and 69Vdc
@ Ta=20°C and 70Vdc, 8r0 should be around 4C/W
two channels becomes 2C/W and I used ~0.9 to 1C/W
If you have been following my posts you will see I consistently recommend doubling the National recommendation for heatsinking the 3886.
Last edited:
Seems reasonable to me...
2W RMS output = 0.5A RMS into 8R
With 35V rails, then dissipation is 33 * 0.5 = 16.5W
For 2 channels that's 33W.
if your h/s in 1C/W and the 2 pcs of LM3886 in (thermal) parallel are 0.6 C/W JS. Then you should expect Tj to be 33 * 1.6 = 52.8C above ambient i.e. ~73C.
TI's recommended h/s is designed to keep Tj under 150C. Of course you will need to X2 Ti's recommendations for a stereo amp!
2W RMS output = 0.5A RMS into 8R
With 35V rails, then dissipation is 33 * 0.5 = 16.5W
For 2 channels that's 33W.
if your h/s in 1C/W and the 2 pcs of LM3886 in (thermal) parallel are 0.6 C/W JS. Then you should expect Tj to be 33 * 1.6 = 52.8C above ambient i.e. ~73C.
TI's recommended h/s is designed to keep Tj under 150C. Of course you will need to X2 Ti's recommendations for a stereo amp!
No.
Usually they state a delta T of around 70C degrees to 80C degrees.
delta T = Ts -Ta
Tomchr gives a good run down. it's also given in a National ap note.
Page 18 of the 3886 datasheet shows the method of determining the maximum dissipation, but unfortunately that's not what is required when power outputs are different from ~ 1/3rd of maximum output.
see the graph and compare 21W into 8r0 with 63W into 8r0.
The 21W value is quite close to maximum dissipation.
And they might consider Ta of 45C.
Anyway, I can see that nothing I can say is of any possible use to you.
For the other members...
Please see equations 1 and 4 here:
http://www.infineon.com/dgdl/an-948.pdf?fileId=5546d462533600a40153559ed34311b1
At 2W, VRMS = 4V. IRMS = 0.5A
Power drawn from the supply = 15.75553553W
So minus the 2W into the load = 13.75553553W
(I don't know why Andrew didn't post this himself, instead of just telling me I'm wrong)
Please see equations 1 and 4 here:
http://www.infineon.com/dgdl/an-948.pdf?fileId=5546d462533600a40153559ed34311b1
At 2W, VRMS = 4V. IRMS = 0.5A
Power drawn from the supply = 15.75553553W
So minus the 2W into the load = 13.75553553W
(I don't know why Andrew didn't post this himself, instead of just telling me I'm wrong)
Because I don't know an accurate method of calculating chipamp dissipation.
I just recognised that the simplified method you proposed would not give an accurate answer for most operating conditions.
Instead of making up some answer I gave some ideas on where to look for good info.
The Infineon ap note does not give the correct answer for a chipamp. It is only considering the output stage.
Last edited: