As long as you are happy with the sound quality, Vunce, don't change it......
If you do change it, you have to redress the value of the active CCS base resistor (R119 on upper graphic which is listed at 925R).
HD
If you do change it, you have to redress the value of the active CCS base resistor (R119 on upper graphic which is listed at 925R).
HD
Is there a more cost-effective and physically-efficient way of removing 100+ watts from a 2 square cm patch of metal? You don't need the fans, but you do need the heat pipe thermal spreaders. They can go to a passive radiator as is done on a nice passively cooled computer case shown earlier in this thread. Even with a pure copper block, there is not enough thermal conductivity in metal to spread the heat over a wide enough area for the passive fins to start working.
To use one CPU cooler, you would need the MOSFETs to be laid out so that they are right next to each other as I did on the veroboard proto.
X, For the 20W amp, I'm planning on using two Dell CPU coolers per channel, passive; if I read what you noted above correctly, should do the trick, shouldn't it?
Should give it a Star Wars appearance, for an out-of-this-world amp.
Thanks,
Last edited:
I am still waiting to see someone trying this out with closed loop water coolers. There are so many out there that are cheap and perform excellently (for CPU cooling at least). And honestly, the slight sounds of the pump are much less of an issue in my view than the fan noise, and probably inaudible over that as well... 🙂
I dont think that can be stated as a matter of course - it is simple and cheap to replace stock fans with quieter alternatives, and the ones i've had experience with were already very quiet 120-140mm fans. Cooling is achieved by recirculation of the coolant through a heatblock fastened to your device and a passivee radiator, which radiates the heat to the air (with the help of a fan usually).
With water coolers you would add liquid to the mix, complicating things out of sight. More complexity, more cost, less reliability, and while more flexibility in space saving arguably unless using a different pcb layout with both outputs close to co-located to exploit only two heat transfer points rather than four clearly much more cost.
Nice idea - I did this many years ago with a 56W single ended stereo amp using six output mosfet pairs - I strongly endorse X's heat pipe system which seems to be as efficient at moving heat as any liquid option and is simpler and cheaper than any other system. His approach is actually cheaper than good quality, large sized passive heatsinks. It doesn't get any better than this approach......
Hugh
Nice idea - I did this many years ago with a 56W single ended stereo amp using six output mosfet pairs - I strongly endorse X's heat pipe system which seems to be as efficient at moving heat as any liquid option and is simpler and cheaper than any other system. His approach is actually cheaper than good quality, large sized passive heatsinks. It doesn't get any better than this approach......
Hugh
xrk will use 4 fans running at or close to minimum rpm, for most good aircoolers that would mean inaudible, burried in noisefloor
Hi Hugh, X,
please find last version schematics attached, please check.
Some tricky modification to do on PCB (input part middle top of layout), but not afraid at all.
JP
please find last version schematics attached, please check.
Some tricky modification to do on PCB (input part middle top of layout), but not afraid at all.
JP
You can have a single radiator per water cooler witha 140mm fan running close to minimum rpm as well, there really isnt much difference on the fan side of things between closed-loop water coolers and the coolers demonstrated by X. I can see your point with added complexity, but i honestly think reliability of current CPU water coolers is high enough that the difference is negligible - at least that has been my experience using them in PCs. Cost-wise, yes it is definitely cheaper with air coolers, i think the main advantage you could have from water cooling is space, since the surface area required for efficient heat transfer is smaller, and because you can mount the radiator away from your output devices.
That said, i was not trying to argue that it is necessarily better, but in the spirit of this thread, which seems to be as much about coming up with new cool ways to cool hot amps, i thought it'd be a fun thing to try 🙂
That said, i was not trying to argue that it is necessarily better, but in the spirit of this thread, which seems to be as much about coming up with new cool ways to cool hot amps, i thought it'd be a fun thing to try 🙂
Thanks Hugh for the very good explanation !
I've simulated the recommended lower current 1.46A Alpha20 8R (0.44r 0.12r) with the two options for the fb CCS resistor:
with equal current swing:
520r fb CCS: clipping at 0.72Vin +-18Vout
In the FFT a nice declining profile: H3 8db lower than H2
with H3 at -79db
output half the current swing of the CCS:
390r fb CCS: clipping at 0.7Vin +-17.4Vout
In the FFT less declining profile: H3 4db lower than H2
but lower distortion: H3 at -82db
So the one with equal currents has a nicer FFT declining profile and a little more power,
the one with half currents has lower distortion but less nice declining FFT.
So it's up to personal preference, me I go for the nice declining FFT profile 🙂
The H3 is already very low.
Attachments
Last edited:
Thanks Danny_66 for the simulation comparison.
So, if I try this for my ALPHA20 the control resistor (R131) should be .44R, current sense resistor (R132) should be .12R and R128 should be 520R.
Hi JPS,
I have studied your schematic and I can see no problem at all based on the original circuit. Thank you very much for this hard work!
I have just answered Danny and Vunce concerning the dynamic CCS values. R145/6/7 are the CCS sense resistor and my choice would be different; I like the total of 0.44 being 1.32R for each resistor (1.32/3= 0.44) but I will check with X. The advantage of higher current is both more drive and ability to drive <8R speakers with ease. In that case if R145/6/7 is changed from your values (I find it difficult to read the values in fact, so small on a printed page) we will need to change base transistor controlling resistor, R119 925R to 520R to ensure the output devices see the same dynamic current swings.
We will wait until X is online, he should arise in the next couple of hours in DC!
Danny:
You have plumbed this circuit well and your curves from LTSpice are pretty much what I get. Well done, this is not particularly intuitive.....
I like to ensure that the current swings are identical, so that when they both bottom out at zero mA we have maxed out the voltage output, and therefore the power. As you have found, this amp does clip on the negative half cycle first, a less than ideal situation and quite different to a PP Class A or AB. Like you, I prefer this approach not just because the output is maximised but also because the harmonic profile, which Nelson does talk about briefly but does not go into detail, it is monotonic decreasing with the equalised current swings. When you can achieve a linear decrease of the harmonic profile from H2 down to H5 you deliver a good amplifier, but to be sure you have to have a log scale on both X axis AND the Y axis. Another level of excellence for a good audio amplifier is the ratio of THD for H2+H3+H4 to THD for H2 + ........... + H15. You can select the number of harmonics in the four command line on LTSpice, like this:
.four {freq}15 8 V(Vout)
.four {freq} 4 8 V(Vout)
The top line gives you 15 harmonics, the second, four. If you can achieve second THD divided by top THD higher than 99.5%, then you have a very good amplifier. The figure of excellence on this amplifier is 0.027421%/0.027503%, which is 99.7%. This figure is at +20dBU output, which is 14.14Vpeak output into 8R at 1KHz.
Cheers,
Hugh
I have studied your schematic and I can see no problem at all based on the original circuit. Thank you very much for this hard work!
I have just answered Danny and Vunce concerning the dynamic CCS values. R145/6/7 are the CCS sense resistor and my choice would be different; I like the total of 0.44 being 1.32R for each resistor (1.32/3= 0.44) but I will check with X. The advantage of higher current is both more drive and ability to drive <8R speakers with ease. In that case if R145/6/7 is changed from your values (I find it difficult to read the values in fact, so small on a printed page) we will need to change base transistor controlling resistor, R119 925R to 520R to ensure the output devices see the same dynamic current swings.
We will wait until X is online, he should arise in the next couple of hours in DC!
Danny:
You have plumbed this circuit well and your curves from LTSpice are pretty much what I get. Well done, this is not particularly intuitive.....
I like to ensure that the current swings are identical, so that when they both bottom out at zero mA we have maxed out the voltage output, and therefore the power. As you have found, this amp does clip on the negative half cycle first, a less than ideal situation and quite different to a PP Class A or AB. Like you, I prefer this approach not just because the output is maximised but also because the harmonic profile, which Nelson does talk about briefly but does not go into detail, it is monotonic decreasing with the equalised current swings. When you can achieve a linear decrease of the harmonic profile from H2 down to H5 you deliver a good amplifier, but to be sure you have to have a log scale on both X axis AND the Y axis. Another level of excellence for a good audio amplifier is the ratio of THD for H2+H3+H4 to THD for H2 + ........... + H15. You can select the number of harmonics in the four command line on LTSpice, like this:
.four {freq}15 8 V(Vout)
.four {freq} 4 8 V(Vout)
The top line gives you 15 harmonics, the second, four. If you can achieve second THD divided by top THD higher than 99.5%, then you have a very good amplifier. The figure of excellence on this amplifier is 0.027421%/0.027503%, which is 99.7%. This figure is at +20dBU output, which is 14.14Vpeak output into 8R at 1KHz.
Cheers,
Hugh
Last edited:
Like you said, only the pump is added, which is why over here they test watercoolers at 40dB and 50dB noiselevels and aircoolers at 30dB and 40dB.
Only pump reaches 30dB already, some aircoolers absolutely can't reach 40dB level and normally run around 18dB, well below pumpnoise.
I am not sure which water coolers you are referring to, but something like the Corsair ones are rated around 30db fan noise, and that can easily be reduced with a cheap lower noise fan. Pump is only audible when the fan is not running at all, nowhere near 30db.
All of them. Hardware.info tests them only at 40dB and 50dB because all of them reach 30dB or over 30dB with just the pump running. Spec by chinese or marketing spec are useless, the best watercooler they tested lately was nzxt x72, nztx spec says max 34.5dB noise with all fans at max speed, in the test the usual 40dB and 50dB tests are done and no trouble reaching that level although according to nztx it is well above maximum LOLOL, 30dB being impossibly low.