Is it possible to make a heatsink that has the shape of a "tree"? That is, a tall skinny "trunk" that is the same footprint as the TPA chip, tall enough to clear all the neighbor components. Then it can "bloom" out at the top. Or maybe it's possible to build the trunk out of heat pipes (like CPU coolers)?
FWIW, with my moderate listening levels, I've never felt even a naked TPA chip get beyond warm to the touch. With nearfield 8 ohm load, the chip is not even warm. Although there were reports in the big thread about the big heatsink on the Sure 3116 getting warm.
FWIW, with my moderate listening levels, I've never felt even a naked TPA chip get beyond warm to the touch. With nearfield 8 ohm load, the chip is not even warm. Although there were reports in the big thread about the big heatsink on the Sure 3116 getting warm.
I thought about putting the TPA3116 on the bottom side of the boominator card - problem is the chip is barely 1mm high, which means the PCB will end up 1mm from the plate. You'd have to use fully SMT parts on the board, or cut through-hole leads flush with the board so they don't touch the plate.
If you want to put a heatsink on this card, I'd say just find some thick brass or copper sheet, cut it/bend it to whatever shape you need to clear everything, and solder it straight to the top of the TPA.
Anyway, here's the latest. AVCC decoupling added.
If you want to put a heatsink on this card, I'd say just find some thick brass or copper sheet, cut it/bend it to whatever shape you need to clear everything, and solder it straight to the top of the TPA.
Anyway, here's the latest. AVCC decoupling added.
Just thinking aloud~
using both polymer and electros seem a little extravagant (looking at yer chart above) if you could combine them into one part wouldn't that help clear space for a proper heat sink? ( into a 8mm electro eg low esr "tall boy" )
or use a heat spreader shim on the bottom chassis plate to increase the chip distance to match the bottom PCB 2 mm standoffs.
using both polymer and electros seem a little extravagant (looking at yer chart above) if you could combine them into one part wouldn't that help clear space for a proper heat sink? ( into a 8mm electro eg low esr "tall boy" )
Last edited:
PaulChen TI's answer is more "open to" than his answer to gmarsh a year later for 3118, I don't see discussions btw. Here on diyaudio I read differences down to 0.5V gave added audible noise, so below 0.5V noise performance is improved? Or just not audibly worse? People hear all kind of improvements, did someone measure if chip does improve? What is voltage difference 10R/1uF on like 15V dc input and on 24V dc input? I read FB in R position took away added audible noise problem too, is this also measured improvement or just heared by some. I remember reading advise to remove FB's from DCinput EVM when some noise problems occurred, what would cause added noise problems using FB on DC input?
I don't get the question? The static voltage difference is 0, while the dynamic difference is a function of current. It's mainly a lowpass dv/dt limiter with hf rejection. You could also go for an LC, an RC is cheaper and sufficient.
gmarsh, why not rotating the avcc cap by 90 deg?
gmarsh, why not rotating the avcc cap by 90 deg?
Last edited:
Don't know that is why I ask, to me the 10R resistor and the impedance of the capacitor look exactly like a voltage divider and that would alter voltage I am guessing ?
The PCB layout is coming along very nicely. I'm ready for the GB. 🙂
Another alternative that can work well is to solder 1/4" faston connectors to the ground plane / heatsink. Of course provisions are needed in the PCB. I don't think they are needed for this board though.
Another alternative that can work well is to solder 1/4" faston connectors to the ground plane / heatsink. Of course provisions are needed in the PCB. I don't think they are needed for this board though.
Last edited:
I'll give that a try. I crammed that cap in there while attempting to move as few parts as possible, if I move the surrounding parts further away I'll be able to make it fit.
The 0603 MLCC is only 0.2mm thinner than the TPA (I wouldn't depend on that gap) and the 0805/1206 X7Rs are both taller. You're modifying a heatsink for that anyway.
All of this is why I'm going with the 3118. The PCB ground plane acts as a heat spreader, increasing the hot surface area and allowing a thick thermal pad to be used to pull away heat. With normal music listening into 4 ohm speakers, or running full scale tones into 8 ohm speakers, you probably won't even need the pad - the PCB by itself should be adequate.
And yes, the decoupling is extravagant, that's the point of it 🙂 The polymer cap is 82uF and the wet electrolytic is 470uF.
I'm gonna try to make the wet electrolytic 13mm diameter if I can make it fit, so I can get more capacitance while keeping the board height sensible. Currently the tallest part is the 470uF cap (20mm H) followed by the output inductors (17mm) and input film caps (13mm).
yer going the wrong way by adding larger caps for decoupling, at some point the power supply impedance dominates at low freq. esp. for regulated supplies ( negative feedback enhances the transient response and output impedance) a SMPS closed loop BW is typically ~ 1/20*Fsw and the output capacitor used to filter the switching ripple voltage determines the impedance after that. In fact adding too large of load capacitance spoils the loop dynamics. for unregulated supplies the caps are usually ginormous anyways.
In my experience you get the most performance with decoupling caps with smaller lead spacing eg 'tall boys' a 8mm x 20 mm electro should equal the SRF of the polymer. the 3.5 mm lead spacing of this can take the place of both parts EKZH350ELL391MH20D United Chemi-Con | Mouser
In my experience you get the most performance with decoupling caps with smaller lead spacing eg 'tall boys' a 8mm x 20 mm electro should equal the SRF of the polymer. the 3.5 mm lead spacing of this can take the place of both parts EKZH350ELL391MH20D United Chemi-Con | Mouser
Last edited:
With 6 Ohms speakers I make do with a small heat sink equivalent to tpa3116 chip area + legs. Heatsink becomes only luke warm when pushed! Heat is not a problem with this chip when the layout around the tpa3116 chip is optimal.
Matt Garman, I have a heatsink all the way into the alu casing above^^. (self made copper column which is completely overkill!). But I've used "a tree" in the past with success.
There is also no need to solder on the heatsink, superglue in 1 corner and the rest with paste is more than enough.
Also, is it possible to fat'n up the last part of the speaker traces just a bit more??
The board is coming along nicely and its looking glorious !
Matt Garman, I have a heatsink all the way into the alu casing above^^. (self made copper column which is completely overkill!). But I've used "a tree" in the past with success.
There is also no need to solder on the heatsink, superglue in 1 corner and the rest with paste is more than enough.
Also, is it possible to fat'n up the last part of the speaker traces just a bit more??
The board is coming along nicely and its looking glorious !
Hmm:
EEEFT1E821AP - PANASONIC - CAP, ALU ELEC, 820UF, 25V, SMD | Farnell element14
820uF 25V.
For 35V 560uF are avail at the same height:
EEEFT1V561AP - PANASONIC - CAP, ALU ELEC, 560UF, 35V, SMD | Farnell element14
^
the cap in my link just above, has less capacitance yet has almost half the esr of both of yours , plus it should stay lower at higher frequencies due to narrow lead spacing.
the cap in my link just above, has less capacitance yet has almost half the esr of both of yours , plus it should stay lower at higher frequencies due to narrow lead spacing.
Damnit, just lost a couple hours taking in power supply schematics (how much C does this guy use on the output? hmm...) and notes on compensation. You've got a point, maximizing the onboard C probably isn't necessary, and maybe I shouldn't fire 2200uF of low ESR capacitance onto an off the shelf PSU... I'll keep the 10mm diameter cap and 470uF for now.
Still trying different caps in the impedance calculator.
PSU's I use indeed have reduced dynamics when 4 2200uF's KZH on two monoboards are on one PSU compared to one board one PSU. The lowimpedance speakers I use do however make even 4 2200uF's with one PSU sound more dynamic than 4 1000uF or 4 330 uF. 8 ohm speakers might not have any dynamic difference, I don't have any. PSU's are off the shelf regulated smps that have two 3300uF low impedance Elna in output, desktop, 145w. Laptopbricks I have seen inside mostly had two 470uF low impedance chinese/taiwanese caps on output, but also a little 2A 30watt one with two 1500uF Rubycon MCZ ultralow impedance LOL so design laptopbricks varies wildly too I guess.
let's hear it for the 'tall boys'
mounted normally yes,
besides, IDK about some secret height requirement.🙂
decoupling should match the load current and dynamic ripple voltage specs seen at the power device ( oops class D switcher ) using some shorter time constant. FWIW this isn't class AB with unregulated rails. if you hear bass notes changing with bigger decoupling caps, something is wrong with yer DC supply.
Last edited:
I'm aiming to keep the board with the "official" parts list at <20mm H, for no reason other than it fits nicely in a Hammond 1455L box with a couple mm to spare for capacitor venting, and I figure other builders may appreciate the low profile.
But the parts list isn't going to be strict, if you're buying a bare PCB or "SMT only" PCB I'm not going to stop you from installing your own parts. I'll even customize a fully built card if you want...
Thinking ahead to the group buy, here's what the options are going to be:
#1 Bare PCB. Order your own parts.
#2 Bare PCB with TPA3118 and minimum SMT installed to allow testing of TPA3118 (bootstrap caps, GVDD cap, close-in decoupling caps, etc). Order the rest of the parts yourself.
#3 PCB with all SMT installed/tested. Optionally available with a kit of TH parts to complete the board, or order your own TH parts.
#4 Fully built PCB.
Kitting SMT is a pain in the *** and I'd rather not do it. Options #1 through #3 will be shipped in an envelope, #4 (and #3 w/parts kit) will be shipped in a box.
I'll do customization of anything, within reason. If you want different power caps, different output LC parts, different input coupling caps, a board configured for slave operation, parts left off the board, etc... I can do all that.
But the parts list isn't going to be strict, if you're buying a bare PCB or "SMT only" PCB I'm not going to stop you from installing your own parts. I'll even customize a fully built card if you want...
Thinking ahead to the group buy, here's what the options are going to be:
#1 Bare PCB. Order your own parts.
#2 Bare PCB with TPA3118 and minimum SMT installed to allow testing of TPA3118 (bootstrap caps, GVDD cap, close-in decoupling caps, etc). Order the rest of the parts yourself.
#3 PCB with all SMT installed/tested. Optionally available with a kit of TH parts to complete the board, or order your own TH parts.
#4 Fully built PCB.
Kitting SMT is a pain in the *** and I'd rather not do it. Options #1 through #3 will be shipped in an envelope, #4 (and #3 w/parts kit) will be shipped in a box.
I'll do customization of anything, within reason. If you want different power caps, different output LC parts, different input coupling caps, a board configured for slave operation, parts left off the board, etc... I can do all that.
- Status
- Not open for further replies.