Amplitude does change, are you guys saying it can't or are you talking about something else, a note isn't same as dynamics but could be language?, I thought dynamics was talked about? Depending on electrolytics and amount of PSU's used for 2 monoboards driving 2 ohm or maybe even 1.5 ohm speakers (impedance drops with frequency, above 8k with signalgenerator ampboards switch off below 1 watt output, with music I haven't noticed), but even on zobelled 3.6ohm fullrange cone speakers dynamics decrease just less dramatic. I thought it was 10A PSU being insufficient, could be if TI advices 8A per ampboard, but here I read too much capacitance could reduce dynamic capabilities of regulated psu, so I thought maybe that is what is happening here connecting 2 boards on one PSU? Or part of it might.
my rough guideline for adding external capacitance on a regulated SMPS is don't add more than 100% of the PS internal value (on that secondary output). The power supply will handle most audio dynamics if you haven't exceeded any limits , PS max current , chip amp internal protections, wiring drops etc.
edit> of coarse some PS are bad quality / designs not fit for daylight, so it pays to check things out closely.
edit> of coarse some PS are bad quality / designs not fit for daylight, so it pays to check things out closely.
Last edited:
And if you've got access to a scope, I'd suggest getting a handful of power resistors and banging on/off loads on the supply to check the transient response with the added capacitance. Make sure it's overdamped, the supply should dip/peak when you add/remove a load, and return to regulation with zero overshoot/undershoot.
Ringing is very bad - if you play a tone at 1/2 the ring frequency through the amp, causing current draw at the ring frequency, you could get the supply oscillating and blow it (and your amp) up.
I haven't looked at laptop power brick in the lab, but really what can you expect of $10-20 100W flyback > all the energy storage is in the gapped transformer and output capacitor, they don even bother filtering the mains voltage, expect bad 120Hz ripple and poor load transient. so .. gets what you payfor. edit > those things are designed for charging a DC battery, not HQ audio !
Last edited:
at a min get a real power supply ~ 25 dollars Meanwell Power Supply Products
safety marking > yes
3 year warrantee > check
manual select 110 V / 220 V > yeah ok not bad
hold up time 20 msec > real line storage / filter caps
safety marking > yes
3 year warrantee > check
manual select 110 V / 220 V > yeah ok not bad
hold up time 20 msec > real line storage / filter caps
LT108x doesn't like lots of capacitance if you're relying on the internal compensation, add a cap from ADJ to OUT to push the phase margin back if you're loading it up.
I built a simple MOSFET switch and resistor load. The gate is driven from a function generator. works great.
here's a laptop power adapter reference design. ( see link at bottom) Look at the 3x 4700 uF output caps. also notice no big caps after the AC input bridge so the flyback is essentially off during every 8msec zero crossing. Basically all the energy needed during this off time is stored at the output. More load = more caps needed or/ more line ripple. This also affects loop dynamics during the zero crossing = unregulated part of the time.
They call it "inherent power factor correction" like it's a great feature. LOL Its a really cheap-n-crappy way to build a power supply if you ask me.
http://www.onsemi.com/pub_link/Collateral/AND8397-D.PDF
Last edited:
the ripple to noise for 24v = 120mVp-p
That's not so good.
I got mine debugged down to 55 mv at 24V 5amp.
That's not so good.
I got mine debugged down to 55 mv at 24V 5amp.
yea it's all relative 120mv/24 V = 0.5%
I reckon its light years ahead of the laptop adapter solution.
Class D + SMPS = common mode hash city anyhow
how to measure anything lower than 55mV without limiting the bandwidth and using true differential probes.
here's a lil taste https://www.youtube.com/watch?v=BFLZm4LbzQU
I reckon its light years ahead of the laptop adapter solution.
Class D + SMPS = common mode hash city anyhow
how to measure anything lower than 55mV without limiting the bandwidth and using true differential probes.
here's a lil taste https://www.youtube.com/watch?v=BFLZm4LbzQU
Last edited:
FET + function generator will certainly do the job.
That laptop design is definitely no worse than an unregulated linear supply - it's dumping current into the output caps for a longer "conduction angle" than you'd get on the linear. Plus laptops don't really need clean/ripple free 19.5V, as they only step it down to whatever other voltages they need instead of directly using it for anything important. Overall, the adapter seems pretty well designed for the application. And hey, it could be worse, it could be a ringing choke converter.
Current PCB. AVCC cap is still vertical but I like the layout a lot better. Someone on here wanted fat traces on the output, decided to overkill that, now the board looks like it needs fat copper on the right hand side 🙂
A few things left to do:
- revisit comparator circuit. I'm thinking there's some clever way of using a power supervisor/sequencer IC to do the job more simply, and with an absolute startup/shutdown threshold on PVCC.
- ponder if there's some way to make the gain switches work in slave mode.
- add pads for pogo pins to allow testing partially built boards, ponder a test jig design
- come up with BOM/price/etc.
yes they work fine for their intended apps eg laptops with their internal buck converters. I wouldn't say they're all that great for audio stuff unless you add some extra clean up circuitry and know what to watch out for.
fatter traces always look nicer, sometimes they couple more junk into ground planes "CMOS ground bounce". but hey that's why we use differential I/O's everywhere.
if your layout starts becoming a lot different than TI 's suggested layout artwork, you may have unintended consequences to deal with later. I don't think our gang here is any smarter than TI's.
fatter traces always look nicer, sometimes they couple more junk into ground planes "CMOS ground bounce". but hey that's why we use differential I/O's everywhere.
if your layout starts becoming a lot different than TI 's suggested layout artwork, you may have unintended consequences to deal with later. I don't think our gang here is any smarter than TI's.
Last edited:
here's a ESR meter thingy I came across http://ludens.cl/Electron/esr/esr.html
I like the transformer idea!
I've used the same principle (a voltage divider with known resistor) with a swept signal generator. finding the minima impedance also determines ESL by series resonance. I 'd probably redesign it for a factor of x10 lower range with a sig gen. input or switched to an internal NE555 at 100Khz.
I like the transformer idea!
I've used the same principle (a voltage divider with known resistor) with a swept signal generator. finding the minima impedance also determines ESL by series resonance. I 'd probably redesign it for a factor of x10 lower range with a sig gen. input or switched to an internal NE555 at 100Khz.
Last edited:
+1 for small ferrite cored trafos 😎
I'd probably redesign this to have switchable frequency and force/sense connections on the probes. I'm concerned about repeatability when the probe wires themselves might be subject to proximity effect losses.....
Last edited:
at 1/10 range the contact resistance would need attention think HP gold plated test fixtures spring loaded thingy.
the transformer loads the op amp at 2000 ohms I think it needs a new op-amp to drive lower ohms yet. maybe OPA2350 for >>200KHz full power BW?
the transformer loads the op amp at 2000 ohms I think it needs a new op-amp to drive lower ohms yet. maybe OPA2350 for >>200KHz full power BW?
Last edited:
On a 2-layer 1.6mm FR4 PCB, there's about 15pF of capacitance to ground per square inch of top layer copper. The fattened up outputs already have 0.68uF to ground, and slew rates are pretty slow at that point, so I'm not too worried there.
I don't plan on making planes for the OUTx outputs on the TPA though.
Lately measured the output-slewrate at our 3132D2 running filterless, having 6-8ns at PVCC=20V.
Thought about that in the past but rejected due to costs. (HV-supervisors are expensive, $3-4, LV-supervisors needs some adaption.. so in the end.. comparator)
Thought about that in the past but rejected due to costs. (HV-supervisors are expensive, $3-4, LV-supervisors needs some adaption.. so in the end.. comparator)
Last edited:
- Status
- Not open for further replies.