Best 5V SMPS ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Unfortunately there is no other than the Amplifier-Power Supply section at DIY-A.

I give it try over here with this topic.

***

I'm looking for a very good (DIY-) 5V SMPS ( and/or tweaks) out there. Yep - no linear supply!!!

Applications are endless for these.

RPI, HAT devices, routers, PC peripherals, you name it.

Right now I'm quite happy with the iFi iPower. With 2.5A it also delivery plenty of power.

iFi claims 1uV noise. Which obviously can be considered quite a low value. However. iFi excludes mains and switching related noise. Hmmh. And they limit it to the audio band. Hmmmmh I'd be interested to see what's left of that 1uV - if you take the exclusions out.

Anyhow. I still think and know from experience that the device performs really well.

I tried to put a Meanwell RS15-5 (@5.5V) + TPS7A4700 up against the iPower. That combo didn't perform on the same level.

Any ideas for approaching the subject are welcome. Maybe also ideas of how to get e.g. the Meanwell noise (80mV) down would be good starting point.

Cheers

BTW: I added some mods to the iPower. I shortened the cable down to around 12 inches and I also added a large buffer cap at the end of the cable.
 
Last edited:
For stuff like a Pi or a router, power supply noise is irrelevant. If it has 5V input, internally it must run on 3V3 and other lower voltages, which means it has internal regulators, most likely switching.

What is relevant:

- If target device is connected to an audio system through cables, then common mode current generated by the power supply will flow into those cables and contaminate grounds. This really is the most important spec. It is really hard to get rid of common mode currents. Power supply design needs to take this into account right from the start.

- Radiated electromagnetic fields

- Amount of noise injected into the mains (and picked up by nearby equipment).

- Fire & electrocution hazard with "too cheap to be true" phone chargers...
 
R u kidding me?

Well, unless there is a big design mistake like :

- using the Pi as the DAC master clock
- using the same power supply both for the Pi and the DAC without appropriate care

Then yeah, it's pretty much irrelevant. Besides, if you measure the Pi's supplies while it works, or even the output of your "low noise" supply with the Pi connected and powered, I'm pretty sure you'd find the usual huge amount of noise associated with digital supplies. Go on, grab your scope probes and post a screenshot, it takes very little time...

> But - good quality psu are never wrong.

Sure!
Just don't believe that because it has "1µV noise" written on it, there won't be 100000 times more noise once you connect a load which draws wildly varying current...
 
Practically , I think that first you need to chose the design of the smps. In my opinion , the best candidate for a 5v (audiophile) solution will be the quasi resonant fly back. You will have less EMIs then any other solution.

Of course then you need to solve the ripple/noise/impedance .
 
It all adds up doesn't it ?

First design the SYSTEM, including calculating what the performance requirements are for each subsystem, then pick the subsystems based on those requirements, this is good design.

The 'Best' is not a meaningful requirement. 100uV RMS noise over DC to 20MHz (The usual measurement bandwidth for most power supply datasheets), between 20 and 100% load, is a possible requirement. Common mode noise currents less then 1mA over DC-20MHz is a possible requirement. Greater then 50ms hold up at 100% load is a possible requirement, MBTF > 100KHrs at 100% load, 60C ambient is a possible requirement, maybe smaller then 8 cubic inches is a possible requirement....

Seriously, 1uV in an uncommonly narrow bandwidth? When powering a high speed processor supplied by the cheapest on board buck converters possible? Who cares? It is not like that noise contribution is even going to be visible in the roar of non synchronised spread spectrum clocks and bucks on the CPU card.

Now if you are designing either a precision clock or small signal analogue then you care, but you start from a completely different place, LC filters and low noise linear regulators having PSRR up to above where the LC filter cuts off, then you start worrying about the grounding, signal integrity and (sometimes) screening, but first you get your matlab on to see how good you need to make things, only once you have the numbers do you start shopping.

Then you LISTEN, and Measure, and LISTEN some more, and have your friends around and LISTEN doubly blind, and measure some more... One thing I quickly realised was that if I could hear it, I could measure it, but that correlating what sounds good to what measures what is a difficult problem, still good design will usually tell, and good design is NOT automatically the use of the most expensive possible parts.

Sometimes you can get sneaky, I did an ADC/Mic preamp design powered by a switcher, I locked the switcher to a harmonic of the word clock so that any switching noise that got into the analogue section would alias to DC and be removed by the digital filters. It effectively made the leakage from the switcher pretty much a non issue, I mean it was 100dB down anyway, but when I fitted that jumper and the loop locked you literally could not see it in the output any more.

ZVS quasi resonant converters are generally MUCH cleaner then bucks or flyback converters, but even they can be tamed if you are smart about it (Cheap ~60W flybacks are often bad news at ~25MHz or so due to the reverse recovery charge and transformer parasitics, this seems to be worse on the class VI efficiency ones, I assume due to less snubber).

I would recommend that anyone building stuff get a few books on RF design, you cannot get away from it these days, and certainly anyone using logic faster then the 4000 series or doing ADC/DAC based things needs to think in terms of RF not audio.

Regards, Dan.
 
SO basically we have 2 choices..ZVS quasi resonant or just plain quasi resonant. The ZVS is a more expensive/bigger design and quasi resonant is cheaper and smaller

On the design that I proposed, first thing will be to add a RC snubber on the bridge rectifier. It will reduce the noise radiated in mains and smooth further the DC
 
The ZVS is a more expensive/bigger design and quasi resonant is cheaper and smaller
About right, depends on how much efficiency matters at high load for your application.

The efficiency advantage of ZVS only really becomes apparent once the savings in cooling start to matter (Usually cooling is the elephant in the room where efficiency is concerned rather then the direct energy cost), I use them in radio transmitters but usually pick something simpler for most audio doings.

Actually if you are smart about your design and layout, even a good quality flyback can be fine, you just need to design based on what you know it will do.

Regards, Dan.
 
Great to see that you guys (@Allo) are working on a PS product. I hope this time a single voltage (e.g. 5V) device.

Some things to keep in mind:

1. Don't forget that the "numerous" cable parameters matter big time This plus the right choice of plugs might have ten times more impact than your observed oscillations.

2. A good balance of risetime plus current delivery has to be found Overregulation, overdamping/-filtering, high ESR, low buffer capacity might be some of the things to eliminate as much as possible. I fought this by adding low ESR buffer capacity to the iFi. 3. I'm not sure if a flexible DC cable length wouldn't have been a better approach on the iFi. Fixed length and then rather long DC cables are not what I'd call the ultimate solution.

4. And the other thing which IMO might not be the best of all approaches is to have the supply sitting right in the mains distribution block. I'd rather see it closer to the device.

5. iFi IMO has done a great job by supplying several uselful adapters. Something you want to think of if you intend to be commercially successful.

Just my thoughts (based on my experiences).

Looking forward to an iPower alternative. ;)

Good luck.
 
Last edited:
Hi

1. Are you referring to better cables (like wire thickness ?)
2. Already taken care of... I am sure you will like what we did :)
3. Not enough space on our casing.. if you referring to retractable cables.
4. Agreed. Taken care of.
5. Included..I am thinking if a specif material (like cooper) will be better than standard.
 
"Better cables"

Obviously that's a topic on its own. iFi uses coax + ferrite bead. There are some advantages by using coax. Decades ago Thorsten Loesch - the man behind the iPower - wrote an article on TNT about interconnects. He applies pretty much the same principles for the iPower DC cable nowadays.

There are probably better choices. But the coax might be a good (or the best) compromise. However. The "targeted" retail price will probably dictate your choice. If you go for a rather cheap and/or rather inflexible solution, I'd appreciate if you'd keep the DC outlet DIY friendly. Meaning: I should be able to rather easily open the case and attach (even soldering is OK ) my own custom cable!

Cheers
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.