New DIY idea: USB power conditioner - first layout

[balou]

original posting below. see post #17 for this prototype

----
Just had an idea while writing a reply in this usb dac mod thread: http://www.head-fi.org/forums/showthread.php?t=244738

This and other DACs (for example the headroom total bithead and the aliendac) both take some or all of their power from the usb port of a computer. Computers don't really have cleanly regulated 5v lines, so a lot of noise will come along.

So my idea I just had: Make a little box with an usb input and an usb output. Pass the data lines straight thru (maybe add some ferrites for filtering out unwanted HF noise). Leave the incoming power lines unconnected (ok or maybe you have to connect them so that it works, then just add a high ohm resistor so that 1-2mA is flowing). And in said box, there's a TREAD or some other PSU, which will put out a nice clean 5v voltage. Input is a wall-wart, 7 to say 12v (you aren't getting more anyway with consumer wallwarts).

Benefits:
-lower psu impedance
-better regulated voltage
-could provide the full 500mA, other usb ports don't always provide this (once it was meant that it was 500mA max for _all_ your usb ports, not per single port)


Just had to post this idea. Maybe this is useless, but I hope someone will find this idea interesting.

 

[MisterX]

I believe externally power USB hubs are pretty common.
I have even seen battery powered ones.

 

[balou]

Yes I know, but most of them will have a rather bad power supply in them. And DIY is more fun

Anyway, this is meant as audiophile grade power supply.. a tread with a usb passthrough and 5v out

 

[Illah]

That's my thread, and I like the idea

This would be used with a powered hub so it wouldn't be *as* bad as from the laptop (which are notorious, even worse than PC's). I could get a premium wall-wart for the hub but not sure what difference that would make.

How about an even simpler hack - just get a decent Elpac regulated 5V wall wart, plug into a mini-power filter like the Monster AV200, and hack the lines into a USB cable? Would that even work?

--Illah

 

[MisterX]

It would be easy to hook up a battery pack and see if "cleaner power" really helps (that way you would know for sure if this is a worthwhile pursuit for your device)

 

[error401]

Quote:

Originally Posted by Illah /img/forum/go_quote.gif That's my thread, and I like the idea This would be used with a powered hub so it wouldn't be *as* bad as from the laptop (which are notorious, even worse than PC's). I could get a premium wall-wart for the hub but not sure what difference that would make. How about an even simpler hack - just get a decent Elpac regulated 5V wall wart, plug into a mini-power filter like the Monster AV200, and hack the lines into a USB cable? Would that even work? --Illah

You don't want to hack the wires directly into the USB cable, at least as I think you mean it. If you connect the +5V from the powersupply to the USB V+, there will undoubtedly be a voltage difference between the two, and the supplies will be pulled up or down, fighting the regulation and probably sinking large amounts of current (which will likely be burned off in the regulators). This is possibly damaging to your motherboard (or hub), and will probably make the DC regulation much worse rather than better.

However, it should be safe to tie the grounds together. You can then cut the USB V+ lead and insulate the side with power coming from the computer/hub, and connect your better supply to the end going towards the device to be powered. There are a couple of possible issues here. As far as I understand it, many USB devices expect the data lines to be connected and online when power comes up, and will malfunction if power is connected and data is not. If powering such a device, you'd need to connect the computer and power supply first, and then connect the device to the cable. I can't think of a simple solution that would maintain isolation.

This follows to what balou is talking about, as the topology would be basically the same. A couple things to say about that. First, you won't get 500mA out of it just because the power supply can provide it. The USB protocol specifies that a device must negotiate with the upstream device if it requires more than 100mA. What we're talking about is a passive device, and wouldn't be able to respond to such a negotiation request, thus the upstream hub will act as if the power were allocated from its supply and may still deny requests. Also, you can't just pop ferrites on data lines blindly and expect things to be okay. Data lines carry digital signals (ie. square waves) that have a lot of power in high frequencies by design. Filtering the high frequencies must be done carefully or you'll break the signal and the receiving device won't be able to properly detect the signal's value. It'll also introduce introduce phase delay versus the rest of the USB bus that may cause issues. As long as the bus works reliably though, you should succeed at reducing noise due the proximity of the power lines.

 

[balou]

Quote:

As far as I understand it, many USB devices expect the data lines to be connected and online when power comes up, and will malfunction if power is connected and data is not. If powering such a device, you'd need to connect the computer and power supply first, and then connect the device to the cable. I can't think of a simple solution that would maintain isolation.

Well, doesn't a usb host powered device need the power first so that it can start and only then it initiates the data connection to the computer? Anyway, if some devices would require that, I can think of a simple solution: Make the data lines a constant passthru, and add a switch which turns the power on/off. I just reread my last sentence.. and this description somehow sounds a so-called "power-switch". Yeah, I just reinvented the power switch

.

Quote:

First, you won't get 500mA out of it just because the power supply can provide it. The USB protocol specifies that a device must negotiate with the upstream device if it requires more than 100mA

According to the specifications, a device should request if it needs a larger current than 100mA, and a usb host may deny to give more than 100mA power if it's not requested correctly. But as I understand, this is one of the most abused usb specifications... there are for example external harddisks requiring 1000mA, and they just have a Y-splitter cable, one completely passive, just to tap off another 500mA (disregarding the 'request first' rule and disregarding the 500mA _total_ power for _all_ usb ports together rule).
But this doesn't really matter. The conditioner will be built in such a way that it can provide 500mA at any time, regardless if the power is correctly requested or not. If the upstream hub denies that request, it wouldn't work either way, with or without the power conditioner. But most usb host devices have responded to the abuse of the specifications in that they will provide 500mA on every port, regardless of the once-in-effect specs.
Quote:

Also, you can't just pop ferrites on data lines blindly and expect things to be okay. Data lines carry digital signals (ie. square waves) that have a lot of power in high frequencies by design.

Yes I remember now. Fourier transformations of square waves leads to an infinite number of sine waves.


A couple of things I'm stil not sure about:
Is there a need for a minimal current draw on a usb port so that the computer correctly detects that there is a new device connected? If yes, there is maybe a need to add a resistor from Voltage in to Ground. 10k-20k should be okay, it will draw 0.25-0.5mA of current.

Must the input ground be connected to the output ground? The power conditioner and it's output ground will be grounded via the power supply. This might be figured out empirically (i.e. just make room for a wire bridge in the circuit between in ground and out ground, and test which way it sounds better)


edit: it seems that the basic modes of operation are set by tying a resistor to ground. So a connection between in and out ground is needed, OR the resistors could be substituted in the conditioner.
still leaves the question, is it a good idea to separate grounds?

edit2: read the usb 2.0 specifications, pages 119 to 148, for information regarding the electrical specs which must be followed. to have separate grounds, it could be necessary to use two high speed buffers in the signal path. I'll have a closer look at it...

 

[balou]

Ok, I think a separation of the ground planes would be far to complex.
This could be another project... a fully galvanic isolated usb hub.

For now I will keep the grounds connected together. Maybe I draw a schematic and make a pcb layout this evening.


edit: Quote:

I could get a premium wall-wart for the hub but not sure what difference that would make. How about an even simpler hack - just get a decent Elpac regulated 5V wall wart, plug into a mini-power filter like the Monster AV200, and hack the lines into a USB cable? Would that even work?

missed that one... a premium wall wart would not do much of a difference after a regulator circuit like the one I'm trying to implement. I suppose you could even use a switched wallwart. Second hack would be an idea, but look at tangents noise measurements for generic wallwarts... one word: bad
The mentioned power filter, on the other side, would most likely not be much help, I think it would only make sense if you need directly 110v, after a low voltage power supply it's effects are dubious (look at the TREAD vs STEPS measurings, the STEPS has ac filtering, the effects are very minimal). yet another edit: I will implement a TREAD, which is basically the circuit directly taken from various LM317 datasheets

 

[fordgtlover]

^

To revive a resting thread, have you done any more on this?

 

[balou]

hi fordgtlover,
split ground planes seem unfeasible with just a simple design, so I abandoned it.
what remains is just a basic regulator circuit with a LM317, which can be found in the accompanying datasheet, or on tangents site as TREAD circuit

Here's the link to the TREAD site: http://tangentsoft.net/elec/tread/
for 5v, the resistor R2 should be 110 ohm. no other modifications are needed

Here's the usb pinout:
http://pinouts.ws/usb-pinout.html
You just have to chop off the VCC line and replace it with the one from the TREAD. connect ground from the tread and from the usb line (on both ends of the cable) together

and use a version of the lm317 with a 500mA current limit - you don't need more for usb

that would be pretty much everything that is needed for this design.
if there's some interest for a small pcb for a conditioner which will run off a dc wallwart (as opposed to the ac wallwart design of tangents board), I'll design one. if several people are interested, we could even think about a pcb group buy

 

[fordgtlover]

Is there any reason that I couldn't take a 12v line from my PC to run the LM317?

 

[balou]

no, not really. pc power supplies tend to be noisy, but the regulator should take good care of this.

you might also try to get the power off the +5v/+12v lines instead of gnd/+12v lines. this would yield to less wasted power and thus to less heat in the regulator. but the resulting 7v might not work well, especially if your computer psu itself already has problems providing enough voltage. you'd have to try


I have enough free time right now, I'll design a pcb which will fit in some hammond case (dunno how big it's gonna be)

 

[MASantos]

but most USB DAC's already have good power regulators inside, see the alien dac for example. AMB did some testing with the Alien DAC using USB power, and external power, and couldn't find a difference in sound.

 

[balou]

MASantos: yeah, the DIY products will most likely provide good enough regulation, but with commercial products you can never be sure

and hey, double regulation couldn't hurt

I'm btw doing the schematic right now. it kinda sucks... search parts, search parts that are in an eagle library ^^, read datasheets only to know as much as before etc

edit: case will be either hammond 1455C801 or 1455J1201.
one problem I still have is that I have to draw the usb b connector... dunno yet how to do that in eagle

 

[balou]

here's a first prototype

the usb out connector isn't there yet, only a placeholder in form of the pins, I haven't thought about heatsinking yet, traces are probably to tiny, and the layout looks rather strange at the moment. but it's bed-time here, so no more updates for 12h or so