Anyone "regenerated" USB power line ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
There is a whole bunch of DACs nowadays requiring single +5V power supply. And the DACs are often intended for USB, so the +5V comes from USB.

USB host controller in its turn gets its power from the computer PSU, and typically the PSU is "dirty", so the USB +5V is most likely dirty too.

So, practical questions.

Has anyone "played" with assessing USB powered DAC sound quality as a function of USB +5V noise + ripple ?

Has anyone tried to improve the situation ? If yes, what was the approach ?

For example, it is possible to double the +5V into +10V, and then to use a linear low noise regulator to get back the needed +5V, but the voltage doubler should not be "evil" introducing more noise with which the linear low noise regulator won't be able to cope.
 
Simplest...

solution if one is using a DAC or USB-SPDIF converter which uses the 5VDC power from USB, is to build a breakout adapter which brings out the 5VDC connection point (and breaks this connection to the computer) and the ground connection (but this must stay connected to the computer) and then apply a low noise regulated supply externally.
Back when I used a Wavelink Async USB interface I made this mod and it made a nice improvement to sonic performance.
 
USB Regeneration.

Nowhere
There are some ideas at the attached link. Despite the protestations from an R.G. member who is an EE,(with views similar to your own posted views in DIYAudio), that it can't possibly make any difference, quite a few were constructed by other R.G. members with very positive results.
It should be just as effective on USB powered DACs as it is with external HDDs, Optical devices and USB memory.

As the highly experienced "Barrows" has already said, the noisy +5V from the PC should be isolated.
There is a simplified version on p.4 of the thread.
Alex

Rock Grotto Audio Forum - For Headphones - Headphone Amps - Amplifiers - X-Can V2 - Musical Fidelity - headphone Discussion - Amplifier Discussion - DIY - Amplifier Kits - Projects - SCHA - Sennheiser - Beyer - Grado - Audio Technica - Headphone amp

Full info on the JLH PSU Add-on is at Greg Erskine's homepage. Click on Power Supplies. Other Super Regulators could also be used instead of the JLH.

Greg's Web Site
 
solution if one is using a DAC or USB-SPDIF converter which uses the 5VDC power from USB, is to build a breakout adapter which brings out the 5VDC connection point (and breaks this connection to the computer) and the ground connection (but this must stay connected to the computer) and then apply a low noise regulated supply externally.
Back when I used a Wavelink Async USB interface I made this mod and it made a nice improvement to sonic performance.

I am not sure I understand what you mean.

In this thread I am not interested in some kind of clean external power supply. I am specifically interested in using USB +5V as primary source and cleaning it. I.e. let's assume I have a design requirement not to bring into the pictures external power supplies, I am only allowed to use USB +5V.
 
Nowhere
There are some ideas at the attached link. Despite the protestations from an R.G. member who is an EE,(with views similar to your own posted views in DIYAudio), that it can't possibly make any difference, quite a few were constructed by other R.G. members with very positive results.
It should be just as effective on USB powered DACs as it is with external HDDs, Optical devices and USB memory.

As the highly experienced "Barrows" has already said, the noisy +5V from the PC should be isolated.
There is a simplified version on p.4 of the thread.
Alex

Rock Grotto Audio Forum - For Headphones - Headphone Amps - Amplifiers - X-Can V2 - Musical Fidelity - headphone Discussion - Amplifier Discussion - DIY - Amplifier Kits - Projects - SCHA - Sennheiser - Beyer - Grado - Audio Technica - Headphone amp

Full info on the JLH PSU Add-on is at Greg Erskine's homepage. Click on Power Supplies. Other Super Regulators could also be used instead of the JLH.

Greg's Web Site

I followed the links. The forum one requires forum membership in order to see full sized schematics, and I am not a member of that forum.

Anyway, I think it's not what I need. If I understand correctly, the folks entertain the idea of an "electronic capacitor". I.e. active circuitry which monitors ripple and tries "to do the opposite".

Alas, this doesn't quite work. Because no amplifier has infinite bandwidth and ripple nowadays is quite wide band, especially from computer PSUs.

I want something else:

1) to increase the voltage using some kind of switching power supply;
2) to filter the output combining capacitors and transistor filter. The latter is the simplest one like at the top of kpdf http://brettcave.net/howto/circuits/Transistor_VoltageRegulators-BJT-Simple.pdf .

HF ripple rejection is determined in the worst case by the transistor collector-emitter capacitance and output capacitor - they form C-C AC voltage divider.

From AC point of view base voltage is constant, and since the transistor is used as emitter follower, emitter voltage is constant. I.e. ideally as long as the transistor is not in saturation, ripple is "passively" rejected (and that's the key). By "passively" I mean that dynamic impedance between collector and emitter is high and output impedance, since it is voltage follower, is low.

I have actually implemented such a circuit and it works. The devil, as always, is in the details. E.g. what to use as the second switching power supply increasing USB voltage.

At the moment I have BJT based voltage doubler. At least, it is better than NE5556 based one I had earlier - less visible wideband spikes.

I can buy a LM2577 based step up converter on eBay. Until I buy it I can't know how the spikes look.

Or I can start looking for stuff like this: http://cds.linear.com/docs/Application Note/an70.pdf (LT1533) - the article is quite interesting, by the way.

But whenever it comes to switching regulator, physical implementation of the board, parasitic inductance and capacitance, etc, becomes crucial.
 
OK...

It seems like it would be easier to just implement an external low noise supply, but clearly you have other needs.
It also appears like you are ready to try this approach, please let us know how it goes. My own feeling is, that the regenerative approach may cause some problems of its own (noise) and you will have to be careful to keep the losses to a minimum: computers supply at most 500 mA via USB, and some do not meet that spec at all. Some USB audio interfaces require almost the full 500 mA, so you are going to have to keep the losses low.
Good luck!
 
http://datasheets.maxim-ic.com/en/ds/MAX743.pdf - nice recommendations on checking whether noise is real or not - pp. 9-10.

Honestly, I have a bad oscilloscope probe - not according to the recommendations.

OTOH, lowering slew rate of the switch is a good way to decrease the wide band spikes (and efficiency :D).

I'm wondering if a capacitor tied to switching transistor collector in LM2577 can reduce slew rate and decrease spikes.
 
Part of the problem is the coupling of the SMPS noise from the +5V Vbus into the adjacent D- and D+ twisted pair. Many people report improvements when using expensive USB cables with improved separation between the power and data leads. Some USB cables have these leads in separate cables, only coming together at the plugs. Look at the +5V Vbus lead at the device end with a CRO, and you will even see low level packets of data in the noise when the device is operational.
BTW, the JLH PSU add-on has been demonstrated by a Sydney member to remove a 1V p/p 100kHz triangular wave on top of the DC from a specialised PSU, into the residual noise level .It has been verified by many people to also work well with +12V and +5V SMPS plugpacks used with external devices.
Improvements such as quietening external Optical readers both from external vibration, and greatly reduced ripping errors when using dBpoweramp have also been reported.
 
Part of the problem is the coupling of the SMPS noise from the +5V Vbus into the adjacent D- and D+ twisted pair. Many people report improvements when using expensive USB cables with improved separation between the power and data leads. Some USB cables have these leads in separate cables, only coming together at the plugs. Look at the +5V Vbus lead at the device end with a CRO, and you will even see low level packets of data in the noise when the device is operational.
BTW, the JLH PSU add-on has been demonstrated by a Sydney member to remove a 1V p/p 100kHz triangular wave on top of the DC from a specialised PSU, into the residual noise level .It has been verified by many people to also work well with +12V and +5V SMPS plugpacks used with external devices.
Improvements such as quietening external Optical readers both from external vibration, and greatly reduced ripping errors when using dBpoweramp have also been reported.

I don't doubt people tested the circuit and saw improvements. But, as I said, "strategically" the approach of chasing HF ripple and trying to compensate for it dynamically doesn't look good to me.

I prefer to deal with HF ripple using passive means. And the transistor filter I described is still passive because it uses huge collector-emitter dynamic (dU/dI) impedance and load impedance as voltage divider. Since load impedance is much lower than collector-emitter dynamic impedance, the voltage divider attenuates ripple significantly. But it is still passive approach.

To put it metaphorically, in this case I prefer wide band passive brute force to electronic band limited smartness ;).
 
breakout adaptor

solution if one is using a DAC or USB-SPDIF converter which uses the 5VDC power from USB, is to build a breakout adapter which brings out the 5VDC connection point (and breaks this connection to the computer) and the ground connection (but this must stay connected to the computer) and then apply a low noise regulated supply externally.
Back when I used a Wavelink Async USB interface I made this mod and it made a nice improvement to sonic performance.
I've made several attempts to input clean battery and or linear psu to my Stello U-3 by interrupting the the USB power from my 2010 mac-mini and inserting the 5v external supply. The Stello's red LED indicates power on but fails to turn green when the music streams. The D+ and D- lines must work by triggering something that the mac-mini's 5v output recognize's. Sure puzzles a body...help anyone?
 
In answer to the origional question: Yes, have a look at some Linear and Texas power supply devices and pages.
The noise in a PC is no primerialy caused by the SMPS, all digital circuitry create noise, the reason digital is used to transmit signals is cos it is unaffected by noise (up to a certain limit, which is way above what the system power has in PC's) and differential signaling is even better at filtering out noise.
 
So, howabout a 15v plug pack, tie knots in the cord, put on a magnet choke, put a series inductor (or K-multi) and power up an MC7912 and then use that to power up an MC7805? The concept works for Ken's phase inverter, for Lightspeed Attenuator, and there are several similar schemes for CD players, mainly all involving the MC7*** reg from Onsemi.

I have always wondered if a highly rated negative regulator followed by a highly rated positive regulator would first clean the ground and then clean the hot? If not, then the really very polluted ground of the computer is a worry or at least there is for improvement. Most of the noise is double sided (a signal) so I believe that approximately half the problem would be in the ground plane.
 
You're allowed to draw so much current from USB (officially 100mA, or 500mA if your chip negotiates it) and I believe that, officially, you should not place more than 10uF capacitance across the supply lines.

However, I could envisage a system based on an 'ideal diode' e.g.
LTC4411 - 2.6A Low Loss Ideal Diode in ThinSOT - Linear Technology

combined with an op amp & MOSFET-based current limiter that could charge up a huge capacitor with minimal voltage drop, giving you a lovely smooth 5V supply to work with. It would also give you time to short your DAC output to ground before the supply collapsed in the event of pulling out the USB cable or your PC turning off - which is reputedly not good for your speakers!
 
A negative regulator is referenced to the same 0V as a positive regulator. PC grround has some noise as do ALL digital systems, the noise is inherent with digital systems where numerous gates are switching simultaneously, it is created by the digital systems switching (The SMPS will contribute some noise, but is not the main culprit). The ground bounce noise (Delta I) is down to the huge instantaneous current draw when multiple gates switch (100s of amps) and the main way of controlling it is correct decoupling schemes, as all power supplies are to slow as are most decoupling capacitors these days, the interplane capacitance and on die capacitance provide the initial charge, the decoupling caps filling these up, then the reservoir caps filling those up and finaly the power supply. Most deoupling is based on rules of thumbs and myths passed down over the years, PIA (power integrity analysis) software is becoming more cost effective and soon shows up the limitations of how digital supplies have been decoupled. Of course the beauty of digital is; it is pretty much immune from a lot of noise so the PC's work (even though a lot of PC's are an example of being designed down to a cost, 6 layer motherboards where 12 woulkd be better with multiple ground planes for return paths for signals and numerous power/ground plane pairs to increase planar capacitance.
Have a look at "simultaneous switching noise" it wil give you a good insight into digial noise.
 
Ground bounce, the ground plane is at many different voltages above and below the arbitary 0V reference point, the voltages dependant on the switching currents, plane impedances etc etc. This noise when controlled by good power plane layout and decoupling capacitor choice does not affect digital as long as the noise is kept withing certain limits, due to the digital signals not being as susseptable to noise like analog signals are. The main problem is the effect this noise can have on analogue signals, any analogue, not just audio. With the ever decreasing system voltages it can be most probmatic when you have ADC DAC devices with low operating voltages for both the digital and analogue sides, DACs working at 1.8V and below can be a pain.
So to answer your question relative to its self, though unless you use power system simulation software you cant realy picture the PDS voltage fluctuations and impedances, and simulation is only cost effective for high end designs where they are willing to spend the money and time on simulation. You have got to remember that switching currents can be in the hundered of amps (for a very small amount of time) so even slight impedance changes will cause changes in nthe voltage at any one point. Using a high bandwidth scope you can also see the effect this noise has on analogue signals, some audio signals from a mixed technology board are quite fuzzy on a 13G scope.
It is the same mode of noise that you worry about when isolating power grounds from signal grounds in amps...


http://www.analog.com/static/imported-files/rarely_asked_questions/groundNoiseMeasurement.pdf
 
Fuzz, yes. I'd sure like to dodge that. It is most annoying (with headphones).
The Circlophone amp, with its very mild noise gate function, somehow manages to play back from an ordinary inbuilt computer sound chip, without the noise and without the fuzz. However, my headphones don't fare well when directly connected to the computer.

First try was a Burr Brown USB chip. Not only was the fuzz worse than the Soundmax/Xfi/Tremor/Envy cards, but there was no need to worry about regenerating the USB power or other such stuff because that particular USB chip has very loud "non-appreciable" treble. It was simply not good enough to get it "the rest of the way" by playing with the power.

What USB chip(s) / chipset(s) should I be looking for?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.