Continuing pursuit of power supply improvements and improved Dac performance.

[sjoc2000]

In two previous posts work has been done to explore improvement in sq (sound quality) with Dac performance by increasing the quality of ps (power supply) performance.

 

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/stock-power-supply-and-peachtree-dacit-13214/

 

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/peachtree-dacit-and-usb-5-volts-13995/

 

In the first effort a low noise linear 9 volt v (volt) ps replaced the stock marginal ps supplied with a Peachtree Dacit. Secondly the computer ps via USB for the Dacit was replaced with a cost effective regulator and filtering circuit (CLC) with amazing results in inrceased Dac performance.

 

Presently, work has just been completed in adding the JLH Ripple Eater filter to bring any residual noise after the 9 v ps down to uv (micro volts) levels. Of course by placing the USB regulator and filter in parallel with the JLH the benefits will be seen in this circuit as well as it supplies USB processing circuitry in the Dacit.

 

 

The filters were feeding 5 and 9 volt supplies to the Dac with approx. 1-3 mv (millivolt) of left over noise, as far as it is possible to measure with the equipment here. The upper circuit in the following pic is the JLH filtering the output of the new 9 v. supply. Some reports on the JLH place it's filtering capabilities at 4-20 uv for lower ends of the audio spectrum. At upper frequencies 300 uv has been reported.

 

The lower circuit is the CLC filter and regulator ps for Dac USB implementation.

 

 

The resulting scope trace follows at 50 mv increments. Although I have tried amplification of the o'scope input, software o'scopes purporting to be able to measure at the uv level, and borrowing more expensive voltmeters, none of them resolved the problem of measuring microvolts in this circumstance accurately.

 

 

Eventually I just put it all together, and just listened. I was so frustrated by not being able to do an effective measurement, put my expectation bias at close to zeroe. :0)

 

But, I was pleasantly surprised at a further improvement in sq. Previous to the JLH, the Dacit seemed flawless in it's performance, but with the JLH addition further detail and harmonics pleasantly added a further depth and richness. If you are interested in such pursuits you will likely be rewarded in some big ways...

 

Jim

[sandyk]

Hi Jim

As the shunt output impedance of the original design was quoted as around .02 ohms, and this version, especially when using 2 parallel 4,700uF capacitors per supply rail, is likely to be even less, it seems likely that when you build it up on a suitable PCB with shorter leads such as 7.5A mains cable, that it may be even more audibly effective.

Nice presentation!

Kind Regards

Alex

[sjoc2000]

Opps, I hit the post button before I finished. I became informed about the JLH and CLC circuits by SandyK, and was greatly motivated in pursuing this subject because of some of the work he has done. In addition, I was able to get his advice when I had some questions while building these circuits. Thanks Alex from myself, and others here who benefit greatly from your contributions.

 

Jim

[sjoc2000]

In the interest of objective data and measurement, the following may be useful. Scope is set at 50 mv increments. The first pic is the scope trace line with probes grounded. It can generally be assumed that a decent scope will have some internal noise of 100-150 uv (microvolts)

 

 

The second pic represents the trace measuring any ac, or in this case noise, from the final filtered power supply point before the Dac. Notice the very slight bow in the trace, caused by a very low freq. residual (probabaly much lower than 25-35 Hz.)

 

 

One can assume that it is very likely that ripple and noise is likely well within the uv range. Where precisely, don't know, but likely below 300-400 uv, depending on the freq. we're concerned with. :0)

 

I can tell you subjectively, that this Dac performs very well, as was reported by Chris' review, especially when ps to main and USB are well controlled for spurious voltages.

 

Jim

[sjoc2000]

The following represents further analysis with Visual Analyzer (VA software, available on the web free as o'scope and analyzer)) of the output of the power supply to the Dac and it's USB input, after the addition of the JLH filter.

 

Scope;

 

scope.bmp

 

This measurement is done with the Dacit operating as a normal load situation. The increments are .001 volt, or 1 millivolt. The avg. voltage output is around 200-400 uv (microvolts). It is difficult to know if the actual output from the filter is less than 200-400 uv and that this level is just inherent noise in the measuring equipment, with the actual signal from the filter being enveloped in this noise.

 

Following is spectrum voltage from ps. The slight aberation described earlier is seen at approx. 10 Hz.

 

spectral.bmp

[sjoc2000]

Opps, this system doesn't like the bmp format, here's the pics in jpg

 

The following represents further analysis with Visual Analyzer (VA software, available on the web free as o'scope and analyzer)) of the output of the power supply to the Dac and it's USB input, after the addition of the JLH filter.

 

This measurement is done with the Dacit operating as a normal load situation. The increments are .001 volt, or 1 millivolt. The avg. voltage output is around 200-400 uv (microvolts). It is difficult to know if the actual output from the filter is less than 200-400 uv and that this level is just inherent noise in the measuring equipment, with the actual signal from the filter being enveloped in this noise.

 

 

Following is spectrum voltage from ps. The slight aberation described earlier is seen at approx. 10 Hz

 

 

Jim

[sjoc2000]

Was being lazy. Measured inherent system noise (noise of system without measuring), was at 130-140 uv, so conclusion can be made that ps output noise levels with JLH in this circumstance is somewhere from 200-400 uv. Not difficult to build, easy on the pocketbook, and very friendly to the sq of your system.

 

Jim

[sandyk]

Hi Jim

As you already realise , it should improve further when off the breadboard and on a proper PCB.

It would be interesting to see further screen grabs then.

I wonder how John Linsley Hood managed to read noise levels in the 4uV region so many years ago ?

Perhaps some of the more qualified members can give us a few tips ?

Regards

Alex

[sjoc2000]

Set up a new means of testing. Instead of the Asus pc I toke out a battery powered net book computer (quieter) I have here, and set it up with the VA (visual analysis) software. The new setup was calibrated using a signal generator, verified as putting out a 1k sine wave at 114.45 millivolts per a soundcard through a 3 ft. cable terminated with 3.5 mm mic connectors.

 

BTW, have given up on the o'scope at these levels. Trying to objectively measure the width of a trace line eventually leads to insanity. :0)

 

Discovered a couple of transistors (Q1,Q2) on the JLH had opened up, and needed to be replaced. After carefully checking all the other components and performance of the JLH, I started testing with the net book setup. This system is much more stable, and makes uv testing much easier, at least to the degree that it is possible. Following is a final representation of what the JLH can do;

 

 

Inherent system noise was found to actually be 170-190 uv with the net book and it's mic cable. Anything above this level was immediately reflected in the readings. The JLH tested at the system noise levels of 170-190 uv, meaning that it's filtering capabilities are somewhere below the testing equipment noise levels. Resolution to test exact uv levels is limited by this factor, but it is clear that this filter reduces uv ripple and noise levels at or below 170-190 uv for a power supply.

 

As noted in the pic, the 10 Hz remnant disappeared with the replacement of the transistors, and is now relatively flat across the audio range. Avg. db being approximately -100 to -110. Residual noise appears to be randomized with no apparent patterns, at least at this level (50x) scope magnification.

 

The sound of the Dacit with this filtering to the 9 v. main power input (< 170-190 uv), and working in concert with the CLC filter (again < 170-190 uv) for the 5 v. USB power requirements allows a level of detail and fineness to be heard, that was previous to these changes, not present.

 

This is the last post, unless there are any comments...

 

Enjoy the music,

 

Jim

[sjoc2000]

My least favorite subject, grounding. Discovered that house grounding was insufficient, as well as how grounds were connected between parts of this filtering system, the computer dac, and ps. Reworked all the grounding and placed all of the system on a very adequate HF radio ground setup I have. Good thing there's no woman around here there's a wire running through the kitchen to the ground. :0) As I suspected it cleared up a great deal, as well as the noise of the netbook computer for measurement. Net book was recalibrated before measurements. Calibration is very much an issue at these levels of measurement, I have done the best I can with what I have, although measurements are in accordance with previous measurements of the JLH filter over the years

 

 

As can be seen, noise and ripple is now <50 uv. Level spectrum, avg. @ -96 dB from spectrum across audio freq.. Computer o'scope image indicative of levels measured.

 

 

Trace on regular scope almost like grounded probe trace. Again scope increment set at 50 mv., best this scope can do. Sq increased further with resolution of grounding issues. That means that these old ears heard differences between 185 uv and <50 uv of ripple and noise from a ps for music sq. That is, for this specific situation, and equipment.

 

The cost of this project was approximately $60

 

Sincerely,

 

Jim

[4est]

Amazing how delicate and instrument our ears are in discernment/differencing. Just imagine all of the filtering going on in our brains to keep us from being overwhelmed. Thanks for the read Jim.

[sjoc2000]

Amazing how delicate and instrument our ears are in discernment/differencing. Just imagine all of the filtering going on in our brains to keep us from being overwhelmed. Thanks for the read Jim.

 

Thanks for your comments 4est.

 

The circumstances here had to do with a main ground, but grounding between systems as well. The Peachtree actually puts out a continuing signal of some kind per the ground back to the computer. If it is disconnected at any time, the computer will immediately fail to recognize the dac. This makes it difficult to get rid of this dirty ground, and get it back to the cleaner source ps.

 

Tried isolating this circuit while allowing a signal to get back to the computer with diodes. This compromised sq by quite a lot. A simple loop back to the pc did not work since the signal must engage the dac for the info. to carry back.

 

Finally ended up doing what I did not want to do, that is tying the two grounds together (could get away with it in this instance), and then dropping both of these to main ground. I still need to measure the number of ma being drawn by this bad ground, but it's effects are mitigated to near zeroe with the procedure.

 

Yes, our ability to differentiate these differences is quite amazing. Although a great deal of time is spent in polarized discussion about the reality of these differences to some, it does not come from imagined or magical realms. If converters, PLL's, and other aspects of a Dac are operating in parameters of nano and pico seconds, one would expect that the power driving these systems would need to not be carrying ripple or noise at levels of 10-30 milliseconds to realize the ultimate potential of these designs.

 

As with the endless cable discussion, what seem to be silly claims to some are more often than not concerned with issues of impedance matching. The solution may not be a $500 cable, but simply an appropriately shielded inexpensive cable that is matched properly. In some cases, 6" off with the snips, and "voila" a $20 cable will sound as good as a $150 cable.

 

As you, SandyK, and others here know; endless contentious discussion about these realities will get you no where, but resolving them will, and does not always require big $$$ expenditures.

 

Of course there are those differences that occur that we don't understand or cannot account for yet. To figure those out, whether they reside in the observer or that being observed, that's what keeps me getting out of bed in the morning. It ain't magic though...

 

Jim

 

p.s. Nothing against magic, I like magic. :0)

[4est]

Well said Jim!

 

I oft say to acquaintances, whom wonder why I expend so much effort on sound reproduction, something akin to: It is a geeky sort of hobby that also incorporates art, philosophy and alchemy in the pursuit of music.

 

You are right, it is not magic, but at times it would appear as thus. Hence my use of "alchemy".

 

I too like magic btw!

[sjoc2000]

Well said Jim!

 

I oft say to acquaintances, whom wonder why I expend so much effort on sound reproduction, something akin to: It is a geeky sort of hobby that also incorporates art, philosophy and alchemy in the pursuit of music.

 

You are right, it is not magic, but at times it would appear as thus. Hence my use of "alchemy".

 

I too like magic btw!

 

4est;

 

The Alchemist. a good symbol of the prodding pursuit of the truth, steeped in all the trappings of the ongoing Renaissance. My favorite Alchemist would be Sir Isaac Newton, not bad company to be in. :0)

 

Jim

 

P.s. Your package will be arriving just after I get one here. Just pm me if there are any questions.

[4est]

Jim: Tee hee hee! Or is it Ho Ho Ho - Merry Christmas to me- regardless when it gets here! Please let me know if I can return the favor, and thank you!

[sjoc2000]

Following is the nearly completed project on the power supply's for the Peachtree Dacit. Have to finish the enclosure, darn enclosures...

 

 

 

 

On the left printed circuit board (pcb) is the JLH filter for the Teradac 9 v. power supply. After this filter stage the filtered 9 v. is supplied to the Dacit. Measurements are < 100 uv, but the accuracy of these readings is suspect because of the limitations of my measuring equipment for uv, and no reference voltage for calibration at these levels. They could be higher or lower than this...

 

 

 

 

The pcb on the right is connected in parallel to left board, receiving the filtered 9 v to a 7805 (5 v. regulator) which is then filtered by another JLH and then is used to power usb implementation in the Dacit. The effects on final sound quality is dramatic.

 

This project is not difficult with basic do it yourself (DIY) skills, and will pay off in big benefits to your audio system. I hope this little thread will encourage some to try it, or at least to begin to pay big attention to ripple and noise levels when purchasing a Dac.

 

Sincerely,

 

Jim

[sjoc2000]

A few continuing notes on grounding and isolation of ps';

 

1.) 9 v. power supply and it's JLH have an increase in noise uv levels of approximately 20-40 uv when the Dacit is powered up. For some this is of little consequence. For those wanting to deal with this issue, a LDO placed between the Teradak and it's JLH may alleviate or eliminate this problem.

 

2.) To isolate the 7805 ( 5 v. regulator ), and it's associated JLH is a bit more challenging. The +/- 5 v. leads from the JLH is connected to the +/- cables, Dacit side of the split USB cable. We have now isolated the 5. v ps used by the dacit for usb implementation.

 

 

 

 

Unfortunately, the pc USB controller (I think it's the controller monitoring this) will detect that it is not receiving a ps current back from the dacit, and will then shut down the port. Therefore it becomes necessary to give the pc neg. side of the cable a ground, which satisfies the controller that the device is still connected.

 

This of course exposes all of our work to the dirty ground of the pc. I have used a diode to re-isolate the ground. The diode allows the controller to see a current from the Dacit, while not allowing the pc dirty ground to contaminate the new ps. Orientation of the diode is + side toward new ps.

 

 

 

 

3.) You would think that the positive lead from the pc could just be tapped and forgotten. Sandyk emailed me about considerations of a hot line coupling with the data lines of the USB cable. He is correct. This has been mentioned several times in the forum. Take a matchbook cover or some thin cardboard, cut a piece out to fit when folded over itself, and slide it over the female +5 v. connector in the usb pc side. Then replace the male side. This will eliminate the +5 v. from the pc getting on the cable line, thereby precluding any coupling.

 

All of these changes resulted in further improvement in sq.

 

Jim

[sjoc2000]

QUOTE=sjoc2000;199714]A few continuing notes on grounding and isolation of ps';

 

 

2.) To isolate the 7805 ( 5 v. regulator ), and it's associated JLH is a bit more challenging. The +/- 5 v. leads from the JLH is connected to the +/- cables, Dacit side of the split USB cable. We have now isolated the 5. v ps used by the dacit for usb implementation.

 

 

[ATTACH=CONFIG]3646[/ATTACH]

 

 

Unfortunately, the pc USB controller (I think it's the controller monitoring this) will detect that it is not receiving a ps current back from the dacit, and will then shut down the port. Therefore it becomes necessary to give the pc neg. side of the cable a ground, which satisfies the controller that the device is still connected.

 

This of course exposes all of our work to the dirty ground of the pc. I have used a diode to re-isolate the ground. The diode allows the controller to see a current from the Dacit, while not allowing the pc dirty ground to contaminate the new ps. Orientation of the diode is + side toward new ps.

 

 

[ATTACH=CONFIG]3644[/ATTACH]

 

 

Jim

 

 

*** A further amendment to the above.

 

As has been stated we do not want to contaminate the new ps' clean output by tying it to the pc ground with it's EMR and RFI anomalies present from pc functions.

 

I used the diode as a means to accomplish this. This may work in some systems, but here there was a failure of this configuration after 4-6 minutes, when audio would simply stop. I tried several other seups with varying positioning and resister-diode combinations, which would all fail after 1 - 12 minutes.

 

Had to give this some thought. I think the problem here is reference voltages, or in this case reference grounds. The new power supply has a reference ground, as does the pc. If we just tie one ground to another they stabilize to a common reference, and all works albeit with the dirty ground. If they are completely isolated the two different references create two slightly different voltages.

 

With digital electronics the definition of the basic bit is established by the presence or lack of it by voltages. If one system see's 3.2 volts and turns on a transistor (a bit ) based on one reference ground, but in the other system where the reference is different we may have only 2.7 volts. This gets a bit complicated, but buffers, registers, etc. will operate differently in one than the other.

 

This is why a failure takes a while to occur. Eventually one system will stop because it is overwhelmed with many errors.

 

See here for a little discussion on this problem, reference from Sandyk;

 

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

 

 

Sandyk and I have been working on a simple resistor setup that will "attenuate" the effects of tying to the pc ground. This method will lessen the effects of EMR and RFI present from the pc ground without changing the reference voltage enough to cause a failure.

 

In this system values = > than 100 ohms will cause a failure in time. I have been working with 10 ohms, without a failure in 62 minutes. Will experiment with greater values short of 100.

 

A difference in sq can be heard with the implementation of the 10 ohm resistor, but this is not isolation, but attenuation. Possibly someone will come along who has some further ideas on improving removal of the effects of the pc ground.

 

Sorry for the length of this...

 

Jim

[sandyk]

Jim has previously referred to a link in another forum where we are investigating these issues.

As this is a combined effort between members of both forums,with a few other C.A. members also about to dip their toes in these USB power waters, I have attached a reply that I just made in the other forum.

Regards

Alex

 

This is the thread that my friend Jim from Iowa has started about experiments with a JLH and USB PSU improvements.

 

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/continuing-pursuit-power-supply-improvements-and-improved-dac-performance-14165/

Jim is presently using 10 ohms again, and will try increasing up to 100 ohms where he struck problems.

However, we shouldn't normally need to go much higher than that

as the earth lift resistor in most amplifiers is around 10 ohms, or in a recent Silicon Chip magaxine design 22 ohms, because they obtained slightly better measurements due to the proximity of a toroidal transformer.

The reason being that with a typical PC as distinct from a laptop, we are seeing a good old fashioned earth loop with devices such as USB memory sticks for example.The -VE supply from the PC is internally connected to the shield in USB memory sticks.

You can easily verify this by measuring between the black wire

( 0 volts line) and the screen with the device plugged in.You will see a short circuit. Unplug the USB memory stick and the s/c will disappear.

I was previously aware of this.

Regards

Alex

[sjoc2000]

QUOTE=sjoc2000;199714]A few continuing notes on grounding and isolation of ps';

 

In this system values = > than 100 ohms will cause a failure in time. I have been working with 10 ohms, without a failure in 62 minutes. Will experiment with greater values short of 100.

 

A difference in sq can be heard with the implementation of the 10 ohm resistor, but this is not isolation, but attenuation. Possibly someone will come along who has some further ideas on improving removal of the effects of the pc ground.

 

Jim

 

A resistor of 50-60 ohms has performed best with no failure after 2 hours. Decided to add a RF choke (inductor - available at radio shack - $3-4, don't need a gold plated one ) 100 uh in series with the resistor to block any residual RF that may be lurking on the pc ground. This combination has also been tested for 2 hours. The addition of the choke allowed upper freq. audio to have a bit more presence, without any harshness. With this little addition there is noticeable improvement in sq with this system and these ears. Music is a bit smoother-silkier.

 

Will be looking at a few more ideas, but this setup nearly equals the function of the diode or isolation without shifting reference levels enough to cause failures. Will be looking at differences on a scope to illustrate these findings.

 

Jim

[sandyk]

Something for the DIY person to experiment with.

I have found that .wav files saved to a Corsair Voyager USB 2.0 memory stick sound noticably better after either of these modifications. In theory, the LED and Resistor version has the potential to sound a little cleaner sounding,

although I didn't find much difference between methods.Both of these versions sound better than with just VBus disconnected at the PC end of the USB cable when used with an external Linear +5V supply.

Erin and Jim have worked with me in this research, and Erin suggested the Resistor and LED version

That version does not however work for Jim's DACiT, but the resistor in the power return lead does..

Without that resistor there is a potential earth loop.

Regards

Alex.

 

 

P.S.

Be warned that doing this to a USB-A plug is quite a fiddly job.

 

IF you intend to reply in this thread simply to argue about the merits of these changes,

please don't bother.

It is posted to stimulate people findings these things out for themselves.

YMMV.Not all ears and equipment are created equal.

 

[sandyk]

If anybody who has read Jim's (sjoc2000) experimentation and results using the JLH PSU add-on PCB,and is interested in trying a couple for themselves, a limited number (100) are now available from Rock Grotto Forum. They are non commercial DIY PCBs, and non profit. Information is at the attached link.

Regards

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

[4est]

Thanks for the heads up Alex!

[sandyk]

Hi Forrest

Due to the limited number in this run, and expected likely high demand, you are unlikely to be able to obtain the quantity you asked for. We have now supplied over 800 dual JLH PCBs in it's various revisions worldwide in several Group Buys.

If demand is way more than expected this time, another run further down the track is possible.

Kind Regards

Alex

P.S.

Greg Erskine's very nicely done section on the JLH PCB is not quite up to date. Further info is in various threads in that forum, or can be supplied by me on request.

 

Greg's Web Site

Jim (sjoc2000)will also be able to offer valuable advice on putting them together and testing them, and perhaps where he sourced some of the components in the U.S.

[4est]

Well, they said 10 was the limit, and a guy can hope can he not? These are for a friend and I, and I thought I would build up as many as I could at once. I am a simple man, and will accept whatever is decided as per my post.