Why a linear power supply?

[zyzyx]

I'm looking into options for replacing the switch mode power supply in my 2010 Mac Mini. As most of you know, these things aren't cheap. The minimum price I've found for a purpose built linear power supply and dc cable is around $500. So that got me to thinking—instead of a high quality linear power supply for the Mac Mini, why not a HIGH QUALITY purpose built SMPS? According to the experts, these power supplies when designed correctly are very good. Good enough that they've found a place in some very expensive equipment.

 

Is it crazy to think of replacing the stock SMPS with a higher quality SMPS that doesn't inject noise into your system?

 

Does anyone know of a vendor for high quality switch mode power supplies? I've had good luck with connexelectronic SMPS power supplies for class D amps in the past. I wonder if they'd be interested in designing a power supply for the Mac Mini?

[tranz]

There are a few threads regarding low noise SPS.

 

Also check out the Keces LPSU threads as I got mine from Taiwan for just over $200.

[Paul R]

I think John Swenson is building a new PS for Mac Minis - might want to PM CA member Superdad about it. I am looking forward to hearing about it and seeing what one can do here. Depending upon the price of course.

 

-Paul

[JohnSwenson]

I think John Swenson is building a new PS for Mac Minis - might want to PM CA member Superdad about it. I am looking forward to hearing about it and seeing what one can do here. Depending upon the price of course.

 

-Paul

 

Last weekend I took a prototype over to SuperDad's place and we tried it out. This supply has two regulators, one normal and one with "Kelvin sense", this runs the regulators sense inputs to the load rather than the output of the regulator. It attempts to compensate for any effects of the cable in between the PS and the load.

 

First pass we were listening to a commercial lab supply, not the mini SMPS. This trip I never got to hear a comparison to the factory setup.

 

Then we connected the regular output(non Kelvin) to the Mac mini, the result was definitely improved sound, but not spectacular "WOW" type thing. We listened this way for quite sometime. Then we hooked up the Kelvin sense (it's a board that goes in the mini) and it's separate coax to the supply, now THAT was getting to the WOW factor. It significantly cleaned up the sound over the non-Kelvin version. What we were hearing was a much cleaner sound where subtleties of intonation were much more audible.

 

Strange as it may seem the effects from the Kelvin sense were less than the "base supply", but they are what added the "WOW".

 

This whole thing is a problem for me as an engineer, this was the COMPUTER, not a DAC or preamp. At this point I know it works, but no idea as to the exact mechanism. I spent a lot of effort on this design cutting down on the amount of noise sent back down the line, but the Kelvin sense is just cleaning up the power going to the load. (shaking head in disbelief)

 

This was just a listening trip, no measurements were taken so I have no clue what actual electrical differences occurred. ( didn't want to haul around several hundred pounds of test equipment)

 

This was pretty close to the final version, I have to make a few tweaks to get things to fit in the case Alex chose.

 

John S.

[Paul R]

That's interesting indeed! Are you using something like a 4-wire Kelvin clip? If so, is it possible you are (inadvertently) setting up some kind of RC/LC filter in the connection, with the power supply side providing a bit of capacitance? Probably showing my ignorance there.

 

-Paul

 

 

Last weekend I took a prototype over to SuperDad's place and we tried it out. This supply has two regulators, one normal and one with "Kelvin sense", this runs the regulators sense inputs to the load rather than the output of the regulator. It attempts to compensate for any effects of the cable in between the PS and the load.

 

First pass we were listening to a commercial lab supply, not the mini SMPS. This trip I never got to hear a comparison to the factory setup.

 

Then we connected the regular output(non Kelvin) to the Mac mini, the result was definitely improved sound, but not spectacular "WOW" type thing. We listened this way for quite sometime. Then we hooked up the Kelvin sense (it's a board that goes in the mini) and it's separate coax to the supply, now THAT was getting to the WOW factor. It significantly cleaned up the sound over the non-Kelvin version. What we were hearing was a much cleaner sound where subtleties of intonation were much more audible.

 

Strange as it may seem the effects from the Kelvin sense were less than the "base supply", but they are what added the "WOW".

 

This whole thing is a problem for me as an engineer, this was the COMPUTER, not a DAC or preamp. At this point I know it works, but no idea as to the exact mechanism. I spent a lot of effort on this design cutting down on the amount of noise sent back down the line, but the Kelvin sense is just cleaning up the power going to the load. (shaking head in disbelief)

 

This was just a listening trip, no measurements were taken so I have no clue what actual electrical differences occurred. ( didn't want to haul around several hundred pounds of test equipment)

 

This was pretty close to the final version, I have to make a few tweaks to get things to fit in the case Alex chose.

 

John S.

[sandyk]

This whole thing is a problem for me as an engineer, this was the COMPUTER, not a DAC or preamp. At this point I know it works, but no idea as to the exact mechanism. I spent a lot of effort on this design cutting down on the amount of noise sent back down the line, but the Kelvin sense is just cleaning up the power going to the load. (shaking head in disbelief)

 

Hi John

That's a similar reaction to that of my Sydney E.E. friend about what vastly improved PSUs can do.

I very much look forward to seeing those user reports coming in when they are let loose in the wild.

Recently a solitary C.A. member reported that playback sounded better with the cable modem unplugged.

I wonder if a far better Linear low impedance PSU would also help to reduce those unexplained anomalies due to interaction between various areas of a PC ?

 

Alex

[Audio_ELF]

John, I have to ask this; but were you doing any of the comparisons blind or always knew what PSU you were listening to?

Last weekend I took a prototype over to SuperDad's place and we tried it out...........

[sandyk]

Perhaps they could also try doing a rip of a high quality CD from a standard Mac Mini, followed by the same from a fully tricked out John Swenson Linear PSU version?

[Superdad]

That's interesting indeed! Are you using something like a 4-wire Kelvin clip? If so, is it possibly you are setting up some kind of RC filter in the connection, with the power supply side providing a bit of capacitance? Probably showing my ignorance there.

 

-Paul

 

Paul:

I'm sure John will answer with a more proper technical explanation (even if the mechanism for the actual sonic differences are not clear), but I can tell you a couple of things:

After we wired the boards up (attaching regulators and diodes to temporary heatsinks, etc.) in my garage, we put a bank of load resistors on the supply to check the voltages and see how hot the devices would get (1/4" copper bar attached to an old Hovland RADIA heatsink worked great!). Without the pseudo-Kelvin sense line, the voltage at the load end of a 5 foot 18awg dropped about 0.8V (when we had the PS set for 12V as I recall; there are settings for 5,7,9,12V BTW). With the sense line connected, there was no droop at all.

In listening (and BTW, I can report like others that a linear versus the stock Mac mini SMPS offers much better bass!), what I heard with our supply without and with sense was that without it, transients and complexity seemed to tax the system. With it, there was a total evenness and nothing could fluster it. I'm not describing it very well, but it did seem to jive with the purpose of the sense line--to keep up with the load and instantaneous demand, even with the PS being at a distance. In a typical piece of audio gear, say a preamp or power amp, a designer would never want the power supply to attached by several feet of thin cord. I may make up a 12AWG cord and see of the sense line makes as much of a difference.

 

What's neat about it is that the board we will offer for the mini (with the correct SPoX connector allowing use of the stock internal motherboard>SMPS 9-wire cable--no cutting or soldering) will have the voltage-set resistors on it--and 2.5mmx5.5mm barrel jack and SMA jack for the sense. So installation is really easy (Mac mini surgery is actually much easier than the helpful pictures make it out to be), and one does not have to use the sense line. Any PS can hook up. Even if someone uses our supply, they can switch the sense circuit on an off (at the back of the PS) without moving any wires.

 

Maybe some pics will help:

 

 

This next one was just a short test version of the Mac mini PCB. We are still dimensioning for a board with both jacks at the end that will fit in the vacant holes.

 

A few people have PM'ed me for pictures of the chassis, so here you go:

Overall dimensions are 9 inches x 9 inches x 3.25 inches tall (including the height of these really beautiful feet I found). The fit and finish of this enclosure from Japan is far nicer than anything I could find from China (I looked for months!), though it does cost a bit more. I won't be able to calculate total cost or sale price until I can send CAD files for heatsink drilling and back panel machining/printing to the chassis supplier. We hope to be in production before mid-year.

Chris wants me to be careful about not doing commercial promotion in the forums, so I won't say much more about this.

 

I will say that John and I had a lot of fun this past weekend. We listened to a lot of different things. I even played him an A/B of Ethernet cables from the machine across the room that shares the tracks with the dedicated music mini. So much fun to baffle even an open-minded, brilliant engineer!

 

Ciao,

Alex Crespi

[Superdad]

John, I have to ask this; but were you doing any of the comparisons blind or always knew what PSU you were listening to?

 

 

Perhaps they could also try doing a rip of a high quality CD from a standard Mac Mini, followed by the same from a fully tricked out John Swenson Linear PSU version?

 

 

Come on folks, must EVERY thread end up at these places? This one is just barely 7 posts in...

[jtwrace]

What is the projected cost of the board?

[sandyk]

Alex C

Very nice.

I posted before your post. If I had seen it I would not have shot back at the Elf.

It annoyed me that such a question came right on cue from one of the usual suspects.

I understand the awkward position you are in with respect to the commercial side of things here.

Alex K.

[gmgraves]

I think John Swenson is building a new PS for Mac Minis - might want to PM CA member Superdad about it. I am looking forward to hearing about it and seeing what one can do here. Depending upon the price of course.

 

-Paul

 

Is Swenson Mojo Audio?

 

Personally, I'm skeptical (but then, I haven't heard the difference, so I can't say for 100% certain), but I have looked at the different voltage rails on my Mac Mini (last generation. still has USB2, but no optical drive) using my Tektronix 350 MHz, 4-channel, 2456A 'scope. I see no visible noise on any of the rails at any resolution or TB.

[jtwrace]

Is Swenson Mojo Audio?

 

No, that's Benjamin Zwickel.

[Superdad]

What is the projected cost of the board?

 

If you mean just the little 0.75" x 3" conversion board for inside the mini, the answer is as little as we can have it made for plus a little to cover basic handling and transaction costs. But I don't feel comfortable guessing yet. Whatever it is, nobody is going to wince in pain at the price. I figure that people who use it might then be interested in hooking our power supply up to it and taking advantage of the Kelvin sense connector.

 

If you were referring to the cost for the PS boards themselves: It will be sold only as a complete supply in a chassis, not as a DIY. I am calculating for 50 and 100 unit runs, and even then the parts and labor add up. For this product I still think we will go direct (versus dealers/distributors), and when I officially announce I will be forthcoming about built cost, technical details, and performance. I want to test out for my business what is known as "radical transparency." I think it is a way to engender trust and an appreciation of value--and it fits with my personal ethics.

Whoa, how did I get off on that tangent?…

 

BTW, our plan is to also offer external Kelvin-sense modules (it's just a pair of resistors chosen for the particular voltage desired at the end of the cable) for use with other gear (DACs, CAPS, etc.). These will consist of a female SMA coax jack and female DC barrel jack on one side, and a male DC barrel plug (or later maybe product-specifc plugs) on the other. Think little 1-inch cubes.

 

Thanks all for the interest. I'll try to keep people posted--without stepping over the forum/commercial lines too often.

 

AJC

[gmgraves]

No, that's Benjamin Zwickel.

 

Thanks.

[jtwrace]

So it's not going to be this expensive?

Mac Mini Modifications : Mac Mini Mod Kit Pre Assembled

 

Same idea though. Right? Will you have any objective data posted?

[Don Hills]

... using my Tektronix 350 MHz, 4-channel, 2456A 'scope. I see no visible noise on any of the rails at any resolution or TB.

 

Since the effect of the supply is apparently audible, I suggest also trying audio-oriented measurement techniques such as spectrum analysis, or even your ears. (For listening to PSU rails, use a blocking capacitor and potentiometer fed to an amplifier. Remember to turn the pot down before connecting/disconnecting.)

[Superdad]

So it's not going to be this expensive?

Mac Mini Modifications : Mac Mini Mod Kit Pre Assembled

 

Same idea though. Right? Will you have any objective data posted?

 

Uh, yes and no. First off, our little board should be less than $100, not $450! Secondly, the stock internal MB>PS cable will plug directly to our board and the DC barrel and SMA jack (for Kelvin sense) will be mounted on the opposite end of the board--and will be what secures/supports the whole thing inside the mini. So I think installation will be easier.

 

Third and most importantly, John feels that putting a bunch of capacitors inside the mini, AFTER the regulated supply, either does little or does harm (depending upon the values and total ringing of the cap array). I'm sure he will jump in later to elaborate or correct me.

 

As for measured performance data: Yes, that is a goal of mine. John has explained a lot of good reasons for doing a choke-filtered, regulated supply versus all the fancy discrete super regulator methods that others employ. From a sales/marketing perspective, it will be great to have real scope shots showing just how well the supply rejects line noise, provides good regulation and low ripply, is quiet, and can handle transient loads. Much to do before we get there though.;-)

[jtwrace]

Uh, yes and no. First off, our little board should be less than $100, not $450! Secondly, the stock internal MB>PS cable will plug directly to our board and the DC barrel and SMA jack (for Kelvin sense) will be mounted on the opposite end of the board--and will be what secures/supports the whole thing inside the mini. So I think installation will be easier.

 

Third and most importantly, John feels that putting a bunch of capacitors inside the mini, AFTER the regulated supply, either does little or does harm (depending upon the values and total ringing of the cap array). I'm sure he will jump in later to elaborate or correct me.

 

As for measured performance data: Yes, that is a goal of mine. John has explained a lot of good reasons for doing a choke-filtered, regulated supply versus all the fancy discrete super regulator methods that others employ. From a sales/marketing perspective, it will be great to have real scope shots showing just how well the supply rejects line noise, provides good regulation and low ripply, is quiet, and can handle transient loads. Much to do before we get there though.;-)

 

I'm intrigued where you're going with this so far.

[acg]

but I have looked at the different voltage rails on my Mac Mini (last generation. still has USB2, but no optical drive) using my Tektronix 350 MHz, 4-channel, 2456A 'scope. I see no visible noise on any of the rails at any resolution or TB.

 

George, the vertical sensitivity on that scope is what 0.5mV, 1mV, 2mV? Serious question, I don't know, but perhaps the noise-floor of that CRO is too high to measure the noise that is certainly on those lines.

 

I have a 0.5mV vertical sensitivity on my CRO and a friend is building a pre-amp so that we will be able to measure power supply noise down to a noise floor of about 100nV, which should start to be useful.

 

Anthony

[Superdad]

I'm intrigued where you're going with this so far.

 

I guess I should get our web site going so I can post some of the pages of John's detailed explanation of his power supply design and where he also exposed some oft-believed myths about this area.

 

BTW, I see in your signature that you have a pair of those Rythmik Audio subwoofers. John just got one recently (the F12G) and I persuaded him to bring it to my place this past weekend. I already get outstanding bass (tight and ample into high-20s) in my very rigid but small (19' x 12') room, but that sub really took big performances (Peter Gabriel w/New Blood Orchestra!) to a whole new level of impact! What always amazes me is how even the midrange and top-end seem to benefit from a sub--just more reality of presence I guess. And that was without using the high-pass filter… I know what I want for my birthday! Sadly, the only place it really worked well was front and center--right in front of my amplifier.

[jtwrace]

I guess I should get our web site going so I can post some of the pages of John's detailed explanation of his power supply design and where he also exposed some oft-believed myths about this area.

Yes you should! So what's the business setup? John is the engineer and your the business guy? Does John do this for a living or?

 

 

 

BTW, I see in your signature that you have a pair of those Rythmik Audio subwoofers. John just got one recently (the F12G) and I persuaded him to bring it to my place this past weekend. I already get outstanding bass (tight and ample into high-20s) in my very rigid but small (19' x 12') room, but that sub really took big performances (Peter Gabriel w/New Blood Orchestra!) to a whole new level of impact! What always amazes me is how even the midrange and top-end seem to benefit from a sub--just more reality of presence I guess. And that was without using the high-pass filter… I know what I want for my birthday! Sadly, the only place it really worked well was front and center--right in front of my amplifier.

 

Yes, they're very good. I have two of them and also use two bandpass subs for a total of four. It's the only way to fly if your retentive and measure but that's a whole nutha thread..

[jtwrace]

Is this the same John Swenson?

Q&A with John Swenson. Part 1: What is Digital? | AudioStream

[sandyk]

George, the vertical sensitivity on that scope is what 0.5mV, 1mV, 2mV? Serious question, I don't know, but perhaps the noise-floor of that CRO is too high to measure the noise that is certainly on those lines.

 

I have a 0.5mV vertical sensitivity on my CRO and a friend is building a pre-amp so that we will be able to measure power supply noise down to a noise floor of about 100nV, which should start to be useful.

 

Anthony

 

Anthony

My own CRO ran out of grunt when looking at noise on PSU rails and the noise output of my preamp, so I needed to build a small battery powered 10X low noise preamp in a screened case.

Alex