A novel way to massively improve the SQ of computer audio streaming

[LTG2010]

1 hour ago, flkin said:

Ok, that's an interesting test, I'll have a look. Will it matter to you if I keep the mod switch there?

Firstly thanks for the detailed write up, especialy listing all your musical choices.

I think the key here is how good is the Antipodes USB output? Keeping the switch should not matter.

How good is the USB out when it is re-clocked by the txUSB Ultra. Can the Mutec components be added to this re-clocking chain somehow, and then on to the DAC?

The ultimate question is then which sounds better, ethernet out to Ultrarendu and or the trifecta/ reclocking chain?

Or USB out to reclocking chain?

Some of us have ditched our renderers for USB, do we need to go back to ethernet now? it never ends

[LTG2010]

2 hours ago, RickyV said:

what do you mean by “atx mode”? Is that bios setting? 

You can power the entire board from the 4 pin 12V connector. When you connect the 24 pin connector you are connecting 5v standby plus power ok line etc, signals to processor plus the seperate voltages, this is what I meant by ATX mode ,the board receives 'ATX' supply signals, there is no special Bios setting.

1 hour ago, RickyV said:

Btw i noticed the zenith se uses a 20 pin atx connector and not a 24 pin.

The last 4 pins on a 24 pin ATX connector supply the PCIE slots and are design to deliver the extra current for current hungry graphics cards etc. If not using PCIE slots or have externally powered USB cards etc, then I think you can do away with those 4 pins and use a 20 pin connector.

[LTG2010]

13 minutes ago, austinpop said:

Is there a trick to opening up the tX? I unscrewed the 4 x T8 screws each on the front and back, but how do I get the top off to expose the boards?

If it's like the sMS200 Ultra (I think they share the same box?) the front face lays flat inside and you pull the whole assembly through from the rear, taking care to note the exact slots the boards are located so that you can slide them back in reverse.

[LTG2010]

4 minutes ago, austinpop said:

Peeking in from the front panel opening. It looks like the sCK-EX board is held in place by some kind of caulk like glob. Do I have to apply pressure to detach that glob?

If its glued to the side of the case I would think so, with the sMS200 Ultra it simply slid out when you pulled the back cover there was no glue on the case.

[LTG2010]

6 hours ago, flkin said:

That blob of glue seems to serve no purpose except to check whether people have opened their unit. This may be for warranty purposes?

Most likely just to stop the board from sliding about in the case especially in transit.

Probably a drop of hot glue, good for damping vibrations. Harmless to remove it.

[LTG2010]

2 hours ago, seeteeyou said:

Could that be more hype than anything else?

No it's definitely not hype. I put one a year ago in my laptop, moved the Hitachi HDD over for storage- the improvement wasn't subtle, clearer, more spacious, warmer.

Problem was playing music direct from the Samsung NVME was superior to playback from the storage HDD and due to the lack of space I found myself moving files over and deleting etc. Soon got fed up with this though and accepted the sound as it was from the HDD, still pretty good once you adjust.

1 hour ago, lmitche said:

FWIW, hearing a major improvement in SQ moving  from a NAS to local playback seems reasonable to me. I think there a better ways to accomplish this than with a Samsung SSD, but nevertheless it does seem like a reasonable claim.

This is HDD via your USB card? (for storage I mean)

[LTG2010]

6 minutes ago, Jiffi32 said:

Hi @LTG2010

Has your power supply arrived yet?

Keen to hear your impressions if it has

Hi, it's due tomorrow, I'll be setting things up over the weekend and will post my impressions. We had a couple of days delay due to bad weather with the case and the transformer manufacturer had flue so we lost around 10 days there. It's all finished now so looking forward to receiving it.

[LTG2010]

My Linear ATX power supply by sean Jacobs arrived today. It needs a few weeks for component burn in but it's already sounding great, large soundstage, great dynamics, silent background, very detailed, neutral, good bass, very pleased, took just under 3 months mainly due to transformer manufacture.

It's a twin transformer design with 9 rails, 3.3v, 5v, 2x12V for the motherboard/CPU, 2x5v for ssd and hdd and 3x9v for SOTM trio.

It comes with a single umbilical connection at power supply and gland that feeds into the server. The wires are colour coded, but bare ended as Sean has an exclusivity agreement with Innuos. Heres my ScLK-ex server.

ATX wiring information is readily available here's a link with some good detail - this is how I did it.

https://en.wikipedia.org/wiki/ATX

I used ATX extension cables by Silverstone, 24 pin and 4 pin for the seperate CPU rail, simply cut off the male connectors to expose the wire and group the ground, 3.3v, 5v & 12V wires together and solder these to your seperate power supply feeds. The ground wires are all common on the motherboard so these can be all connected or daisy chained.

Modern motherboards don't need negative -12v voltages (pin 14), you can ignore these. A linear power supply does not have a 3.3V sensing voltage (pin 13) you can snip that off. You can't turn off your linear power supply from the pc so no need for the power on feed (pin 16). Pin 9 (purple) is the stanby voltage and is 5v, so connect this with your 5v feed (red wires).

Pin 8 (usually grey) is the power ok signal. The atx power supply sends a 5v signal to the motherboard once its all powered up and stable. We don't have this on a linear power supply so we need to 'fool' the motherboard into thinking its receiving a power ok signal, otherwise it won't boot. So we also connect this to the 5v feed. The power ok signal should activate between 100ms - 500ms after the supply is switched on, some motherboards might not boot if this criteria is not met. I had no problems with the supermicro X10, if your's does not boot you can easily get around this with a cheap delay relay from ebay, there are other methods also.

I originally made up some ATX connectors with silver plated copper cables and for powering the hard discs, these sounded awfull, very bright and distorted, so I switched to the Silverstone leads that appear to be tinned plated copper. For the SSD and hard disc I used Neotech OFC copper, amazing how a small piece of cable can clean up the sound. If making up your own SATA power connector, the middle pin is +5v and a pin either side of that is ground. Similarly I used Neotech copper for SOTM boards directly soldered to the board behind the external connectors, thus avoiding external connections and jack plugs.

Alec

 

 

[LTG2010]

5 minutes ago, Johnseye said:

 

Excellent!  Very nice.  Fix those picture links so we can get a better look.

 

Wondering how you attached the sCLK tray to the side of the Streacom case.  Did the screw holes line up and did the screws fit?

Can't seem to edit the post anymore there's only 3 pics, 2 of the case and 1 of the server, dont know where those other links popped up from.

2 of the 3 holes on the Sclk- ex lined up with the holes on the Streacom case, I used some screws that were used to hold the USB extension headers on the case as I removed these, they seem to be standard size and matched up.

[LTG2010]

7 hours ago, tims said:

I'd be most interested to read more about this as I've run out of PS outputs from my LPS so possibly using one rail (or not!) to feed my txUSB-exp > sCLK-ex > txUSBhub-ex would help a lot.

You can do it with just one 9V rail supply to txUSBexp. SOTM supply cable/board connectors that connect to sCLK-ex and then onto the USBhub. I used the sPS500 set at 9v with a y-cable from Ghent audio, one connection  to txUSB-exp linked to sCLK-ex and one connection to txUSBhub-in

[LTG2010]

5 hours ago, lmitche said:

Thanks for this great post. I'm am heading in the same direction with lt3045 lpsus as a source for everything but the CPU power. The information you shared here is every helpful in that regard.

You're welcome Larry

 

4 hours ago, Jiffi32 said:

Curious to know which case you are using for the pc? 

That's a Streacom FC9

[LTG2010]

49 minutes ago, ElviaCaprice said:

Thus it would appear that phase noise is not the only issue being resolved with the sCLK-EX, barring some other negative feature besides clocking in the ISO Regen that would explain it's detriment.

Phase noise is measured as frequency phase error -short term oscillator  instability, there is also jitter - timing error. Jitter or timing error causes a loss of detail and focus '3 dimensionality' This is where the sCLK ex excels. Syncing to a OCXO master clock, reduces these timing errors further. Sometimes measurements alone will not predict how the component will effect the 'sound'

[LTG2010]

11 minutes ago, RickyV said:

No, re-generation and reclocking is good for SQ,  determined by this thread.

Yes, re - generation and re clocking or daisy chaining the same clock has been shown to improve sound - reduce errors further, this seems to indicate even better clocking at the source could be attained

[LTG2010]

10 minutes ago, Johnseye said:

So how does the reference clock influence timing from one sample to another as a digital audio stream is transferred from one device to another? 

As far as I'm aware, the master clock provides a 'reference' but cannot replace the actual clock it needs the signal from the source clock but can accept multiple signals from various clocks. These clocks become 'slaves' to the Master clock, thus their performance ( in the timing domain) draws closer to that of the Master clock. Eg. your sCLK ex clocks take on board the characteristics (improved timing) of the OCXO Master clock in the MUTEC.

[LTG2010]

5 hours ago, rickca said:

We don't have any measurements that support this statement, so I don't think we can say that this is why sCLK-EX sounds so good

Well the sClk ex is just a clock board, a clock replacement. Clock = timing. If it's not providing a more accurate clock signal than that which  it replaces, what else is it doing?

[LTG2010]

54 minutes ago, Johnseye said:

The power contributes to phase noise, does it not?

Yes especially switching supplies. Spikes/ peaks, poor voltage regulators induce deterministic jitter (and therefore phase noise) as does EMI / RFI. Also there is crosstalk- induced current from adjacent conductors, the increase and decrease causes jitter. There is also random jitter (from less obvious sources) eg. overheating of components and non uniformity.

So the specification of a crystal is not the only factor.

[LTG2010]

2 hours ago, jean-michel6 said:

Thank you for this very interesting feedback. 

I have a dual pc set up similar to yours . I am using also the same Supermicro mother board as you. 

On the player side I have also a sCLK ex board for pc lan and system clock and pcie SOTM usb as well as the txUSB board. 

 

I have a separate 4x5volts psu for ssd os and music ssd. 

Sotm pcie card and tx usb as well as sCLK board are powered by two sotm sps500. 

I have also a dual rail Sean Jacobs 12v psu for the mother board and I am powering the mother board with the 4pin atx plug only. 

 

I have already extremely good music from this system. 

 

However you went even further than me and I am extremely interested to know if you did compare powering your server with the 4pin atx only and 4 pin +24 pin atx. 

 

If not I would be extremely interested if you could do it to have an idea of the progress brought by using both atx 4and 24 to power the Supermicro board.

 

Could you give the 24 pin assignment that you used for the atx connector. 

According to your feedback , for the super micro board it was quite easy to set up an atx linear psu which is not the case for other mb. 

I've given a link to ATX pin layout in my previous post, its a view from the underside of the connector, if you have any specific question let me know I'll try to help. I forgot to mention the -5V /0V (pin 20)  is also not required.

It's not as simple as pulling out the 24 pin connector to test as both my 12V rails are 1.5 AMP which is insuffiecient to power the board, but previously I did power the motherboard using a 12V meanwell medical SMPS, a LPS1 for SSD and sPS500 with a Y cable for the SOTM trio.

The step up in sound quality is significant, in all aspects, not sure if it's mainly due to the quality of Sean's power supply or the multiple rails, I would think it's due to both factors.

Basically the principle of a seperate clean regulated supply to as many components as possible yields benefits. You are powering CPU, LAN, PCIE ,USB, RAM, with seperate rails, although the motherboard is further stepping down voltages it still seems to work well.

I was a bit sceptical with the SOTM trio but there is a definite step up in quality with 3 rails, - imagery, seperation, detail, etc. all improve.

For example comparing the motherboard USB output (already known to be very good)  to the txUSB exp, there is a big improvement in the stereo image, solidity and naturalness with the tx, the digital edge dissapears. Moving onto the txUSB-HubIN there is an improvement again but its more subtle, I suspect the USB link between the 2 components could be improved also. What is clear is that the SOTM items really sing with Sean's PSU.

The atx specification / 24 pin should be the same for all boards, you might have 6 pin or 2 x 4 pin etc for the CPU rail.You can also use a 20 pin connector as the last 4 pins are for higher power delivery meant for large PCIE graphics cards, I left it as 24 pin. You seem to have a very good set up as is, so the step up in quality might not be as pronounced as it was in my case.

 

 

[LTG2010]

14 minutes ago, Johnseye said:

I suspect it's the quality of the supply.  Even though the supply has multiple rails it still shares a single transformer.  In our limited testing of my multi-rail SR7 against the SR4 there was a sonic benefit to using two completely different supplies.  It was audible, and it bothers me because of its implications.  Further testing would need to be done to validate, which Rajiv and Eric are planning.

There are 2 transformers, I originally planned to use just the 12v connection as I also have a jetway board and opted to power the board with its own transformer and SOTM and SSD/HDD with a seperate transformer. Then it expanded to an ATX Supply and the one transformer has the 2x12V rails.

That was suprising to hear regarding the SR7, I wonder if the same regulators are used for both supplies? a higher amp supply will need higher powered components, which might have a knock on effect sound quality, it will really need to drive a difficult load to 'sing.'

[LTG2010]

1 hour ago, Johnseye said:

 

There are 3 regulators.  Rail 1 has a pre and post (aka DR).  Rail 2 has its own.

Does Paul use the same regulator circuit for both SR7 and SR4 ? One supply is rated at 10/5 AMP and the other at 2 amp, the 2 amp supply would usually require less or lower power components, less noise, etc. When powering a low powered device, the 2 AMP supply will sound more transparent, but will lack the ability to drive a difficult load. Unless the same circuit is used in both supplies and the transformer is smaller in the 2 AMP supply.

A test with the SR7 using 1 rail only vs SR4 will be useful.

 

40 minutes ago, sandyk said:

I found this when using 2 separate transformer windings to provide 2 low noise PSUs for USB.

 It didn't sound as good as when using 2 separate supplies.

You can never achieve 100% isolation between rails and both rails will be using the same core, but to stretch to 9 transformers would be financially a bit impossible for me ( I'd need a mortgage ) I'd actually thought 2 or 3 rails was the maximum feasable soundwise, but Sean seems to have managed it, with 8 rails in the Innuos Statement prototype.

[LTG2010]

6 hours ago, jean-michel6 said:

I am extremely interested to know if you did compare powering your server with the 4pin atx only and 4 pin +24 pin atx. 

 As per my previous reply to you I could not use the 12V - 4 pin alone due to the low Ampage but what I could do is test the 24 pin connection alone without the 4 pin, therefore omitting the 12V 'CPU rail'. I could hear a reduction in quality, less seperation and a slightly more grainier sound, not huge but noticeable.

[LTG2010]

1 minute ago, Cooler said:

Can you provide any photos, i dont understand, you connected some ATX PSU to 24pin and at the same time some another PSU to 12V mb plug, like this:

The 12V is a 4 pin atx connector on motherboard.

[LTG2010]

15 minutes ago, seeteeyou said:

Not sure why but it didn't really look like 12V to me

Looks like its linear powered at 9.3V with internal Supercaps and 18V charger/ energiser ?

[LTG2010]

On 12/05/2018 at 11:24 AM, vortecjr said:

It's not clear if the external power supply the new Hugo TT 2 uses acts like a "charger / energizer". Calling the external power supply a "charger / energizer" might imply that is has a switching banks of ultracapacitors. If it doesn't have a switching banks of ultracapacitors then the unit just uses the ultracapacitors in some other way.

Yes it is purely guesswork on my part, the original TT2 had a battery supply and charger so I assumed the battery has been replaced by Supercaps. I also note that the supply is rated at 5 amps peak so it would be interesting to see how they have achieved that with 6 supercaps.

[LTG2010]

1 hour ago, mozes said:

one more question, does upsampling to PCM 384 require similar processing power like upsampling to DSD512 does?

I can do it with my celeron j1900 @1.9ghz runs at 15% It can also do DSD 128 but it runs at 85% sound is as Johnseye reported a bit softer, but I'm not using hqplayer, where I assume the differences would be more apparent.

[LTG2010]

5 hours ago, vortecjr said:

They can just be caps after the regulators.

 

5 hours ago, Superdad said:

Much more likely that Chord is using them before the regulators.  Just as extra filter capacitance, not as an actual disconnected-from-mains power source as we do in our UltraCap supples (I guarantee they are not using a bank-alternating scheme—that takes a WHOLE LOT of circuitry!).

Here's some more information from Rob Watts:

'' So now we are getting into the more interesting stuff with TT2 - the output stage and PSU. When you turn on TT 2, it enters a charge mode, where a large controlled current is fed into the supercaps to charge them up and this takes about 10 seconds. The charge process is via a FET switcher and a very large inductor; this has two benefits - the current drawn from the supply is large and constant (3A) but initially is converted to ten amps going into the super-caps as the voltage is low. The second benefit of the inductor is that, together with other other inductors, ferrites and capacitors, removes noise from 20 Hz (actually the filter starts at 1 Hz) to 5 GHz. The power unit supplied is actually much more capable than it seems, and the PSU structure has been built around this particular power unit. I went through a lot of power supplies and a couple of design revisions, to ensure that when connected there was no measured change in performance at all. So this power structure has the benefits of ultra low noise from batteries - without the worry of battery life - but also huge dynamic current capability.

The OP stage was radically re-designed - initially I was using a beefed up Hugo2 design; but ended up using the design from the power pulse array project as this had been running in parallel to TT2's development. This OP stage is capable of large amounts of current, and more importantly, is extremely low distortion with current delivery - as you increase current load, distortion barely changes, and then only by increasing 2nd harmonic only.

But there is no point in having an output stage that has no distortion with current, if the PSU can't deliver the current. This is where the super caps comes in; audio requires bursts of current, and super caps are perfect at doing this; so the effect of the large inductor and the supercaps with the 1 Hz filtering, means that the PSU is only seeing DC current - no audio components.''