Building a DIY Music Server

[Marcin_gps]

OK @Nenon ,  I will post a link to the shootout when the review is published. 

 

 

 

[dctom]

Thanks Exocer, I have the same phSR7t with internal silver wire, ph silver cables- which hxplex ATX are you using is it the 800w. Can you say which usb cable you use to your as well please.

l have signed up for the taiko atx so wondered about the comparison.

are you streaming with roon etc? Or using local files.

I find local files stored on a nvme m2 using hqplayer sound much better than  roon-buffalo-ER-hqp.

 

 

[Exocer]

5 hours ago, dctom said:

which hxplex ATX are you using is it the 800w. Can you say which usb cable you use to your as well please.

The 800W HDPLEX was used. I probably should have mentioned that. 

 

USB Cable is the Sablon 2020.

 

Euphony Stylus is what I am testing with (no Roon components involved).

 

Both streaming from Tidal and local Redbook flac. Music files currently stored on Optane 800P.

[Nenon]

12 hours ago, Exocer said:

Back to Taiko DC-ATX testing with PHSR7T.

 

Let's label the configurations A and B. Where:

 

A. Taiko feeding entire server, fed Paul Hynes SR7T 19v 10a rated single rail. Ghent Gotham JSSG360 from Taiko Dc-ATX to EPS on motherboard.

B. Taiko feeding ATX only. EPS powered directly from a Paul Hynes SR7t 12v 12a rated rail. All using the best of Paul Hynes best silver DC cables. No JSSG.

 

Solution A:

 

Pros

1. Amazing transparency.

2. Sense of space is more open.

3. New microdetails are heard and are surprisingly at the forefront. These details cannot be unheard when switching back but did not draw attention to themselves in the same way.

4. The sense of more information retrieval. There will be moments in your listening where you are shocked to re-listen to a very familiar song and will hear a series of notes in the background that you did not know existed! It is something to experience for sure.

 

Cons

1. The sound is a bit leaner and i'm a sucker for a "full bodied" sound. (my preference)

2. More noticeable sibilance over time. I am convinced this has more to do with the EPS cable than the Taiko DC-ATX.

3. Not as much sense of ease. Perhaps I was approaching the 10A limit or not staying within a comfortable range.

4. Down a few notches in slam/impact.

 

Solution B:

 

Pros

1. The sense of ease you get with a dedicated clean 12v rail is apparent.

2. More slam and visceral effect in hard hitting songs.

3. Sibilance is completely not an issue, especially with female vocals and sharper sounding songs.

4. Fuller bodied sound with no obvious loss of speed.

 

Cons

1. The sound does not have the same openness as Solution A although I would not consider the soundstage smaller or closed in.

2. Ultra fine details are not as audible and do not draw as much attention to themselves although they are present (after they jump out at you with Solution A and you cannot unhear them).

3. You kind of miss the magic of solution A depending on the song

 

Summary

In terms of musicality it is a toss-up. Please keep in mind that the cables are not entirely equal and I have had sibilance issues with the Ghent cable in my system before. With the cables I have on-hand I would prefer Solution B for longer sessions, but Solution A has a Lot of potential and it has this way of drawing you into songs you know very well. I would be curious to see how Solution A would work out with a higher capacity rail and/or an equal quality EPS cable.

 

Is the Taiko DC-ATX worth it as an upgrade from the HDPlex? In my particular setup I took a risk and ordered it not knowing what to expect...and it was a major step in the right direction. The HDPlex is good, the Taiko is astounding in comparison, even for ATX alone I feel it is a worthwhile upgrade.

 

Follow-up: In the near future I will be building another EPS cable made from the same wire we will use for our unregulated linear power supplies. This cable is known to be very neutral so the hope is that I can conduct a fairer comparison between solutions A and B. 

 

Hopefully you all find this useful!

 

That's great feedback @Exocer. I just wanted to add a couple observations of mine as well. If you have read my big post a few pages ago, there was this section there:

On 4/14/2021 at 1:04 PM, Nenon said:

Replacing the HDPlex (which was only feeding the 3.3V and 5V) with the Taiko ATX was a big improvement. But Emile told me the strength of the Taiko ATX is in powering the 12V EPS rail. I put together a higher current higher voltage LPS (mine was 12V which was not enough for the Taiko ATX) and while experimenting I managed to destroy one of the rails on the Taiko ATX. Oops! That happened on the day I received it. Emile added protection in the design, so all the units he is selling have that now. But it took a while to send me a new unit. However, I managed to bypass the ATX on the Taiko ATX and just use the 12V EPS rail on the Taiko ATX.

 

Let me expand on that a little. The first month or two after I received the Taiko ATX I could not use it for the ATX section because I destroyed one of the rails and was waiting for a replacement unit. But I used the PS_ON jumper and hooked it up directly to the EPS connectors, supplying only 12V to my dual CPU ASUS Sage motherboard. 

On 4/14/2021 at 1:04 PM, Nenon said:

Did that sound better than my pimped out high-current LPS? No, it did not! At least not at the time. I rolled up my sleeves and started working on this unregulated LPS with some help from Emile. I quickly realized how transparent the Taiko ATX was. Any change you make to the unregulated LPS, and you can immediately hear. I played with different transformers, different rectifiers (including mosfet rectifiers), different capacitors, bypass capacitors, inductors, etc. I have been doing this for about 6 months.

 

There was a few things that I learned at this time. Here are some comments: 

 

1. Although, the ATX input works with any voltage between 16V and 48V, I found that higher voltage sounds much better. I would say that 16V-19V did not sound very good, and the magic happens at higher voltage - 24V - 35V in my opinion is the sweet spot. With the transformer I picked, I am at around 30-31V DC at the moment. 

 

2. For some reason, I prefered feeding it with an unregulated LPS instead of regulated LPS. I don't know why, but there is something really good about feeding the GaN Fets with unrestricted current and unregulated linear. One thing to mention, though, I did not try a 30-35V DC regulated LPS. I could have tried one but did not feel the need. I got excited that we could have such excellent results from a relatively inexpensive unregulated LPS. Perhaps we should try a 30-35V regulated LPS, but I strongly believe at this time that my server benefits more from unrestricted current than regulation before the GaN ATX. 

 

3. All my experiments were with two 85W TDP Xeon CPUs. Those are much harder to power than a single 65W TDP AMD Ryzen CPU or a 95W TDP I9 9900k. And that is where this power supply arrangement really shines. I am not sure how well it works on lower power CPUs. I am guessing on low powered CPUs a good LPS would work better. But this dual CPU computer is a different story. Those are some experiments people would need to make. There are specific applications in which we want to use the Taiko ATX. It's not a miracle device that would work best in every single application. 

 

4. I find the GaN Fet ATX to be extremely fast. Too fast actually. Adding some big capacitors in front of it makes it fuller, more dynamic, and more expansive. It also slows it down a little, although it is still faster than a LPS. Seems like this is one of the things @Exocer ran into. 

 

Just to get an idea of how sensitive the GaN ATX is on what is feeding it... I did an A/B comparison between different ring terminals that connect the high current capacitors yesterday. That's the connector on each end of the 3-parallel white wires link on the picture below:

 

I compared a regular tin-plated copper ring terminal from my local hardware store to a high quality Mundorf copper ring terminal:

 

Well, it turned out everything matters, so I will be buying more of those Mundorfs. I would not say the difference was shockingly big, but definitely worth the few extra bucks in my system. I guess it never ends :). 

[seeteeyou]

On 4/4/2021 at 2:51 AM, Topk said:

However, I noticed 2 very interesting posts from Energy and Seetoyou on the Gigabyte Aorus Master MB. It’s the first MB on the market that has 16 direct power phases. Power phases/VRM down convert and regulate the 12V and provide power to the CPU and chipset. 
 

Therefore, the idea naturally came that the Aorus Master might be the best MB for audio PC since it has the best power phases (along with 6 layer PCB, 2 oz of copper etc.) therefore benefiting from robust, low ripple, stable, more dynamic power to the CPU. Also high quality VRMs overclock better, generate much less heat etc. This is entirely measurable and all proven via benchmarks. In fact the Aorus master is considered one of the best MB for the AMD Ryzen in the PC world.

 

FYI - Intel fans might wanna take a look at this Z590 baby

 

https://www.asrock.com/mb/Intel/Z590 OC Formula/Specification.asp#Specification

Quote

- 16 Phase CPU Power design + 2 Phase Memory Power design
- High Density Glass Fabric PCB
- 12 Layer PCB
- 2oz Copper PCB

 

It didn't look like a fanless motherboard, though

 

https://wccftech.com/asrock-unleashes-z590-oc-formula-motherboard-aimed-at-overclockers/

Quote

The CPUs are powered by dual 8-pin connectors & fed by a monstrous 16 Phase SPS power delivery that makes use of 90A power stages and a high-grade (50A Dr.MOS) MOSFET implementation. The motherboard features a 12-layer server-grade and low-loss PCB.

 

And then each 8-pin EPS could very well be fed by this fairly affordable GaN-based PSU (up to DC 12V / 10A a piece so having a pair of them could deliver 240W to the CPU)

 

https://www.banggood.com/ToolkitRC-P200-Mini-30V-AC100W-DC200W-10A-GaN-Adjustable-Mini-Desktop-Power-Supply-USB-Output-p-1825686.html

https://www.toolkitrc.com/P200

[Exocer]

5 hours ago, Nenon said:

3. All my experiments were with two 85W TDP Xeon CPUs. Those are much harder to power than a single 65W TDP AMD Ryzen CPU or a 95W TDP I9 9900k. And that is where this power supply arrangement really shines. I am not sure how well it works on lower power CPUs. I am guessing on low powered CPUs a good LPS would work better. But this dual CPU computer is a different story. Those are some experiments people would need to make. There are specific applications in which we want to use the Taiko ATX. It's not a miracle device that would work best in every single application. 

 

4. I find the GaN Fet ATX to be extremely fast. Too fast actually. Adding some big capacitors in front of it makes it fuller, more dynamic, and more expansive. It also slows it down a little, although it is still faster than a LPS. Seems like this is one of the things @Exocer ran into. 

 

@Nenon, thanks.  Do you have any interest in testing the unregulated supply + Taiko solution with your i9 rig (if its still around)? It would be an interesting data point for those who plan on sticking with single CPU builds for a while. 

 

In the future I will test whether the unregulated linear supply + Taiko Dc-ATX will surpass the two high quality PH SR7T rails. That would not be an apples to apples comparison since my SR7T only outputs 19V DC, so considering the impact voltage has on the Taiko DC-ATX it would only serve as a way to compare what most would have on-hand (at least a single high quality 19V regulated DC rail) to a dedicated unregulated supply voiced to work with the Taiko DC-ATX/Asus Sage C621/Dual Xeon Scalable CPUs. This is still a worthwhile test to conduct for the community.

 

Another test to conduct would be powering just the CPU from the PHSR7T and ATX from the Unregulated supply + Taiko. I would think this would get a bit tricky and may introduce grounding issues if not careful. The more tests we conduct, the better.

 

With the right EPS cable perhaps Solution A (Taiko powering everything) will work better for me. I am very picky about the high frequencies. Maybe an issue elsewhere in my chain is revealed by the Taiko. There are so many variables to keep in check and to take into proper context when testing and making comparisons.

 

I have switched to Qobuz today and I will be conducting a similar test to the previous one in the near future. I will probably keep both Tidal and Qobuz since Qobuz did not have some of my favs. 

 

Cheers,

-Rob

 

 

[The Computer Audiophile]

4 hours ago, seeteeyou said:

didn't look like a fanless motherboard, though

What do you mean fanless motherboard?

[seeteeyou]

Regarding those "flagship" AMD motherboards, they're meant for extreme overclocking and therefore almost all of them would come with active cooling by default

 

https://www.reddit.com/r/Amd/comments/btkx56/currently_gigabyte_aorus_extreme_is_the_only_x570/

 

Recently they managed to produce a fanless flavor as follows

 

AMD Reportedly Launching New X570S Chipset with Passive Cooling
https://www.thefpsreview.com/2021/04/06/amd-reportedly-launching-new-x570s-chipset-with-passive-cooling/

 

---

 

And then Intel's Z590 might seem to be the opposite of AMD's X570, passive cooling should be the norm while some outliers might require quite a few fans for starter

 

https://www.tomshardware.com/features/intel-z590-motherboard-and-chipset-overview-45-rocket-lake-boards-detailed

Quote

But we’ll see the continued use of formidable VRMs and large heatsinks to keep these power-hungry CPUs in check. Some Halo products may include active cooling on the VRMs or even integrate water blocks to cool the VRMs and CPU.

 

---

 

Fanless or not, ultimately maybe it's more like a balancing act?

 

https://audiophilestyle.com/forums/topic/58164-building-a-diy-music-server/page/83/?tab=comments#comment-1122925

On 4/3/2021 at 7:04 PM, guiltyboxswapper said:

Simply put, running hardware comfortably within its specification thermally rather than letting it run to its limits can in some cases outweigh the vibrations from the fan itself. This will vary from hardware to hardware.  

 

And then the OS / software player etc. would also play an important role, some of them could actually reap the benefits by scaling right up to the performance provided by a (much) higher clock speed.

 

In the end it's just a matter of the interaction between the hardware side and the software side, for instance the whole point of a dual-socket motherboard should be pinning HQP to Xeon Silver #1 and NAA to Xeon Silver #2 with the right CPU affinity.

[genvirt]

Have a question:

unfortunately Euphony OS meanwhile not supporting temperature monitoring of the AMD Ryzen (my MB is an Asus ROG Crosshair VII), so is here some way/trick to monitoring temperature externally(some small external display/panel that can duplicate on board status diodes)? 

To load Win OS with monitoring software this is too annoying and not comfortable of course.

Thanks

[dctom]

20 hours ago, Exocer said:

The 800W HDPLEX was used. I probably should have mentioned that. 

 

USB Cable is the Sablon 2020.

 

Euphony Stylus is what I am testing with (no Roon components involved).

 

Both streaming from Tidal and local Redbook flac. Music files currently stored on Optane 800P.

 

Thank you Rob for the info.

 

Yes I use the same USB cable, performs well in my set up.

 

It would be good if you could compare the PH SR7t to the Taiko PSU, maybe powering a single CPU would give different results? After waiting so long for the PH7 seems a shame to give it up!

 

Have you compared local files and Qobuz using the Taiko?

 

From what you and Nenon have written it seems there are no predictable results with the Taiko. With the Taiko ATX being so transparent, cpu, motherboard, file source etc will all influence results in some way.

 

 

[MarcelNL]

a fan-less MB is nice, it is possible to use fan management as for the CPU on some MB;s so depending on the use case there may not be a reason to look for a fan-less MB. The Aorus MB I'm currently using has a fan-stop feature, in my usage the chipset fan never comes on, the CPU fan also hardly does.

[guiltyboxswapper]

15 hours ago, seeteeyou said:

And then the OS / software player etc. would also play an important role, some of them could actually reap the benefits by scaling right up to the performance provided by a (much) higher clock speed.

 

I sometimes wonder if high clock speeds = much more likely to disable the CPU's C-State (power saving) transitions which means hopefully a more steady state power consumption leading to better SQ potentially.  No doubt about it, disabling C-States for me results in a very noticeable uptick.  This can be done in software (Linux Kernel has an option to simply disable sleep states).

[MarcelNL]

that should be easy to check for someone, enable and disable C states while overclocking and listen. For me the improved SQ is noticeable with C states disabled (in Bios and in Daphile-Linux)

[Exocer]

3 hours ago, dctom said:

It would be good if you could compare the PH SR7t to the Taiko PSU, maybe powering a single CPU would give different results? After waiting so long for the PH7 seems a shame to give it up!

Sure. I will conduct this test. I do not think you will need to give it up if you are sticking with a single CPU. The Taiko DC-ATX + PH SR7T is fantastic. You have the option to keep it and I think you will be pleased going that route. I think the Unregulated supply is more of a necessity for the Dual CPU build per Nenons tests. Will keep you posted.

 

  

3 hours ago, dctom said:

Have you compared local files and Qobuz using the Taiko?

No, I have not. I can test and report back my findings. Porting over my favorites to Qobuz as we speak :)

[Nenon]

2 hours ago, guiltyboxswapper said:

a more steady state power consumption leading to better SQ potentially

Steady power consumptions is a key factor in many high-end servers. It's one of the many variables that helps sound quality. And there are multiple ways to control that from software and hardware. 

 

2 hours ago, MarcelNL said:

The Aorus MB I'm currently using has a fan-stop feature, in my usage the chipset fan never comes on

That's a good feature. 

 

2 hours ago, MarcelNL said:

the CPU fan also hardly does

Ugh, I would stay away from any computer fans. But that's me. 

 

4 minutes ago, Exocer said:

Sure. I will conduct this test. I do not think you will need to give it up if you are sticking with a single CPU. The Taiko DC-ATX + PH SR7T is fantastic. I think the Unregulated supply is more of a necessity for the Dual CPU build per Nenons tests. Will keep you posted.

We still need to test all that.

I would not be surprised if "unregulated LPS --> Taiko ATX --> ATX+EPS" sounds better than using a separate regulated LPS for the EPS and TaikoATX on the ATX on single CPU servers too. You are mixing up ground planes, restricting the current, etc. You haven't really tried yet. And I haven't tried it on a single CPU server either. But it would be nice to test all that.

 

In my tests, the Taiko ATX really shined on the EPS, not on the ATX. But it needs to be fed by higher voltage properly designed unregulated LPS. It also takes some time to burn-in... gets better over time.

 

As soon as the Noratel transformers get delivered to Taiko, we would have everything needed to build an unregulated LPS. 

@Exocer I may even build one and send it to you to try. Or I can send you the parts needed (or just those you can't find) and you build one yourself. We can discuss the details on PM when the transformers are ready. 

 

[seeteeyou]

FYI - for those of us who aren't interested in running both HQP and NAA on the same dual-socket Xeon Scalable motherboard, we could find specific single-socket ones (i.e. Skylake-SP and Cascade Lake-SP) with Sub-NUMA Clustering that's supported by relevant BIOS options

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/616/?tab=comments#comment-1014096

On 12/16/2019 at 7:37 AM, romaz said:

"Well RAM is a topic on its own, to start with, the 2 cpu’s are split into domains (NUMA / SNC), so you really have 2 x 6 dimms, 6 for each CPU, they are not shared. Music services have their own cpu/dimms and the OS has its own cpu/dims. So its sort of a core and endpoint into a single machine going beyond just core allocations for individual processes.

 

https://software.intel.com/content/www/us/en/develop/articles/intel-xeon-processor-scalable-family-technical-overview.html

Quote

SNC mode has to be enabled by BIOS to expose two SNC domains per socket and set up resource affinity and latency parameters for use with NUMA primitives.

 

I just DAFS on Google and only found motherboards from maybe a dozen vendors with such BIOS options

• Cisco
• Dell
• EVGA
• Fujitsu
• H3C
• HPE
• Huawei
• Intel
• Lenovo
• NEC
• Oracle
• Supermicro

The vast majority of them could only be purchased as a complete server system (usually the form factor itself ain't even close to SSI-EEB / E-ATX / ATX) instead of a standalone motherboard, therefore I just narrowed them down to these guys

• EVGA SR-3 Dark
• Supermicro X11SPL-F
• Supermicro X11SPD-F
• Supermicro X11SPA-T
• Supermicro X11SPA-TF
• Supermicro X11SPG-TF
• Supermicro X11SPi-TF
• Supermicro X11SPW-TF
• Supermicro X11SPW-CTF
• Supermicro X11SPH-NCTF
• Supermicro X11SPH-NCTPF

IMHO it would be quite interesting to try SNC with a single 20-core Xeon Gold 6138 (125W TDP might not need any fans if we weren't upsampling with HQP?) and see if that were getting anywhere close to a pair of 10-core Xeon Silver

 

https://ark.intel.com/content/www/us/en/ark/products/120476/intel-xeon-gold-6138-processor-27-5m-cache-2-00-ghz.html

https://www.ebay.com/itm/Xeon-Scalable-Gold-6138-SkyLake-20-Core-2-0-GHz-3-7-GHz-Turbo-LGA-3647-SR3B5-b/254921475866

 

---

 

I have no idea if this seller in China were legit or otherwise, though 20-core Xeon Gold 6222V @ 115W TDP could be found below for 4,300 RMB / 660 bucks

 

https://item.taobao.com/item.htm?id=641024936013

https://ark.intel.com/content/www/us/en/ark/products/193962/intel-xeon-gold-6222v-processor-27-5m-cache-1-80-ghz.html

[Downtheline]

14 minutes ago, seeteeyou said:

Supermicro X11SPL-F

I have this motherboard and the silver xeon 4210. I'm also waiting for a taiko dc/atx and I have a 2 rail ph sr7t. 

 

I just received 4gb x 12 ecc wt ram from Apacer for my upcoming dual cpu build... I could try these with my current server and use these bios customizations... but I'm a noob on these things. I'm open to instructions! 

 

 

@Nenon, with the latest info from extreme-land regarding improved player software and usb connection decreasing need for network upgrades, do you think that delta between dual cpu and single cpu systems might decrease?

 

Since the dual cpu system is designed to decrease cpu use of network resources for the "endpoint cpu", perhaps most of that could be taken care of with software. 

 

Either way, this is somewhat testable but a big investment. 

 

 

[Exocer]

9 minutes ago, Downtheline said:

1 hour ago, seeteeyou said:

Supermicro X11SPL-F

I have this motherboard and the silver xeon 4210. I'm also waiting for a taiko dc/atx and I have a 2 rail ph sr7t. 

Would be interesting if you tried it with the Xeon Silver. You'll have less cores overall but there may still be benefits to doing what@seeteeyou suggested. This would be a great data point.

[seeteeyou]

Part 1 is BIOS setup as shown on page 80 of the Supermicro X11SPL-F manual

 

https://www.supermicro.com/manuals/motherboard/C620/MNL-1950.pdf#page=80

Quote

Chipset Configuration → North Bridge → UPI Configuration → SNC → Enable

 

Part 2 is the OS setup and Windows 10 should look like this

 

https://www.windowsdigitals.com/how-to-set-process-cpu-affinity-priority-permanently-in-windows-10/

 

Process Lasso and dAffinity could also handle something like that but not all features seemed to be free, though the freebies could already cover the basics

 

https://bitsum.com/howfree/

https://www.d7xtech.com/daffinity/

 

However, we're talking about 2 clusters on the same 10-core Xeon Silver 4210 processor and therefore I dunno if the primary one were occupying CPU 0 through CPU 4 and secondary one were occupying CPU 5 through CPU 9 respectively.

 

It doesn't hurt to try different combos and compare the results, though.

[MarcelNL]

2 hours ago, Nenon said:

 

6 hours ago, MarcelNL said:

the CPU fan also hardly does

Ugh, I would stay away from any computer fans. But that's me. 

 

I definitelyy plan to get rid of the fan, so far I haven't found a solution that I trust for passive cooling but the fan does not kick on until 55'C core temp which I hardly reach....current temp with a FLAC playing is 48'C, the Xilinx card runs hotter!

 

[dctom]

15 hours ago, Exocer said:

Sure. I will conduct this test. I do not think you will need to give it up if you are sticking with a single CPU. The Taiko DC-ATX + PH SR7T is fantastic. You have the option to keep it and I think you will be pleased going that route. I think the Unregulated supply is more of a necessity for the Dual CPU build per Nenons tests. Will keep you posted.

 

  

No, I have not. I can test and report back my findings. Porting over my favorites to Qobuz as we speak :)

 

Thanks, look forward to seeing your results.

 

Do you know if Taiko are contacting prospective purchasers yet. I did submit one of their contact forms.

[Exocer]

8 hours ago, dctom said:

Do you know if Taiko are contacting prospective purchasers yet. I did submit one of their contact forms.

Hi dctom,

This is a better question for @JayM who is here representing the Taiko brand and providing updates.

[Nenon]

Had a quick chat with @paulhynes and got some updates from him.

Although he had closed the custom build service to new orders, there are currently 7 custom builds available for transfer to new owners. There is a last chance for 7 people to get the otherwise unobtainable custom dual regulated SR7, which is now called DR7.

 

Per Paul:

Quote

The custom build schedule is now moving ahead again as I have managed to acquire the lease on another workshop unit next to PHD Ltd were I can work in isolation clearing the outstanding backlog of custom builds without fear of contracting covid-19. I am moving furniture and equipment into this unit over the next week and should be operational the following week.

 

Just passing the message here... 

[Exocer]

Hi All,

In light of the recent dialogue around power supplies I decided to reach out to Paul to get his take on unregulated supplies. As a happy SR7T owner I am still open to trying the new Unregulated supply with the TAIKO DC-ATX and chances are I will be keeping both power supplies around regardless of the outcome of my tests. 

 

Here it is:

 

"...

Lets go through your questions :-

 

1. Fast transient response requires speed from the regulator circuit design. Mosfets are blazingly fast and have much wider bandwidth compared to bipolar transistors so I use Mosfets in the regulator output stages. Driving Mosfets at very high frequencies can be tricky but with a little care in the design of the driver circuitry exceptional transient response and bandwidth can be achieved. Also keeping the Mosfets and the driver circuitry stable at frequencies approaching many hundreds of Mhz can be tricky. The circuitry I use for this has been carefully optimised to provide very fast transient response, whilst minimising transient overshoot and ringing (critically damped).
2. Many voltage regulator designs utilise a number of gain stages to increase the available feedback to use for output error correction. This can be a can of worms as the more stages used the more frequency compensation is required to keep the design stable. This has an adverse effect on operating bandwidth, the transient response time and the settling time. The regulator design used in the SR7 has only one gain stage in the error drive amplifier directly feeding the output Mosfet. This provides a tight loop around the error amplifier and output mosfet that does not require frequency compensation for stability and the resulting phase margin is excellent and such that the design is load tolerant and not prone to parasitic oscillations. I have used this regulator circuit to drive anything from large low ESR capacitor banks to inductive loads like field coil loudspeakers and I have not managed to initiate instability. It works very well with both analogue and digital loads (as you are finding with your SR7T).
3. The very low wide band output impedance from the regulator output stage is capable of excellent load damping so HF noise feeding back into the regulator from the load is well suppressed.
4. The load regulation performance is very good as the error amplifier is fast and can drive the Mosfet well. One point I will add here is that many power supply products on the market have rather abrupt current limiting applied to the output stage. This is fine if the load never demands currents exceeding the current limit but can cause all sorts of problems with loads demanding high transient current delivery. Transient current starvation is one likely cause for problems with motherboard boot sequences and program dropouts when the current limit of a supply is exceeded. The rated load current in your SR7 high current rails would be specified for say 19v 10A or 12v 12A continuous depending on which modules have been fitted. The voltage regulator error amplifier is capable of providing enough voltage between the gate and source terminals of the output Mosfet to deliver transient currents up to 50A into the load (I specify 40A transient to allow for device manufacturing spreads). The regulator circuit design is easily scalable for higher continuous and transient current delivery although the thermal management does require rather larger heat sinks (and larger mains transformers) than the ones in the SR7T chassis. Such products are in the production engineering phase at PHD Ltd at present.
5. I specify the PSRR of the SR7T voltage regulator as >80dB (once again to allow for device manufacturing spreads) from DC to 100 KHz with a gradual roll off above this frequency. This is not the whole story of course as the mains transformer is configured as a balanced isolation transformer providing good common mode rejection of incoming RF interference and the energy storage capacitor bank also has a very low impedance, especially at higher frequencies due to the very low ESR of the capacitors arranged in a parallel array. The SR7DR (double regulated) power supplies move the PSRR up to around 150dB from DC to 100 KHz.
6. The PSRR specifications apply to ripple breakthrough and also form part of the noise specification. There are a number of steps taken to reduce the regulator circuit noise. Firstly a very low noise voltage reference is used to feed the error amplifier via a noise filter which provides an exceptional low noise reference. I specify the error amplifier devices having a noise rating of 0.5 nanovolts root Hz but this is conservative as they typically measure around 0.25 nanovolts root Hz. This is a lot quieter than the Linear Technology voltage regulator chips that are currently popular in power supply designs. The composite error amplifier/output stage has no noise increasing voltage gain so the noise levels are typical across the range of output voltages of the regulator modules.

 

I have been through a phase of working with high performance unregulated power supplies back in the 1990s which can work well especially when compared to older style integrated circuit voltage regulators with limited bandwidth, transient response time and settling time. However historically I have found that it is possible to design voltage regulator circuits that offer better performance than unregulated linear supplies generally, although knowledge increases with time and better components and materials can make a difference so I would be reluctant to state categorically that active linear supplies can be a better solution. I would have to conduct extensive testing and listening trials before forming any conclusions currently.

 

Out of interest I have spent 41 years specialising in audio power supply design and I can say from experience that inadequate design of voltage regulator circuitry can severely limit performance. I touched on this subject in a series of articles published in underground audio magazine “Common Ground” during the 1990s. I have attached an updated copy of these articles for you to read. You might be surprised by the overall conclusions discussed in the articles as they also discuss power supply system design for the lowest inter-stage interaction. Unfortunately it would be unlikely that current motherboards could be adapted to this philosophy on a DIY basis although separate EPS power connections from the ATX power connections are a positive move in this direction.

 

Also of interest, over the 41 years I have been researching power supply systems I have found that no amount of audio component “bling” can compensate for an inadequate voltage regulation circuitry although it can enhance performance of a good voltage regulator design.

 

Heavy load current changes on the SR7 outputs only produce small output voltage changes and static circuit noise remains the same at all output currents within specification. Nothing is perfect and it would require an output impedance of zero Ohms at all load operating frequencies to have no output variations, which is currently not possible with any available technology I am aware of.

 

I hope this helps.

 

Regards

Paul"

[NanoSword]

28 minutes ago, Nenon said:

which is now called DR7.

He is offering now  DR7T  turbo upgrade 🥰.