Buffalo GS20xx - Bypass 12V stage via 3 external PSs

[tgb]

hi,

I propose to feed this thread with ideas & modds regarding the power supply of GS20xx switches using 3 PSs at the voltage used on the PCB : 3.3 / 1.5 / 1VDC.

 

This thread is not about the 12VDC PS, but the step further : bypassing this stage of downregulation from 12VDC to the 3 lower voltages.

 

Please find here the rough info I posted yesteday eve on another thread :

2nd attempt, I finally managed to boot & stream music with my GS2024 using 3 external PS, bypassing the internal regulation from 12 to 3.3/1.5/1V.

That's was just a rough test. It works.

Next step : find the right way to give it a clean current.

The trick is : as you all notice, the heat sink is quite hot (the 2nd one on my GS2024 is mild vs the common one), thus nearly all the power goes thru these 3 tiny voltages, thus : high current.

Please find here my report => http://forum-hifi.fr/thread-16891-post-378333.html#pid378333

If Google translate get stuck with my French, I'll post here the ENG version. Just ask.

 

These are the photo to make it clearer :

This is how I did it :

PS is a basic 5VDC USB charger, it can deliver 2.1A max

Downstream : 3 Droks modules (see : www.droking.com ; cheap & efficient Buck converter). The Droks I have can manage high  current but can't go lower to 1.25VDC, thus as I was not sure how to manage the voltages  I used a MPaudio module downstream the Drok on the 1VDC line, to split in 2 the voltage drop from 5V to 1V.

 

Below the way I power. Rough work as I said, it was just to confirm that the switch can work this way :

 

@ RickyV, I post here the reply to your post (https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/?do=findComment&comment=1067930)

- 2A reg stage is effectively to short.

the Drok on the 1.5V line was a bit hot after 10min ; quite unsual for these very efficient modules

the MPaaudio SLS was @ 40°C after 10min msuci streaming => I stopped the experiment there. This module is design for 2.5A max current. I guess we are not far. A 5A MPaudio module should be the right one to power the Buffalos.

- as you mentioned, I should have removed the coils, but it was just a experiment to see roughly if it can boot & work this way & how high are the currents (Sunday eve, no time to go further). 

And please notice that my GS2024 is still not burnt, and without coils I can't complete the burning simply using the 12VDC ! 🙂

Next step is to cut the coils of course & find the right spots to power the PCB.

- liek you say : the key here is the current, quite high, thus we need to get rid of a lot fo heat. To reach the 1V, I think the only way to process is too use a 2 steps down-reg to split the heat between 2 modules.

 

I did the bypass on my Tp-Link SG5412F : night/day after the modd.

Pretty sur it can be the same with the Buffalos.

Rgds

[tims]

Hi tgb

Thanks for starting this thread. 

I'm waiting for my gs2016 from Japan and will have to replace the 100v mains supply and feed 12vdc directly to the pcb to get it up and running. 

Any idea if the 12v input is strictly necessary or can it be lower as the 3.3 / 1.5 / 1VDC regulators surely don't need a 12v input. 

Or maybe there are other regs on the board that need the 12v?  

Thx.

[Middy]

Has anyone Emailed buffalo tech support, its always worth a try..

[tgb]

Hi Tims,

if you simply want to replace the internal PS, just plug yours on the 12V pins.

 

This thread is related to power the switch directly with the 3 voltages that are used on the PCB : 3.3 / 1.5 / 1VDC

No 12VDC PS needed in this mod, simply because the purpose is to bypass the regulation steps.

Please find below this photo with the voltage zones & regulation zone to see how it works =>

The size of the PCB of my GS2024 is a bit bigger than 2008 & 2016, but not much.

There is a 2nd heat sink on the GS2024, but if you find photos of the PCB of 2008 & 2016 you'll see that you have the same zone of down-regulation (there are not strictly at the same place vs GS2024 but the overall story is the same :

this is the GS2024 =>

this a rough comparaison of 2024 vs 2016 PCB =>

 

 

[RickyV]

This is the BS-gs2016.

 

 

with temporary smps input.

[tgb]

Cool !

Does it sound better with the LDOVR ?

[RickyV]

8 minutes ago, tgb said:

Cool !

Does it sound better with the LDOVR ?

Yes, this config already removed some harshness but you have to know that my audio switch is the ER and that I go via planettech, fiber, planettech to buffalo to modem router. Now I have a lps that powers the router and the buffalo with 13V and the 3A ldovr goes to 12V. 
Experimentation is still ongoing but at the moment I am rebuilding my NUC server power supply.

In the novel thread I already posted some of what I did to my buffalo 🐃.
 

No animals where hurt during this modd. 😀

[zerung]

[tgb]

thanks for the photo zerung (I saw it somewhere).

Is it yours ?

 

Unfortunatly, this mod is not about the 3 voltage supply.

12VDC seems to be pretty well done, but then it's the regs on the PCB that still work.

 

The Pink Faun clock add-on : really cool. But unfortunatly we don't how it has been done. I asked the owner to know more or have a more precise photo where the clock has been soldiered : "I paid, fine for me. The technical stuff behind that, that's not my cup of tea"... a pity we get stuck to this photo without explanation...

Rgds

[zerung]

Sometimes a picture speaks for itself. Or not in this case.

 

A group of us are doing this.

 

There will always be a debate on 'discrete and not' LPS power. This takes the former route. The intervention of sub-intervention in the power path is another issue worthy of debate. Your intervention is welcome from that point of view (Which we are not doing). The sonic result will be interesting for sure-to know.

 

The clock in the switch is another point for debate. Which you have decided not to partake in (Perhaps wisely?).

Using a high quality clock on my Ether Regen gave my network a turbo boost - perhaps should state - completely calmed the network down and pushes the music thru in its most ethereal state (Pun intended) ....So it is, that I am doing the same for my Buffalo switch.

 

Yes, many have decide not to add clock and implement the best power they can provide into their system. Like you, and I understand that is wonderful too!

[tims]

3 minutes ago, zerung said:

Sometimes a picture speaks for itself. Or not in this case.

 

A group of us are doing this.

 

There will always be a debate on 'discrete and not' LPS power. This takes the former route. The intervention of sub-intervention in the power path is another issue worthy of debate. Your intervention is welcome from that point of view (Which we are not doing). The sonic result will be interesting for sure-to know.

 

The clock in the switch is another point for debate. Which you have decided not to partake in (Perhaps wisely?).

Using a high quality clock on my Ether Regen gave my network a turbo boost - perhaps should state - completely calmed the network down and pushes the music thru in its most ethereal state (Pun intended) ....So it is, that I am doing the same for my Buffalo switch.

 

Yes, many have decide not to add clock and implement the best power they can provide into their system. Like you, and I understand that is wonderful too!

Hi Zerung

With you say there was a large boost to SQ your 12v LPS power supply compared with the stock switching supply?  How would you describe the changes?

[zerung]

9 hours ago, tims said:

Hi Zerung

With you say there was a large boost to SQ your 12v LPS power supply compared with the stock switching supply?  How would you describe the changes?

@tims The unit with out the LPS is a computer interface. Mediator of data, 'A post man'. Best in its class. With the LPS it becomes a musical instrument and starts to be the facilitator of music. The noise dies down and it becomes quieter. The nuance in the music comes through.

[tgb]

Hi Zerung,

sorry about my last post, my point was not to be rude towards you or this kind of mods. Absolutly not.

My point was just to know some technical things about your modds :

- how & where the Pink Faun for instance is plugged ?

- is it plugged on the part of the PCB that works @ 3.3VDC / 1.5VDC / 1VDC ?

I know that that kind of mod leads to a big jump in SQ. Totally agree vs this mod.

 

For instance, I received transfomer & LPS block (LT2045 based) from MPaudio yesteday : night/day vs the internal SMPS. Amazing that GS2024...

Rgds

[zerung]

@tgb No offence taken.

 

@Soul Analogue Has just joined CA. He is our resident DIY wizard, the proponent of this intervention. Perhaps I can ask him to chime in on his clock design.

[Soul Analogue]

Hello all,

 

yes, i have the same idea of powering the buffalo by three lps rails.

 

the intervention post by zerung was my work, most cost effective and keeping the original buffalo chassis. The result of 12v lps and the clock mod is promising.
 

i tried to measure the current taken by the 1v on the gs2016a, i believe it takes higher than 10a..... to build a lps for it is not that easy. And you need a big chassis and heatsink to accomodate a over 10a lps. 

the 3.3v and 1.5v rails should be pretty easy. I have yet to measure the current taken by these rails yet.

 

The separately powered clock mod is essential for the sound quality. But i could just make it work by an smd oscillator 25mhz. I could not make it pass the power on cycle with ocxo. Tried two ocxo, same frequency 25mhz, no avail. 
 

Did someone say there is mod by pinkfaun ocxo? I would be interested

[tgb]

Hi Soul Analogue,

Very nice work.

 

Regarding the Amp @ 1V : I don't see how it can be so high.

For instance, I power the GS2024 at 12VDC => 0.8A max, thus : 9.6W at max.

How could ti be possible that the 1V rail uses more power than the device as a whole ?

Sure it is high current, but I'm like you => I'll check again & more precisely the 3 currents :-)

 

Do you know why the OCXO you tested did not power on ?

 

I think it is me that talked about the Pink Faun mod, saying that we do not see how they plug the clock under the heat sink.

So, based on your photo, you proceed like this : green wire is "clock" ? & black one ?

[Soul Analogue]

Green wire is gnd, clock signal is black

 

no idea why it could not be powered on with ocxo... even the manufacture of the ocxo cannot explain... he just ignored me after i show him the pics

 

there is no way an ocxo be placed “under the heatsink”... ocxo physical size must be substantial... 🙄

[tgb]

the clock can be placed aside the heat sink, like Pink Faun did.

With 2 wires like on your photo. No !?

[Soul Analogue]

Can you show me Pinkfanu’s work on buffalo?

thanks

 

i would also be grateful if you can check the current drawn by the 1v....

i failed to do so in my attempt..

i used a 10a lab variable psu (linear) set to 1v and a fluke dmm in series to check the current...but when powered up... the voltage only got 0.6v at the buffalo end.. and obviously it didnt boot...

[tgb]

Strange you did not manage to boot it.

See my post above => 

the Buffalo booted.

The issue then... was that the 1V required too much amps for my MPaudio block. But : it worked !

Knowing that the MPaudio block I used is limited at 2.5Amax. The 1V circuits should require 2.5A at very max. Maybe 1.5 or 2A..

Since this trial, I received this => https://fr.farnell.com/tenma/72-2985/multimetre-pince-amp-auto-100a/dp/2667180 ; now the Amp measurement will be more precise & easy to do. Very simple & efficient device.

 

The photo of PinkF work was on the other thread => see my post for instance, here (the pb is that we see nothing "interesting" => 

 

[tims]

My two GS2016 will hopefuly arrive soon.  To get them going I need to remove the 100vac switching PS and replace them with a 12v linear supply direct to the pcb's but I've only got one 6 amp linear PS available for both switches. 

Will this be detrimental to SQ by a loss of galvanic isolation due to sharing the power supply with both switches?

Thanks

 

[tgb]

hi tims,

my GS2024 runs at 12VDC and uses 0.9Amax

so with 6A available you have much more than enough amps to power your 2 switches ! A Y cable & it gonna be fine. No pb.

Rgds

[RickyV]

So I Frankensteined this today and it works,😀.
 

 

So I had to fall back to a smps to deliver the power, lower right corner.

From 7.5V I dropped it to 3.6V, that heatsink gets 60 degrees C. From this 3.6V I fed a 1A 3.3V module and a 2A 1.5V module. I used 2A module because dissipated heat/ power. 
The 1V, 5A was still a problem I started out with 4V on the input but that the regulators got about 80 degrees C, too hot. Than I tried it one diode, dropping 0.8V that dropped the temperature to 60 degrees C, still to hot. Than two diodes dropping 1.6V so now the 1V/5A regulator has 2V on its input and is about 50 degrees C now.

The heatsink’s from the first regulator and the diodes get about 60 degrees C. Even the smps get 43 degrees C.

So now that everything is working I will put it, temporary, in my system upstream of my Etherregen, connected by fiber.

So now I have to figure out what I need for a power supply. 
7.5Vx 5A= 37.5W    37.5W/3.6V= 10.4A at 3.6V ???

 

 

[tgb]

Well done Ricky !

 

Regarding the amps, do you have a wattmeter or something to know how much the power SMPS deliver ?

the best is to use an amperemeter to check the amps at each voltage. It is on my todo list, this week.

7.5V/5A it is its max output ? not what it really deliver ?

Then, as your way to deliver the power is a chain 3.3 > 1.5 > 1, we can't guess the amps at each voltage this way (I think).

 

When I power it, I'll use a 4ports USB charger, 1 port per voltage, and will use a amperemeter to check the amps at each voltage. I'll do it quickly because last time the buck converter on the 1V got hot quickly (like you).

Apparently we face high currents (not too high anyway), thus, the final setup I see is : custom transformer (MPaudio can do that ; 4-6weeks delay ; around 70eur).

A "custom" transfo to get a tight dV at the regs (1V should be efficient vs regulation & not too big to keep the heat at minimum).

For instance / the 1VDC : I'll need 1+1=2VDC upstream the regs => with 2*0.4V=0.8V voltage drop in the diode bridge & the ratio 1.41 => (2VDC + 0.8) / 1.41 = 1.985VAC. So a rough 2VAC transfo output would be ok to get the 1VDC with a tight dV=1V at regs.

Hope this rough calculation is ok.

I think using a custom transfo is the only way to manage thoses V & A, and avoid at max the issues regarding the heat transfer at regs due to too high voltage drop.

 

You plugs the GRD of each voltage to the caps next to screws ? Interesting. Equipolarity by doing so ? I'll use your method, thanks.

Rgds

[RickyV]

15 minutes ago, tgb said:

Well done Ricky !

 

Regarding the amps, do you have a wattmeter or something to know how much the power SMPS deliver ?

the best is to use an amperemeter to check the amps at each voltage. It is on my todo list, this week.

7.5V/5A it is its max output ? not what it really deliver ?

Then, as your way to deliver the power is a chain 3.3 > 1.5 > 1, we can't guess the amps at each voltage this way (I think).

 

When I power it, I'll use a 4ports USB charger, 1 port per voltage, and will use a amperemeter to check the amps at each voltage. I'll do it quickly because last time the buck converter on the 1V got hot quickly (like you).

Apparently we face high currents (not too high anyway), thus, the final setup I see is : custom transformer (MPaudio can do that ; 4-6weeks delay ; around 70eur).

A "custom" transfo to get a tight dV at the regs (1V should be efficient vs regulation & not too big to keep the heat at minimum).

For instance / the 1VDC : I'll need 1+1=2VDC upstream the regs => with 2*0.4V=0.8V voltage drop in the diode bridge & the ratio 1.41 => (2VDC + 0.8) / 1.41 = 1.985VAC. So a rough 2VAC transfo output would be ok to get the 1VDC with a tight dV=1V at regs.

Hope this rough calculation is ok.

I think using a custom transfo is the only way to manage thoses V & A, and avoid at max the issues regarding the heat transfer at regs due to too high voltage drop.

 

You plugs the GRD of each voltage to the caps next to screws ? Interesting. Equipolarity by doing so ? I'll use your method, thanks.

Rgds

I have a multi meter but I haven’t measured the current. 
All the voltages have there one regulator and the 3 regulators get power from the regulator, top left corner, and that regulator gets power from the smps.

 

I am not sure if you can get a 2Vac transformer.