Buffalo GS20xx - Bypass 12V stage via 3 external PSs

[Exocer]

On 9/1/2020 at 12:05 AM, tgb said:

Hi,

update vs my GS2024 mods :

I installed a MPaudio PS (ALD-HPULN) + its transformer 2*4.2VAC => to power the 3.3 & 1.5VDC.

The Buff is not the same anymore : night-day. 3D, precision, crosstalk...

 

Issue : I wanted to keep things cheap, thus I still powered the 1VDC thru SMPS & buck converter.

In previous tests, the 1VDC did not seem to play any role on the SQ.

Now, with good PS on 3.3 & 1.5 => current from SMPS on the 1VDC pollutes completly the SQ, clear to hear harshness on "notes attacks" & its related harmonics.

Thus, as current is limited to 2.3-2.5A, I will upgrade the 1VDC with another MPAudio PS (ALS-HPULN), combined with a custom transfo 2.1VAC.

 

Not a big deal to fullfil the job on the 1VDC & pay more than expected, cause the mod on the 3.3 & 1.5 changes completely the Buff... very surprising...

More news in 3-4weeks time when I got the custom transfo.

Rgds

@tgb very much looking forward to your updates!

 

I am stil chugging along here with two BS-GS2016P with stock SMPS and I think they sound great. I have some plans to try the LPSU at the input with 1 rail but will probably try what you've done when funds allow.

[tgb]

Hi Exocer

You're, you should try a LPS instead of the stock SMPS. Jump is right there.

You have PoE version. Then, if you move to LPS, I think you'll have to power to the righ pins on the PCB, to power the switching part only, & no power to PoE. Otherwise it is useless & pricey to put a LPS to power the PoE part. Just a guess, as I have a gs2024, without PoE.

 

To be honest regarding the 3 external PS : when it will be ready, the Buff will remains without its hood/top. Then, to make it boot & work, because voltages are low and currents are high, all wires have to be as short as possible (from transfo to Mpaudio reg step AND from MPaudio bloc to PCB). Espacially with a transfo as PS, with a SMPS it is less true. But anyway, few centimeters are ok, >20cm is not. So, you have to find a way to put the mpaudio bloc & transfo close to eachother & close to the PCB... DIY stuff !

I'm curious about the result... Hope this PS based on transfo will work, but without testing I'm still not sure it will boot... It works ok on 1.5 & 3.3, but 1VDC @ 2.5A is a different story.

Will see :-)

Rgds

[Exocer]

18 minutes ago, tgb said:

Hi Exocer

You're, you should try a LPS instead of the stock SMPS. Jump is right there.

You have PoE version. Then, if you move to LPS, I think you'll have to power to the righ pins on the PCB, to power the switching part only, & no power to PoE. Otherwise it is useless & pricey to put a LPS to power the PoE part. Just a guess, as I have a gs2024, without PoE.

 

To be honest regarding the 3 external PS : when it will be ready, the Buff will remains without its hood/top. Then, to make it boot & work, because voltages are low and currents are high, all wires have to be as short as possible (from transfo to Mpaudio reg step AND from MPaudio bloc to PCB). Espacially with a transfo as PS, with a SMPS it is less true. But anyway, few centimeters are ok, >20cm is not. So, you have to find a way to put the mpaudio bloc & transfo close to eachother & close to the PCB... DIY stuff !

I'm curious about the result... Hope this PS based on transfo will work, but without testing I'm still not sure it will boot... It works ok on 1.5 & 3.3, but 1VDC @ 2.5A is a different story.

Will see 🙂

Rgds

Here are the internals.

 

I circled the 12V input and the power input for the POE board.

 

Planning to switch to an LPS and remove the SMPS in the next month or two. 

 

I'm hoping I can simply remove the POE board on top and power it exactly the same way one would power the non-poe version. Will try and report back...It was virtually impossible for me to locate a non-POE version.

 

See innards below:

[tgb]

Waouh ! Nice picture ! The 1st of a PoE version I think.

Very different design due to the PoE !?

 

Just based on your pic, the PoE seems to be a simple add-on feature to a GS2016.

If you disconnect the 2 connectors from the PoE board (with the sticker "1G" on the wires), does the switch works perfectly ?

If it works, then you can simply take off the PoE board, the internal PS & simply mod it as a standard GS2016... I hope it can boot & work without PoE...

[Soul Analogue]

On 9/6/2020 at 6:25 AM, Exocer said:

Here are the internals.

 

I circled the 12V input and the power input for the POE board.

 

Planning to switch to an LPS and remove the SMPS in the next month or two. 

 

I'm hoping I can simply remove the POE board on top and power it exactly the same way one would power the non-poe version. Will try and report back...It was virtually impossible for me to locate a non-POE version.

 

See innards below:

Yes.. just take off the POE board and it works exactly as non-poe version

[Exocer]

21 minutes ago, Soul Analogue said:

Yes.. just take off the POE board and it works exactly as non-poe version

Thanks, you're right. I have already validated this on my end.

[zacho]

Has anyone tried modding the BS-GS2008 in any way?

[RickyV]

On 9/1/2020 at 6:05 AM, tgb said:

Thus, as current is limited to 2.3-2.5A, I will upgrade the 1VDC with another MPAudio PS (ALS-HPULN), combined with a custom transfo 2.1VAC.

 

Not a big deal to fullfil the job on the 1VDC & pay more than expected, cause the mod on the 3.3 & 1.5 changes completely the Buff... very surprising...

 

Hi @tgb how is your 1V MP-audio and transformer comming along?

[tgb]

Hi RickyV,

it's coming ! Still some days to wait for testing ! :-)

A bit long that mod but :

- according to my "basic" calculation (VDC output = VAC input*1.41 -0.8), I needed a 2.1VAC transfo => but Micheal from MPaudio told me that at : as we have very low voltage & high intensity, the best is to have a transfo that deliver a higher voltage that needed. This extra-voltage will be "lost" somehow with 2 stuffs : the length of wires (at very low voltage, you loose voltage vs wire length), and as the transfomers are "slow" (by design) & we have high intensity, it may deliver a bit less than th expected voltage (2nd steps leading to a lower voltage than expected).

So : he ran some tests => conclusion was => a transformer with 2.7VAC output would be safe.

 

Then, I asked for a module ALS-HPULN (to deal with the 3.5A), without heat sink => the point is that the heat sink of standard ALS is too small to deal with the watts to dissipate (I found a nice one).

Problem... on Monday I received the parcel with modules & transfos... but the modules had their standard heat sink tapped ! :-) (mistake from my side... I changed so many times this order...) => so => today I'll post back to MPa the 2 nice ALS with heat sink & just have to wait for 2 new ALS without heat sink.

Let's wait for 1 or 2weeks (max !)...

[OAudio]

On 7/24/2020 at 7:50 AM, Soul Analogue said:

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...

Hi Soul Analogue,

 

Appologies in advance if you are aware of DMM current range "burden voltages" and have already taken this into account.

 

Is the 0.4v drop you are seeing is just the "burden voltage" for the 6ish Amp current and the meters 10amp measurement range on your Fluke ?

 

There will be specs for the meter that explain how to calculate the voltage drop across the DMM when measuring current in a current measurement range. A 0.4v drop for a 6 or 7 amp current is in the right area for burden voltage on many meters.

 

If it were the normal drop across the Fluke and you have a second meter to to hand to measure the output voltage being applied to the "1v" rail, you should be able to wind up the voltage on your 10 amp bench supply to get to the desired 1v rail voltage. The meter measureing current will read correct current value but the voltage drop across the meter will have been compensated for.

 

Here is an example form fluke for a mA range measurement. You may need to look up the sense resistor size for your 10A range.

 

https://www.fluke.com/en-us/learn/blog/electrical/can-you-live-with-the-burden

 

Hope this might be helpful.

[Soul Analogue]

10 hours ago, OAudio said:

Hi Soul Analogue,

 

Appologies in advance if you are aware of DMM current range "burden voltages" and have already taken this into account.

 

Is the 0.4v drop you are seeing is just the "burden voltage" for the 6ish Amp current and the meters 10amp measurement range on your Fluke ?

 

There will be specs for the meter that explain how to calculate the voltage drop across the DMM when measuring current in a current measurement range. A 0.4v drop for a 6 or 7 amp current is in the right area for burden voltage on many meters.

 

If it were the normal drop across the Fluke and you have a second meter to to hand to measure the output voltage being applied to the "1v" rail, you should be able to wind up the voltage on your 10 amp bench supply to get to the desired 1v rail voltage. The meter measureing current will read correct current value but the voltage drop across the meter will have been compensated for.

 

Here is an example form fluke for a mA range measurement. You may need to look up the sense resistor size for your 10A range.

 

https://www.fluke.com/en-us/learn/blog/electrical/can-you-live-with-the-burden

 

Hope this might be helpful.

 

It was not the burden voltage issue... its truly was the limitation of the variable LAB PSU 

I already changed the method of measuring the current on the rails

I built a regulator...and placed a resistor in series between the regulator and the PSU...the current levels of the rails were calculated by the voltage drop across the resistor

 

[OAudio]

1 hour ago, Soul Analogue said:

 

It was not the burden voltage issue... its truly was the limitation of the variable LAB PSU 

I already changed the method of measuring the current on the rails

I built a regulator...and placed a resistor in series between the regulator and the PSU...the current levels of the rails were calculated by the voltage drop across the resistor

 

I figured you might have the burden voltage ruled but thought to mention it just in case 🙄. Decent current at a volt is tough with a linear solution, great that you have it solved.

[Soul Analogue]

Finally I have tamed this freaking tricky bastard with full linear psu and regulators to a smooth power on and working state

pictures worth more than a thousand words

[RickyV]

1 hour ago, Soul Analogue said:

Finally I have tamed this freaking tricky bastard with full linear psu and regulators to a smooth power on and working state

pictures worth more than a thousand words

wow Soul it has become an elaborate switch.

[tgb]

Great SoulAnalogue !

All in a custom box.... nice work !

 

Funny : I took some time yesteday too, to put in place the MPaudio modules for full linear on mine (GS2024).

I have no patience to design a custom box... I'll post photo (different "style" ahead ! )

 

SoulAnalogue, what's its weight ?

Mine is 3kg (standard) => I took off the hood & the internal PS to make it lighter => but now it weights 6.2kg... strange... :-))))

[tgb]

On 11/28/2020 at 11:16 AM, Soul Analogue said:

Finally I have tamed this freaking tricky bastard with full linear psu and regulators to a smooth power on and working state

pictures worth more than a thousand words

Hi SoulAnalogue,

what's the job of the 2 wires that slide below the heat sink (on left side) ?

Is it a specific external PS for the clock ?

 

I didn't think about this mod... but pretty smart.

How SQ improved thanks to this PS on clock ?

Rgds

[tgb]

Hi,

please find here my own 3ple LPS on Buffalo (gs2024).

By far less smart than SoulAnalogue ;-) because based on dedicated "basic"  (transfo+MPaudio module) per each voltage

=>

Profile view =>

 

I posted a few minutes ago this post about this mod => http://forum-hifi.fr/thread-16709-post-430796.html#pid430796

Hope Google Translate can manage....

 

I stated in that post that thru netword & Buffalo, SQ was behind thru USB optical cable (& USB>spdif interface)

=> I must admit that after some 10n minutes, SQ thru the Buff improved...

I'll let it run & see @ d+1  if it changes

Rgds

[RickyV]

On 11/28/2020 at 11:16 AM, Soul Analogue said:

Finally I have tamed this freaking tricky bastard with full linear psu and regulators to a smooth power on and working state

pictures worth more than a thousand words

Hi soul, what clock are you using and on what voltage is it? Is it possible to use a crystek cchd-575?

[tgb]

Hi,

so, FIY, my lovely Buffalo in photo above is out of the hifi setup, all dismantled, back to its box by tomorrow morning :-)

 

Reason is : I ran a test this eve, after burning the Mpaudio ALS (the one to power the 1VDC). The rest (ALD module & Buffalo were burnt for a long time).

Test result = my "new" setup based on USB optical cable+Singxer SU-1 beats this Buffalo (with triple LPS). See signature vs current setup & this thread regarding the USB optical cable => https://audiophilestyle.com/forums/topic/60825-usb-optical-cable-stunning-if-powered-implemented-correctly/#comments

 

 

 

To be clear, I use a digital amp, a rather good one : Classé 2200i.

This amp has a built-in digital interface, for USB / network etc...

So, when I use the Buffalo, I have this setup :

server on PC (Minimserver with embedded convolution at 1e6taps) =>

=> FMC (fiber converter)

=> fiber to switch TP-Link SG5412F (this switch has only SFP ports & is modifed with : LPS on 3.3V & 1.8VDC, and SMPS on 1VDC)

=> fiber to Buffalo

=> I enter the Buff thru a SPF port

=> I get out of the Buff thru RJ45

=> to the RJ45 input of the Classé

FIY : the TP-link switch is the "core" of the optical network, all SPFs & fibers are single-mode.

 

The test was Buff vs :

basic flac file on PC => played by MP-HC => convolution at 16e3taps thru EqualizerAPO => 

=> Corning 3.Optical (MPaudio 5VDC PS <= key is there : very good current !!!)

=> Singxer SU-1 (modded)

=> coax output to Classé

 

So, test regarding SQ between these 2 channels :

1. vs 3D depth / with scene : quite equal

2. vs crosstalk & stage precision : better thru Corning-SU1 ; Buff seems to be "confused" in comparison (espacially vs stage precision)

3. ... the big diff in favour of Corning-SU1 is : analogue sound. It's "fluid", pleasant to hear (no complaint from my tricky hears!), sounds "natural". In comparison, the Buff sounds "digital".

 

The Buff, with 3ple LPS is very very good in his own. This mod woths it. No complaint about the Buff.

By the other setup is nicier to listen.

 

Then, I could go futher with the Buff :

- external clock (although the 3ple LPS helps a lot)

- not to use the internal network interface of the Classé but add a Ered-dock to enter the amp thru spdif

2 options that are too expensive vs the potential improvement (just a guess of course) => reason why I stop there my funny experience with e Buffalo & will try to improve a bit my setup based on the "crasy" USB optical cable.

 

The overall reason I'l go further with Corning-SU1 is that the setup ergonomics is so cool... No complaint vs SQ & cool/easy/no buggy setup : what else... :-)

Rgds

 

PS : thanks to Micheal (MPaudio) to have created these tiny/fancy/smart modules. Because although they are top-notch, when you change your setup you can recycle them thanks to the tiny trimmer !  :-)

[Soul Analogue]

2 hours ago, RickyV said:

Hi soul, what clock are you using and on what voltage is it? Is it possible to use a crystek cchd-575?

It is a CW ocxo 25mhz...make sure it is lvcmos output 

I havent seen a cchd575 with that frequency...but if you can find one.. i believe it will work

 

Cheers

[Soul Analogue]

18 minutes ago, tgb said:

Hi,

so, FIY, my lovely Buffalo in photo above is out of the hifi setup, all dismantled, back to its box by tomorrow morning :-)

 

Reason is : I ran a test this eve, after burning the Mpaudio ALS (the one to power the 1VDC). The rest (ALD module & Buffalo were burnt for a long time).

Test result = my "new" setup based on USB optical cable+Singxer SU-1 beats this Buffalo (with triple LPS). See signature vs current setup & this thread regarding the USB optical cable => https://audiophilestyle.com/forums/topic/60825-usb-optical-cable-stunning-if-powered-implemented-correctly/#comments

 

 

 

To be clear, I use a digital amp, a rather good one : Classé 2200i.

This amp has a built-in digital interface, for USB / network etc...

So, when I use the Buffalo, I have this setup :

server on PC (Minimserver with embedded convolution at 1e6taps) =>

=> FMC (fiber converter)

=> fiber to switch TP-Link SG5412F (this switch has only SFP ports & is modifed with : LPS on 3.3V & 1.8VDC, and SMPS on 1VDC)

=> fiber to Buffalo

=> I enter the Buff thru a SPF port

=> I get out of the Buff thru RJ45

=> to the RJ45 input of the Classé

FIY : the TP-link switch is the "core" of the optical network, all SPFs & fibers are single-mode.

 

The test was Buff vs :

basic flac file on PC => played by MP-HC => convolution at 16e3taps thru EqualizerAPO => 

=> Corning 3.Optical (MPaudio 5VDC PS <= key is there : very good current !!!)

=> Singxer SU-1 (modded)

=> coax output to Classé

 

So, test regarding SQ between these 2 channels :

1. vs 3D depth / with scene : quite equal

2. vs crosstalk & stage precision : better thru Corning-SU1 ; Buff seems to be "confused" in comparison (espacially vs stage precision)

3. ... the big diff in favour of Corning-SU1 is : analogue sound. It's "fluid", pleasant to hear (no complaint from my tricky hears!), sounds "natural". In comparison, the Buff sounds "digital".

 

The Buff, with 3ple LPS is very very good in his own. This mod woths it. No complaint about the Buff.

By the other setup is nicier to listen.

 

Then, I could go futher with the Buff :

- external clock (although the 3ple LPS helps a lot)

- not to use the internal network interface of the Classé but add a Ered-dock to enter the amp thru spdif

2 options that are too expensive vs the potential improvement (just a guess of course) => reason why I stop there my funny experience with e Buffalo & will try to improve a bit my setup based on the "crasy" USB optical cable.

 

The overall reason I'l go further with Corning-SU1 is that the setup ergonomics is so cool... No complaint vs SQ & cool/easy/no buggy setup : what else... :-)

Rgds

 

PS : thanks to Micheal (MPaudio) to have created these tiny/fancy/smart modules. Because although they are top-notch, when you change your setup you can recycle them thanks to the tiny trimmer !  :-)

Glad that you found a nicer option

But in fact...rather than how you split the rails.... The sound quality is directly the outcome of the quality of the power supply

In my dictionary and experience, monothilic regulators suck in terms of musicality and they suck even more to parallel several devices to achieve the current output capability 😅

 

[tgb]

Hi SoulAnalogue,

"split the rails" ?

no...

what did you see ?

I splitted no rails.

In the Buffalo tweak : 1 transfo > 1 AC/DC converter, per each voltage (1 - 1.5 - 3.3)

Tweaks vs transfo & AC/DC converter on the Corning & SU-1 are dedicated to. Thus no "split"

Rgds

[Soul Analogue]

27 minutes ago, tgb said:

Hi SoulAnalogue,

"split the rails" ?

no...

what did you see ?

I splitted no rails.

In the Buffalo tweak : 1 transfo > 1 AC/DC converter, per each voltage (1 - 1.5 - 3.3)

Tweaks vs transfo & AC/DC converter on the Corning & SU-1 are dedicated to. Thus no "split"

Rgds

I meant the way you supplied 1v, 1.5v and 3.3v, instead of the single 12v rail

 

What i wanted to say is that the paralleling LT3045 does not do well for audio

[tgb]

1 minute ago, Soul Analogue said:

What i wanted to say is that the paralleling LT3045 does not do well for audio

I don't see your point...

The trick of 3ple LPS is basically to bypass the internal regs that do 12>3.3 12>1.5 > 12>1VDC

When you feed the PCB with  1/ 1.5 /3.3VDC ahead of these reg step and if if cut the coils (linked to its steps) then, each voltage zone is independant... thus no brainer to feed there.

I don't "parallel" LT3045 (each are more less like a "ray vs star").

"paralleling LT3045 does not do well for audio"  :  I don't see your point... for "audio" !? what's the consequence on SQ ? 

Rgds

[Soul Analogue]

3 minutes ago, tgb said:

I don't see your point...

The trick of 3ple LPS is basically to bypass the internal regs that do 12>3.3 12>1.5 > 12>1VDC

When you feed the PCB with  1/ 1.5 /3.3VDC ahead of these reg step and if if cut the coils (linked to its steps) then, each voltage zone is independant... thus no brainer to feed there.

I don't "parallel" LT3045 (each are more less like a "ray vs star").

"paralleling LT3045 does not do well for audio"  :  I don't see your point... for "audio" !? what's the consequence on SQ ? 

Rgds

Mp audio modules parallel several 3045 to achieve the current output capability... 

The consequences of paralleling is the damage to frequency coherence, bandwidth and impedance, these are essential for the flow of music 

 

Nevermind as you have found your nicer option instead of this buffalo...just enjoy 😉