Buffalo Dac build

steve2701 · « **on:** 18 Dec 2010, 03:48 pm »

Hello to all.
I contacted Paul some time ago and requested some details on modules for a DIY dac build - the 'Twisted Pear Buffalo II'.
I also said that I would be happy to post the build process on here and keep a type of 'blog account' of how it all goes.
This is a long term build - it is simply to expensive for me to buy everything at once and put it together, but I am very much looking forward to it.
I am looking to do this to the best of my ability (somewhat limited electronics wise) and get the best I can out of the components without going totally overboard on modding the individual boards supplied from anyone. I know there will be some - but it will be limited.
This will end up as either a two or possibly three box design.
Crazy possibly, but the intention is to build it dual mono - fully balanced, with totally seperate power supplies for both analogue and digital (= lots of transformers) and probably keep the dac and output stages in their own seperate 'quiet room' which will also probably be shielded.

I have today been one of the lucky folk to be online and get a couple of the dac boards from TP - so they should ship at the end of January. Until then I can take serious notes of everything else I will be needing - casework, transformers, power supplies output stages, connectors and cables etc - along with the fine detail of front panels, rear panel inputs and outputs - all a bit daunting really.
I know I am going to need help - so please feel free to offer advice if you see me heading for major foopar.
I am looking to input data from my iMac via USB (hopefully on the upcoming asynch usb input also from TP) & have the dac output in balanced AES/EBU for input into my Simaudio Moon system. It might also be fun to have inputs available from a couple of sources like a CD player and DVD that have digital outs - but that decision can wait for a week or so - as complex switching could end up confusing.

OK - thats enough for a first post here, and I'm looking at this being a fairly long haul.

steve2701 · « **Reply #1 on:** 21 Dec 2010, 07:55 pm »

I had an interesting chat with Paul today, and some good suggestions were made as to how I might proceed, and where I would like to end up.
Followed that with another interesting conversation with a transformer manufacturer who may be able to make what I am looking for, and also hopefully help with the actual requirements I may need, like will single output transformers give a performance boost to above those having dual etc. This could lead to a lot of transformers being needed, but if it's worth the effort - I will be visiting it for certain.
This is going to keep me occupied for ages - that's a certainty.

steve2701 · « **Reply #2 on:** 1 Jan 2011, 04:35 pm »

I finally got round to ordering another couple of bits today. Two 'Legato' full kits and two additional boards.
My intention is to build the two 'kits' totally stock, and the two boards will be populated with what I want them to be - ie, somewhat 'uprated' components.
The kits will allow two things - 1) a working dac fairly quickly & 2) a reasonable comparator when the other boards are finally completed.
Two more things got ordered as I didn't want to lose them - LME49720 'metal can' op-amps that will replace (the same but non can) LM4562.
Feels like something is happening at last, not certain when the Legatos will get here but the Dac is due to be shipped at the end of the month (Jan).
The shopping list for the bare boards is already looking rather expensive, z foils, tants, Mundorf silver/gold in oil, os cons, tx2575 etc, but it is an interesting list - I just hope that it returns a rather nice sounding signal.

The designs for the PSU's are coming together, along with the thoughts of Paul on just how these are to be implemented / wired so that every single ground plane is kept seperate and silent.

NicMac · « **Reply #3 on:** 1 Jan 2011, 08:25 pm »

Where do you hope/think "boutique" components will be of most benefit on the Legato's?
Thanks,
Nic

steve2701 · « **Reply #4 on:** 1 Jan 2011, 09:32 pm »

Hello Nic,
A very good question and one that has had me reading literally thousands of replies on hundreds of threads about this stage.
My single aim is to get a signal that is as un-polluted as possible , ie, I want no colourations whatsoever if I can possibly get that far.
From that standpoint virtually every component has some importance to it - so getting them all as close to possible to not altering the signal is the aim.
If there is a component that is better than another at this and it happens to be cheaper - fantastic - it will be used. Unfortunately this rarely seems to be the case.
For those that know the circuit my current (and may well yet change) if incomplete list changes the op amp to the metal can version, its associated 100nf caps to mundorf ag/au in oil, the 4 15nf output caps to mundorf ag/au oil probably 5nf or lower, the 1w resistors to be z foils / tx2575.
I am still studying if other caps need changing - those chosen already will need some imagination in their fixing to the board due to physical size. The actual power supplies via Paul may well impact on some of the values already chosen - - I dont know yet. As I said at the start - this is a long term progect so I have s fair amount of learning to do yet for certain.
I am also certain of another thing - the actual supply of power to these modules is going to be very special indeed - but more on that later. That will probably have a far greater effect than a great many of these component swaps, but will hopefully make them even more necessary.

steve2701 · « **Reply #5 on:** 21 Jan 2011, 12:23 pm »

At last,
Something arrives and I can at least start to do something physical.
Two complete Legato kits, and two spare boards for messing with. I'll post some pics later of what the full kit is.
Parcel force charged £13.50 to take the £8.50 Vat off me for the UK government... (Clearance fee they call it)
Well packaged and looks totally professional I must admit.

Paul Hynes · « **Reply #6 on:** 22 Jan 2011, 01:01 pm »

Hello Steve,

Here is the drawing you requested during our telephone discussions.

http://i700.photobucket.com/albums/ww7/paulhynes/PR3PPRAP.jpg

For those who are not up to speed with the ideal power supply arrangement I have included a brief explanation below.

There is a tendency nowadays to power all sub circuit sections of a piece of audio equipment from the same master power supply to save production costs. This automatically produces voltage errors in the ground system because all the ground return currents for all the sub circuits return to the same ground on the master power supply, often mixed along the way by using common ground tracks or planes, which exasperates the interaction between the sub sections using the ground system as a ground reference. This is problematic when using series regulation for the various sub circuit sections, with varying load current draw, as it causes signal related AC and DC voltage shifts in the power supply grounds. Voltage shifts also happen when shunt regulation is used to regulate the sub sections as the constant load current these regulators present to the power supply introduce a DC error that is relative to the actual DC resistance of the ground return. Any sub circuit section that requires a tight voltage reference will have this DC error added to its voltage reference.

The PR3PPRAP (PR3 pre and post regulator application) shows the ideal connection system for one sub circuit section load. Using two regulators in series (Pre and Post regulation) provides a double layer of mains power supply line interference rejection making the equipment less prone to performance variation when the mains power is dirty during high grid usage.

One PR3PPRAP is used for each sub section of the equipment circuitry. Circuit sections requiring a negative voltage rail would use NR3PPRAP constructed using negative voltage regulators. The 0V connection from the post regulator to the load should be connected to the ground terminal of any local load decoupling used in the subsection to be powered. The post regulator +VE connection should go to the positive terminal of the local decoupling. This keeps the ground return currents of each section out of the main earth reference system on the board allowing a very clean earth reference for the whole system. This makes a big difference to the resolution of dynamic scale and micro detail presentation. Timbre and tempo are improved and there is a reduction in confusion in the soundstage particularly when the music gets busy. Music just sounds more like real people playing music.

Setting up the power supply properly where the equipment to be powered has many subsections can be rather expensive and time consuming. Few will venture along this path. Steve is one of the few.

Regards
Paul

steve2701 · « **Reply #7 on:** 22 Jan 2011, 09:32 pm »

Thanks for that Paul, I'll be printing that off on Monday and studying it.
I am so looking forward to travelling this path - the final destination is very tempting.
A couple of pics just for starters of the kit & how it arrives (Legato) and my first basic layout board. The idea for me is to build the two full kits as completely stock, then start the two bare PCB's as the best I possibly can with no holds barred components where I can.
As outlined by Paul above this is probably going to be a fairly monumental build for me, especially as I have not done anything like this for a while.

I promise the pics will get better, my garage was freezing and I need to get some brighter lights so I can get the lens to work properly.

steve2701 · « **Reply #8 on:** 25 Jan 2011, 07:57 pm »

Well it's been 35 yrs since I wielded a soldering iron to do work like this - so this board is going to teach me a fair bit. Some close up photos show that a little more solder was needed in some cases - but that can soon be put right. A bit of flux cleaning may be in order as well I guess. Close inspection showed no dry joints or crossed traces - so I'm happy for now.
A couple of hours this afternoon saw all of the resistors in place and a couple of caps.
Tomorrow , if I can grab an hour or so will see me attempting the first multipin opamp ever, they look a bit close together do those legs. I need the eyes I had thirty years ago..

This shows where this board will sit, the next will be to the right. Hopefully the clear l/h side and front will be enough for a couple of psu's.

tull skull · « **Reply #9 on:** 25 Jan 2011, 09:36 pm »

Thank you for including us in your project Steve. I look forward to your sharing of your time and experience. Much appreciated!

steve2701 · « **Reply #10 on:** 26 Jan 2011, 07:47 pm »

Those op-amp legs are just a bit small, but a small bit in the iron and my strongest glasses - all went to plan, I think.

Time allowed me to not only complete the first board but alos the second as well. Mind you, a bit of practice soon gets speed up.

Next is to put a couple of connectors on a back panel for signals in & out.
Then comes the interesting bit - power for the things.
I intend to get power to these and the dacs with modules that can then be used with the final build.
Hopefully it becomes a bit more interesting from here on.

steve2701 · « **Reply #11 on:** 2 Feb 2011, 05:24 pm »

I managed to spend a couple of hours over the weekend knocking up a basic back panel for input / output. A pair each of RCA's & XLR's for output (I usually run fully balanced) and a 75 ohm BNC for spdif input duty.
Even trying very hard I managed to reverse one set of xlr wires, but spotted that when staring at them through the camera lens...
I have the basic list of PSU's to get this running, and the dac boards are due for despatch now.

NicMac · « **Reply #12 on:** 12 Feb 2011, 09:30 pm »

Hi Steve,
I'm working on a very similar project - also long term. If you don't mind I could post my progress in this thread and we could share experiences. For tasters I have attached a photo of one channel of the dual mono build. The other channel is on the listening and measuring test-bench.
Cheers,
Nic

steve2701 · « **Reply #13 on:** 15 Feb 2011, 04:31 pm »

Hello Nic,
I have no worries about you posting on this thread, as you say, it could be very useful.
My dac boards finally arrived late last week (no duty this time - go figure!!) I'll post a couple of pics up shortly of those in their 'bare' format.
What supplies are you using from Paul - the ones listed above? I see your way of getting them onto the board - I like that.
I'm just saving my pennies for the initail build - and getting very itchy fingers in the process.
Are you going to just supply the dac with SPDIF or are you considering USB as well?
I'm guessing that your not using Legato boards? They appear to have changed resistors & resistor makes not so long ago - this is something I intend to seriously look at in the final build.
I also need to be thinking about the actual voltage of the balanced output - my pre 'expects' to see 4v, not the 2v that these put out as std. I don't think this will make a jot of difference when all is said and done, as it just means less attenuation, but it is all interesting to me.

Intersting looking picture there - I just hope I can get to the same level soon.

Cheers,
Steve.

NicMac · « **Reply #14 on:** 15 Feb 2011, 10:35 pm »

Hi Steve,

For the Buffalo I'm using:1x S17LN-1v2 (DAC), 2x Z17LN-3v5 (AVCC, I think set for 20 mA) and 2x Z1703v3 (DAC and XO, set for 120 mA). All these are powered by a single PR3G2-5v5. I think some use Z1A-5v5 for this job but with my (limited) understanding I fail to see any good reason to power a constant load (like shunt regs) with another shunt reg. I might also try a 1v2 shunt reg in place of the current S17LN-1v2, maybe a Trident from twisted pear.

For the I/V-conversion I'm testing/comparing IVY III and Legato 2. At the moment I actually prefer the IVY III but this is maybe because I have been listening to this for much longer and have gotten used to it. Anyway, both are powered in the same way with 3 set of S17/37LN15 (2 sets when not using the onboard Bal/SE stage which is mostly the case for me). I'm using balanced loads; UcD based DIY power amp or head-phones (when family imposes it). My loads do not tolerate DC-offset well and as a fully DC-coupled discrete amplification chain is my main design goal I have added two twisted pear Ventus modules that perform the Bal/SE conversion and allow me to adjust down DC-offset when I bypass the output caps on Legato. I'm not using a pre-amp and do attenuation in the digital domain (virtually a free lunch with a 32-bit dac). For the Ventus stages I'm also using sets of S17/37LN15 for power. All 4-5 sets of S17/37LN15 are powered by a set of P/NR2G3-20. I might try a shunt reg for this purpose but the current draw is significant and I already will have quite a bit of heat generated in the yet-to-be-build cabinet. If Paul H would have a Z1A-20 with on-board rectification and where I could easily tune the shunted current down to a reasonable level I might very well try it. I like the idea mating shunt and series regulators in dual regulation circuits as it simply makes sense to me. The DAC will have a single high-quality custom made transformer (for both channels) with a single primary and 4x 20V + 2x 9V secondaries. This may be a weak point but I doubt it with two levels of regulation at all points. I would like to try Z17/3715 shunts in place of the current S17/37LN-15 series regulators but I believe there may be prohibitive heat dissipation issues. I have measure the current draw on all rails and if interested I could list the.

My IVY III and Legato 2 are only moderately tweaked as the components provided by twisted pear are already of very good quality. Also, improving power supplies in my experience alway improve more that resistor swapping and should therefore be done first. I can post some details/photos if you are interested. The gain can be adjusted on both Legato and Ventus. With the standard configuration I will have 2.4V (with Legato) and 4V with IVY III - I believe.
I'm using the Buffalo with S/PDIF (mostly a SB Touch, 24/96 capable) or I2S sources. Of the latter I like very much the 32/192 capable UC192 USB2->I2S "transport" from Kingrex, but will also soon test a jkeny modded HiFace.

My final goal is actually a three cabinet solution; a "pre-DAC" (i.e. a digital input selector and DAC controller), the DAC and the power amp. The umbilical cord between the pre-DAC and the DAC will be a "HDMI" interface with differential transmission and galvanic isolation of I2S, DSD or S/PDIF. This interface also carry I2C so the Buffalo can be controlled (attenuation etc.) from the pre-DAC.

Cheers,
Nic

needsp · « **Reply #15 on:** 16 Feb 2011, 08:34 am »

Hi Nic

I wouldn't necessarily assume that the shunt regs will give the better result in a particular application- try them and see! For example Leo told me a while back that he thought he preferred S17LN regs for AVCC duties over Z17. This was only a preliminary observation, and he may have changed his mind. But if so, it might be because the S17LN regs are even lower noise that the shunts- Paul H could confirm?

Have you listened to your dual mono setup next to the usual single Buffalo approach? Is all that extra complexity really leading to a better sound? I commend you for trying it out- but a lot of us would find an honest assessment of its worth very helpful!

Re the transformer. Are you at least using a separate tap- L and R- for each individually regulated stage? I ask, as using the dual mono configuration gives you the opportunity to keep ALL audio grounds separate from power grounds- which- as I'm sure he will confirm- Paul H has found to be extremely beneficial to sound quality in other circuits

I’ve made a number of modifications to my Legato, which- I think- are highly beneficial to sound quality. I have a second board which I’m going to modify back to exactly the original configuration. Then, when I’m feeling strong (!) I’ll reinstall it and try and give an honest assessment of the differences. It’s important, if sometimes painful to do this, IMO. What seemed a series of incremental improvements at the time sometimes add up to very little!

Best wishes

Paul (N!)

steve2701 · « **Reply #16 on:** 16 Feb 2011, 07:38 pm »

As promised some pics of what you actually get with the Buffalo II dac.
They arrive very well packaged in a folding strong cardboard box – wrapped in anti static bags.

The main dac board, AVCC module and mounting hardware.

The AVCC module.

Main DAC board

The object of most of the forthcoming hard work:-

That dac chip is just on 9.5mm square - I would hate to try re soldering that thing.

needsp · « **Reply #17 on:** 16 Feb 2011, 08:41 pm »

Hi Steve

I wrecked my first Buffalo (32S, predecessor of the BII), and did unsolder the DAC chip and fit a replacement- successfully! Bloody tricky though. No specialist gear- just a fine tipped soldering iron, fine solder, and a jeweller's eye glass fixed to some reading glasses by a rubber band!
Best wishes

Paul

Paul Hynes · « **Reply #18 on:** 16 Feb 2011, 09:55 pm »

Hi Nic, Hi Paul,

A few answers to some questions.

The S17LN reg noise is around 4 times lower than the shunt reg. The error amp device has a wideband noise spec of 0.5 nanovolts root Hz with a noise corner of 1 Hz. The voltage reference is a low noise type and it is well filtered to reduce its noise contribution.

Keeping the ground return currents for each individual stage out of the signal ground reference/ground plane is beneficial even with shunt regulators. This is because all regulated power supplies have bandwidth limits (even mine although somewhat higher than usual). Electronically derived constant current sources that drive typical shunt regulators may have relatively high impedance at low frequencies but there will be a point where, if feedback is used, the feedback drops, reducing the current source impedance, or semiconductor junction capacitance can also reduce the impedance. This allows higher frequency interference to break through to the shunt regulator circuitry and the constant current draw is no longer constant with this additional interference current flow circulating around the system.

The amplifiers typically used for the shunt element of the regulators have high gain at low frequencies but this gain will start to drop once the high frequency compensation kicks in. This drop can begin at surprisingly low frequencies. I see many regulators run out of loop gain at around 1 to 10 MHz. This means they have stopped regulating at this point and might as well not be there. As there is no effective regulation above these frequencies, high-speed signals propagating through the system cause all sorts of transient current disturbances that pass through the local decoupling capacitors and into the ground return system causing intermodulation problems.

So you might have assumed that your shunt regulator will draw a constant current from the supply but this is not true at high frequencies. It stands to reason that any attempt to provide regulation at these frequencies will require extraordinary operating bandwidth, not to mention speed, to have any benefit especially when the digital circuits like the ESS Sabre chips are clocking at 80 to 100 MHz.

The last thing you want is all this interference running amok in the signal ground reference/plane. If you use a common supply for all the regulated signal processing stages, the interference currents have no choice but to mingle, causing inter-modulation between all the stages. Providing a separate galvanicaly isolated supply for each signal stage patched into the circuit right on top of the load point provides a tight return loop for each section’s circulating supply currents that does not include the signal ground reference/plane in it’s path. Inter-modulation between stages is reduced significantly with less disruption of voltage references and reduced power supply induced jitter. The results are easy to hear.

Regards
Paul

NicMac · « **Reply #19 on:** 16 Feb 2011, 11:21 pm »

Hi Paul,
To be specific, If I'm having 2-5 sets of bipolar power supplies composed of S17/37LN-15 regulators connected to a circuit with a common ground do you recommend a single ground connection (to the pre-regulator) or 2-5 ground connections back to the pre-reg (effectively generating 2-5 ground loops!)? Advice very welcome.
Cheers,
Nic