DIY - Summary of observations on non-oversampling dacs

[Kit Taylor]

Last Updated 4/6/05

This projection is on hold indefinitely, as I'm now using a Panasonic SA-XR70 digital amp.

---------------------------

I've spent loads of time buggering about with digital filterless DACs, and reported the results in these DIY room threads: http://pinkfishmedia.net/forum/sear...d=132837&sortby=lastpost&sortorder=descending

Here's a summary of the most useful and closest to objective bits, with lots of more detailed and quite subjective round ups in the posts below.

------------------------------------------------------------------------------------------

- The 3D layout is best, with the TDA1543 glued upside down on top of the CS8412 and eveything earthing to a foil groundplane between the two.

- 1k0 on each CS812 digital out pin

- As SPDIF coupling caps, 1206 NPO 10nF sound sharpest

- One reg per power supply pin.

- The CS8412 digital supply doesn't seem terribly important

- LM317 sound more airy and fluid that LM78XX, which are a bit fuzzy and overthick by comparison

- Bypass the LM317 ADJ pin with Sanyo SEP 330u 16v for more of the same goodness

- Using LM317T or LM78XX, the DAC sounds best with no pre-regulation, more "free" and open.

- Good decoupling caps sound better without bypassing. A 330u Sanyo SEP is my current favorite, and may be as good as it gets. 33nF LCR Polystyrene caps also sound good, louder, rougher and present than the SEPs. Think of the big caps as offering a softer, lower contrast sound, more subtle.

- A smaller decoupling cap is sharper and more defined, whilst the bigger caps the big one does give a more impressionistic and heavy sound that sounds more "alive." To avoid tearing your hair out, settle for one sound or the other and don't bugger about trying to find some magic halfway house

- Filtering is best used only before the regulators, but does improve things quite a bit. Keep things after the regulator as simple as possible, adding inductors only to taste.

- If you do want inductors, the best all rounder the Murata 1210 47uH/1.3R, clean and punchy. Put two before each reg, one at each end of the feed line.

- As loopfilter start with 470R and a Wima polyester 0.47u, this being plain superior to datasheet arrangement. A Sanyo SEP 22u 20v is less bright, adding a bit of grease to the sound and a 330u 16v sounds lush and thick, though both don't sting as hard as the Wima. Don't get too paranoid about what value and type to use.

- The supply voltage to the TDA1543 isn't that critical, provided it matches the individual chip (tolerance -/+15%) and the I/V resistor value. Go for +8v as that seems to be the popular standard, though some have said +6v sounds less harsh.

- The I/V resistor type and value has a large influence on the final sound. Cheap 5% carbon resistors from Farnell are my current reference, but the 1p each types can be just as good for the job as the £2.00 a pop stuff. Everything is worth trying except Welwyn RC55 and the 1206 SMD resistors. Higher wattages sound better.

- Smalll filter caps seem to have more of eveything than big ones. A single Rubycon ZA 220u 35v or Sanyo SEP 330u 16v does the job, the SEP being my pick.

- A laptop external SMPS works very well as a DC source.

 

[Kit Taylor]

A detailed overview over at ZeroGain:
http://zerogain.com/forum/showpost.php?p=110493&postcount=11

Modestly priced nonos DACs are available pre-built from Scott Nixon, Ack! Industries, DIY Fidelity and sometimes on E-bay.

Peter Daniel, whose posts on the diyAudio.com digital forum threads have proved very useful, plans to offer a TDA1543 based DAC soon through AudioSector.com.

47 Labs provide a more expensive but much cooler looking alternative that's well regarded, and squillionaires can go the root of Zanden or Audio Note.

 

[Kit Taylor]

The Nonos Sound

Tom Alves reckons his Scott Nixon DAC sounds better than a Rega P9 turntable: http://pinkfishmedia.net/forum/showthread.php?s=&postid=96766#post96766


Non oversampling DACs offer good sound for little money, and are capable of quite extreme dynamics and bass power. The powerful bass seems the most commented virtue of the nonos approach, but the sound overall is more determined by power supply, grounding and output than the absence of digital filtering. There is a measurable but not audibly obvious roll-off of the extreme treble, but if perceptible at all this manifests as a small fraction of the sound being sliced off rather than an overall muffling effect. This may lend a clearer, less cluttered sound.

Coredump offers his thoughts here: http://pinkfishmedia.net/forum/showthread.php?s=&postid=47668#post47668

It's normal for the DACs to make a high pitched whining noise when not locked on to a digital signal.

The Classic CS8412/TDA1543 Nonos DAC Circuit

A functional nonos DAC requires only a CS8412 s/pdif receiver chip directly connected to a TDA1543 converter. No output stage is needed with passive I/V via a single resistor to ground. The most common kind of DIY nonos DAC, an extremely simple design that's easy (if rather time consuming) and cheap to build and capable of vibrant and highly enjoyable reproduction.

Active I/V via a single opamp allows longer cables to be driven, but consensus is that this either degrades or does not affect sound quality.

General Philosophy of Design

47 Labs have taken a lot of flak for selling a very simple circuit built with inexpensive parts at a high price, but the best sound really does seem to come from keeping the circuit as simple as possible, so long as care is taken not introduce electrical properties that cause components to behave abnormally and perform below spec. When in doubt, use the lowest value component the DAC can still function with.

The CS8412 and particularly the TDA1543 are outdated parts that measure unimpressively, and there is a limit to the performance that can be squeezed out of them. Once a certain level of performance is achieved the rest is flavoring, and there doesn’t seem to be a magic combination of passive components that trumps everything else. Keeping things simple avoid complex interactions and makes tuning to taste easier.

The beauty of DIY nonos is how little fuss is needed to get a basically good sound, and complexity rather defeats the object.

 

[Kit Taylor]

Peter Daniel’s 3D layout is very compact and convenient, and offers the shortest signal and ground paths.

The best board for building on is from Roth Electronik. Sturdy and durable with nice round solder pads. Cheaper board with square pads falls to bits easily.

The DAC sounds cleanest on a hard surface. Mine sits on a pine chopping block on the floor.

 

[Kit Taylor]

Voltage Supply

The TDA1543 supply voltage can be set between +5v and +8.8v.
I use +5v, following a less is more approach. I think this sounds softer, smoother and little grainier than higher voltages.

My original view was that lower voltages sounds somewhat thinner and airier, whilst +8v sounds warmer and bloomier. At least some of this is due to higher volume. However, this may be due to inconsistency between the I/V resistor value actually used and the theoretically ideal value, as tweaking my resitor values has made the sound much cleaner, leaner and sharper.

Passive Resistor I/V Values

UPDATE 20/5/05: Currently using 3 x 1k0 0.5w 1% metal film without problems, on a +5v supply.

The TDA1543 supply voltage determines the necessary I/V and Iref resistor values. Thorsten Loesch explains the calculations in this diyAudio.com post.

With +8v I'm currently finding 2k5 I/V to sound very clean and sharp without sounding sharp. Careful, this will sound too bright in the wrong system.

The DAC functions at +8v with 3k0 I/V and 1k5 Iref. Whilst it clips and compresses it sounds incredibly warm and thick. Also note that a little compression can be helpful, getting more of the impact of the music across at more modest decibel levels. So perhaps try putting the resistor value a little higher than ideal.

Note that there's quite some variation between individual TDA1543, and really you ought to tune the resistor values case by case. See another Thorsten Loesch post on this, this time over at ZeroGain.

I/V Resistor Brands

Ordinary 0.5w 1% metal film resistors have unpredictable results. They can sound harsh and shut in, but their effect is neutral in the current version of the DAC (20/5/05), if slightly smoothed over.

Caddock TF020 are meant to be as good as it gets, but I've not heard them. Rikens come close behind, apparently.

Welwyn RC55 resistors sound rather dark, grainy and harsh. Not recommended. For a less brutal version of this more aggressive sound try SMD resistors from Multicomp, 0.125W 1% 1206.

Audio Note tantalum resistors have a lush, chunky sound without much air.

Vishay bulk foils sound luminous and very detailed, to the point of sounding a bit fuzzy and soft.

Cheap carbon resistors from Farnell.com seem to impose the least sound, leaving you to tweak the DAC elsewhere, and the tiny 0.125W version fits into awkward places. Try them if you like but can't afford Vishay bulk foils.

Two articles with info on the sound of different resistors:
http://www.sasaudiolabs.com/theory10.htm
http://www.borbelyaudio.com/upgrades.asp

Output Coupling Caps

Current favorites are LCR 100nF polystyrene axial caps from Farnell.
These sound similar to SEP 330u 16v. The polystyrenes have a bit less air and sound more rounded and full, aswell as a bit stiffer. With the Sanyo's, the sound change is analogous to turning up the contrast on a TV.

Though film caps are generally thought to be best, when using high quality electrolytics for decoupling try the tha same type and value as coupling cap. Rubycon ZA work well, in spite of steel leads. I use 4.7u 35v. Same value Sanyo OSCON SH sound similar. 1uF LCR polypropylene work very well, broadly lean and agile. 1uF Multicap RTX are very highly regarded.

Connect the positive pole of output coupling caps to the TDA1543. SCSI ribbon cable does not seem to harm the sound when used to hook DAC to cap. Never bypass coupling caps.

Coredump has this to say on coupling caps:

Lucky me I still had a couple of metres of DNM shielded SC cable, which seems to work well in this context. I decided to go for a really good coupling cap and tried Mundorf MCaps (0.47uF, as recommended by the board's designer), but found both the sonic signature and t
he amount of capacity lacking in this context. So I went for MCapZNs, as these are said to have a very low loss and 'quick' and lively sound. I tried 4.7uF and was immediately pleased. They inserted a certain crispness into the sound that made rock music fun again. At the same time the bass gained a bouncy, well-defined character. They fit so well that I cannot be bothered to try other caps now.

 

[Kit Taylor]

Pin 19 squirts out HF noise, so put a 10n NPO 1206 cap across pins 20 and 21. This makes the sound lusher. The type and value of the cap doesn't seem to make any difference. A 4.7u Y5V 1206 does the job well for not much money. Try X7R too, as these are up to 2/3 less microphonic.

Update: Following Peter Daniel I've removed the cap on pin 19, giving a faster sound.

To start with, boost the CS8412s rudimentary reclocking by changing the loopfilter on pin 20 to 470R and 220nF. There's a broad consensus that this is fundamentally better than the datasheet values. Go further following the formula of quadrupling the cap value each time the resistor value is halved.

Before you decide, see also:
http://www.diyaudio.com/forums/showthread.php?postid=525420#post525420

The DAC works with a 330u 16v SEP cap on the loopfilter. It does take a couple of seconds to lock on and you get a blast of white noise first. The tone seems silkier, but the sound overall is chunky and dynamic, with an alarming amount of energy in the low end.

Separate the analog and digital grounds for a drier sound, combine for a warmer, lusher sound.

Don't put an inductor or ferrite bead after the regulator, as it degrades the sound. The digital supply is much less important, but a single Sanyo SEP 22u 20v does sound a bit cleaner than a 1uH/1.3R & 0612 X7R 100nF LC filter.

SPDIF coupling caps do influence the final sound. As well as proving a very compact coupling solution in tandem with a 1206 resistor, 10nF NPO 1206 caps sound sharper than 10nF polypropylene.

If you're using a cheap DVD player as transport, replacing the SMPS caps with the largest Rubycon ZLs that fit adds weight and thickness

Put a 1k0 carbon resistor on pins 11, 12 and 26 for a meatier, more slamming sound. This happens due to something terribly technical to do with earthing and the digital outputs’ intolerance of capacitance. Resistors in the 100R range have little or no audible effect.

See also: http://www.diyaudio.com/forums/showthread.php?s=&threadid=49351&perpage=10&pagenumber=1

 

[Kit Taylor]

Grounding

A good general introduction to grounding can be found here:
http://www.krm.se/dok/Grounding_rym4_Doc.pdf

Connect all the grounds together using a strip of 3M self-adhesive tinned copper tape on top of the CS8412. The adhesive is fairly weak, so a drop of superglue is handy. Trim the edges of the tape to leave a small "finger" that fits over each earth pin. Solder this to the pin for a one-join connection to ground.

Solder capacitors directly onto the groundplane, with the earth lead as short as possible.

SCSI ribbon cable to star earth each individual ground sounds drier and less explosive. Go for the 3M tape.

Bulk DC Source

LC Audio provide an off the shelf solution (well, it needs a box):
http://zerogain.com/forum/showpost.php?p=109063&postcount=46

I use a laptop external SMPS, which contrary to my initial reservations sounds better than a linear wallwart w/schottky diodes, warmer, more open and a little drier. Don't put high capacitance straight after a diode, as apparently it causes the diode to emit high frequency noise.

With the SMPS, a Murata BNX002-01 filter works very well in place of a filter cap. Smaller instruments and less drive, but a much more wide open sound with good timing. The soundstage is notably deeper.

A single, smallish filter cap sounds clearer and has better timing and more slam. A single SEP 330u 16v is superior to 4 x ZA 220u 35v alternated with Murata 47uH/1.3R inductors, sounding less compressed and less hard.

Putting two 33R 1W carbon resistors in parallel between diode and filter cap for extra smoothness without loss of energy.

If you're only using LM78XX or LM317T don't use pre-regulation. These monolithic regs close in the sound.

Don't use a common mode choke as it thins and grainyfies the sound.

Put two Murata 47uH/1.3R before each regulator for a cleaner and more open sound, perhaps at the expense of some tonal weirdness. Avoid complex filters. But if you insist...

For some ideas on passive filtering before regulators investigate the schematics on this diyAudio.com thread. If used at all complex filters are best put before the local regulators, where they can make a surprising amount of differenc, which is not the same thing as improvement. An LCLC filter of 4700uH/20R, Panasonic FC 220u 25v and 1u X7R 1206 can lushen up up the sound, but this is a subjective improvvement only.

An LM317T tracking pre regulator (TPR) from Avondale Audio before the main filter cap can clean the sound and add slam, as well as perhaps some hardness.

Local Power Supplies

Use one LM317T per IC supply pin, and decouple all thee regulator pins. These sound more fluid and open than the more bunged up sounding LM78XX. LM317T need an output bypass cap to reach their datasheet noise specs. 220u capacitance and 0.1 ESR is apparently ideal, the same specs as a Panasonic FC 220u 25v. These caps can sound a bit soggy, astringent and lifeless and as such are unsuitable for use on their own. However the results can be effective if used to compensate for dry, wooly, uptight sounds coming from elsewhere.

Two useful TNT-Audio articles on capacitors for LM317T, Noise and Impedance.

 

[Kit Taylor]

An orthodox post regulator solution seems a series inductor followed by a cap to ground (LC filter). Guido Tent has a popular article outlining this:
http://www.tentlabs.com/Info/Articles/Supply_decoupling.pdf

I prefer a resistor to an inductor, the sound subjectively more natural and the constant resistance more ideologically appealing.

A 1R resistor between reg and decoupling cap adds some of the sound the resistor gives when used for I/V into the mix, and as such is a useful tuning device.

Carbon 0.5W resistor from Farnell are my choice, intimate and organic. Metal film 0.5W resistor sound smoother and cleaner but can ultimately sound shut in, small and harsh, I think depending on the balance of the rest of the DAC.

Current favorite cap is a 1000pF NPO 0805, with no resistor, which seem to have little sound of their own, giving a nice balance of attack and clean tone without getting overbearing. Perhaps a bit lean for some.

 

[Kit Taylor]

There's a useful thread on diyAudio.com discussing alternatives to Black Gate capacitors.

Rubycon ZAs are probably the best "pretty" sounding cap, a cheap and cheerful way of imitating the sound of expensive commercial gear. They have a pale, fluid and luminous "boutique" sort of sound with a strong sense of acoustic space, and plenty of thwack. They seem to get denser the higher the capacitance, so go for 4.7u for a drier, snappier sound, or 220u 35v at the other end. Whilst the cap has a bit of glow to it, it's not obviously "colored." See also: http://www.diyaudio.com/forums/showthread.php?postid=298592#post298592

BC 128 seem to have very little sound of their own, sounding softish, warmish, clear and dryish.

Panasonic FC are very cheap and have clean, slightly astringent sort of sound that some find a bit sterile. Their hairshirtness can sound a bit lifeless, but if you think of them as having a somewhat soggy and metallic sound, when combined with caps that sound too stiff or dry they can produce very pleasant results.

Sanyo OSCON G types have an odd, "fat" sound that's beefy but ultimately artificial. Avoid.

Black Gate FK 220u 25v with NX HiQ on the chip pins sounds punchy and extremely thick and dark. Ultimately suffocating.

Ceramic SMD caps work work well on the TDA1543, such as a triplet 4.7uF X7R 1206 / 220nF X7R 0805 / 100nF X7R 0612. The sound is too weedy without the addtion of a Panasonic FC 220u 25v. This comes from the detailed research posted here: http://www.diyaudio.com/forums/showthread.php?s=&threadid=4562.

A less fiddly solution is a single 1206 X7R 4.7u 16v, which has less finesse than the above combo but makes for loose, lush fun. Note that X7R caps are physically larger than Y5V, but up to 2/3 less microphonic.

LCR 33nF polystyrenes have more air and room sound than the SEPs, and a generally louder, more explicit, high contrast feel. Goinf to the SEPs is like turning doen the contrast and sharpness on a TV, a more subtle and ultimately more satisfying effect. I have to listen harder with the SEPS, whereas the polystyrenes seem to do all the work for me and get a bit overbearing.

Earlier on (<7/5/05) I said:

I've swapped the SEPs for 33nf LCR polystyrenes, a one million percent capacitance difference! The type is called FSCEX or something similar, silver with a blue band at one end. They're meant to be very good.

The sonic difference is supringly modest, but the decoupling LCRs definitely have more warmth, space and room echo.

Taking the SEPs off the ADJ pin of the LM317T loses some of the "air", but also loses a certain tiring metallic quality. The sound is browner without them.

The sound is fuller and does have a certain undamped quality, so maybe a bit of euphonic distortion is going on.

Reading around, 22uF seems a popular decoupling cap value. I compared the 3SEP 30u 16v and 22u 20v thus:

The sound [of the 22u] is less smooth but equally explosive. The timing is more individuated, the instruments don't sound locked together, and phrasing is more defined, as is pitch. The sound is warmer, drier, harder and more forward, and overalll more tightly controlled. I think a few "ringing" artefacts have crept back in, but overall I think these smaller caps offer a profitable tradeoff. http://pinkfishmedia.net/forum/showthread.php?s=&threadid=10809