Yet Another - I want To Make a DAC

[Jmertz]

Hello everyone. As I am sure many of you have done. I also have been reading tons and tons of posts on building a DIY DAC. However, I have been frustrated with the shear number and differing opinions among the DIY audio community.

This thread is my own selfish attempt to help "weed" through the growing forest of DAC threads. To hopefully, "see the forest from the trees". I.e. the big picture. What really matters most. Most "bang for your buck" so to speak.

It is my hope that I will come up with a list of clear and somewhat agreed upon techniques for making a successful (both sonically and measured) DIY DAC design.

I do not want to make a thread which is 1000 posts long. I simply want to compile a list of "Design guidelines" for achieving an excellent quality DAC for a modest price. $200-$300 without enclosure.

Originally this started as a personal email to ThorstenL, as I have read many of his posts and I respect his opinions and experience. He recommended that I start a thread; and so, HERE IT IS. Let's hope this thing doesn't explode out of control.

I think it will cover such topics as follows:

1) DAC chip selection: is there a single chip which gives the best chance of great sonics? regardless of implementation?
2) Power supplies: are shunt regulators needed for great sound?, are they worth the extra complexity?
3) Is an ASRC good or bad? are you better off without it? In what conditions is it good? When is it bad?
4) Decoupling caps: do we really need to use film to get great sound? Whats wrong with ceramic?
5) How important is implementation (artwork) compared to parts selection?
6) etc. etc.

To kick it off, here is my original email to ThorstenL (his reply in italic) and his answers to some of the questions, hopefully this will help some others on the forum as I am hoping it will help me. In the end, I hope to have some concrete design descisions made and I will offer up my results to the DIY community.

So all you experts, lets get it on! Keep in mind, I want this to be short and concise. A simple design guideline for a great DAC. Not a long debate.

Design Goals:
1) natural and transparent sound, but not overly transparent (I like a bit
> of softness/warmth). My current amplifier is a Passlabs Aleph 2 clone, which
> I absolutely love although my experience with other high-end gear is very
> limited. I just like it to be very musical and non-fatiguing.
> 2) something a "cut above" the mainstream products that I can afford (which
> is not much).
> 3) SPDIF input with balanced and single ended analog outputs.
> 4) transport insensitive (in other words, my CD player is not great). It
> needs to sound good regardless of the transport.
> 5) All components easily obtainable (mouser, digikey, etc.)

Have you looked at Nelson's whole DAC? It delivers all that and well.
Might even be possible to get one 2nd hand at sensible money and
restore/tune it.

> Here's what I was planning:
> 1) WM8804 SPDIF receiver

If you want to use 176.4KHz you need to control it via MCU. In that
case 8805 allows multiple inputs, might as well use that. Correctly
implemented it is currently the best "low jitter" receiver.
> 2) SRC4192 ASRC chip (to "clean-up" jitter)

I'd just leave it out. Not worth bothering.
> 3) PCM1794 with Pass Labs D1 I/V stage (IRF610 as common gate stage)

I'd use BJT's (or diamond transistors like OPA861) for the I/V in
common base instead, but I doubt I'd use PCM1794.

Differential PCM1704 is worth doing if you want to be able to play
higher resolution/sample rate recordings.

> 4) Pass labs Aleph P preamp stage and resistor ladder attenuator for volume
> conrtol

Why not combine I/V and Volume control? You could use the I/V resistor
for coarse attenuation and fine attenuation after that.
> 5) regular three terminal regulators for each supply rail

Nothing is better than it's powersupply. The common 3-Pin regulators
as sole PSU are not very good.
> Is the ASRC chip really that bad of a sonic degradation?

In my experience? In most cases yes. In some cases (CD's) it does
improve the sound. On anything better than CD it is really bad.

> Is the PCM1704 worth the money? (I originally wanted to use this, but the
> cost was a bit scary without knowing if I would like the result).

IMHO, yes.
> Can you give me an example circuit for how to implement a secondary PLL
> instead of the ASRC for jitter removal? Do you just need to keep the PLL
> loop bandwidth low? If so, how low?

Nelson's DAC had a secondary PLL, with the 8804/05 you can probably do without.
> How big of a difference does a shunt regulator make on the sound? i.e. are
> they worth the extra circuit complexity. Where are they most important?

Powersupplies are a huge topic. There is no single "silver bullet"
regulator. Important is to select the circuit in the powersupply so it
is optimal for a given application, including decoupling and bypass
capacitors, ferrite beads/inductors where useful etc. et al.
> Can I decouple the DAC supplies as recommended (cermaic caps), or do I need
> to use Film caps as many poeple say on the forum. I am very skeptical that
> the decoupling cap on the supply of the DAC can be audible.

They are quite audible, I'd not worry too much about the ceramic caps,
keep the for very high frequency noise. Use suitable "good sounding"
caps on the analogue supplies of a DAC and any reference pins. I like
Elna Silmic best. And yes, this stuff is quite audible.
> Are AC coupling capacitors as bad as some people say? I would guess not,
> since most tube gear and Pass Labs stuff has them.

They are as bad as "they" say. What "they" often omit is to note that
"coupling" capacitors are not always in series with the signal. They
may sit in power supplies, in servo circuits and so on. This is one
area where well implemented shunt regulators can help, they can
replace the "coupling" capacitors between ground and supply. Equally,
sometimes a simple high quality coupling capacitor is less damaging
sonically than all the stuff that is needed to avoid it, sometimes
it's not.
> So basically, I am asking if you have a simple recipe for success for a DIY
> DAC.

If you can live with 16/44 get something using TDA1541A, no
oversampling or filtering, 8804/8805 receiver, I2S attenuators and DEM
reclocking as per ECdesigns Schematics. I'd use a Tube stage with that
(I designed one that can be used), but I would also consider using a
pair of OPA861 per channel as common base diamond transistor I/V and
Buffer. You can also use the OPA861 as powersupply regulator to give
the TDA1541 (or other chips) +/-5V, using it for the -15V needs a
little more effort.
If 16/44 is not enough, things get more interesting. Keep the
receiver, use a suitable logic to split the signal for 2 (or 4 pcs
balanced) of PCM1704, suitable analogue stage (tubes or other non nfb)
and supplies.

Sorry for the long post, they will be shorter from now on.

[Jmertz]

Here are my replies to ThorstenL,

Have you looked at Nelson's whole DAC? It delivers all that and well.
Might even be possible to get one 2nd hand at sensible money and
restore/tune it.

Yes, I have the D1 schematic. It is too hard to get the PCM63 chip. I was planning to use the I/V stage though.

If you want to use 176.4KHz you need to control it via MCU. In that
case 8805 allows multiple inputs, might as well use that. Correctly
implemented it is currently the best "low jitter" receiver.

96kHz is all I will ever need.

2) SRC4192 ASRC chip (to "clean-up" jitter)

I'd just leave it out. Not worth bothering

No problem leaving it out. Note: I do have a CMAC OCXO with state of the art low jitter (<1ps RMS over the audio bad). Would this give any sonic benefit if I used it as a master clock for the ASRC? or just over-kill?

Why not combine I/V and Volume control? You could use the I/V resistor
for coarse attenuation and fine attenuation after that.

I also would like to use this as a preamp. I need to have RCA inputs to the preamp stage. So I need both the I/V and a preamp stage. I will have relays to select the input as RCA external or DAC direct.

In my experience? In most cases yes. In some cases (CD's) it does
improve the sound. On anything better than CD it is really bad.

I listen to all CD. Would the ASRC be an improvement in this case?

I'd use BJT's (or diamond transistors like OPA861) for the I/V in
common base instead, but I doubt I'd use PCM1794.

Differential PCM1704 is worth doing if you want to be able to play
higher resolution/sample rate recordings.

Please explain (technically and sonically) why are BJT's better? Pass Labs D1 uses IRF610 and has great reviews. With just regular CD, is the PCM1704 still worth the cost?

This is one
area where well implemented shunt regulators can help, they can
replace the "coupling" capacitors between ground and supply.

Are you saying that the shunt regulators should not have capacitors at the output? You depend on the regulator to be low impedance at high frequency?

If you can live with 16/44 get something using TDA1541A, no
oversampling or filtering, 8804/8805 receiver, I2S attenuators and DEM
reclocking as per ECdesigns Schematics. I'd use a Tube stage with that
(I designed one that can be used), but I would also consider using a
pair of OPA861 per channel as common base diamond transistor I/V and
Buffer. You can also use the OPA861 as powersupply regulator to give
the TDA1541 (or other chips) +/-5V, using it for the -15V needs a
little more effort.
If 16/44 is not enough, things get more interesting. Keep the
receiver, use a suitable logic to split the signal for 2 (or 4 pcs
balanced) of PCM1704, suitable analogue stage (tubes or other non nfb)
and supplies.

My sources are all 16/44, but I think TDA1541 is too hard to get. I do like the PCM1704, cost is a big factor I need to consider.

[ThorstenL]

Hi,

Originally Posted by Jmertz

Yes, I have the D1 schematic. It is too hard to get the PCM63 chip. I was planning to use the I/V stage though.

You can keep the rest and use the PCM1704 instead of PCM63. Wilte not pin compatible (plus the 1704 has less output current) both DAC's share more than they have apart.

Originally Posted by Jmertz

96kHz is all I will ever need.

640K is all the memory anyone will ever need... ;-)

I find I now need 176.4 & 192 KHz.

Originally Posted by Jmertz

No problem leaving it out. Note: I do have a CMAC OCXO with state of the art low jitter (<1ps RMS over the audio bad). Would this give any sonic benefit if I used it as a master clock for the ASRC? or just over-kill?

It will certainly not do any harm compared to lesser oscillators.

Originally Posted by Jmertz

I also would like to use this as a preamp. I need to have RCA inputs to the preamp stage. So I need both the I/V and a preamp stage. I will have relays to select the input as RCA external or DAC direct.

You could threat the Inputs as current inputs (with voltage to current converting resistors), just for arguments sake...

Originally Posted by Jmertz

I listen to all CD. Would the ASRC be an improvement in this case?

I find the ASRC a disimprovement in 95%+ of all CD's. If it is a choice of having it or not, permanently in the signal path, I'd leave it out. If I cannot have a remote to turn the ASRC on/off I would leave it out.

Originally Posted by Jmertz

Please explain (technically and sonically) why are BJT's better? Pass Labs D1 uses IRF610 and has great reviews. With just regular CD, is the PCM1704 still worth the cost?

Technically BJT's often have lower emitter impedance and a more linear emitter impedance than Fets source impedance. BJT's are very to compensate for temperature changes. I prefer BJT's in "common base/source" applications, Fets in "common emitter/source" and usually Fet/BJT Sziklai pairs for "common collector/drain" applications, all else being equal.

Originally Posted by Jmertz

Are you saying that the shunt regulators should not have capacitors at the output? You depend on the regulator to be low impedance at high frequency?

Most shunts will need some capacitance to remain stable. Often this capacitance can be fairly low value. In 3-Pin regulators I like to have VERY LARGE value capacitor with fairly high ESR after regulator, with most shunts modest value film caps are possible, or very high quality Electrolytics that do not come in very large values.

Originally Posted by Jmertz

My sources are all 16/44, but I think TDA1541 is too hard to get. I do like the PCM1704, cost is a big factor I need to consider.

The TDA1541A is now starting to become difficult to get. My first TDA1541 DAC was made with a chip pulled from a very dead and ultr-cheap Philips all plastic Chassis CD-Player.

On CD (and only CD) I find the TDA1541 still delivers a realistic and analogue sound that is not matched by any more recent design.

Ciao T

[Kurt von Kubik]

Originally Posted by Jmertz

Hi there
I´ll try to see if I can provide something usefull.

I don´t think so.
In most designs the DAC chip is by far the best component, so I´d pay attention to its surroundings and the load of it.
I think you can achieve outstanding results from both AD, CS, TI, AKM, Wolfson, Sony or whatever you choose.

I´d choose shunts for the analog circuitry, you´ll hardly believe how difficult it is to design a PSU allowing your amplifier to become transparent, unless you choose the shunts.
I think the problems is mostly caused by the capacitors, which are lousy everyone of them. The best results with serial regs though I achived using no electrolytics, but film caps only after the reg itself.

This is a pretty good question, in my experience the ASRC brings in natural proportions in the sonic result, giving you much more natural ambience and 3D in especially AB stereo recordings - it is actually dazzling.
It also takes away what a lot of people call digital sound, because it in some strange way eliminates artificial emphasis on transient start ups

4) Decoupling caps: do we really need to use film to get great sound? Whats wrong with ceramic?

Ceramics sound terrible, that is the only reason.
Use ceramics where absolutely needed and nowhere else.
Any analog need for decoupling, that be the reciever, or the DAC itself as well as your analog stage, is to be done with the very best film caps, which are without any doubt PPS caps.

5) How important is implementation (artwork) compared to parts selection?

A lot, you can really make good parts sound lousy by implementing them poorly.

6) etc. etc.

One good advice I can give is to keep in mind, that resonances are the main cause of listening fatique and distorted sound, so think of low Q supplies and amps all the time

Design Goals:
1) natural and transparent sound, but not overly transparent (I like a bit
> of softness/warmth). My current amplifier is a Passlabs Aleph 2 clone, which
> I absolutely love although my experience with other high-end gear is very
> limited. I just like it to be very musical and non-fatiguing.

The most funny thing is, that the more transparent you design your gear, the more musicality you will achieve. It is a myth that electronics are to hide or ad anything to become musical.
Precission in sound reproduction will reveal the character of the instruments in a both smooth and natural way, because real instruments are actually very smooth and beautyfull to listen to

> 2) something a "cut above" the mainstream products that I can afford (which
> is not much).

Generally electronics are not that expensive

> 3) SPDIF input with balanced and single ended analog outputs.

OK - the balanced interface is hardly used anywhere

> 4) transport insensitive (in other words, my CD player is not great). It
> needs to sound good regardless of the transport.

Which is impossible

>

5) All components easily obtainable (mouser, digikey, etc.)

Good idea

> Here's what I was planning:
> 1) WM8804 SPDIF receiver

Seems OK

> 2) SRC4192 ASRC chip (to "clean-up" jitter)

Also OK

> 3) PCM1794 with Pass Labs D1 I/V stage (IRF610 as common gate stage)

It is a very good chip, but in puts out quite a lot of current

> 4) Pass labs Aleph P preamp stage and resistor ladder attenuator for volume
> conrtol

You ought to do a preamp instead

> 5) regular three terminal regulators for each supply rail

They are not very nice

> Is the ASRC chip really that bad of a sonic degradation?

No

> Is the PCM1704 worth the money? (I originally wanted to use this, but the
> cost was a bit scary without knowing if I would like the result).

I like the PCM 1702/1704 a lot

> How big of a difference does a shunt regulator make on the sound? i.e. are
> they worth the extra circuit complexity. Where are they most important?

They are indeed

> Can I decouple the DAC supplies as recommended (cermaic caps), or do I need
> to use Film caps as many poeple say on the forum. I am very skeptical that
> the decoupling cap on the supply of the DAC can be audible.

All the caps are audible, but I would use ceramics on the digital supplies and film on the analog supllies

> Are AC coupling capacitors as bad as some people say? I would guess not,
> since most tube gear and Pass Labs stuff has them.

With tube gear you will hardly know the difference, but coupling caps in good SS gear are best avoided.
But it is not easy to design a DC servo circuit, that is completely transparent. But one of the rules of thumb is a very low cut of frequency, and use 1. order only.
But if you choose 1 point servo´s you might have a chance of implementing 2. order servo´s without the usual problems.


[/QUOTE]

[Jmertz]

Excellent advice. thanks guys! Let me digest this some and I'll post my response and any more questions I may have later.

I think a simple list of basic guidelines is not too far away.

[Jmertz]

Ok just to summarize what I have so far:

1) PCM1704 is a great sounding DAC. Worth the cost over the newer and cheaper DACs from BB, ADI, CS, Wolfson, AKM, etc.. However, it may be possible to achieve great sound with almost any of the new high end chips. If you only want 16/44 the TDA1541A is as good as they come, but is pretty hard to get.

2) Shunt regulators are well worth the extra effort due to the most times bad sonic signature of almost all large value electrolytic decoupling caps. The shunt regulator circuit should have low output impedance over a wide bandwidth and only incorporate small value high quality (film) capacitors at the output. These should be used for the analog sections and the DAC reference if available.

3) ASRC is usually a bad thing, but may in some cases yield some benefit. Certainly if the incoming jitter is very high the ASRC could be useful to clean it up. However, a 2nd PLL could also do the same thing without adding it's own signature to the signal. Putting it in the design is fine, but a BYPASS option should be included.

4) The WM8804/5 is a cut above the rest in terms of jitter rejection at the SPDIF interface. In fact, a secondary PLL may not even be necessary when using this reciever due to its very low jitter clock recovery architecture.

Let me know if I got this right?

[Jmertz]

a few more questions:

1) good point about >96kHz sampling rate. Can you give me an example or two of where one might need to have the higher rate capability? I am not very familiar with all of the audio formats out there. As far as I know my CD player will not send anything higher than 16/44 unless I turn the upsampling on, which is a bad idea.

2) I understand that BJT's are better suited electrically for the common base I/V stage, but does this translate to better sound? Does the particular BJT used matter much? or does it just need to have enough bandwidth and power handling?

3) What about DF1704? Do I use this with PCM1704? or NOS

[Jmertz]

The OPA861 is very interesting. I have not seen this part before.
How does it compare sonically to a discrete BJT in common base I/V?

I will have to read up on the DEM reclocking and attenuated I2S. I am not familiar with these. Can you give me the basic idea behind them?

[tritosine]

Sadly those bashing ASRC-s never give solid evidence, and the 'net is full with the contrary . Upsampling as popular term appeared with the dCS 972 , so I suggest reading those reviews. It was about the time when pcm1704 was new. Also the dCS thing is somewhat unique piece of gear because not only it does resampling , but user can set many types of noise shaping , and highest order (7? -9? ) was found to sound best, especially with PCM/multibit converters. I think the ESS has only 4th order , so this might be not the best . : ))
Also I think Im goin to sell some 4-5 pc's of pcm1704K in a few weeks, old-old datecodes, pulled, no ROHS.

[EUVL]

Low cost and good sound ?

Just use one piece CS8414 in slave mode, and a ES9022 with a 50MHz Crystek clock.
Simple 5V & 3.3V supply (LT1761s). No IV circuits etc to worry about.
Less than $50 without power supply (battery).

See

Anybody using the new ESS Vout DAC (ES9022)?

and PM John. He'll tell you stories.


Patrick