Simple good enough Clock with batery/cells

[Eldam]

Hi,

I need please some advices about the way to go or not with clocks for a DAC or tweaking CDplayer.

If I understand a good clock is for the most important things, please correct me if i'm wrong:
- good quality of the crystal oscillator
- clean power supply
- stable voltage
- layout : very short traces between components/crystal and dac and good choice about caps.

So I know, i can't go wrong with a tentlabs, diy kwak n°?... but is it always good to tweak already in place clock in a old Cd players like the Philips or Marantz? Old crystal oscillator have advantage : good stability ?

For a DIY DAC, what is a good brand for crystal oscillator (Tentlabs, but what else ?) if i want to do myself the most simple clock :
crystal oscilator : VCXO, TCXO (do i really need that?) : which brand for quality ?
stable voltage : LT-1236 (noise< 1ppm) high precision, 5V
clean voltage : batery, which type ? (low resistive), my idea i for a 1,8 V clock( for ECdesign project)
layout, caps : sanyo SEP but Can i do alone an enough goog layout.

Sorry for the number of questions, but tell me if the ideas are just naives (I'm not a tech) but I want to understand if it's possible to have a good clock with a battery/cells and better than a regulated one from an alone transformer power supply with good reg à la Dexa or others? Maybe cheaper (but no compromise with quality parts).
Or i forget the 1,8 V clock and i go for a easy solution with ian FiFO... but do i need that for a multibit DAC like the AD 1865 or TDA1541, PCM56...?

thanks for your answers or designs, and forget if the que question is too large.. Not a problem if you hijack this thread for sharing about the upon ideas about the pertinence of the clock design in the final result of the good sounding of a DAC... asIi know nothing I can just learn here !

 

[twest820]

You can beat Tent with a Fox Xpresso XO that costs under USD 2. Crystek has some XOs that are better than the Xpressos but usually it comes down where the part falls in the lineup and whether you're willing an order of magnitude more for a chance of maybe dropping the phase noise by 10dB. For crystals look at TXC.

The LT1236 is a voltage reference, not a regulator, and lacks the drive capability to power most XOs. If you want low noise in a relatively easy to solder package the ADP150 and 151 are good starting points. The TPS7A4700 and TPS7A4901 are worth considering as well. OSCONs are probably overkill but if you want polymer 'lytics look at Nichicon FPCAPs---higher performance and cheaper. All four regulators I just mentioned are ceramic stable and will therefore tolerate polymer 'lytics as well.

Ian's FIFO is primarily of interest if you want to try for better than datasheet performance. The datasheet figures come from the parts' eval boards and the eval boards aren't isolated or buffered. In practice there are enough complexities in knitting multiple boards together obtaining a better solution's not exactly trivial; easy to end up with something that's actually worse.

 

[Eldam]

thanks for your inputs

 

[Eldam]

Hi,

If a batery or cell is better than a filtered AC source, do i have to use a bulk cap of 47 uF for ewample between the cell (little lead batery=lowest resistance ?) and the little power SM caps (ceramic/film)of the reg (2 x 1 uF for the adp 150/151) ? I understand that the reg has to be as near as possible of the crystal oscillator.

I don't understand if polymer could be worse that ceramic :

In my mind this bulk cap is here as decoupling caps because of the to high resistance of a cell and for the high frequency noise because we talk about XO and because of the distance between XO & cell (more than 1 or two inchs). Am I right or wrong ? Because of the value I thinked about a polymer (good enough in high frequency and small esr) but a SM of the same should be better here ?

In less critical IC I know some guys who isolate a lead batery with 1 Farad Supercap (very high ESR but less resistive than the lead battery) and put a film cap near the load, here a OP for a CD player... I think a Super cap is too bad for a XO because of the high frequency and the noise due to the high pulse, but which value coul be good for this XO and a cell ?

Sorry for the theorical questions but am I right to benchmark between noise of a AC powersupply and the noise due to the too high resistance of a cell in a XO IC ? IS a cell + caps + reg + XO good enough for a Master clock in a cdplayer or DAC ?

thanks a lot

 

[twest820]

The oscillator market's unusual in that most manufacturers don't provide applications guidance, reference designs, or eval boards. Pretty much the sole exception I'm aware of is SiLabs, whose eval boards use a lab supply with the default 100nF MLCC for local bypassing or something like a MIC5301 with the usual uF MLCCs on the output. I would be surprised if their competitors are doing anything different, so it should be safe to assume the datasheet jitter figures aren't relying on exotic supplies. If you have the phase noise measurement capabilities to determine if a battery or supercap yields improvement or degradation over point of load regulation with clean input supply to the regulator (which the preregulator and point of load LC filtering can deliver pretty cheaply) I wouldn't bother implementing other than point of load regulation. Your effort's probably better spent ensuring ground bounce between the XO and DAC is neglible (it's an unbalanced interconnect after all), terminating well to minimize ISI, and providing vibaration isolation to avoid vibration lifting the jitter above datasheet---quartz oscillators are piezoelectric, after all, so you might end up getting better performance in practice with a sillicon XO even if it has somewhat higher headline jitter spec.

You can simulate various supply networks to better understand how choice of bypass capacitance and layout parasitics affect the supply impedance as a function of frequency. In practice it's unlikely the oscillator is slow enough a polymer 'lytic will make much difference in the supply's output impedance at the oscillation frequency. XOs are big enough the limiting factor on bypass performance is the package bond wires and not the supply plane pours, vias, or capacitor ESL. At the upper end of the clock range used in audio the supply impedance gets to be dominated by on die capacitance (and possibly regulation) and not the external components. If you want to try for better than datasheet performance you could look at 0306s or X2Ys instead of 0603s but, as with the supply, doing so is not so useful without phase noise measurements to determine whether performance improves or degrades. Series resistors or higher ESR parts like tantalum might help, but the sims I've run all suggest it's rather difficult to mitigate on die/in package LC resonances with off package components.

 

[Eldam]

Hi, thank you very much for the time you spent for this instructive and long answer. It helps and i feel like to try with the parts you list.

Yes I read somewhere than vibration is one of the greatest enemy of XO. Thanks again for the silicon tip about the XO.

No I have any measurement tools but the simpliest (for capacitance and voltage-amper-homs). I can't simulate but it's on my to do list to learn : i saw many guys who talk about lt'spice... How to say that : i'm very specialised in caps soup. I can do great improvment on devices just by cooking a good caps soup. It's incredible how a device can be bended just with caps to sing better when you have a long experience of that... but of course it's not science... it's just cooking soup with a ladle, and i want to learn more (how to make a cauldron for example).

That why my questions seems to much simple and out of the context perhaps. I'm sorry for that. In my basic understanding, things are simplier than they are in practice because i think theorical but I hope that DIYA is not only for the owners of oscilloscope ... I understand in the base line it is for everyone and I am grateful and glad with answers like yours.
After all digital clocks is the heart of an audio digital device (I don't say the soul...) but maybe not a problem to make for most of people here ?! (i don't talk about the special project of Ian... what a super & impressive work!).
I wanted just to try to do a little better than the standalone XO in cd player but less expensive than the very impossible to do alone LC audio clock for example... you see it: these clock is like a computer board!).

Is there a wiki function on diya for that ? A wikipedia of audio engeenering nearby the forum for the keen, enthusiastic lad...

 

[twest820]

You might look at the CDCE62002 or LMK04800 and their related design collateral, then. Rather simpler and cheaper than, say, the LC XO 3, and the performance is fairly well characterized. u.fl is a good default for interconnects if you can ensure 50 ohms on the boards.

There's not a DIYA wiki, though it'd be useful to have something along those lines. Well, except that getting DIYers to agree on much of anything is rather like herding cats, so keeping it from getting smothered in contradictory opinions would be a bit of a nice trick. The measurement gear and ABX testing needed to verify results is often well beyond DIY capabilities and getting around the objectivism versus subjectivism thing isn't exactly trivial---a middle ground does exist where subjective experience does a good job of generating design requirements, but it's usually lost in the noise generated by more positions more towards the extremes of the continuum. Jitter, in particular, is a tricky one as it's not really known how to determine if a given distribution of phase noise is audible. For most audio components it's not too difficult to state either design requirements for subjective transparency or obtain intrinsic technological limits but I'm not aware of any ABX test results between XOs and jitter cleaners. It wouldn't surpise me if they're subjectively equivalent given a good implementation but I'm speculating here.

 

[Eldam]

I read the AN-1939 Crystal Based Oscillator Design with the
LMK04000 Family PDF. The TI site has a good design to learn...good for me.

I was almost joking for the wiki because I see there is a tree for that on DIYA... but it's empty... Sure the most important is what you learn from the answers of contributors, wiki can be find elsewhere...and in the answers too.
Most of the time, people like me doesn't know where is the balance between the level of a tweak or shematic and its "good enough" result for ears... And we spent money and a substantial time to control with ears. Because it's trial-and-error method, even if the part of error reduces with the experience (my caps soups are excellent!)
And about the clocks in audio, there are maybe just audio engineers or good technician who have time and money enough to benchmark many conf.
Some people like T Loesch, G Tent, ECdesigns, etc, know extacly these balance. But for the majority of diyers, it's not easy to know... it's political : 170 euros for a clock ! half of a washing machin but may be with no result for cleaning the jitter in the CD player or DAC you are tweaking ! Or may be it works equally with a 3 dollars XO .... or maybe be here the layout or DAC is not good enough for a Masterclock to improve the sound.
That's why your informations are usefull.
Tomorrow reading, but know Jussi Bjorling & Puccini make my day... and T Monk to be sure.
thank you Twest820

 

[tomchr]

It's been a while since Jim Catt and I wrote AN-1939, so its content isn't at the forefront of my mind right now. But I seem to recall it dealing with how to replace the expensive, $10, VCXO with an inexpensive, $1, crystal-based solution and getting performance rivaling that of the $10 solution. As I recall, that app note is pretty specific to the LMK04000-series, though, many of the concepts and tradeoffs apply for any crystal oscillator. For even better performance, I suggest looking at the LMK04800-series or the LMK00301 if you don't need the PLL. The noise floor of the XO core is ridiculously low (see the datasheet for the parts) so I suspect you won't need the PLL to clean it up.

But to your original question. There are many ways to screw up high-performance circuits. Lots of jitter can be added through improper termination, poor supply bypassing, poor layout, etc. Henry Ott's book, "Noise Reduction Techniques in Electronic Systems" is recommended reading and will explain many of the issues that affect precision circuits. Walt Jung's "Op Amp Applications Handbook" covers a lot of these issues as well. I think Jung's book is actually available on-line on Analog Devices site. After reading a few chapters, you'll know more than you ever wanted to know and be amazed that any precision circuit works at all.

I seem to recall Ott's book being a bit easier to read, whereas Jung's book is better for those who are more comfortable with op-amp circuits already. Both are well written.

~Tom

 

[twest820]

The noise floor of the XO core is ridiculously low (see the datasheet for the parts) so I suspect you won't need the PLL to clean it up.

Figure 25 of the LMK00301 datasheet, right? Looks like the same figure is repeated in the LMK00304 datasheet which is a bit curious(it's also figure 25), but the 304 might be a better choice for DIY. Somewhat fewer pins on the package and thereby less overhead for the one LVCMOS output that's plug and play with audio DACs. Either way I can't think of any XOs I've seen with close in phase noise that low, including some righteously expensive Crystek and Wenzel references, and the 30x crystal driver's compatible with a good set of 256 and 512 fs clocks. Nice part, mate.

One question, though; any thoughts on how to pull down the spurs? The main trendline performance's great but the top of spur envelope's not all that much better than, say, an Xpresso. The 40MHz spurs look like they might be harmonics off 50Hz mains, which is kind of interesting as I'd have expected the datasheet to be written in 60Hz land.

 

[Eldam]

Thanks for the refs Tom.