Arcam Alpha 5 CDP upgrades

Yes that's right this never ending thread is active again!

I'm getting all excited again about the possibility of making some further refinements to the Alpha 5.

One of the possible upgrades I've been thinking about is carrying out a more sophisticated re-clocking arrangement. Some time ago I fitted an upgraded Net Audio clock unit which is powered by an Audiocom clock PSU.

In the Alpha 5 as standard the clock feed from the crystal goes to the SAA7220P/B where it is then distributed to the decoder chip and the DAC. The SAA7220P/B is known to be very noisy and 'pollutes' the clock signal and so I gave the decoder chip its own separate clean clock feed.

The clock feed from pin (16?) of the SAA7220P/B goes to pins 2 and 4 of the TDA1541A. This is the called the 'Bit Clock' (BCK).

In theory goes that it is better to give the TDA1541A DAC chip a bit clock feed from the master clock rather than from the SAA7220P/B. However you can't just take a clock feed directly from the master clock as the TDA1541A needs a bit clock feed of half the frequency of the master clock 5.6448 Mhz as opposed to 11.2896Mhz.

What you need to do then is to divide the clock signal frequency by half before feeding it to the TDA1541A. This can be done with a flip-flop IC.

I found this circuit on the audioworkshop blog (Credit: Ventsislav Simonov).
http://blog.audioworkshop.org/low-jitter-clock-for-cd-player-upgrade-module/#.U4CPHyhN330



Here is a photo of the finished installation shown on the Audioworkshop blog. The clock is on the right and the flip-flop circuit is the small vertical board centre screen.



It all looks very neat and I'd like to give it a try.

The two yellow film caps and the Oscon aren't shown on the schematic. My best guess is that the Oscon is a decoupling cap for the +5v supply to the 74HC175N IC?

There is also the possibility of performing a 'DEM re-clock' but that's another can of worms for another day.
 
I've ordered a 74HC175N and a sheet of perforated board to build the flip-flop circuit with.

I can take a +5v supply for the 74HC175N conveniently from the DAC board as there is a +5v supply which comes up the vertical pins from the servo board. As standard this +5v supply is used for the SAA7220 but my player now has a dedicated PSU for the SAA so the original +5v supply is redundant.

Alternatively If the 74HC175N would benefit from its own voltage regulator I could tap into the +9v unregulated supply from the servo board?

I'm a bit unsure what decoupling arrangement I should use with the 74HC175N. I read the data sheet but it didn't say anything about this. Any advice would be appreciated.
 
I've been reading through the epic TDA1541 threads over on diyaudio with great interest. Most of it goes way over my head but I've picked up few ideas which I'd like to try, particularly from member 'ecDesigns'.


Here's are a couple of extracts from his posts on the 'Building the ultimate NOS dac using TDA1541' thread.

Active divider decoupling also depends on DEM clock. Like mentioned before, DEM clock oscillator is able to "lock" on (on-chip) BCK crosstalk. By selecting DEM clock capacitor values that produce (almost) exact multiples of fs, the DEM clock will "lock" to BCK and jitter amplitude would drop significantly. The often used 470pF and 680pF won't result in a "lock" and produce maximum DEM clock jitter amplitude.

Here are some capacitor values that result in a "lock" condition, capacitance might vary slightly due to TDA1541A tolerances:

235.2 KHz, 350pF
256.58181 KHz, 318pF
282.24 KHz, 300pF
313.6 KHz, 260pF
352.8 KHz, 250pF
403.2 KHz, 203pF
470.4 KHz, 180pF
564.48 KHz, 145pF
705.6 KHz, 122pF

The capacitors need to have tight tolerance (1%).

Tuning the capacitor is the easiest way for lowering DEM clock jitter amplitude, but jitter won't be as low as with an low jitter external DEM clock (DEM clock oscillator intrinsic jitter). So this modification improves sound quality over the 470 or 680pF, but still isn't optimal.



This sounded worth a try to me and much easier than building a DEM clock divider circuit. Not all of the capacitor values listed above are available so I chose 180pF as these are available as an exact value match. Today I fitted a 180pF 1% tolerance mica capacitor between pins 16 and 17 of the TDA1541A to replace the 680pF poly cap which was fitted as standard (C223)

I've read several times that the decoupling capacitor values on the three voltage supplies for the DAC (+5v, -5v and -15v) can have a big influence on the sound. It seems that many people feel the +5v supply has most influence.

On the giant thread over on diyaudio member ecdesigns posted:

I had best results with approx. 3000uF (multiple caps in parallel) for power supply decoupling (TDA1541A).

With that in mind I decided to modestly increase the value of the three power supply decoupling caps again. out went the 220uF Rubycon ZLH which I had previously fitted (up from 100uF as standard) and in went some 470uF Panasonic FC (C219, C221 and C225).

Here is a pic of the new DEM capacitor and the three new Panasonic FC capacitors.

 
While I had the DAC board out today I also took the opportunity to tidy up my wiring a little.

By re-routing and twisting rather than plaiting the signal and return wires from the clock I managed to shorted the wires by about 3 inches each which must be good thing.

I also managed to find much better locations to solder the power supply wires to the SAA7220P/B which are much nearer the Vin (pin 24) and the ground plane respectively.

Finally I couldn't resist swapping out the two 22nF ceramic disc caps near the RCA output to some Panasonic polypropylene ones of the same value which I had spare. On reflection I don't know if this was a good idea or not. Probably Arcam knew what they were doing! If it was a bad idea please let me know and I'll swap them back again!

 
Over the next couple of days I plan to upgrade the 7805 voltage reg on the main/servo PCB (Z402) to a LM317T.

I might also swap the 7905 (Z403) while I'm at it.

While I've got the main PCB out is there anything else I should look at improving at the same time? For example should the decoupling cap(s) on the SAA7310 be improved? Or the power supply caps beefed up?

At present the 7310 gets its +5v feed though a 2U2H coil followed by a 3R3 resistor. It has a 100uF electro and 47nF ceramic cap as decoupling caps.
 
It has been p*s*ing it down here this evening. No chance of a bike ride or going climbing so I decided to crack on and do the 7805 to LM317 voltage reg conversion on the main board.

Initially I tried to 'zener-ise' it but I kept getting a lower than expected voltage output (still no idea why?). In the end I gave up and used a pair of resistors instead.

I did fit a 33uF bypass capacitor on the v-adjust pin though ;)

It's not pretty but it works perfectly.



How does it sound after this evening's efforts?

To be honest I wasn't expecting any discernible difference but WOW it actually does sound better. Quite a significant difference too!

Martin, if you still have that Naim CD 2 and Rob, if you still have that Meridian 206 I'll happily send you one of those 180pF silver mica caps to play with.
 
Last edited by a moderator:
It still sounds just as good today even without the added benefit of red wine. That new DEM capacitor is a great little upgrade, well worth a try.

I've now got all the parts needed for that bit clock divider circuit. I'll post again when it's finished.

I think fitting the bit clock circuit should only require me to lift a single resistor so it'll be easy enough to reverse the mod if I don't like it.
 
I found in the CD2 that the dem cap was an off value and quite large; the reason came clear once I started poking about with the scope - the DEM oscillator had locked to the 4Fs word clock at 176Khz....I'm pretty sure it's deliberate, because I recall discussion of Naim adjusting this value inproduction of their better Philips-based players early on. (back from when JVereker used to contribute robustly to his own nacent webforum...)

Some bits and bobs from an old Philips datasheet that might interest you, not least the idea that pushing the DEM oscillaor up to higher frequencies actually increase THD+N%. I suspect this might be down to noise feed-through on the chip susbstrate, but no way of testing that so only supposition. Anyway, I believe this is Philips data..:


PhilipsDEMinfo2.jpg


demfvsthd.PNG
 
Superb info, thanks for sharing!

So if I'm understanding all this correctly (thinking out loud here)....

The Naim engineers knew the significance of the DEM capacitor and 'tuned' the capacitance value until it resulted in a 'lock' frequency, hence you found a non standard value in the CD2.

On my Arcam the 180pF gave an improvement over the 680pF because it gave a 'lock' frequency however I'd be better off still with a lower frequency than the 470.4 Khz which I've presumably locked onto now.

If I had a scope I could stack small value caps until I achieved a 'lock' but without a scope I suppose my best bet is to use the list of frequency/capacitor values posted by ecdesigns and aim for the lowest lock frequency of 235.2Khz which requires a 350pF capacitor.

I've just checked the available capacitor values on the Farnell website and 350pF is a non standard value but I can get to the desired total capacitance by putting two caps in parallel such as 330+20 or 150+200.

Frustratingly I've just put a Farnell order in so it might have to wait until I need enough bits to scrape a £20 minimum spend order together.

I don't suppose you can remember what the DEM cap value in the CD2 was?

All very exciting :)
 
It was over 500pF, which accords with the formula above (which suggests 530pF, ballpark)

I suspect the reason you get a better result with 180pF is because the DEM residue is even further outside the audio band, and so it gets better supressed by the output stage / filter. The Alpha uses a 3rd order filte IIRC. Naim went as far as a Bessel 7th order stage - takes a walloping bite out of noise above the audio band, -100dB ish attained by 176Khz, but a more complex design.

- It's also possible the higher oscillation freq. makes the DEM decoupling slightly more effective, until HF considerations like lead/decoupling path length get in the way. Depending on which pin these are on the decoupling caps see Fdem or Fdem/2; but since the impedance of even a 0.1uF cap is pretty low up around 200Khz, you get better decoupling of the residual ripple of differences between the four current sources contributing to this 'dynamic element match'.

I wouldn't worry much about pursuing 'better values' for the DEM cap. Note especially the stated large tolerance on 'Rint' - it's an impedance formed on-chip by some unspecified means (probably a BJT curent source, but not a great one.) This means the optimal cap for lock on your specific chip may vary quite a bit, too, so reasonably close is as good as you can do without trimming with a scope - and I'm not sure it's worth it even then.
 
Last edited by a moderator:
The more I learn about this stuff the more confused I'm getting.

I haven't got the patience for making one modification at a time as it just takes too damn long and yesterday I made several modifications all in one soldering session, after which the player definitely sounded better to me. I feel the sound is cleaner/crisper and it's easier to follow very subtle details in the music. I'm very familiar with the sound of my system so I'm pretty confident in saying the improvement was genuine and not 'psychoacoustic'.

It could be that it was the other mods which gave the improvement and not the DEM capacitor at all, the DEM capacitor might be a red herring.

It's just that I wasn't expecting such a noticeable improvement from the other mods and the post from ecdesigns talks about "significant reductions in jitter amplitude" if you can get the DEM clock oscillator to lock.

I might resort to a bit of A/B comparison testing using a variety capacitor values temporarily soldered onto to some PCB pins so that I can solder the caps from above with the circuit board in-situ. Oh what fun.....might save that one for a very very rainy day.

Thanks again for the technical input, much appreciated.
 
I do think the talk of DEM 'jitter' is very misleading*. The DEM oscillator is a free-running oscillator and its only purpose it to drive the shift register that averages the multiple internal reference current sources in this dac. (If those ccs turn out from production tolerances to be very closely matched anyway, as intended by Philips, the DEM does very, very little!)

So it's actually got no relation to conversion accuracy, except that any unintentional side-effects (eg a little on-chip HF noise fed through the chip substrate unintentionally into the analogue output) disappear as an artifact of the maths if the DEMclock is also a multiple of the word clock. That's it.

Put another way, the 1541 datasheets gives the performance acheived without any such messing-about, and it's still full 16-bit recovery of audio resolution, and 19bit noise floor.

Stick another disc on, enjoy the music :)


*ever notice how so many of the really popular threads over at DIYaudio are also about marketing ..?
 
I've been at it again.....

I'm still trying to get my head around the DEM oscillator and what (if any) impact it has on the sound.

The 180pF Mica cap which I had fitted seemed to improve the 'crispness' of the higher frequencies but I can't be sure as I did make several other small changes at the same time.

Over on diyaudio there has been much discussion about the DEM oscillator and the most popular set up seems to be a very simple circuit consisting of a 470pF capacitor and two 6K8 resistors:

  • 470pF capacitor between pins 16 & 17 of the TDA1541A
  • 6K8 resistor from each leg of the 470pF capacitor to -15v (pin 15 of the TDA1541A)

This simple circuit is very easy to install.


I haven't had a chance to do much listening but initial impressions are that it sounds a little smoother than the 180pF.

I've also fitted an LM337T on the main board as an upgrade over the 7905 regulator (z403). I made a small circuit on perforated board and mounted it on a PCB standoff. As usual I fitted an electrolytic bypass capacitor across the voltage set resistor on the adjust pin.

 
I have also repositioned the PSU module which supplies the +5v to the SAA7220. By turning the module through 180 degrees I've managed to bring the Vout of the module MUCH closer to the Vin pin of the SAA.
To get everything to fit I needed to use a shorter filter cap on the module so you'll notice this has changed from a 1800uF tall/thin Elna to a fat blue coloured 2200uF which was in my spares pile.
I also re-routed the AC line from the transformer behind the metal chassis further away from the DAC board.



Not a big change but hopefully another tiny improvident all the same.
 
I'm rather enjoying the Alpha 5 at the moment. The sound is really starting to come together:)

But...I was trawling some old forum threads over on PFM and I found this discussion about power supply decoupling for the TDA1541A.


IDM
20-12-06, 07:32 AM
So on decoupling do you suggest

10uf Oscon on +5v, 10uf Oscon on the -5V, 10uf on the -15V and a separate 10uf Oscon from the -5 to -15V?

Cheers,

Ian
martin clark
20-12-06, 12:52 PM
Yup, thats about the size of it; and up to 47uF of oscon seems to work well on the +5v pin.

BUT

Oscons that size alone will not 'play nice' with LM317/337s (too little ESR), so a little fettling/experimentation may be required. Adding about 0R25 to 0R33* in series with the LM317/337 output critically-damps the output impedance into a 47-10uF Oscon(respectively) - and the 1541 doesn't seem to mind the extra series resistance at all, decoupling and low noise seems to be more important (to my ears, anyway)
Click to expand...

On my Alpha the +5v and -5v supplies both have a 470uF Panasonic FC bypassed with a 47nF Mica for HF duties. The -15v supply also has a 470uF Panasonic FC but the bypass on this is a 0.01uF Wima poly.

So a few quick questions:

  • Are Oscons likely to be superior to the decoupling caps which I have fitted now?
  • If I want to fit Oscons should I take out both the electro and the Mica and just have the single Oscon for decoupling on each supply? If I do this presumably I'll need to add the series resistors to damp the regulator output.
  • Or alternatively, can I just take out the low value bypass caps (Mica and the Wima) and fit the Oscons in their place leaving the electrolytics in to damp the regulators? A small electrolytic should have an ESR of about 0.1Ohm, will that be enough to damp things?
 
I wouldn't worry about it.

Firstly those Panasonics have a useful, none-zero amount of ESR, 50-68mR given in the datasheet I found for the 470uF sizes - which is sufficient damping at those capacitance values.

Secondly- look at the graph I posted a while ago:

dietz.gif


- by the time you have a 470uF cap on the output of the reg, there is no peaking - it moves so far down in level and frequency it's not happening because it falls in the region where the reg still has plenty of gain (below 1khz). Finally - you have the Vadj pins bypassed and this decreases the effective output inductance significantly (about 4x at the voltages you are running. So no problem at all :)
 
Thanks Martin. It's good to hear some reassurance that I've already got the TDA1541A power supply decoupling pretty well sorted.

Curiously the Alpha 5 comes with a 47nF decoupling 'bypass' cap on the +5v and -5v supplies but has a 22nF ceramic for the -15v supply (which comes in via a 33uH inductor). My player now has a 0.01uF Wima fitted instead of the ceramic as this is what came in the NET audio upgrade kit which I fitted many moons ago. The -15v supply is also shared with the op-amps if that's of any significance here.

Over the last year or so I've been slowly collecting TDA1541 chips for future projects. I now have a modest collection of four chips with a fifth one on order.

So far I have:
1988 HSH
1993 HSH
1994 HSH
and also 1997 HSH 'Taiwan'

The Taiwanese made chips from 1997 to 2000 are thought by some experts to be the best of the bunch. The theory is that production tolerances on these later chips are much tighter than older chips meaning that many non-crown Taiwanese chips from this era can easily match earlier single crown S1 chips for performance.

My 1997 chip is now sitting in pride of place in my CD player fitted into a high quality turned pin gold plated beryllium copper IC socket.

 
Over the weekend I did some experimentation with different DEM capacitor values.

I wanted to be able to make quick A/B listening comparisons and so in the end I fitted a small DIP socket to the DEM capacitor location on my "development hack" spare Alpha 5.



Here is the selection of capacitors which I had. From left to right:

680pF, 470pF, 400pF, 330pF, 300pF, 180pF and finally a 3pF for 'fine tuning'.



After a few hours of testing I came to the conclusion that changing the DEM capacitor value only has a very very subtle effect.

Unless you are seeking perfection it isn't worth bothering with but for anyone interested in such things I would summarise my listening impressions as follows:

  • I found that higher values (470pF and 680pF) gave a rich/full sound with the best bass.
  • The lowest value of 180pF gave the high frequencies a pleasing 'crispness' but low frequencies to suffer seemed to suffer very slightly.

This seems to fit with the Philips literature provided by Martin in post #108 which states that at high DEM oscillator frequencies distortion increases and that this distortion mainly effects low frequencies. My favourable impression of high frequency 'crispness' when running a low value capacitor (high oscillation frequency) may be explained my Martin's suggestion of better HF output filtering (see post #110).

From all the reading I've been doing on the subject I've come to the conclusion that the ideal DEM oscillator frequency for my Alpha ought to be 176.4 kHz.

Philips themselves originally suggested a DEM oscillator frequency in the range of 100-200kHz and 176.4kHz fits nicely in this range. It is also of course 4x the sampling frequency of 44.1kHz.

Using the formula Fosc=1/(5.9xRintxCext) gives me the magic number of 533.5pF.

I suspect that's more or less the same as Naim fitted in the CD2 judging by what Martin says.

Obviously you can't buy 533pF or 534pF 'off the shelf' so I've ordered a 1% tolerance 500pF Mica and another 2% tolerance 33pF from Mouser. This ought to get me sensibly as close as possible to 176.4kHz.

They ought to arrive later this week so I'll report back with some listening impressions.
 
Last edited by a moderator:
mjp200581 said:
Over the weekend I did some experimentation with different DEM capacitor values.
...
After a few hours of testing I came to the conclusion that changing the DEM capacitor value only has a very very subtle effect.
Click to expand...

Nice little experiment I've not tried. Thanks.

I was having a sort-out recently and found my spare -S1 single crown Holland-produced chip (bought a decade or more ago just.in.case for the CD2.) I'll happily lend it to you to play with if you like, could be interesting to plug in and see what you make of it /whether it's worth the fuss against the later chips you have to hand :)
 
Hi Martin,

That's very kind of you and yes I'd love to hear what an S1 chip sounds like in my Arcam. I think you've got my address already in an old P.M. Let me know what I owe you for postage.

Earlier this week my 500pF and 33pF Mica capacitors arrived from Mouser. I have them fitted in parallel to give a total capacitance of 533pF which will hopefully give my a 176.4kHz DEM oscillator frequency and result in a 'lock' to the word clock.

Here they are fitted:



I haven't done enough listening yet to say for certain whether the 533pF sounds any better than a 470pF but it certainly sounds at least as good.

I tried it with and without the 6K8 resistors tying each leg of the cap(s) to the -15v supply on pin 15 of the TDA1541A. To my ears it sounds better with the resistors in place and subjectively I'd say the resistors make more difference than the DEM capacitor.

"These result in approx. 3Vpp clock signal amplitude making it less sensitive to on-chip switching noise. This in turn reduces noise and deterministic jitter on the DEM clock."

(From ecdesigns over on diyaudio)
 

Latest posts

Back
Top