Arcam Alpha 5 CDP upgrades

Neatly done.

If you want to get closer to 5.0v try swapping one green led for an equally-cheap Red one; the forward voltage will be about 0.1-0.15v lower; c. 1.8v for red leds rather than c 1.95 for green :)
 
I just stumbled across this thread which seems to be very relevant to my efforts with the Alpha 5:

http://www.pinkfishmedia.net/forum/archive/index.php/t-29579.html

The bit I'm really interested in is this:

Well first off, the complement of all the signal current out of the TDA1541 also appears superimposed on the +5v pin, so decoupling here is critical... and a superreg would be an ideal supply to keep impedance low here. The supply on this pin is pretty much just pure DC, plus the signal components. It is the most critical in terms of direct effect on sound quality.

The -15v supply is the only pure analogue supply and should be quiet. It does not appear particularly sensitive to supply impedance, but low HF noise helps; a capacitance multiplier as a filter on the raw supply, followed by an LM337 does well here.

The-5v supply is the least critical, again a well-sorted LM317 is fine.

However - the internal current reference in the chip is between the -5V supply and the -15v supply, and tying these supply pins together with a cap really helps. 10uF oscon (+ to the -5v pin, remember) is a great start. Tweak your regs so that 10.0v is held between these pins, +/- 0.1v.

After tweaking the audio board power supplies with LED's and Zeners the -15v supply to my TDA1541 is running at -14.31v and the -5v is running at -5.27 giving me a voltage difference of 9.04v. According to the advice in this post I should tweak the supplies to get as close to 10v as possible.

I have some 47uF Oscons in my parts stash. Would this be suitable for connecting across the -15v supply (pin 15?) and the -5v supply (pin 26?) or is 47uF too much capacitance? 10uF Oscons seem to be quite difficult to come by these days.
 
I'd be wary of using a 47uF oscon because it ties the two regulator outputs together with very low AC impedance and I've not tested that for stability...whereas the 10uF works fine becasue its not enough to upset things, (the lossy caps on the two regs is enough to keep things stable) - and it seems more than enough to deal with HF noise. You might try other caps if you have them to hand - something like a 10uF tantalum could be ideal [since tants have a bit of ESR - very useful for a little inherent damping - and fairly low leakage current]. Don't worry much about that 10v number, it's not essential providing the supplies are 'quiet'.

Certainly try 47uF on the -15v supply to 0v (watch polarity - the new caps -ve terminal goes to -15v); the 337 will accept this providing you leave any other electrolytic cap on this pin in place for a bit of damping.
 
Well finding LED's with the right forward voltage is proving to be a headache. I've bought a variety of red and green LED's from Maplin and off ebay but so far they've all have a forward voltage of 2v or higher. Farnell specify the forward voltage of the LED's they sell which makes it easy to select the right component but they have a £25 minimum spend and I don't need anything else right now so it may have to wait.

In the meantime I've been reading up on 3 pin regs. Specifically Felix's articles on the acoustica site and this article: http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html

Here is a circuit diagram of the PSU module which I am using for the SAA7220 (rectifier omitted).


imgur

If I've understood the above articles correctly they may be some benefit in swapping C2 to a larger value electrolytic, maybe 220uF. Similarly it would be beneficial to swap C3 to a larger value electrolytic, again something like 220uF.
Both C2 and C3 should be standard ish caps rather than very low ESR types.
 
It's not worth £25 to swap at all! Pennies, maximum...

The other way round to drop the voltage a tad with LEDs or zeners as the 'reference' is to increase R1 from 100R or 240R or whatever to (say) 1Kohm. This also has essentially the same effect as bumping the Vadj-to-0v capacitor up in size by a comparable %age change in magnitude.

But the key question is, are you enjoying music more? Don't let that soldering iron get in the way ;)
 
I'm not enjoying any music right now because the CDP is it bits (again) LOL.

Actually that's not true, the turntable set-up is working better than ever.

I have plenty of spare resistors of varying values so I'll tweak the voltages as best I can.

Is swapping out C3 on my diagram for a larger value cap a good idea? The existing component is a 0.1uF film (I think) cap.
 
No, not really; the cap that really matters is always the one doing the bypassing right at the load - your new 47uf Oscon in this case.

The LM317 regulator does not require a cap close-by on its output for stability - and putting a high-quality film type one there actually causes more problems than it helps in at least two ways; leave it out :)
 
I found some 5mm red LED's with an almost ideal forward voltage. I have used them in the PSU for the SAA7220 which now has an output of 5.05v. Almost perfect!

I swapped the 22uF capacitor on the adjust pin (C2) for a 100uF which I had spare.

I also removed the 0.1uF film cap between the output and 0v (C3) and replaced it with a 100uF electrolytic. As far as I understand it it's a bad idea to have a low ESR capacitor in this position and higher capacitance can be useful to help lower the impedance of the PSU but feel free to correct me if I'm wrong.


image hosting 15mb

Things didn't go quite so well with the audio board.....

I still want to find some 3mm (T1) LEDs with the right forward voltage to get the +5 and -5v supplies closer to the ideal voltage. I'm trying to avoid 5mm LEDs as there isn't much space. For now the green green LEDs which I have been using are back in place which gives me +/- 5.3v.

I rearranged the 47uF Oscon which is tying together pin 12 and pin 24 of the SAA7220. You can see on the pic that I have trimmed the legs right down and mounted it vertically to make use of the ground plane and hence slightly shorten the loop area. It still isn't as good as a surface mounted capacitor but hopefully it's a slight improvement.

Incidentally the blue film caps which you can see on the pic were moved to the underside of the board in order to create enough space around the op-amps to allow me to take them out of their sockets and solder them directly to the board. I needed to do this as the top edge of the op-amps was catching on the lid of the casework each time I removed the lid.

I also installed a 10uF tantalum capacitor under the TDA1541 from the -5 pin (pin 26) to the -15v pin (pin 15) (visible bottom right on the pic).

However I'm afraid I had stupidly assumed that tantalum capacitors are non-polar. Within the first few seconds of powering up the CDP again this assumption was proven to be dramatically wrong. A funny crackly noise and the smell of something getting very very hot was the result of this mistake but luckily the tant itself seems to have been the only casualty. The offending capacitor has now been removed and the player seems to be non the worse for wear.

If anyone has a 10uF Oscon they'd like to sell to me please get in touch.


screen shot windows 7
 
Reading around on other forums it seems that other people have successfully used 25uF Oscons between the -5 and -15v pins of the TDA1541A. I've found a source of 15uF Oscons so I've ordered some and I'll give it a try, hopefully 15uF will be OK.

I have also been reading up about the use of different capacitors for the decoupling the TDA1541A on (pins 7,8,9,10,12,13 one one side of the chip and 18,19,20,21, 22 and 23 on the other side).

The datasheet for the TDA1541A states 100nF on all of the pins and this is indeed what most players use. However many of the most highly regarded TDA1541 based players use 220nF on all pins.

Arcam went a stage further and used 470nF tantalum capacitors on the last two pins on each side (12, 13 and 18, 19).

There seems to be general consensus amongst TDA1541 enthusiasts that higher capacitance on pins 13 and 18 sounds better. These are the MSB pins (most significant bit) and the general consensus seems to be that 1uF is optimal.

There also seems to be a general agreement that the exact type of capacitor used has a very noticeable effect on the sound and in this respect everyone seems to have their own favourite flavour of cap.

This evening I tried a pair of 1uF 50v Wima MKS capacitors on the MSB pins 13 and 18. I chose these because they are very compact. I left all of the other pins untouched.

What I can say for certain is that changing the caps on pins 13 and 18 definitely changes the sound. In this particular case the new Wima caps sounded more 'polite' than the original tantals, maybe a bit smoother (non necessarily a bad thing) but overall just too reserved for my liking.

I'll give them a bit another listen tomorrow but as things stand right now the Wima caps will definitely be going.

The obvious option would be to put the original tantalum caps back in place but I'm tempted to try a few other types first. At the moment I like the look of Epcos 1uF PET (+/- 5% tolerance) which I think will fit in the space available.

I'd welcome any suggestions/advice on this. I'm a bit baffled by MKP vs MKS vs MKT etc...etc..
 
What I can say for certain is that changing the caps on pins 13 and 18 definitely changes the sound.
Click to expand...
Certainly it does!

These are the Most Significant Bit pins, from which this dac actively subdivides each bit-level current - a process of constant re-calibration.

The significance is - noise accidentally impressed on the filter pins affects the 'transfer function' of the dac - it's linearity and goodness in other words.
With a little insight into how the dac works you'll find these caps do not need to be very large at all, which means the best designs use film-type caps throughtout. But because the filter pins are dealing with switching noise at the DEM pin oscillator rate (176.6Khz in the alpha5) good RF properties are also required. Nice, very compact layout and appropriate choice of caps then required. If one had a free hand to layout a pcb these days I suspect surface-mount 1206-size film caps would be the part of choice, right underneath the DAC chip to a ground plane tied to only the dac's 0v pins.

Naim CDPs using this dac used Siemens stacked-film caps of 100nF, and this turns out to be a very solid choice. Here's a pic from my CD2 (before I messsed with it)

na-cd-small.jpg


- in summary my points being:
1) no I would not use tantalum caps on these pins even though Arcam did; these are too 'leaky' and leakage current means nonlinear dac performance. You might like what this contributes, which could make a swap to a different tech sound odd / explain what you hear. Against this - you only need to please your ears of course (at least half the point of DIY).
2) No need to go mad with filter cap values, especially if that means fitting physically-large caps, or axial caps. Both of those are a Very bad idea from an RF perfomance point of view. I've seen pics of some things suggested on other sites which beggar belief - huge axial caps used just because they are polypropylene; black gate electrolytics (ugh, leaky, never the same thing twice, expensive wrong tool for job) and so forth. Poorly-informed choices. Me? I've used up to 10uF of miniature film cap in paralell on pins 13 & 18 as above; made virtually no perceptual difference. I'd rank it *a lot* lower in benefits than simply providing 'quiet', well regulated supplies as previously discussed; poor value in comparison. Probably because the inital design is so solid ...but I still had to try it ;)

3) Make sure the ground plane you tie the caps to is electrically 'clean;, not used for anything else (look at the Naim pic closely - the caps land on a low-inductance [= flat-wide] isolated ground plane that ties directly to the reference's 0v pins)

Also one's general technique needs to be clean. If playing with these caps consider swabbing the board locally with alcohol then distlilled water, remove excess flux, all that sort of thing. Food for thought: in the strictest measure, at LSB (internally-divided bit 16) the 'bit' current is just 32nA. 32-billionths of an amp. A maximum of half that is therefore the target leakage from any of the filtered pins (the top 6bits per channel) on the PCB. It's very, very, VERY difficult to achieve. An electrolytic cap on a filter pin (tantalum included) will blow that error budget by a factor of at least 1000x.

(This sort of thing is why manufacturer's hated the expensive to implement Multibit dacs and fell in love with bitstream /delta-sigma devices ...for good, bad or indifferent results.)

have fun!
 
Last edited by a moderator:
Hi Felix,

Thank you for another excellent contribution to this thread.

In hindsight I was way to hasty to dismiss the Wima MKS capacitors which I had fitted to pins 13 and 18. After a bit more listening I came to enjoy the sound with the Wimas fitted more and more.
I don't know if the new capacitors needed time to settle in or if it just took me a little while to be accustomed to the different presentation (I suspect the latter). Anyway, after a few days of listening I'd come to the conclusion that I actually preferred it to when the tantalum caps were fitted on the MSB pins. I'd describe the sound with the Wima film caps on the MSB pins as smoother, sweeter and more refined.

I have now swapped the remaining 0.47uF Tantalum capacitors on pins 12 and 19 to Evox 0.47uF MMK polyester films.

Despite liking the sound of the Wima capacitors on pins 13 and 18 I have also swapped these to Epcos 1uF MKT as they are dimensionally slimmer and fit more easily. MKT also seems to get a better press on the forums than MKS although to the best of my knowledge they are both polyester film.


screen shot software

I'm still waiting for the 15uF Oscon to arrive which I intend to fit from the -5V input pin to the -15V input pin.
 
That's all very neat!

And no, I wouldn't be surprised if it takes a few days to get a grip on what has changed and, more especially, what you think about it; one should always take time to listen critically, absorb and then trust your own tastes and judgement; even undo the last mod and listen to how it was before as a check.

... because, if you succeed in properly refining the engineering then successive differences really should get smaller, then disappear. Do stop when you're almost-entirely-happy else you'll drive yourself mad %)


(PS If I haven't put you off, I have some small 3.3uF Wima MKS-2 that might fit if you want to play, not that I'd expect or suggest they'd bring differences worth having. pm me an adddress and I'll happily post you a couple to play with. they're quite useful for general decoupling if you don't have a direct use.)
 
Last edited by a moderator:
Hi Felix,

Thanks for the offer of the Wima caps. I'm think I'm unlikely to get much improvement over the film caps which I have just fitted so I'll leave it for now thanks.

Initial listening of the new capacitors on pins 13,18,12 and 19 is very promising.

One of the first mods which I did was to remove the two muting transistors (Q2 and Q102). But what about the transistors used in the de-emphasis circuit? Are these just as bad?

As far as I understand it only some very old CD recordings are were mastered with pre-emphasis and it's pretty unlikely that I will have any of these recordings in my music collection. Indeed quite a few modern top end players do come with a de-emphasis facility any more.

I'm tempted to remove the transistors in the de-emphasis circuit just to see what difference it makes, I can always put them back in.

Here is the schematic:

picture share
 
Yes deemph is rare but I wouldn't worry about those, the way they are used the strays are completely swamped by the circuitry around the active device (Q1).
 
Thanks, no need to worry about those then.

The new film caps around the DAC sound great! Quite a noticeable difference.

I did disconnect the digital output though from the SAA7220 though by removing R228 just to eliminate the possibility of the exposed RCA connector from picking up RF noise and bringing it back into the audio board.

I dug out my spare Alpha 5 recently for another listen (the one with the NOS conversion in the first part of this thread). I much prefer the sound of the player which retains oversampling (the one featured in the second half of this thread).
 
O.K. time for another update on the Alpha 5.

Since my last couple of posts I've been having increasing problems with an intermittent hiss/crackle coming through the speakers. I quickly narrowed it down to the CDP by trying other sources.

To begin with it wasn't so much of a problem because it would only last for a minute or so after turning on the CDP and then it would abruptly stop and everything would be fine. Unfortunately it's been getting steadily worse. It was also initially only on the left channel but now seems to be intermittently on both channels although it remains far worse on the left.

I initially suspected a failing DAC chip so I purchased a new DAC removed the old one and fitted a high quality IC socket to allow me to swap them easily. Changing the DAC makes no difference (at least I now have a spare :)).

Unfortunately I think that diagnosing the fault may well be beyond me as I don't have a scope.

All is not lost though as I have my spare Alpha 5. I have therefore liberated the spare CDP of it's audio board and set about upgrading this to the same spec as the faulty audio board. I've nearly finished doing this now, just the DC blocking capacitors at the output and the 'LED trick' on the power supplies left to do.

This audio board was previously fitted with a non-oversampling conversion. I therefore had to remove the SIL strip where this non-os module was fitted and I elected to fit a high quality IC socket in its place just in case the SAA7220P/B which I had purchased turned out to be defective. I didn't fancy having to de-solder it again, I'm not sure how much abuse the solder pads can take.

My 15uF Oscons have also finally arrived so I put one of these on the underside of the board to tie the -15v and -5v pins of the DAC together.....and yes I did double check the polarity this time!

This is the audio board which I previously fitted a bunch of fancy resistors to around the op-amps and output. It's debatable if this was worth it in the first place but at least they are being used again.

I've also been busy adding a few other refinements:

The 9v toroid powering the +5v PSU for SAA7220 was still making the heatsink on the reg awfully hot so I've swapped this for a 7v transformer. This measured at 8.5vac so should still have plenty of voltage headroom when rectified to DC. I have also added an in-line fuse before the transformer, yes that's right it wasn't fused separately before, only at the plug (I know, I know). It's currently using a 250mA fuse as that's what I had in my spares pile. I could do with some advice on the best fuse rating to use here.

I've also re-routed the AC line to the clock PSU externally. It now runs out of the back of the case, underneath and pops back up next to the clock PSU.

I also added some screw terminal connectors to the clock output to make dismantling the player much easier and had a general tidy up of the wiring making sure that everything is neatly braided etc.

Here is how it all looks now (sorry about the crappy photo).


online photo storage

I currently have some Burr Brown OPA604's fitted in all four locations. I fancy swapping these to OPA627. RS seem to be offering OPA627AP at a good price at the moment. I'd appreciate it if someone could confirm that these are the correct version of the OPA627 for the A5.

I'm also looking at various capacitor options for the DC blocking caps at the output. On the other board I have some Mundorf M-cap MKP axial caps mounted vertically and squeezed in. Suggestions for suitable other caps are welcome. The best possible sound quality is the aim but I'm very limited on space.
 
OPA 627 are excellent opamps but they are verywide-bandwidth (50MHz + at unity-gain) and that can cause other problems that degrades the effective result unless real care is taken in implementation - tight, specific local decoupling, care over gain structure - and preferably, a fast 'scope available to check that things are happy.

The OPA604s are not the very quietest but do the maths and you find they are *not* a noise penalty on the A5 circuitry.
More importantly - the fet-input 604s are quite difficult to make misbehave in any way.

Bottom line - I think you have to have very specific results in mind to want to change opamps (after which, one needs to re-engineer the circuit to properly exploit the part you've chosen.) This is entirely typical of any audio device, not just the A5 output stage! (and also why I'm no fan of idle opamp swaps. In my own cd player the one bit that *hasn't * been changed is the opamps... which has been a bit of fun at bake-offs ;)
 
Hi Felix, thanks for the reply.

I was tempted to try the OPA627 because their use in the A5 seems to be reasonably well documented. Net Audio offer either the OPA604 or the OPA627 in their A5 upgrade kits. Of course the '627 option is considerably more expensive.

Having said that I don't want to throw away £80 on a set of OPA627's unless I'm going to hear the benefit. I don't have the equipment (an oscilloscope) or the expertise to be able to optimise the circuit for them.

As standard the A5 uses OP27 nearest the DAC and NE5534 nearest the output.

Is Wima MKS2 3.3uF 50v a sound choice (pardon the pun) for DC blocking just before the output?
 
Yes they work fine. 3.3uF will be OK providing the preamp has an input impedance of at least 10Kohms (most will be)

MKS2 are polyester caps; they are a more-linear choice over most electrolytics for coupling duties, but given a free hand I'd prefer polypropylene for that purpose - staying with Wima for a moment, try the MKP4.

And here's a useful thing - I've found that Maplins are selling a grab-bag of misc. film caps at the moment that usually includes a pair of the large, red, MKP4 in 2.2uf (these should be obvious when you have a look at the pack!). It also contains a bunch of other small polyester caps useful for decoupling/bypass uses, and also 10-20 axial polystyrene caps in very useful values like 1nF. For £4.99 it's a bargain for the bits-box - I have bought a couple in the last 6months ;)

(2.2uF is fine providing your player 'sees' a load of say 22K or more. )
 
felix said:
OPA 627 are excellent opamps but they are verywide-bandwidth (50MHz + at unity-gain) and that can cause other problems that degrades the effective result unless real care is taken in implementation - tight, specific local decoupling, care over gain structure - and preferably, a fast 'scope available to check that things are happy.

The OPA604s are not the very quietest but do the maths and you find they are *not* a noise penalty on the A5 circuitry.
More importantly - the fet-input 604s are quite difficult to make misbehave in any way.

Bottom line - I think you have to have very specific results in mind to want to change opamps (after which, one needs to re-engineer the circuit to properly exploit the part you've chosen.) This is entirely typical of any audio device, not just the A5 output stage! (and also why I'm no fan of idle opamp swaps. In my own cd player the one bit that *hasn't * been changed is the opamps... which has been a bit of fun at bake-offs ;)
Click to expand...

Bloody good post!
 

Latest posts

Back
Top