Mods to Naim CDS3

[Jo Sharp]

Hi Ron,

By the way, how much do the chokes cost?

John

IIRC they were 30H...and around £300

 

[James]

A question on one of the above comments: why would a (tantalum) capacitor of a given rating have faster transients than a.n.other type?

Just to piggy-back this question, how are tants used on Naim amps? As coupling caps, bypass caps or something else?

James

 

[martin clark]

(James - both for coupling and bypass, i.e. decoupling).

Tant caps for all their folklore faults have one very, very nice feature: the ESR - often an ohm or so - dominates the response. This means that rails decoupled using tants will rarely 'ring' - the stable ESR provides very useful damping well above audio and into the HF (because the cap intrincially is pretty low-inductance). And nicely-damped things rarely display unwanted overshoot or 'ringing' as can easily be provoked for example by unthinking exchange for film caps (with very low-ESR)

I suggest if one wants something that might contribute to 'a sense of musical timing' it surely includes 'good time-domain response'...

 

[john.luckins]

A suitable series (or parallel) resistor will damp the ringing of any capacitor, including a film cap. Now you don't see that being done very often, but it would allow them to be used say in the load or bypass of a 3 terminal variable regulator such as a LM317. Film caps can be wound or stacked to ensure low inductance as well, though at a price.

I wonder how a properly damped low inductance film cap would stand up sonically to a Tantalum in the Tantalums optimum environment? I suspect it will be both non polar and more linear and thus be technically better. Its time domain response could be made to be the equal of the Tantalum, but would it actually sound better? I know it would be much more expensive.

I believe that Tants are very good value for money in the right places just as Naim use them, but even there they are not the very best possible solution, just the easiest.

John

 

[James]

(James - both for coupling and bypass, i.e. decoupling).

Tant caps for all their folklore faults have one very, very nice feature: the ESR - often an ohm or so - dominates the response. This means that rails decoupled using tants will rarely 'ring' - the stable ESR provides very useful damping well above audio and into the HF (because the cap intrincially is pretty low-inductance). And nicely-damped things rarely display unwanted overshoot or 'ringing' as can easily be provoked for example by unthinking exchange for film caps (with very low-ESR)

I suggest if one wants something that might contribute to 'a sense of musical timing' it surely includes 'good time-domain response'...

Thanks, Martin.

How might the substitution of film caps with tants at the input and/or output of my active crossover change the sound?

James

 

[martin clark]

I use a small-value series R a lot. Adding parallel R would usually imply horrific power dissipation for the same effect. (Thevenin doesn't care, so might as well save the power)

I do this because I prefer the result to tants - and it'll never need replacing . Appropriate values of R are easily calculated if you know or can measure, even guess the equivalent inductance you are working with:
(LM317 example )

 

[trancera]

James,

Does your pre amp have an output cap? If so the best input cap on the XO might be a piece of resistor offcut ? Carl seemed keen on this on input caps to power amps for example.

Should be able to easily measure any DC to check.

Brian

 

[James]

Does your pre amp have an output cap? If so the best input cap on the XO might be a piece of resistor offcut ? Carl seemed keen on this on input caps to power amps for example.

I suspect not, as I once noticed a glitch when I measured DC going into my loudspeakers, and traced it back to the preamp output. That 'fault' could not be replicated since. By that measure, I also suspect that one of my amplifiers is DC coupled (otherwise, how would it have amplified DC) but the other pair probably has input caps.

I have retained coupling caps at both ends of my AXO, just so that if the fault should recur, at least my amps and loudspeakers are protected. Perhaps I could remove the input coupling caps, but I'm most keen to retain the output caps.

James

 

[trancera]

Ah I see. Thread trashing but IME different caps can change the sound quite markedly. Films going from bright to just perfect. There is something in those there tants and maybe just might suit you. (I like how you are being dragged into the electronics more )

 

[James]

I like how you are being dragged into the electronics more

Since building my first active XO, I have little choice.

It's also much easier to tweak an AXO than a PXO. I just want to know I haven't compromised the 'quality' of upstream signal with second rate mid-stream componentry.

I can't remember the values of the coupling caps, but it should be a simple matter of swapping them out to hear the difference(s) first hand.

James

 

[martin clark]

Working into highish impedance, like coupling caps in an AXO are, will be much less sensitive than passives in a PXO. If it sounds good.. it is

 

[rontoolsie]

With the chokes located between the rectifiers and the main filter caps, there is no measureable (within two decimal points) change of the voltage either going into or coming out of the voltage regulators. What however does changes is the amount of measured ripple coming out from the voltage regulators (down to about one tenth of the original) and the spectrum of the ripple. These measurements have been done both without and with load (i.e. with the CDS3 hooked up).

As you suggest, having ringing at a value much less than 1Hz is of no consequence. The ripple spectum seems to be 120Hz and harmonics.

A much more important calculation is that the choke offers an impedence of over 10,000 ohm to 120Hz ripple, yet only about 0.25 ohm to the DC it does transmit.

Hi Ron,

I've been practicing with lower power chokes on preamps and done a few calculations on your design on the XPS2.

With a 40Henry choke with a 0.25 ohm resistance and 18800uF capacitors your looking at a very high Q filter Q = 188. Normally you would want to have the Q below 0.58 here to avoid ringing. Interestingly the ringing, if you can call it that, will be at a very low frequency of about 0.18Hz. Maybe that is OK as the pulses of feed current are of much higher frequency so the voltage never gets a chance to wander.

Perhaps a bit more worrying is the effect using the choke may be having on your voltage headroom for the regulators further on. The output voltage of the choke/capacitor filter will be 63% of the voltage you had with just the capacitor filter. That may be OK but my guess is that it is too low for the regulators that follow. They normally need 3 volts or so of headroom and it would be unusual for Naim to design with much much more voltage headroom than they need here.

If I were you I would check the voltages across the regulators in your XPS2.

By the way, how much do the chokes cost?

John

 

[john.luckins]

That's good news on the regulator voltages Ron. I still can't get my head round how you achieved such a high inductance for such a low DC resistance. I fancy giving this a try myself on my preamp supply. I've had some success with a low value choke already. How did you find someone with the required expertees and what were the specifications? Maybe there are such people over this side of the pond?

John

 

[rontoolsie]

I was in contact with the lead design engineer in the largest inductor/choke factory in the world. With the specs I provided- Henry value, physical size and DC resistance- they said it couldn't be done. But the guy who did all these CDS3 mods had some interesting ideas. The prototype he built had a measurable DC resistance of 0.23 ohms, a value greater than his meter could read (30 Henries maximum) and would carry 10A plus continuous. The design is very proprietary and comes heavily encapsulated not only to provide screening and insulation, but to discourage curious eyes as to what lies inside.

I have had these 'wonder chokes' installed into XPS2s (three of them), a 555PS, three NAP500PS, two 552PS and an Epona SPS power supply going to a Prefix. The results in each case has been very large subjective and objective improvements.

Each choke requires several hours of hands on machining time, with additional time spent potting it (the external assembly is some plumbing accessory made from PVC and is good insulation to many kilovolts). The cost is appropriately high given the assembly time and the fact that no other chokes of even remotely similar specs can be had anywhere for any price.

The problem with other high value chokes is their inordinate high DC resistance (which in some power supply designs is not a terrible thing, and can add another degree of noise rejection) but more importantly their inability to pass more than a few mA of current, and therefore cannot be used between the rectifiers and the filter caps- where they seem to work the best.

Also deliberately putting these high Henry chokes in places with high current draw guarantees a very high additional energy storage, which typically exceeds the number of Joules the filter caps are storing. Think about it, the Joule value is 0.5 current (maybe 1-2 amps) times the value in Henries^2. At least in the case of the CDS3 where the head unit is pulling large amounts of current.

A commercially available Hammond choke for instance in a 30H value has a DC resistance of almost 600 ohms and a current rating of only 40mA. Entirely inappropriate to be used where I have had these chokes placed.

 

[Juhleren]

Dear Ron and all other contributers,

Very interesting reading this thread!
I´m planning on making some XPS´s for naim CDP´s and came across this thread in search for the experiences out there, and have been very taken by the reports on the effects of the chokes mentioned.

Has anyone tried jensen or DNM four pole caps, and care to report their experieces on the audible effects of these type of caps?
-Would it make sense to apply chokes before this type of caps, as their construction means separation of input and output terminals detaching the feed from the source by the film in the caps?

When I first read the stuff about the special chokes, I first thought Ron ment 30 to 40 mH as these numbers seem to be what can be found elsewhere (at greater sizes and/or slightly higher DCR, though). Can it really be that these chokes offer around 1000X what other coil types can do and keep such low DCR figures and smallish sizes? I´m really not doubting the effect of these devices, but if the specs are as claimed, they would potentionally revolutionise the coil industry, as for instance my donut shaped 8.2 mH coils on my woofer is around 4-5 inches in diameter and still only offering around 0.1-0.12 ohm DCR and not being entirely cheap either...

Any comments would be greatly appreciated.

all the best,
Juhleren

 

[rontoolsie]

Juhleren...trust me-these chokes are at LEAST 30 Henries (not mH)...the inductance meter on the DVM tops out at 30H...and these chokes are in excess of them. How much-who knows..50, 80H or even higher. I guess I could hook up 3 or 4 in parallel and then see what the reading is and multiply that by 3 (or 4) for the correct value. What I do know is that the ripple is greatly reduced (100x or more going into the filter caps and 10x or more coming out of the voltage regulators) with very low but measurable DCR. And as I mentioned before the chokes are situated between the rectifiers and the filter caps, and before the voltage regulator boards.


I too ended up with an 8.2mH tape foil inductor in my after market DBL crossover, made by Alphacore...and it was the size as you described-as wide as a saucer and weighing in at over a kilogram or three, with a cost of circa $120 USD. BTW Jensen now offer toroid inductors of that value with a DCR of less than 0.01 ohm. This potentially can greatly increase the damping factor of your speaker, and indeed in the DBL allowed for a quicker tighter bass with these in situ as compared to the Alphacore one with 20x the DCR.

As far as the main filter caps go...the standard BHC/Aerovox have been replaced by much physically smaller ones with the same nominal value but with lower ESR (Nichicon I believe) with the amount of capacitance for the digital feed doubled over the original.

 

[rontoolsie]

Wow...6 years since any mods were done to the CDS3. Anyway-one of the modded units after ten years needed a new transport. Fortunately I had purchased the last ten VAM 1250 transports when NANA dissolved, so I have a few to spare.

This seemed like a great opportunity to effect some additional mods. There are four PROMISIC CO31 caps, two of which seem to be right after the DAC.
You can see them here

So.....
Out with the old

And in with the new

The replacement caps are oil-filled 47uF and are bypassed with some much smaller value Cardas caps.

Results-Much more expressive, articulate and extended bass and pitch accuracy. For control there was an otherwise identical CDS3 side by side, that had all the mods except these latest cap replacement ones.

Next round will be to replace all 52 resistors in the signal path with equivalent value tantalum ones.

 

[Yomanze]

Just chucking in 'premium' resistors like tantalums isn't a good idea IMO - sure you can change the flavour, but different resistors can work better in different areas, and a lot of the time for a much cheaper price.

 

[rontoolsie]

I should also mention that the four 47uF caps measured closer to 25uF...which surely would cause bandwidth limiting as well as phase anomalies around their cut-off point. Both of which would significantly contribute to lowered performance. Also, it looks like all the previous images are no longer available. Eventually, I will reupload them to a current image hosting site and restore all the lost images.

 

[rontoolsie]

The Naim resistors are specced as 1% tolerance (see diagram below). But based on upgrading resistors in other Naim products-the Superline for instance-there are great potential improvements out there.

Here is the circuit the analog portion of CDS3-showing exactly where these resistors could be upgraded and the values of them.