advertisement


Naim CD3.5 mods (not on Acoustica)

Nothing done with the 3.5 of late as I have been busy with amplifiers and speakers. Still running all the power supplies (4 transformers, 6 secondary windings) , alwsr, flea powered tentlabs clock, flea on dac pin 28, TPR4's, cap mulitipliers, acoustica style LED modded LM317's.

Even the PFM mosfet gyrator (piglets dad) is still doing sterling service in front of all the LM317's in the digital section prior to the dac. :rolleyes:

Thoughts have returned to the 3.5 of late. Started reading diyaudio.com again. I do it in fits and starts as much of it is over my head and fatigue sets in. Anyway, I was reading some random thread when Jan Didden popped up in the conversation and mentioned his fairly new silent switcher psu. I remembered Jan had something to do with Jung super regs and ALW used to mention him from time to time.

So, been looking into the https://linearaudio.nl/silentswitcher which is a switch mode psu specially designed by Jan for audio. Thinking about giving it a try in my 3.5 for what ever it is suitable for. The thought of getting rid of all those linear psu's (instead of adding yet more) is quite refreshing yet I am sure there would be some problems to over come.

Also I worry about how long a SMPSU will last compared to a linear one which can still be going 30 to 40 years later. Also have to get past the bad reputation these switch mode things have gathered with regards to dumping noise on the mains and RFI etc.

Seriously considering getting one or two :oops:
 
Putting a silent switcher on the digital gubbins before the dac is not impossible because the switcher can handle the load. The problem is two of the LM317 regs on the back panel output 9V for a couple of TDA7073AT power driver chips and one of these regs also acts as a pre reg for the one in the swing out tray (which has 5V output).

I need to figure out if anything bad would happen from lowering the input voltage to these two TDA7073. The data sheet says 5V is typical and anywhere up to 18V ! Even if I could do that then that would leave the reg in the swing out tray with no voltage overhead to work with. The silent switcher benefits from being as close to the load(s) as possible and that isn't going to happen if I rip out the reg in the swing out tray and power directly from the switcher.

On a positive note, the SS will definitely be suitable for the dac, clock and anologue stage. I am loathed to pull a flea powered tentlabs clock and feck about with it so maybe I would stick to just powering the output stage and dac with the SS.

Cant decide what to do. It may even turn out to be a step back from what I have now.
 
I’d not bother for the digital bits before the dac; someone here - might have been Ced - tried using superregs for each of those supplies some years ago and reported, yep, no change (as you might actually hope...)

[NB the digital filter in this later player has none of the psu sensitivity of say the old SAA7220, because 1) it's a modern SMD package with pins optimised by location for decoupling and 2) does not perform any clock division or generation ]
 
Hey Martin.

Thanks for the input. Yes i realise the digital part before the Dac is said to be immune to psu improvements. Ced had the alwsr before the lm317's which he thought might be messing up any improvement. The silent switcher has some pretty awesome regs on its output so would mean bypassing all the lm317 in the 3.5 and who knows what might happen. A lot of work for possibly no return.

Apart from any possible sonic improvements (unlikely) it was a way of getting shot of a linear psu and all the space taken up, emi and rectification noise that goes along with it. But then maybe better the devil you know.
 
I’d be very interested in knowing the results of you trying...but seems a lot of work for questionable/unknown gain? :)
 
Hey Richard.

I intend giving them a go for something just not entirely settled on exactly what. The SS is a split rail affair (on one of its outputs) and unfortunately the CD3.5 is positive voltage only so half of its current capability is wasted in this case. At +/-24V the available current is only about 100mA so if only using the positive half then that is 50mA. As the voltage comes down the available current increases and at +/- 5V it is about 1A . Current draw of the CD3'5 purely digital section (before the dac and not including flea powered clock)) is roughly 400mA max and the reading is up and down all over the place. That includes all the LM317 and the current they draw for house keeping. Take all the 317's out of the equation and the current required will be lower but not sure how much.

The fly in the ointment are the two back panel regs with 9V output. If they could be reduced to 5V output without anything going bang then jobs a good un. Except the LM317 in the swing out transport draw which is 5V output and pre regged by one of the 9V output regs. Pulling the LM317 in the pull out draw and using a SS many inches away will be no good at all I should think.

The other option is to set the SS up for +/- 12V and leave all the local regs in. That will give enough voltage headroom for the 9V LM317's and maybe the available current from the SS will be enough at that voltage. Obviously the LM317's are no way as good as the SS regs but then if they have to stay they have to stay. I could always do some two green LED acoustica style mods on all the 5V digital LM317 like I have done on Dac supply pin 1 and 10.

The Dac power supply is a prime candidate for removing all regs and powering from the SS. Clock also but not so sure about ripping out a flea.:( That leaves the anologue output stage opamps, relay etc. At 18V (which is what Naim ran them at) a SS would have enough current available to make it work.

So I could run the whole digital section from one SS at 12V plus the other SS output (they have two voltage out points) at 5V could do the clock for example. A second SS could power the whole anologue stage and its other output could power the dac. In that case there would be no transformers at all on the CD3.5 and no connection to mains !

Yes it is a lot of work for unknown results but the prospect of replacing four psu's consisting 500VA toroid, 200VA toroid and two 5VA pcb mount EI transformers with a pcb no bigger than a Naim 321 board is very appealing.
 
martin clark said:
I’d not bother for the digital bits before the dac; someone here - might have been Ced - tried using superregs for each of those supplies some years ago and reported, yep, no change (as you might actually hope...)

[NB the digital filter in this later player has none of the psu sensitivity of say the old SAA7220, because 1) it's a modern SMD package with pins optimised by location for decoupling and 2) does not perform any clock division or generation ]
Click to expand...

Thread here https://pinkfishmedia.net/forum/threads/cd-3-5-mods-part-2-diy-‘xps’.5557/ from CED with his glorious failure in trying to improve the digital and transport psu

Another one from CED here https://pinkfishmedia.net/forum/threads/a-non-os-naim-cd3-5.5043/ .

Essential reading for you 3.5 bodgers out there. I only came across the non-os one recently. The TDA1305 is seen as the bottle kneck of the 3.5 so why not replace it ? Not much of the 3.5 left at that point but hey you will have had a lot of fun by that point. :)
 
Arf!
Glancing at that second thread - 14years ago! - then I moved to a TDA1541-based device, and forgotten most of what I knew about the 1305...
 
Yes Martin, 14 years ago is an age. I had not even joined pfm at that point, didn't even know what a resistor was.

You still use a CD2 ? or CDS2 ?
 
Don't you just love the glow of green led's in the evening :D

P1000154.JPG


I intend to do the acoustica green led mod to the rest of the 5V LM317 regs in the player. Not so much still in circuit in this 3.5 anymore after the recent bypass of the anologue stage (see here https://www.pinkfishmedia.net/forum/threads/bypass-cdp-anologue-output-stage.220308/) .

There are 3 LM317 at 5V output for the decoder, one for the RAM chip, one for the swing out tray sensor and one in the swing out tray for the laser. Cant wait to see all them green led's light up like a christas tree. There one on each flea as well :)
 
Still enjoying the sounds from my 1st order output stage :D

Further reading of the data sheet suggests that the dac has a Zo (dynamic output impedance) of 10R, output load resistance of 3k and output load capacitance of 200pf.

Not sure how the output resistance and capacitance effect anything i put on the end of this dac. The output impedance of 10R (which just so happens to be the figure Naim quote as the Zo for the whole cdp) suggests to me that i dont need a buffer as 10R is low and should be no problem with a resistive volume control like a stepped attenuator.

Please some one put me right if i am wrong about this.
 
Hi Stu,
Can you detail the LM317 led mod for me please? I get the idea but I’ve not come across this one before.

Cheers mate,
Stephen
 
It's easy: instead of a resistor to 0v on the Vadj pin, try 2 cheap green LEDs in series instead. the resistor from Vout to Vadj can remain anything convenient, but 220R down to 100R is nice if you can spare the current (gives about 5-10ma through the LEDs)

Why?: because at this range of current, the dynamic impedance of the LEDs is about 20ohms broadband, and flat, down to DC. Which means the '317 performs at the very best of its datasheet capability (as the 1.25v reg it really is, just stood-off by c 3.9v from two green leds) without some of the in-band phase shift effects that come from using a large bypass cap instead. And it's cheaper than a large bypass cap!

(longer waffle - start here: http://www.acoustica.org.uk/t/3pin_reg_notes1.html )


NB: another simple tweak that does need mentioning, esp with dynamic loads: ensure the close local bypass cap on the Vin pin of the LM317 is of a decent size - 100uF say. While the datasheet names 0.1uF as a recommendation this is the minimum for stability, not best performance: in actual fact, load steps on the reg output are obvs reflected through to the input as a current demand not much attenuated: so a reasonable-size, local input reservoir really does help the dynamic response. Try it and see.
 
You must log in or register to reply here.


advertisement


Back
Top