advertisement


Superregs are rubbish inside my CD3.5 ... why?

nickcase

Enlightened Member
I am trying to understand why for me the results of replacing the LM317's powering the analog stages of my CD3.5 with SR's (as described on Martin Clarke's page on Acoustica) were so poor that I have since removed the SR's and gone back to the original LM317's.

With the SR's in place the sound completely lack life and vitality. All the 'noises' seemed to be there however something was not right. After going back to the 317's the sound was back, instruments were instruments again and music returned.

This was not what I was expecting and clearly differ from others experience of this modification. Something must be wrong with my implementation and I'm polling for suggestions.

My setup:

External Dual rail OSCAP with one SR per rail set for 27.3V.
Custom 'mains SNAIC' PSU interconnect.
Link plug removed (of course)
SR's inside the CDP set for 17.2V each.
Measured output of LM317's (now back in-situ) 18.3V
All analog stage op-amps are OPA627's
The rest of the changes are on the digital and mains power side.

Thoughts:

1. Are the SR's in the *CAP and those inside the CDP conflicting? Are there any amendments to the SR's I should make when 'daisy chaining' them like this?

2. Is the 1.1V difference between the SR Vout and LM317 Vout enough to cause this problem?

3. Not checked the data sheet but maybe the OPA627's require a greater Vout than 17.2V?

4. I've gone deaf.

Things I'll try:

1. Putting the CDP back on the link plug (i.e. disconnecting the *CAP) and refitting the internal SR's. Will rule out SR -- SR interaction.

2. Changing R8/R9 on the internal SR's to increase Vout - 19V? more?

3. Changing LM317's for LM1086T's. According to the data sheets I need to add approx 22u - 150u of low ESR capacitance across the output for stability. I like this option as I've parts to hand and it'll be quick to do. Any body tried this?

Help, comments, abuse, all welcome.

Nick
 
If you just remove the LM317s and solder the output of the SR into the vacant output hole of the 317, doesn't that take the SR output to ground via R1 and R2 of the 317 circuit?

Perhaps removing R1 and R2 (if not already done) would improve matters?

Or, I could have it all wrong and be talking bollocks!

Merry Christmas!
 
Can't find a mention of removing R1 & R2 anywhere else but looking at the 317 data sheet what you say appears to make sense to me as ADJ is no longer connected.

Lifting the leg of R1 would be enough to take the old 317 resistor circuit out of play.

Or, maybe talking bollocks is infectious!

Still, it all sounds superb anyway - especially since the wife and kids are out at the moment I can realy shift some air :)

Nick
 
I'm a bit puzzled because it definitely should be an improvement!

On the voltage ouptut front I ran mine at 17.2ish and moved up to 24v when I went to 627s - purely because twiddling the Hicap I use easily provided a 30v 'raw' supply. There is the very slightest of improvements with upping the supply voltage , but it's not enough to worry about. Similarly, the cascading of SRegs should not be an issue.

This really leaves implementation. Are all the 0v connections ot the SR boards connected together at ONE point?
sr-connections.jpg
The ideal spot is on the ground trace running between the two audio output tantalum coupling caps ( I need to update my page with better photos!). Perhaps you should try the SRs without remote sensing - the improvement it offers is more theoretical than tangible, and it is a bit of an invitation to oscillation - which, at a guess, is the problem from the description you give.

PS removing R1, R2 really doesn't matter. If anyting, leave them - they draw a little bit of current which is never a bad thing for getting regs into the 'sweet spot'
 
Puzzled me too Martin :confused: Dissappointing too as I'd made a lovely heatsink and mount plate for the SR's out of some salvaged steel.

I did indeed have the 0V returning to a single point however I had them returning to the top of the rightmost tant leg in your picture. I can't see this being the problem but I'm willing to try it in the alternative position you describe.

Firstly I'll re-fit the SR's and run it off the link plug (i.e. without the *CAP) and then I'll try knocking of remote sensing as I like the sound of your oscillation guess.

Before all that though I'm off for a Christmas curry :)

Nick
 
I did indeed have the 0V returning to a single point however I had them returning to the top of the rightmost tant leg in your picture.
That works fine too - the star-ground principle is the important thing.

Before all that though I'm off for a Christmas curry
Ahh - far more more important. God rest ye gentle popadom... ;)
 
nickcase said:
Never tried. In what applications have you tried and preferred SR's without the TPR?

Nick

Powering a preamp (Jung nested f/b scheme using AD8066 and EL2001s), I found the SRs were rather clinical and tended to revert to discrete shunt regs. Decided to try the SRs without the TPR and thought the sound was more relaxed and coherent - very similar to the shunt regs.
 
nickcase said:
All analog stage op-amps are OPA627's

Martin will know best here, but you need to be careful with the PS when using 627's. These opamps are very fast - slew rate of something like 50V/us, and like any fast opamp, care needs to be taken with power rail decoupling to avoid the possibility of instability. SR's are also built for speed, so they are more able to unleash the full potential of the 627's, compared to a 317, but some extra care may be needed with the power feeds.

The spec sheet recommends using 0.1uF ceramic cap's close (very close!) to the power supply pins. Check what's in your CD 3.5 at the minute - there may be none at all.

HTH

Mr Tibbs
 
Thats a very good point, but I've had no problems with the 3.5s existing decoupling here. There are 0.1uF film caps very close by each of the powerpins, which go straight into the ground plane; not worth messing with IMO. Also no probs with these caps being 'seen' by the SRegs.

BTW a nice naim touch here - the ground 'planes for the analgue stage are actuallly split depending on whether they 'earth' signal or power decoupling caps.
 
martin clark said:
Thats a very good point, but I've had no problems with the 3.5s existing decoupling here. There are 0.1uF film caps very close by each of the powerpins, which go straight into the ground plane; not worth messing with IMO.

I have added some 1uF film caps directly to the pins of the OPA627 and got oscillations from the local S-R.

But...
Changing the 1uF film to 4.4uF tantalums and adding 5 Ohm resistor near each of the OPA627 (By cutting the PCB rail) solve the problem, and made a big improvment to the sound of the CD3.5.

Avi.
 
To add my two penn'th worth.

I have similarly modded my CD5: all OPA627 (with 0.1uF blackgate nx hi-q directly across power pins) and local S-R's running 18.0V o/p in local sense mode. The whole lot is powered externally by a standard flatcap2 (planning to add S-R's into this over xmas to give 24V o/p rather than standard 22 and a bit).
S-Rs seem to run OK even though voltage headroom is a little lower than recommended, (I tested them on the bench before installation and LED's and o/p seemed steady at 21 volts plus).
Well to be honest I felt the S-R's gave only a very little improvement over the standard 317's though certainly no-worse, just not much bang for the effort. However, just out of interest I tried running the CD5 off the link plug the other night. Now I expected this to have little effect as the S-R's should be doing most of the work and the modest flatcap wouldn't be adding much, right?
Wrong! from the link plug the sound was very thin and tight (and to be honest worse than I remember it being before I did any mods!), the flatcap now seemed to make MORE difference than before! Confused? well yes as the local supply inside the CD5 is the same 22 and a bit volts as the flatcap.
Well maybe the S-R's are on the verge of oscillation or something and the cleaner external supply is helping out here - I'll try and investigate further, but as the sound is fine from the flatcap I'll pursue the S-R into flatcap mod and report back any findings.
Stu.
 
S-Man said:
Powering a preamp (Jung nested f/b scheme using AD8066 and EL2001s), I found the SRs were rather clinical and tended to revert to discrete shunt regs. Decided to try the SRs without the TPR and thought the sound was more relaxed and coherent - very similar to the shunt regs.

Only just digested what you said S-man. This is interesting as I have just now added the S-R's to my flatcap2. Firstly I tried just the CDP output only and whilst there was some significant improvement in detail and transparency and good old boogie I felt the sound was a little bright and certainly not as relaxed as before. Actually I felt the S-R's in the flatcap made more of a difference than the in the CD5.
Not content with that I then also added S-Rs to the amp output (in my case a humble nait 5) and this time a more modest improvement (perhaps surprisingly?) in detail etc but thankfully not obviously brighter.

Well I think I'll let this little lot burn in for a week or two then perhaps I'll investigate removing the TPR's on the CDP supply side. Nick - for what it is worth I am using LM1086's in the TPR and silmic 2 caps where possible.

Stu.
 
Nick - for what it is worth I am using LM1086's in the TPR and silmic 2 caps where possible.

Hmm...

Silmics are great for some places, but I wouldn't use them everywhere, specifically C2 wants to be a low ESR type (Oscon's are great here), can you be more specific about the actual build standard / component choice?

The LM1086 wants moderate ESR on it's output, not low ESR, so a Silmic should be OK here (47u, C1).

Is C7 fitted (I'd recommend leaving it out)?

Also wiring / earthing is critical, I'd stay away from remote sensing in the first instance, since this makes things even more critical. Remote sensing of the ground lines can be beneficial though.

Are you using AD825's?

Andy.
 
Hi Andy,

OK the build standard:

AD825's
C1,C4,C8 47uF silmic 2's
C2 Rubycon ZA 100uF
C3 Rubycon ZA 22uF
no C7
they are wired in local sense mode both gnds starred to tantalum gnd leg on main board.

I'm guessing now that C2 might be the issue?

By the way after more extended listening the improvements are definitely outweighing the downsides (via flatcap) though perhaps more of an "olive" sound than 5 series sound.

BTW if I need Oscons, is Audiocom the best/easiest supplier, to be honest I was being a bit lazy when speccing the caps as I already had the Silmics (courtesy Richard Hallman) and the rest I could get from Farnell.

Thanks,
Stuart.
 
Stuart,

The ZA is low ESR, and should be fine, I doubt it's the problem. Try changing C3 for a moderate ESR part - even a tantalum is worth trying for a test.

I often use Elna Starget (ROD) there on it's own.

What about C5 - not fitted? I often leave it out (C3 is in parallel), but you can try adding another cap here, usually less C = better sound (within stability limits, 22u being a recommended minimum), unless you have a stability problem, which is very rare with an AD825-based unit.

It may also be worth trying a damped choke (something like a few uH - 3.9uH for example) with a 1R in parallel, in series with the regulator output (at the regulator). You can try something like 15-20 turns of enammelled copper wire, wound around an 8-10mm drill bit, with the 1R in parallel.

Andy.
 
Andy

Only just getting around to looking at sorting out this implementation and I'd like your comments upon the SR build - I've used:

C1,C4, C5, C8 = Elna Starget 47u, 35V
C2 = Rubycon YXF 100u, 10V
C3 = Elna Starget 22u, 35V
C7 - not fitted

Question:

In an oscillation prone environment is my use of YXF's a likely cause?

Previously I've used ZA's for C2 - they are much lower impedance than YXF (0.15 vs 0.9 at 100Khz) - never had problems there.

Things for me try now (in order):

1. Use local sensing on SR output (currently remote sensing)
2. Use local sensing on OV (currenty remote sensing)
3. Change C2 for Rubycon ZA's
4. Place handwound inductor and 1R in series at the regulator.

Anything else worth trying or is that about it?

Thanks in advance.

Nick
 
Nick,

That all seems fine to me, I'd definitely remove remote sensing for the moment on the output side and swap 3 and 4 for priority, I doubt C2 is the problem.

In fact if it were me and assuming the 0V / 0vsense wiring is neat (i.e. the two wires are twisted or run closely parallel to each other and go to a sensible location, the 0V connections for the LM317 resistors is as good as any to start if Naim did the design correctly, I'd try 1-4-2-3.

Andy.
 
Thanks for your input Andy.

I am pretty anal when it comes to wiring so I am happy with the 0V / 0Vsense arrangements.

I'd hope to get the oscillation sorted before I start on option 5 which will be tweaking the OPA627 decouplings.

Cheers

Nick
 


advertisement


Back
Top