advertisement


NAP 250 PSU regulation

Jeff Young

Jeff Young

pfm Member
I've found a lot of information on split PSUs (a la HackerNAP, Voyager, etc.), but very little on the original NAP250 regulator.

If one uses a C-multiplier on the FE, is there any value to something like the original regulator on the OS? Or was its only value to the OS the current limiting?

Thanks,
Jeff.
 
Good question Jeff.

In my opinion the regulator gives the 250 its refinement. It also constrains it, which is why the decent unregulated amps can sound more fun.
The same refinement can be had by filtering the front end as you say. This is why an unregulated amp with FE filtering like the NCC/Hackernap/our Phoenix boards in it sounds more refined than a stock 250. With a decent raw power supply behind them they can walk all over a 250, and often a NAP300 too [We actually did this to a Naim sales rep - oops]

The current limiting sort of works. It's too slow to stop shorted speaker leads killing amps in my experience though.
 
I was thinking of doing a set of boards with MJ15003s and a set with BUV20s so I could compare them.

But then I might as well do a couple different regulator boards too. I'd like to compare split supplies vs. single regulator, so one will be the standard Naim design. For split supplies, is it worth doing both a C-multiplier and a shunt regulator (or are they going to sound the same)? Has anyone tried a shunt regulator with the front end?

Cheers,
Jeff.
 
The BUV20s are obsolete now. However they are much better than the MJE15003 in this application if you can find some real ones.

A lot of this has been done before with the Hackernap and NCC200 threads - you'd do well to have a trawl through them.
 
He he, I don't think I've built an amp yet without at least 1 obsolete part. ;)

I think I read most of those threads, but I'll go back over them. Thanks for the pointer!
 
Here's a relevant post I found from the "Improving NAP135 Transient response" thread:
martin clark said:
.... Filtered/regulated supplies to the small-signal & VAS stages are the way to go first. Since an emitter-follower-based output stage has very high PSRR intrinsically - as PigletsDad has repeatedly posted here over the years - the effort to provide a high-current reg here subsequently adds little except poor-value complication, cost, etc - to no significant benefit. If you have the budget for such parts, there are very likely better ways to spend it.
Click to expand...
 
Good thread that one. It was my Hypex amp that sent John off on that particular mission! He still drops in from time to time so maybe will be able to answer some questions. I know he's rather busier with work than he was at that time though.
 
Thanks Stefan. It was indeed your Hypex Amps that prompted my changes to my HackerNap front end regulation. I've stuck with my rather extreme multi-layered front end regulation for a couple of years now and am very happy with it. I have built more than a dozen Super-regs since for my DAC experiments and am certain that if properly implemented they are a substantial improvement upon all available LDO regs including the LT3042/5.

I found it quite helpful to get a feel for the rate of improvement as the front end regulation was upgraded by extended listening to each config on an Ebay clone board with BUV60 output devices as follows:

No front end regulation
then
RC filters from back end supply
RC filters from separate front end supply
RC filters from front end supply with choke regulation
RC filters from front end supply with dual choke regulation, and transformer snubbers
Add cap multipliers (I used Hackers boards)
Add Fetlington mod (or better FETs that are now available)
Exchange Cap multipliers with HV ALWSR's (DIY Audio also have basic super-reg boards without the pre -reg)
Improvements to the Super-regs.

Every stage brought a sonic improvement. About midway through this process I changed to the better HackerNap layout which doesn't use the T03 devices. I do prefer BDY58/BUV20 and BUV60 output devices though and tried these mounted off board. They really need to be on the board though, nice and close to the driver circuitry. Jeff, this is (just?) one area where your design can surpass the HackerNap.

The process above took a long time and didn't end up with a presentable finished product. It is too big (particularly the front end supply) for a single box stereo amp. Mono is a must. Cases are pricey. The process was very rewarding and I have a good feel for how much further improvement is available vs the effort and cost it will take. I sold my NAP135's after I added separate front end supply as they were not as good, compromised by combining the front and back end regulation. In Naims defence there was no room for separate transformers until they changed to two box power amp.

The alternative route of course is to design a board from the outset to incorporate all of one stage of this process above. I didn't find that the law of diminishing returns was kicking in until the very final improvements to the Superregs so I would recommend the last.

My ultimate NAP 250 would use T03 devices (BUV60's), on board modified ALWSR regs for the front end with much of Hackers layout and carefully routed star earthing. Its choke regulation would be improved by using a constant current source/basic shunt regulator before the superreg. It may also have 500VA R core transformers rather than just the 1200VA toroids that I currently use!

If a compromise had to be made I would suggest skipping the front end dual choke regulation as this is both costly and takes up a lot of space. I wouldn't skip the cap multipliers followed by the improved super-regs. They are the most cost and space effective improvements you can make. Very much depends on what you are trying to achieve.

I might have some BDY58R's kicking around un my garage which you can try and LesW occasionally sells them. Naim used specially selected ones. I used to use them in fork lift truck motors drives.

Sorry to babble on.....

John
 
Hi John,

Great to see someone has blazed some of the same trails in front of me.

Did you find much difference between the BDY58, BUV20 and BUV60? I have some BUV20s on hand (an odd number of them, so I plan to crack one open to see if there's anything in there that shouts "counterfeit"). But I'm also planning on building 3 versions to compare.

I'm not sure how far I'll go experimenting with front-end regulation. I'm also building a pair of JFET/MOSFET/MOSFET Pass-inspired monoblocks with shunt regulation for the front-end. They're going to burn about 350W at standby, so I can run the shunts hot and still be in the noise inefficiency-wise. They use a 3-board design, with the shunts and FE on one board and one OS board for each side of the chassis.

Do you know what Naim was selecting for in the output devices? I suspect Ft, given Vereker's preference for speed over linearity?

Cheers,
Jeff.
 
For those of you running Naim's regulator, I did some SPICE simulations to find out how critical the spec on the 10uF electrolytics is. Annoyingly, Nichicon doesn't give ESR specs for their VX caps, so we can only guess that it's higher than similarly-sized low-ESR caps, which are around 1.4 to 1.6ohms.

If it's as high as 2ohms, the phase margin is 20º. That's pretty awful.

At 1.4ohms things look pretty good at 40º.

Increasing the capacitance also lowers the ESR. A 56uF/50V Nichicon PW clocks in at 0.4ohms. You can now drop the compensation cap to 150pF (which will improve residual ripple) and still have 40º+ of phase margin.

How about a film cap, down in the 0.02ohm range? Negative phase margin and near-guaranteed oscillation. :eek:

The "high water" mark for phase margin is around 0.16ohms (at least according to SPICE). You can get near 40º of phase margin with a 100pF compensation cap, which drops residual ripple from 240uV to 100uV.
 
Naim's 001 transistors were made for them by Semelab. They weren't selected at the factory at all. All I know about them is that they're similar to BDY58 and probably based on the same die after they were made obsolete.
 
The 0R1 rail resistors in the regulator are just there so the current limiting can measure the current, right? If I have no current limiting, I can delete the 0R1 resistors as well?
 
Jeff Young said:
Did you find much difference between the BDY58, BUV20 and BUV60? I have some BUV20s on hand (an odd number of them, so I plan to crack one open to see if there's anything in there that shouts "counterfeit"). But I'm also planning on building 3 versions to compare.
Click to expand...

I'm afraid I never managed to do a direct comparison apart from changing a pair of boards from Naim 001's to BUV60's and really preferring the latter.

J
 
John, what did you end up doing to the ALWSRs in order to get them to do 50V-ish output? IIRC the standard ALWSR tops out around 35V or so.
 
I just reduced the feed voltage to the opamp with a filtered 36v Zener with a 50v+ npn follower. Your gyrator boards are quite suited to this role! I did have to cut the track from the output to the opamp +ve supply terminal to do this. I also increased the voltage of the Zener that connects to the base of the power transistor to keep the opamp output at about mid supply (18v) and used 63v caps where necessary. A cap multiplier on the input slows down the startup and prevents more than 37 volts appearing across the LM317 pre-regulator. Has worked reliably for more than two years.

John
 
I can't recommend enough there is a benefit from filtering the supply feed to the opamp on a Jung-stye superreg, anyway (and have done here previously.)

It doesn't take much - handful of ohms, handful of uF; for the usu AD825 (which remains a fabulous error amp for this kind of thing) 3.3R/3.3uF, or convenient values to at least this equiv time constant.

Bonus points for anyone who spots why ;)
 
martin clark said:
I can't recommend enough there is a benefit from filtering the supply feed to the opamp on a Jung-stye superreg, anyway (and have done here previously.)

It doesn't take much - handful of ohms, handful of uF; for the usu AD825 (which remains a fabulous error amp for this kind of thing) 3.3R/3.3uF, or convenient values to at least this equiv time constant.

Bonus points for anyone who spots why ;)
Click to expand...

The only thing I can think off is falling PSRR with rising frequency for both the op amp and the whole regulator going into a "snake eating it's own tail" ie the op amp is powered from the output of the regulator which it controls so as the performance of the regulator falls with frequency, the purity of the voltage to the op amp falls, which makes it less able to give clean drive to the pass transistor, which....
Related to this is that really fast transient load fluctuations (spikes) would be prevented from getting back to the op amp.

Edit: and maybe as a stability enhancement as when the op amps PSRR gets bad enough the +V pin will become an input in effect...
 
Gold star. AD825 gain falls from 15Khz, so its own PSRR curve follows (as the datasheet makes clear) - and a little bit of filtering obviates that effect to useful benefit in a couple of ways IME.

f=1/ (2.pi.f.c) for values I suggest = 15Khz near enough; it works.

Same is true for any opamp, but the AD825 starts with less open-loop gain than many... *

:D

*(well actually it doesn't; the design trades some LF OL gain for flat OL bandwidth, which is how it has 'flat' gain out to 10Khz+, not that it matters...)
 
You must log in or register to reply here.


advertisement


Back
Top