advertisement


The modded poweramp project for Xmas- OR ‘dooo-in’ it right for my little Redbox’.

C

ced1

pfm Member
So, with xmas fast approaching, there are 5 weeks left to get pcbs delivered in time for a little modding over the winter solstice break. Crikey.
So what to do with a couple of free and self indugent days?

Well I intend to finally get the Naim (style) poweramp well and truly sorted. near enough.


Some background: the current poweramp prototype is a fine mess and all over the place. Its needs to be rebuilt and while we're at it, slip in -on the sly- the latest preamp and regulator mods which are rather nice in that application. Though it looks nothing like it, its actually just a Naim poweramp circuit with regulation. I just build my own boards in a big chassis so theres bags of space to mod. One day I'll have enough, put it in a nice chassis and settle down to rear turkeys or something.

http://www.naimmods.com/pix poweramp project/poweramp prototype top view 1.jpg
http://www.naimmods.com/pix poweramp project/poweramp prototype side view 1.jpg
http://www.naimmods.com/pix poweramp project/poweramp PSU.jpg

Dunno if can hit that deadline (I should be finishing some architectural plans tonight but sick of it), but a little distraction is welcome right now.
Anyway, tried a lot of mods over the last 20 months and the prototype is a mess. Certainly not fit to move, nevermind transport (zaaap/kabooom)- all wobbly wires and shorts just waiting for a light breeze to happen. Bye-bye cat.
However it’s really the amp and regulator boards I want to tidy up and redesign in light of some nice results on the preamp too. Though got to throw in a couple of new bits at the end as well or it'd be no fun.
Besides, its been years since there was any real update on the Naim pages on Acoustica. Apart from a little poweramp teasing, and a big one’s been on the cards for ages so thought I’d run a preview.

Now, the current chassis’ above are ‘rather large’, but I thought I'd make the ‘xmas Redbox’ (got to give it a name- first law of missions) a bit more general, and so able to fit in my baby NAP140 chassis too.

http://www.naimmods.com/pix poweramp project/The Nap140.jpg

(Am I alone in thinking EVERY application Google develop is complete and utter shite? Like the free resources; damn near unusable though.)

In fact the Nap140 will be the test bed. That means there will be some compromises involved, though I expect them to be small.
And when I think about it, I kind of like the idea of an, innocuous-on-the-outside looking, ‘Stealth NAP 140’, heh.

So, to the the Design Brief, in vague order of importance:
1) Give the Nap500 a run for its money, or beat it, sound quality wise? Not power obviously. Lets make a stab at it. I expect the answer to be yes, but ****, who knows.
2) Boards to be a straight replacement able to fit into a Nap 140. That means will fit a 250, 135, 180 also. No rebuilding desired- I don't want to fiddle with the existing psu of the 140 (this is really a proof of concept).
3) Full multi-regulation of all elements except the output transistors. That means regulation of front end, VAS, driver and biasing strings.
Sorry, but the output transistors are probably the least important part of the Naim amps in terms of regulation. Anyone who's tried front end regulation will realise this. Either that or the 250 regulators are real cows but 250s left me pretty unimpressed. Either way it would also require double the chassis real estate which is impossible to do and still meet criteria 2 so that’s for another project. As I said, there are some compromises but they are small. Also I wouldn't necessarily go for straight regulation of the output stage anyway as there are other tricks to try.
And yes, the regulation will be exceptional.
4) Spitzenclasse components. All top-of -the-line parts within my current knowledge. And yes, there are better parts than the MMKs for feedback (if the sound is to your taste).
5) Budget, maybe £250-300 for the whole shebang- amp boards, regulators, parts; both channels. This is not a psu project so we live with the power supply we already have.
6) Ability to experiment a bit with grounding, parts selection, topologies and output transistors. Got to learn something new with each build and have the space to stretch a bit.
7) Extreme ease of assembly and mod-ability. To help 6. above along. And rebuilding does get tedious after a while so anything to cut down on that time is desirable.

So lets have a look at the design criteria in a bit more detail.
There are 4 main elements to consider:
1) amp circuit
2) layout and chassis assembly
3) Regulation circuits
4) parts and components
5) pcb layout, routing, wiring and grounding

I put chassis layout as number 2 advisedly as the regulation could spread like bindweed so chassis space available will be a major constraint on design and will demand very tight layout.


1) The Amp Circuit
So lets have a look at what circuit we're going to use. Nothing very exciting going to happen here. Base building blocks are going to be the classic old Naim circuit:

http://www.neilmcbride.co.uk/output-amp2.pdf

Aaaandd lets learn a few lessons from the Avondale NC200 circuit on the way. I'm going to be quite conservative as these are marvellous sounding circuits and I am, after all, a Naim aficionado.
So lets start with the NC200.

http://www.naimmods.com/pix poweramp project/1 NC200 v1.pdf

Well, as I’m going for very thorough levels of regulation won't need the (apologies Les Worstenholm) pretty crude front end regulation. Basically on the NC200 its an RC filter to protect the the sensitive front end input and VAS sections from the worst excesses of the high current driver and output stage supply demands which make the rail pretty noisy. The added diodes prevents current suckout on big output transients.

So out goes (c2, c11, r11, d3 and their -ve rail counterparts) and we're left with:

http://www.naimmods.com/pix poweramp project/2 NC200 V2 (stripped rails).pdf

Now its no longer a NAP and its no longer an NC200. It’s turning into the ‘Redbox’, named after my first ever silver bumper NAP140 with its lone little red LED. A completely undramatic and un-earth-moving change. Ahh, fond audiogeek memories though.
Anyway, we’re basically back to the NAP circuit with 2 differences:
1) Emitter degeneration on the input differential (the 100R resistors R5 and R8).
2) the output inductor and its damper resistor (L2 and R29).
Now I remember reading somewhere that the operation of the NAP circuit is deceptively complex and works not how it seems (It certainly simulates very freakily). If I remember, the guy on diyaudio.com analysing it said that emitter degeneration would be pointless in the circuit but I didn't understand his reasoning at the time and haven't bothered to really analyse the circuit yet to try and understand it. So, I am going to leave them in. Emitter degeneration generally makes bipolar transistors more linear at some loss of gain so why not? The very respected Leach amp uses emitter degeneration so another ‘yes’ to the idea.
Also, most critically, I'm working on the principle that you can leave parts OUT if you don't want to use them, but its a nightmare to add parts in afterwards without drilling holes, rewiring, jumpers and other horrible bits. So the plan is to work it into the design if you can leave it out easily. Layout is going to be VERY tight in this project so there will be little scope for hard core bodging after the event. I shall call this principle DUMPING. There is much scope to try things out and customise to tastes and prejudices.
I'm also going to leave in the full zoble network as I want the opportunity to eventually try out different cables without worrying about stability and blowing up the amp. Again it can be DUMPed if not wanted.
Ok that’s the legacy ideas over with. Why does the Avondale NC200 sound better than a standard NAP? My guess would be largely due to the good quality Mil spec tantalum capacitor it used to use for the feedback loop (polyesters now I believe?) and to a lesser extent the crude front end regulation. No great magic when you know what works, though Les W might decide to disagree with my sorry ass.

OK, back to the circuit. The final part is to add something a bit extra. Now I don't want to do anything fancy. I could try cascoding in the signal path but in my experience so far that is 'intense', and I'm not comfortable enough in that area yet in terms of musical presentation and balance; don’t intend to trade hyper resolution for musical involvement.


...and this writeup is actually going to take a 'tad' longer than I thought, so more in the next couple of days.

Cheers
Ced
 
Good luck with your project ced - i'll be watching with interest.
I trod a similar path last year - but have now ended up abandoning the old rca architecture altogether.

Jim.
 
ced1 said:
So, to the the Design Brief, in vague order of importance:
1) Give the Nap500 a run for its money, or beat it, sound quality wise? Not power obviously. Lets make a stab at it. I expect the answer to be yes, but ****, who knows.
2) Boards to be a straight replacement able to fit into a Nap 140. That means will fit a 250, 135, 180 also. No rebuilding desired- I don't want to fiddle with the existing psu of the 140 (this is really a proof of concept).
Click to expand...

Is there really a chance of #1 when limited by #2?
 
Ced,

My understanding of the driving force for LesW's changes in the NCC200 is to improve the HF stability margins of the Naim design; the LPT degeneration is part of that, but he also tweaks the compensation round the VAS. And adding the Zobel is part of the same thrust.

Some of these changes are tied into the choice of output devices, which cannot be separated from loop stability settings.
 
To answer midcm1, yes, mulitiple regulation is the plan. 13 regulators per channel including 4 shunt pre-regulators. Also more on this shortly.

Hi Jim,
thanks for the good wishes. Yes I noticed a couple of years back you had a prototype but never pursued it. My thoughts at the time were, don't blame you when you realise the time and effort required in developing these things.

Anyway, re the RCA architecture yep, know where you're coming from. I have a couple of topologies I'd like to try in the future.
However for the moment I'm really coming from a slightly different angle in that really there are several factors involved in making a great amp.- circuit, parts, layout, regulation, grounding, wiring looms, psu, vibration damping to name most. My angle is to develop a modular system with a solid foundation of psu, chassis, regulation system and the other common elements to any good amp (which are most of them e.g Pots). With that in place, swapping out circuits becomes a doddle without having to rebuild the whole system every time. Want to try a aleph style circuit? no problem. More explained shortly on this anyway.
My other thought is really that the Naim circuit is rather good when you really put a bit of welly behind the details. That inveterate bodger Ron has tried a fair few amps and keeps coming back to his NC200s. It does something right thats for sure so I'm really interested in seeing what the max is you can squeeze out of the topology.

PG- yeah, both ideas are pretty standard fare. I'm not denigrating the mods; on the contrary, stability is always a worry which is a good reason for keeping them. What I mean is IMO neither the emitter degen or zobel are major contributors to the difference in sound between a standard naim board and an NC200. I've run both circuits and i'd put most down to a good feedback cap and the bit of front end regulation.

I hadn't noticed any change to the VAS though? the 47pf polystyrene is mainly an availability issue as 39pf items are pretty non-standard. And I believe Les drops one of the phase network polystyrenes but thats it I think. Have I missed something?
 
ced1 said:
T
PG- yeah, both ideas are pretty standard fare. I'm not denigrating the mods; on the contrary, stability is always a worry. But IMO neither the emitter degen or zobel are major contributors to the difference in sound between a standard naim board and an NC200. I've run both circuits and i'd put most down to a good feedback cap and the bit of front end regulation. I hadn't noticed any change to the VAS though? the 47pf polystyrene is mainly an availability issue as 39pf items are pretty non-standard. And I believe Les drops one of the phase network polystyrenes but thats it I think. Have I missed something?
Click to expand...

Yes, but because the open loop gain has been dropped, keeping the same value of compensation cap, corresponds to an increase of phase margin; to put it another way, the unity gain frequency is now lower, and so further from the second+third poles.

The other option would have been to have reduced the compensation, in pursuit of increased bandwidth, at a comparable stability margin to the Naim circuit; LesW did not choose this.
 
Hi PG, yes sorry with you now. Indeed Les mentioned more than once that the NC200 was far less likely to oscillate than the Naim boards (which are borderline) into wider range of loads. I've had some in the past but thankfully not blown anything I'll also have to try to dig up that diyaudio.com discussion again. IIRC it delved into some nuances (to the likes of me anyway) in terms of operation of the differential which was quite interesting. Deceptively simple was the gist of it- I'm sure you'd appreciate it though was over my head and caused a fair amount of squabbling between the EEs thrashing it out. My theoretical knowledge is probably just enought to be dangerous, heh.

Anyway, its late but a bit more before crashing out. And so, back to...

The circuit, part 2

So yes, what extra tweaks to make? Another option considerd was to use a DC servo and dump the feedback cap. However, I want something solid and mechanical like a cap that is unlikely to fail. Also to do a servo right is a shag compared to slotting in a cap. And the ones Im going to be putting in are rather good.

So all I propose to do is
1) cascode the current sinks. In effect 'superreg' the current sinks and make them work with better current regulation to higher frequency, not change them. There are 2 in the circuit; one sinking the input long tailed pair and the other the VAS so these will be cascoded. I think it was Jonathan Carr of Lyra Connosieur considered rock solid CCSs vital to solid bass reproduction. Works quite pleasantly if not dramtically on the preamp but every little counts.

2) The other thing I'm going to do is, firstly split their biasing feeds and secondly take those feeds away from (signal) ground and give both their own voltage regulation. That should decrease noise on the ground (good) and decrease any intermodulation effects between the CCSs too. Never understood the slack way these were implemented though I guess sourcing current from ground would have been less noisy than off the power rails. So I want something like this:

http://www.naimmods.com/pix poweramp project/3 RedBox poweramp V0.5 (recascoded).pdf

Cascoding CCSs
Now one caveat to CCS cascoding is it eats away at voltage swing headroom. The long tailed pair CCS is fine as there’s little voltage swing on it and it doesn’t reach down to the –ve rail. The VAS however, does.
A NAP or NC200 board uses 2 diodes giving around 1.3V off the rails- 0.65V across the current set resistor and 0.65V Vbe drop for the CCS transistor. LEDs are very low noise and additionally pretty (awww). I usually run Green ones at 5ma biasing in CCSs as they give lowest noise of around 0.3uV if I remember some research by Christer on DIYaudio.com. (Probably impedance is way more important but not tested this). But they guzzle around 2.2V of headroom. For a NAP140 we’re only running about 34V rails and green LEDs would devour a whopping 4.5v off the –ve swing. 2 ways to ameliorate this.
Red LEDS are better at around 2 x 1.7V drop. I’ve designed for LEDs but suspect the traditional 2 diodes in place will be the way to go for the NAP140 and possibly 180. That would give 4 x 0.65V
If anyone knows what voltage rails a 180 runs off, I’d like to know.
My upcoming supply (or 250s, 135s) will have bags of voltage headroom to burn for what I plan on. To be honest I suspect some low impedance series voltage references running at about 1.25V each might be best but that’s something to research another day. Capacitor decoupling around D2 and D4 would also be ideal but from first draft layouts, I don’t think I can fit them!! Also it increases the possibility of oscillation. I’ll revisit this later. Now if you look at the NC200 schematic for convenience (NAPS are the same), you’ll see the rough currents for the CCSs and the biasing string. Biasing of somewhere between 2-5 ma is fine depending on absolute power ratings of the parts, but the CCS’s must remain at their values. What diodes, leds or references you use determines the current set resistors you have to use to achieve these currents. Actually thinking about it, power rating of the biasing resistor may may have a significant part to play in the decision to bias the CCSs from ground not the positive rail. I’ll have to work it out.
Either way, CCS cascoding can be easily DUMPED if desired or does’t work out, going back to the traditional Naim circuit.

So to tidy up some points:
a) R9 and R13 are damping resistors to ensure no oscillation. These can probably can be DUMPED, but best keep in just in case until its been thoroughly tested.
b) Q2, Q7 can be just about any transistor, why not BC550s? But with the biasing strings being independently supplied, hfe isn’t so important any more. I’m pondering low capacitance, high ft items for extended bandwidth.
c) Q1 and Q6 should really be high voltage rated- at least twice the rail voltages to be safe. Really Q6 is the main worry and I wouldn’t go less than 120V rating.
d) C8. Why does it exist? My thoughts would be to damp some sort of high frequency oscillation between CCS transistors originally as both CCSs use the same biasing string in the NAP and NC200 boards. I’m leaving it in on DUMP principles but I doubt it will have any function now.

Ok shortly on to some assembly considerations and how to make the whole lot fit and then on the next and most important part of the design- multirailed regulation. This is where it will start to get interesting and includes some development on Teddyregs (gyrators) and how to get more for less.

To be continued again….
 
Lily gilding suggestions for the Redbox.

You could split R10 and R11 (say 5K/10K) and put a filter/bypass cap from the junction to the negative rail - this will wipe out any HF noise on the reference chains.

Q6 is critical, and a nice device for it might be the Sanyo 2SC2911, high breakdown (160V), very fast, VERY low cob, adequate power dissipation, unusually high Early voltage - all the properties for a wideband current source. 69p from Farnell. Would do for Q1, although clearly something smaller would work as well. The complement, 2SA1209 might be good for Q5. As a PNP, the Early voltage is lower, but still up round 150V.

They handle 1W with no heatsink, and up to 10W with one.
 
Hi PG,
yes I had considered a bit of RC filtering- always do it on my preamp CCSs. To be honest can't remember why I scrapped the idea ... scratch head. Think will have to put them back in again, space permitting. Couple of small polyprops should be all thats required as the regs should be effective out to the low Mhz range.
And bang on the head with the transistors. Was intending to go for the 2SC2705/2SA1145 but actually your idea has serious merits in terms of the high voltage. The former are only good to 120V iirc. Those sanyo transistors are excellent on spec though, hey? They have a whole family of fast, low capacitance and mid range hfe transistors which have clear use in a couple of other implementations for example gyrators :) And didn't realise farnell were carrying them so thanks for the tip.
 
....So, yeah....uhh

And I guess the final thing is to split up the power rails, so each circuit fragment is fed from its own regulator. And You end up with this:

http://www.naimmods.com/pix poweramp project/5 RedBox poweramp V1.pdf

Basically D1 to D4 may need to be replaced by 2 x 1n4148 diodes instead. Or try some voltage refs in the 1.2V area. C8 can probably be dropped. C7 could be 39pf for a bit extra bandwidth, if you can find that value part. Though you loose a bit of stability margin.
My growing concern is voltage headroom though, as the regulators will probably eat another 5 volts. So allowing a 140 voltage swings of maybe +/- 28V if I’m lucky and being optimistic. Hmmm. On the positive side my old Nait still gave plenty of welly so maybe not such a terrible issue.

Ok next to deal with the layout.


Layout and assembly considerations

1) dimensions. This project will have to be a very very tight fit to get all the components needed in, but if it is able to fit in a 140, it will fit any of the older style chassis.
Plan view: Now the naim amp boards measure about 128 by 92 mm. We have a couple of mm slack either way in the 140 chassis if need be, but not much so I'll stick to this. Actually I'm going to make the boards 132 by 94 mm as that suits panelling with my preferred pcb fab house. A few mm larger but still no problem and I want every bit of pcb real estate I can get.

The layout is going to have to be very compact because I intend to use some very very large capacitors. This will be the critical bit to make fit – they must go in.

Now there is no space for extra regulation on the boards themselves or anywhere else in the 140 so where to put those regulators? Well, there is actually space where I shove my big caps at the moment- underneath the amp boards. So its going to be an 'Amp Sandwich'. Elegant for several reasons. So lets find out the exact space available to play with as this will be the biggest constraint and may determine following design decisions.
So this is the measured dimensions in the 140:

http://www.naimmods.com/pix poweramp project/6 NAP 140 inside dimensions.pdf

Side elevations: The heatsink determines space available top and bottom as the amp board is fixed to it and this will be a straight replacement for the Naim one. The chassis sleeve has an internal height of 69mm, but the inner tray is 2mm alu, so that leaves 67mm max internal space wehn everything is snug.
So there are 40 mm above the amp board and 27mm below if I’m lucky. Now the tallest component (capacitor) I intend to use on the amp board is 39.5 mm high, which is a minor miracle though it will scrape the top of the chassis sleeve in practice. I have to get this capacitor right though as it’s the feedback cap(s) and critical to sound quality. Knowing my soldering, it won’t sit completely flush or be manufactured to the exact size either which could be a horror- a project buggered for want of a mm headroom. However have noticed the boards actually bend a bit. The solution is to BODGE this with a 'clever' bit of design (god, I love a good bodge. The pcb Mick Parry, where are you, baby?) so that it will indeed fit by sticking it at the far end of the pcb (away from the heatsinks). It flexes slightly so will bend down the couple of mm needed to make it all fit if all is not as measured. So the chassis sleeve will fit whatever. This may make a bit of a demand on the layout, but will be manageable, and rather safe than short.

So assuming there are 40mm at the top, there are a maximum of 27mm left to play with below the amp boards. Though I need to be a bit more exact here:
Standard pcbs are 1.6mm thick and I'll need 5mm standoffs minimum for theregulator board sandwich. Am sticking head in sand at fact that haven't found less than 6mm ones yet but it will work out for sure. Also I'm a bit nervous about this as the bolts holding the TO3 output transistor to the heatsink are probably 4mm deep already. If they push onto the regulator boards or the boards don’t fit I’ll have to have holes routed out to fit them. That is an area I'll have to keep a close eye on later or assembly could suddenly become very difficult. In fact its not really ideal at all so will have to think carefully about how these boards mate.
Now if the regulator board is also 1.6mm deep that gives me about 18.8mm to play with below the regulator board. Which is not much.
So far OK as my regulators will use polypropylene caps, the crucial and largest 1uf ones I intend to use are 16mm tall praise the lord. 1.5uf are 19mm which could also be an option if desperate but I'd rather not. More expensive and not necessary I don’t think.
My only other worry is the driver transistor regulator will need heatsinking so that’s another thing to keep an eye on. I was eyeing up a heatsink 25mmm tall, but will have to look elsewhere. Probably fine for the drivers though. Also worrying TO220 transistors can be from 17.5mm to 20.5mm tall according to the datasheet if perfectly sat upright on the pcb. Its getting trickier again and I really don’t want to have to make up my own heatsinks out of alu bar if at all possible. Some number crunching to do in terms of dissipation and worry about shorting the case to chassis.

Anyway, at least all the parts look to be able to be made to fit in theory and that’s good enough for me. Critical will be being careful about soldering a handful of critical components in practice.

Onwards to:

Regulation

1. Multiregulation

Now as well as regulators of various types, there are also what I call multi-regulator systems. Without wanting to try to teach pink fishers to suck eggs, to my knowledge, not that many high end designers use this type of regulation, all considered to make some of the finer equipment money can buy.
But, working up from basics again, its easiest to look at a preamp first to get a clearer idea. The supercap psu:

http://www.acoustica.org.uk/t/naim/psu pix/supercap.jpg

We see in there are 15 regulators providing power to a preamp. The lyra connoisseur 4.2 uses 24 apparently!

As you go higher up the Naim food chain, there is more regulation and it makes sense. Snaps has 1 or 2 regulators, Hi-cap has 2 regulators, Nac82 can accept 2 hi-caps, Nac 52 has the supercap with 15 regulators and the Nac 552 psu has, dunno, lots too- Lost that picture years ago.

---------------
Now the critical concept: I like to look at a poweramp as a (high voltage) preamp with a high current buffer tagged on the end.
---------------

When you separate and independently regulate the supply going to the 6 boards in say a Nac 32, you get a big lift in sound quality. There is less modulation of the power supplies, i.e changes in current drawn by 1 board don't affect the current going to a different board so you have cleaner power to each board. The currents may only be minute, but they do have a very significant effect on sound quality.

Well you can do that to an individual board such as the 329 gain board too. So multi-regulating the gain boards means that every circuit fragment on the gain board gets its own individually regulated power supply too: The biasing network, the input, the VAS, the output, the current source.

Same can be done with a poweramp. Why? Because no reason not to get the same rewards is the theory. Thats why I've broken up the 5 RedBox poweramp V1.pdf schematic above. Now you want a regulator for each rail. On the schematic I'll need 9 regulated rails; 6+ve and 3-ve rails. And additional 2 raw dc rails for the output transistors of course.

So how does multi-regulation sound? On a preamp, I will say simply that a multi-regulation system is the way forwards. Here's a picture of an early prototype preamp with 16 Teddy reg style gyrators.

http://www.naimmods.com/pix poweramp project/7 preamp proto V2a.jpg

You see 8 -ve regs on top (4 for each channel), underneath those are the gain boards and underneath those are the 8+ve regulators. Sorry the pic is lousy. It’s a bit of an old one. And no you won’t recognise most of the bits because I roll my own.

Anyway, sound quality: going from a simple split rail preamp to multi-regulating JUST the +ve rails brings a big gain in sound quality. Detail and textures suddenly appear. You can almost see and touch bells, brass and similar instruments; the texture of the sound becomes so subtle, detailed, refined and plain just … real. Acoustic spaces also suddenly appear (if you like that round earth stuff- and why say no to more. You CAN have it all: PRaT and round earth stuff) so you can hear if a recording is in a studio, a hall, how big, how much in the way of effects was being applied in mixing, little clunks of a shoe bumping against a drum or whatever. Resolution is seriously upped. Multiregulate the -ve rails brings a similar increase in the same qualities, more or less a doubling. Here is food for the mind as well as the heart. However, does it decrease the emotional impact of the music? No- dynamics are still there, involvement, excitement etc, just more of everything to listen to. Rhythms also become more apparent, more easy to hear and undestand and so more enjoyable. Great stuff and very worth while.

Initial fooling around with this on the poweramp has proved equally satisfactory. Anyone whose played with front end regulation will know this.

HOWEVER, the preamp has moved on with some big gains in sound quality by playing with topologies so the idea is to apply those to the poweramp too. So what am I looking at doing? Well lets start with basics.

2. Regulators
So what regulators should we use for the RedBox poweramp? Simple: gyrators, as they sound exceptional, are cheap and simple to build and understand. Also no feedback so no oscillations to worry about (more or less). You can easily build them for high voltages where you just can’t easily get opamps to make superreg (feedback) type regulators. There are ways around this but they are messy and I'm very short of space.

We must have the best gyrators though. We must have Teddy regs, I hear you cry. Actually….No. And here’s why.

If you have every played with Teddy reg style gyrators will know they are incredibly robust things. They sound great, but fiddle as you may, its murder trying to get them to sound better or worse really. They just sound the way they do. Try playing with different types of capacitors, polyester, polypropylene, ceramic. Different values, 0.1uf, 0.33uf, 1uf, 10uf, 22uf, 47uf. Not really anything in it. Fiddle with resistor values or different transistors. Well you have to move quite far out before things go to pieces. They have a huge sweet spot. Now don’t get me wrong- they are fantastic. First time I tried them instead of an ALW superreg, I was rather pissed off- Superregs are such elegant circuits, they just DESERVE to sound better. But they don't. Not by much mind, but the teddy reg style gyrators are just smoother and more extended in the top end, making superregs sound a bit ragged and edgy at hi frequencies. That moves down the audio spectrum with a slightly cleaner more detailed sound, with no downsides. I was expecting dynamics to lose out. But they didn't. If superregs are 8/10, these were 9.2/10 and so much simpler to build.

Well, here's Trick 1 for a big boost in performance. 'Will it never end?' I hear you cry.

2a. Isolate the gyrator filters.
If you read the original pre-teddy-reg posts by John luckins (I think) on developing the FET gyrator from years ago you will have noted (actually probably missed) a couple of comments he made that went really quite unremarked on and I've never really seen brought up again. Basically, feeding the filter of the gyrator from a separate supply really improved the sound! A lot. Well it does.


OK I’m completely gone here. More later again, night night.
Ced
 
Ced

Very intersting thread......

Could you explain a little on point 2a above. I seem to recall this but can't find the thread or recall the detail.

Cheers
 
surely can midcm1, heh.

Basically, I think John Luckins actually used a separate small psu- transformer, rectifier and reservoir caps- at the time to feed the filter of his mosfet gyrator. I'd guess somehting like a 10Va thing with maybe 1000uf capacitance? So the relatively large pass transistor collector current load demands were completely isolated from the current feeding the filter of the gyrators- stopping that intermodulation rearing its ugly head again.

Theoretically I wouldn’t expect it to make bugger all difference. In practice it REALLY does. So this is the basic concept (without going into values, what sort of regulator is being used etc) :

http://www.naimmods.com/pix poweramp project/8 basic gyrator concepts.pdf


Right, this google webs is really pissing me off. Some webspace on the way, and not before time either.

Anyway, back on topic. So the basic problem with a normal gyrator is current is shared by both the pass transistor and the gyrator filter at point 'A' in the 1st diagram above. So output current draw through the pass transistor collector affects (modulates) the voltage at the filter. The John Luckins idea is therefore to break this intermodulation in some way.

In fooling around to try this out, my quick and cheap-ass solution as an alternative to building a separate supply was to wheel out those old ALW superregs and feed the filters of all the teddy reg style gyrators from it. Essentially to proved the seperate regulator for the filter as in the second schematic above.

Why? Because ALW superregs have a fantastically low impedance and are extremely accurate which means they would in theory kill most intermodulation between the gyrator filters (which should be orders of magnitude less than having to deal with the collector currents anyway). I reckoned the concept would be OK anyway, and even if not as good as a seperate PSU (Actually, from my experience with PSUs, it would be MUCH MORE effective- regulators are the dominant element sound quality wise in a PSU) it should do the job. And it did.

On mulling it over my explanation of what was going on was this:
The ALW superregs don't work to the high frequencies of teddy regs but… they don’t need to. Remember gyrators are low pass filters- the higher the frequencies, the better they work (until they start to give out again due to limits of transistor ft in the Mhz range).
But at low frequencies gyrators are not so good. That’s why you try for a really low corner frequency, well below 1Hz, with teddy regs normally and use a big 1st stage filter capacitor of around 10-33uf. But note, even a low corner frequency of 0.1Hz which sounds good is just that, a corner frequency- -3db down or amplitude reduced by 50%.
But superregs are fantastic at the lower frequencies from DC upwards where their specs beat the pants off teddy regs/gyrators and fall away at mid range frequencies, (perhaps 50-100Khz?) where their impedance and PSRR drop slowly off. So we have a superreg that works brilliantly at low frequencies feeding gyrator filters where they are weak, but that work fantastically at cleaning up high frequency noise. I'm also a great marriage counsellor.
Result either way? Bingo. Big lift in sound quality with more scale, solidity and general impact blah blah blah. Very nice.
And I did try the same trick using a gyrator instead of a superreg to compare and it just didn't do it. I presume the high output impedance allowed too much intermodulation between gyrator filters.

I also note that even the super teddy reg doesn’t separate the feed to the pass transistor collector from the base filter. I would be interested to know if that improved the sound of a superteddy reg.

Now if you ponder it, there are probably dozens of different ways to skin this conceptual cat effectively. Haven’t tried em, don’t know. Some may be better some not. Might all work equally well. You could try for example making a super duper ‘teddy reg’ using 2 pre-'teddy regs’ to feed the pass transistor collector and filter of a 3rd output ‘teddy reg’. But that sounds a bit mental….. and I’m ODing on the damn name now. I guess also the sensible thing would be to do what john luckins did and actually try a separate psu as well as suprereg.
So to summarise, 2 things happening here;
1) feeding the filter part of a gyrator with some sort of seperate supply to stop modulation between the load current demands and the filtered drive current to the base of the pass transistor.
2) For multiregulation, using some sort of high performance feedback regulator e.g an ALW superreg. The low output impendance is necessarty to feed the filters of several gyrators from a single supply withough intermodulation between them.

Now, you obviously can’t use ALW superregs for a poweramp as they can only take about 36V before 'the engines cannae take any muir' and they blow. So we need to go discreet with high voltage parts. Cool. More on this later.

Anyway, not finished with gyrators yet. So to the next boost in sound quality. And I must have tried 30 different variations to stumble on these!

2b. Get rid of the 2 pole, darlington (teddy reg style) gyrator all together, and go back to basics.

And that’s it for tonight. Late again.
Ced
 
Hi Ced.
Good to see you posting I actually built your airguitar pre a while back and very nice it was to.
Very interesting thread, what sort of voltage are you planning to use eventually?
Geoff
 
Hi GWM,
nice to hear about that old air guitariste. Nice some people made it, but God that was so many years ago. Very long in the tooth now:) though still a solid performer.

My plan is that the amp boards will be good for up to +/- 80V or so.
Really it will depend on capacitior selection. For the Nap 140 not a problem.
However I also have a 'big amp' and the new trannies run at about +/-70V raw dc. The weakest link will be the polyester caps which are rated at 50, 63 or 100V. I think I can fit in the 100V sizes but may have to make do with lower ratings. Need to work the fine details out in the next few days.


cheers
Ced
 
+/-80V is 20A into 4 Ohms - you may have trouble with SOA on the output devices. Either you will have to add fierce protection circuits, or use multiple output devices in parallel. It would be a very big amp indeed - 800W into 4Ohms, 400W into 8!
 
nice to hear about that old air guitariste. Nice some people made it, but God that was so many years ago. Very long in the tooth now though still a solid performer.

Yea it blew my 62 into the weeds, still have it around someware but the 47uF MMK now does duty on feedback in the power amp.
Noticed you mention something beter than the MMK not a Wima MKP4 by anychance?
Geoff.
 
Noticed you mention something beter than the MMK not a Wima MKP4 by anychance?
Click to expand...
A MMk 47uF bypassed by a Wima 22uF MKS2-XL ???

In my NAP140 the MMK alone lacks of HF so that the voices loose a litle bit of life.
Adding the 22uF Wima brings life !
I use this combo for all the feedbacks: NAP, NAC and both feedbacks in the 322s ... this combo is marvelous.
But then, as usual, only until You find something better ... so I am also following this thread :)
Olivier
 
You must log in or register to reply here.


advertisement


Back
Top