OK a bit more progress made:The build- amp boards
Well, here we go with the before and after shots. The Redbox amp board against the Naim board:
First thing is to lay out the workspace. Fiddling round looking for parts wastes so much time, and is usually a source of mistakes/frustration so its worth having everything to hand and in order.
A good red or whisky may also be essential kit depending on mood.
Strong lighting and tools required. The tools I like to have to hand are:
snips,
snub nosed pliers,
screwdriver with various heads,
wire stripper,
pen,
multimeter and bizzarely
a file.
Then obviously
the soldering iron,
solder and
solderwick. I have had 2 solder suckers and never ever got them to work even once! I don’t know if there's a secret method, but I'm buggered if I've figured it out.
It also helps to have built up parts bins for when stuff’s missing like the legendary 'specific and odd value resistor'. I have a couple of resistor boxes, 3 caps bins, fixings, couple of transistor bins and some sundry parts cases. Even fixings such as solder pins, bolts and other little bits that don’t seem critical can stop you dead in your tracks if you don’t have them to hand. And I tend to get real ‘resistor rage’ throw-computer-out-of-window type frustrated so I think of them as ‘Daddy’s little helpers’- electronics Valium.
This is also where building a thorough Bill Of Materials (BOM) really pays dividends. You will have bought that little annoying piece that didn’t really seem part of the pcb you were building but nevertheless you need to install it. And similar cockups. I’ve resort ed to ‘the bins’ about 5 times on just the amps boards already, but hey, this is still in development. Part of the appeal of kits. Christ I hate pcb assembly!
Anyway, parts for the build include the boards (stating the obvious perhaps, old mate?), caps (electrolytic and film [polystyrene, polypropylene and polyester]), diodes of various types, fixings and wire, heatsinks, resistors (power, pots and standard) and transistors.
Then lastly but not least the BOMs and schematics. I forgot to print mine off so am working on screen instead. I’ll pass that off as eco-warrior-iness, in saving paper. Good dog.
To business.
First off is the you’re-an-edjit-if-you-don’t rule of pcb assembly- start with the smallest, lowest parts first and then build up to the biggest last. More or less. So, as no ICs, its resistors next.
As am still waiting for one more parts delivery, which is apparently somewhere over the channel right now, I am doing as much as possible on the Amp board first- Fortunately only a couple of power resistors and heatsinks still to come which are completely accessible so this won’t cause problems down the line.
A nice thing is I’ve used Vishay Dales for the amp boards and BCC (Actually Vishay BCC, like all parts these days) MRS25s for the amp boards as they’re good low noise 1% metal film resistors for regs. It makes assembly a lot easier.- The irritating bit with resistors is there are so many and they all look the same so If I can sweep half out the way for the moment, that’s a good start. It also pays to lay them out in value order. I’ll start with the vishays and the amp boards first.
I laid out my resistors in 3 rows:
1 – 1,000R,
1,001 – 10,000R and
10,001R upwards.
Mmm, absolutely fascinating fact.
And then to stuff the boards as per the BOM.
Start with the RN65s first, as they are the 0.5W (1W industrial) large ones.
Then stuff the rest of the RN60s.
Note 2 resistors go under the pcb due to layout constraints. The need to go right flush up against the pcb- no sloppy soldering allowed on these two.
resistors all done.
I always make sure the resistor values are facing upwards so that when they are inserted in the pcb, you can easily read the value. Makes 'flight-checking' and trouble shooting a world easier if a problem occurs.
Now a quirk with this RedBox is that the BOMs vary a bit depending on the voltage the amp rails run at. I, in glorious cowboy fashion, hadn’t quite got round to working out the various voltage ranges so got nailed on some resistor values. Thank god, parts bins. I’ve just done an upgrade to my big amp psu, going to a 4 winding transformer which has a different voltage output that I worked to. Allowed me to relocated the star ground back in the amp where it damn well belongs (it was a pisspoor make do before as I didn’t have the right capacitor boards for the design) which has been an great improvement and reaffirms the Mr Tibbs grounding mod. It also makes me realize just how damn good even the prototype to this build is. If the RedBox delivers as expected, it will be mindblowingly, hair-raisingly good! I’ve been listening on and off all morning to the prototype and Christ, synth bass notes sound so tight and textured they’re like shimmering, inflated balloons appearing and disappearing in front of you. You can almost reach out and caress each one they’re that vivid. And the musical access is just mad, wah. Its all there in front of you like a medieval feast. Choose a rhythm line, follow it, jump to another, ooh a little backing vocal line ever so quiet, just tucked away in the back corner, lets have a nibble at that, ahhh. Little chime tapping away to the left, god that’s so clear. Ahem, er… that might have been a bit mad. Very satisfactory upgrade anyway.
I have to say that I never used to rate poweramps as particularly important in the musical chain sound quality wise, but I have completely changed my opinion. I think my original view was largely formed from listening to various Naim poweramps which, truth be told, all sounded much of a muchness. OK, the 250 was a shade meatier and if you needed extra power for the speakers, then yes, it or 135s were essential as I found out with some Epos ES22s. But really, for the money a 140 is quite satisfactory for reasonably sensitive speakers in smaller rooms. Spend the difference on 3 weeks in Mexico for a way bigger grin.
But, that’s complete cobblers I now realise- do em right and poweramps are every bit as important as preamps!
ANYWAY…, back from that random tangent.
Basically resistor values have to be selected for the CCSinks depending on what your raw PSU rails are. For Nap140 running at around 34V (probably also applies to a Nap180 at about 40V??) raw DC, I want minimal voltage drop. With a 250, 135s, (which run at 56V raw dc due to the power regulator boards), and the higher voltage DIY PSUs (basically any stuff with rails running at over say 50V), you can afford a bit more off the max voltage the amp can swing, so can adjust for more optimum CCSs. Basically, referring to a paper by. .Walt Jung maybe, measurements showed that IC voltage references performed better than diodes as references in CCSinks. Cascodes obviously perform better too (by the numbers anyway, though sonically too in my exerience). So this means the voltage references should be swapped out at the same time as you select the resistors. This is basically what it boils down to:
CCS biasing resistors.
Refering back to the amp schematic
For 35V raw dc supplies, use R10, R11 of 30K. This should give biasing currents of about 2ma on the CCS voltage references and power dissipation of ballpark 150mW in the resistor, within both the mil and industrial max ratings though they should get warm in practice.
For 50V+ Raw DC supplies, I will use 50K resistors which should give around the 200mW mark dissipation. Higher currents would push power dissipation too high for long term reliability in these resistors I think. Current is essentially solid dc with minute ac components through these anyway so I think those values will be fine.
The VAS CCSink.
Whatever you use, the current through R15 must remain at 8.2ma
For 35V RAW DC rails on a NAP140, the regulators are going to eat a fair bit of voltage headroom at least 5V. I therefore want to keep the VAS CCS headroom as low as possible or voltage swing will be poor and the amp may run out of welly to drive speakers. That means making the D3/D4 voltage drop as low as possible.
I am using 2x1N4148 diodes in series to replace each LED (D3/D4), Like this:
Eeach 1n4148 drops 0.6v, giving a 1.2V drop when using 2 in series. Or LM385 1.2V voltage references would also probably be very nice/better. Again R15 would have to be 69R1 to give 8.2 ma through R15.
Now cascoding stiffens up the VAS CCS very considerably. In its standard form the ac component though this is around 0.2ma with a steadyish 8ma bias ie its not a very good current regulator. Casoding this reduces the a.c component to about 3uA in sims; or look at it as current regulation improving about a hundred fold at 1khz and high power. I would be surprised if this wasn’t clearly audible.
If running raw DC rails higher, (NAP250/135s come in at 56V DC for example regulated down to about 40V), there is 16V headroom to burn, nice. In this instance I would feed the redbox regs off the raw dc, let the powerregs just handle the output transistors and voila a fully, from head to toe regulated, poweramp.
In that case use I’d at least use Green LEDs for D3,D4 and R15=162 Ohm.
Or Use LM336, (2.5V references) for D3,D4 and R15=…???- can’t remember, somewhere around 200 Ohm? Have to work it out again. I would guess the LM336s are the best sonically, but that is something for a listening test. So the VAS CCS then starts eating about 5V instead of the standard Naim1.2V of voltage headroom but that would be fine as you can run the redbox reg boards at say 45V and not affect the overall voltage swing of the amp a single iota. Very elegant.
Now, to the LTP CCSink.
R6 must pass 0.9ma whatever the parts selection for the CCSink and its value.
The collector of Q1 doesn’t drop more than a couple of volts below ground in worst case scenarios. This means we can use very good voltage references in theory with plenty of voltage headroom- better CCSs tend to require more volts to function. I would go for the trusty LM329DZs, 6.9V references. These are buried zeners which are super low noise, low dynamic impedance devices. Cascoding drops the a.c element down to about 17nAin sims! holy cow. Also significant will probably be Cob – 6.9 V is quite generous headroom but if you look at transistor datasheet s you’ll see how Cob invariably reduces as Vbc is increased- ie the larger the voltage across a transistor, the lower the ouput capacitance. Lower capacitance should= better speed, linearity and high freq performance as a general rule. A reason why its no bad idea to run highish voltages across signal path transistors when possible. Q1 which takes the brunt of voltage swings from the input diff transistors should also be nicely linear. However, despite this I’m going to start with green LEDs as I know these are stable in this cascode CCS topology. I don’t want to have to deal with too many variables at this stage.
If using LM329s for D1 and D2, R6 must be about 6K5 Ohm.
Otherwise use Green LEDs for D1 and D2, and R6=1K5.
In summary I’m doing this for a NAP 140, so I’m starting with Green LEDs for D1,D2 and 2x1n4148 each for D3,D4. R10, R11 are 30K
So after having stuffed the boards with resistors, and diodes, both normal as Light emitting, as the CCS voltage references, they’re ready to be soldered into place.
But before that, a quick double check that all the resistors are in the correct positions against the BOM. Saves a horror of trouble shooting down the line.
I started by soldering R6 and R100 on the underside of the board. These need to be completely flush with the board as they sit in the sandwich gap and will prevent proper mating of the boards if set badly. The leads must also be trimmed completely flat on the top side of the board as the feedback caps sit above them.
Though through hole plating is a pain in the butt in general, it does mean you can solder from the top of the board. The underside is a thorn bush of leads. And yes this is a bit stupid from a flux cleaning perspective, but I'm a bit lazy about that and never had any problems with corrosion. Hate flux cleaning chemicals too and they seem to create as much mess as they remove. Otherwise I'd do it immediately on all the resistors if I was going to.
I also try to make sure not to accidentally slop solder around and fill up empty plated though-holes- its murder trying to suck out solder once they’re filled. Otherwise it’s a fast process and only takes about 5 mins per board, weee.
Any leads I can’t easily get at to solder from the top, I leave.
Then flip the board and trim off all the leads. I love this, snip snip snip, scissors hands style. I Leave about 1mm of lead sticking up, no more.
Then I finish off soldering by going over each joint and soldering any that look unfilled and finishing off any leads I left unsoldered from the top side.
Double check everything again under a bright light. I’ve tracked down more than one fault to an unsoldered joint when tired and rushing.
Link
I’m going to use T03 ouput transistors so the 2 Link PTHs have to remain EMPTY. This link is only for using T0264 output transistors.
diodes
D5, D6, D7. Solder and snip.
VR1, Is the blue box in the pic above. I place a big dollop of solder on the tip of the iron, fit VR1, hold in place with one hand and solder one leg in place using the dollop of solder on the end of the iron held with the other hand. It will be a dry and faulty joint but it holds the pot in position which is the point. You can then flip the board, solder the other 2 legs and then resolder the bad joint properly. Nice snug fit. I don’t have one of those soldering sponge flip table thingys.
At this point its worth just checking all the resistors values with a multimeter to make sure they aren’t faulty and haven’t been damaged during soldering. Once capacitors go into the pcb, they screw up the readings and its often impossible to measure on the board. Takes 2 minutes.
Polystyene capacitors.
Fit (value facing upwards), solder, flip board, finish soldering, snip.
Note, C7 has to be fitted under the board and must lie completely flat.
Bears remembering that polystyrenes can’t tolerate much heat so mucho care. Don’t touch the body with the soldering iron or you can melt it and possibly damage the capacitor, and I try to be swift but sure in soldering the leads so it doesn’t heat up via conduction.
Anal-retentively I always keep a handful of snipped component legs. They always come in handy as ready tinned link wires.
I gound myself regularly by touching exposed metal on a nearby radiator (Radiators and piping are linked in to your domestic ground) to ensure no static buildup- It can blow semiconductors such as …
Small signal transistors.
Laid out as per resistors on the work table.
Stuff PCB as per BOM, I make a point of pressing firmly all the way in for minimal leg length. I recommend a dab of solder on one leg from the top of the pcb, then flip board and solder rest of legs and snip.
I usually stuff each type of transistor before moving on to the next type i.e do all the BC546s first, then the 2sc2705s etc. Its easy to get them mixed up otherwise which can become an arse-ache to check as the text on the bodies is so damn faint.
Note, the middle legs need to be pushed either forwards OR backwards on insertion into the redbox PCBs. Orientation of the body is the critical element to watch- My PCB CAD program gives you great control over the primitives on the PCB so you can move a component’s pads around in situ to suit routing which I do a lot with transistors.
PCB pins
Then came the vero pins and 4 pcb tabs. The tabs need to be pressed in quite hard, but they're rock solid then.
Almost there; the TO220 devices:
Caddock power resistors.
Again I recommend the ‘solder dollop’ technique- big dollop of solder on the iron tip,
hold the TO220 resistor in position with one hand, flip the board and dollop the solder on one leg to make a bad joint. Then solder the other leg properly, go back and solder the first leg again. I did each resistor one at a time by this technique and they stand perfectly.
Driver Transistors
Did Q10,Q9. also using the ‘solder dollop’ technique; particularly important as these parts should sit straight and even for the heatsinks.
Capacitors
And then my favourite bits, because they are easy, chunky and done in seconds. Also you have a little frisson of nearing completion because suddenly the board’s almost full.
The inductor.
Ok, inductor should be about 0.8- 0.84uH. I’m using 1.5mm enameled copper and I have a 7mm diameter pencil to hand as a winding form which works out that I need 20 turns and should make a 3cm long inductor- I can just fit it on my pcb. Otherwise I'd have to find a smaller form.
The knack is to wind on something slippery and long (fnah fnah); my synthi-pencil (it’s a weird eco pencil thingy made from recycled plastic cups and is almost kinda greasy) and wind directly off the reel so you keep a good grip on both the fixed end of copper and the winding end. Grip one end of the copper around the pencil and then keep rotating the pencil in your wrist so the copper winds round it. Perfect.
Ive used small dowels as formers before and it becomes fiddly and messy. Snip off the ends leaving about 1cm leads then hold the inductor and gently file away the enamel on the leads so you can solder onto the board.
And that’s about it.
I have 3 small capacitors, 2 heatsinks and 3 power resistors still to arrive and then ready to test
After shot; here the new RedBox boards against the old Naim ones. I sure know which look like they sound better, hah.
There are however enough variables, that for once I may use a bench supply to test this thing before hooking it up, instead of my usual gung ho approach. As I don’t have a bench supply, in true DIY spirit I was thinking of using 2 x 9V batteries to make a +/- supply which should give enough voltage to get the board up and running and be current limited- how much current can a 9V batter dump in short circuit after all, hey?
Ok for now.
Ced