In search of synergy…

Slightly off-topic post, but I have written a few times about how I think that system-matching is much more important than any “absolute” sound quality, at least as far as enjoying the music is concerned. Well, today was another reminder that I still think this is the case 🙂

A couple of months ago I got new speakers, trading my old (and much-loved) standmount Sonus Fabers for some floor standing Scansonics that offered a bit more low-end slam. I was quite happy with the trade from the beginning and I have absolutely no regrets, but after a time the inevitable restlessness sets in and you start thinking about change (at least I do…). I’ve been running the Scansonics with a simple 125ASX amp on my Harman/Kardon preamp, but just to try it I dug out another ICEpower-amp from my collection, this time based on the older 200ASC-modules.

Although I would definitely still class the 125ASX as the better amp overall, the Scansonics (which are just a little bit bright) immediately benefitted from the more “closed-in” presentation style of the 200ASC, so as usual after initially listening to half a track I started to go through my normal playlist of tracks I know well and just enjoyed listening to some music that I would normally say I know back-to-front already.

To be fair I am honestly not surprised at this, because I saw the same change when I switched from the even older Elac speakers that much preferred the warmer sound of a 50ASX amp whereas the Sonus Fabers really came to life with the more lively presentation of the 125ASX. However, I still think that it’s nice to be reminded once again what really matters when putting a well-rounded system together and of course experimentation is always fun (although it can sometimes be very expensive as well…)


Building an(other) F5…

Although I recently built a new type of F5 amplifier, I haven’t completely abandoned the original F5 design 🙂 Hiding in one of my many boxes were a pair of half-finished F5 boards and some matching matching fan heatsinks that only needed the last bits of assembly and calibration. That honestly didn’t take long to do once the right parts showed up and I then managed to confirm the boards were indeed working.

The boards were originally bought from ebay and are more or less the same as my original F5 build – nothing special there. I have some matching PSU boards as well, only missing the last few parts which are now in the queue for my next order and that’s going to be a standard C-R-C type thing as well.

The mechanical design is from the same time as my JLH mono blocks, so the idea is also more or less the same. This heat sink profile is too large to fit in most enclosures though, so cracking what to do took some time but I think I have it figured out now. It’s also going to be monoblocks, but much larger ones than the JLHs. From my first tests during calibration of the boards I think a slow-speed fan should be enough to keep the heat under control, so hopefully they will be living-room friendly when they are done 🙂

Improving small DC-DC converters…

I’ve written a few posts about DC-DC converters and I’ve found them very useful for many circuits where a “complex” (= multi-rail with fixed voltages) power supply needs to be replaced with something “simpler” and more flexible (= single-rail with large variations allowed). However, I have previously largely ignored small SIP-type converters because I didn’t believe they were powerful enough to be of much use. As I (recently) realised you can actually get 3W from several manufacturers and even 5-6W from certain others. That makes it’s possible to get 100mA or more at +/-12-15V which is more than plenty for most small opamp circuits such as buffers, preamps and RIAAs, and of course single-rail 5-12V currents more than suitable for small auxiliary circuits in power amps etc.

Now, there are a few drawbacks to these converters: Even the small converters normally still have quite large ripple voltages and I expect there is quite a bit of HF-noise as well, but I’ve tried to use a passive filter to compensate for that. The basic idea is that because the switching frequency of the converter is very high (typically 50-100 kHz) which is nearly 1000 times higher than a linear supply, a simple passive filter is also 1000 times better at removing ripple and noise and so even small capacitor/resistor values for the filter gets you very far. A second drawback is that the converters have limited tolerance for capacitive loading, so it’s normally a good idea to think the power source into the design/build of the consumer circuit. That’s normally also doable though.

The basic SIP-8 form factor is used by several manufacturers so there are quite a few different converters to choose from, both cheap and not-so-cheap. One thing that differs between manufacturers seems to be the allowed capacitance load that the converters will tolerate. Here, the more expensive Recoms and Tracos list considerably better specs than the cheaper converters, so that’s worth looking into before you choose. Given how the converter works, this restriction mainly applies to higher voltages of 12-15V or higher – at 5V the load margin is likely to be fine even for the cheaper converters.

The boards I’ve made are both a single and a dual version with the same form factor and they both work as expected. However, after I received the boards and assembled the prototypes I had a couple of ideas to improve the filtering a bit so I’m going to hold off making the board files public until I’ve tried those ideas 🙂

Building a different F5…

As I have mentioned a few times, the First Watt F5 is one of my favourite amplifier designs (and of course I am not the only one who likes it). It’s very simple to build, it’s reasonably priced and it sounds exceptionally good. The only drawbacks are the heat and the relatively low power (which is why I sold my original build), but with both new speakers and a new room comes new opportunities so I wanted to try the design again.

I actually have a few F5 clone boards more or less done, but that’s a story for another time because the original F5 design has spawned a few variations. One of them by diyaudio-user Juma is based on using several smaller output devices in the form of Toshiba 2SK2013/2SJ313 (which of course are obsolete…). For reasons I don’t really pretend to understand these devices are very linear and so the sound of this F5-version should be even more special – we’ll see about that I guess.

I’ve looked at this particular F5-design before and it’s not exactly new, but sometime you have to wait a (long) while for inspiration to strike and in this case it only did a few weeks ago, so the finished boards turned up only this week.

My version has four device pairs in the output to allow a bit more idle current for low-impedance loads. Also included is some additional rail capacitance close to the outputs (mostly because it seemed wasteful not to use the board space for anything), but otherwise it is that same as Jumas original circuit. I’ve only bench-tested it for now and I can’t do proper trimming of idle current and offset until I’ve drilled some heatsinks to mount the board on, but it powers up like an F5 and it responds to the trimpots, so hopefully it should adjust properly when the time comes. For now I’m just excited to have gotten it this far 😀

Project files: PA100 parallel gainclone

What is it?
Board files for my “PA100” parallel chip amp with the LM3886 first presented here.

I’ve used the app. note version of the circuit which is non-inverting and uses low-tolerance components to minimise offset between the two ICs. There is also the Jeff Rowland-derived inverting circuit that is normally employed as a PA150/BPA300 configuration with three ICs per board.

I’ve mosty stuck to the datasheet circuit, but in some areas I have drawn inspiration from Tom Christensens article on the LM3886 IC. I’ve used SMT-components where I believe it makes sense to get a tight layout, but mostly its nice and diy-friendly leaded parts 🙂

How big are the boards?
The board measures 3.9” x 2.4” (app. 99 x 61 mm).

What is the status of the boards?
The files are for board version 1.1. I’ve made the following changes compared to the v1.0 prototype.

  • Mute capacitor footprint enlarged.
  • Mute resistor moved to the center of the board to make space for the larger capacitor.
  • Footprint for the LM3886 changed as the holes were very too small.
  • Made a small space between the large reservoir capacitors so they don’t touch each other.

Note that I haven’t tested the v1.1 (yet – will include them with my next PCB order) but I don’t expect any adverse effects of these changes.

Does it use any special/expensive/hard-to-find parts?
Not really, but the recommended resistors are lower tolerance than what is common (the 0805 resistors are 0.1% and the 0R1/3W output resistors are 1%). Mouser has them all and there should be plenty of other sources. The amp will work with standard tolerances (1% for the SMTs, 5% for the outputs) but if you’re unlucky with the tolerances then performance will suffer a bit (higher DC-offset on the output and higher idle dissipation in the ICs). The recommended parts are not much more expensive so I definitely recommend you stick to them.

Anything else I need to know?

  • The gain setting resistors (the SMD-ones) should be 0.1% tolerance for best performance (see above).
  • Similarly, the load-sharing resistors on the output should be 1% tolerance for best performance (see above).
  • The power LED on the board is only between the negative supply and ground, so it is not a 100% indication that everything is OK.
  • The board obviously works with both versions of the LM3886, but I recommend the isolated (TF) version because it’s easier to mount.
  • Decoupling: My decoupling scheme is somewhere between the datasheet recommendation and TomChrs decoupling scheme. The topside parts are intended to be 100nF MKT or X7R MLCCs which is more or less what the data sheet specifies, but on the bottom there are pads for 1206/1210 SMD caps which you can fill with 4u7-10uF X7R MLCCs. You can also use the SMD pads for 100nF MLCCs and then mount electrolytic on top, but there isn’t much space so be a bit careful.
  • The board should be fed from a DC power supply, linear or switching. The large reservoir caps can be as big as you like, but as my prototype boards are intended to be powered by an SMPS (which is sensitive to capacitive loading) I’ve used fairly small capacitors. If you use a linear supply by all means use bigger capacitors.
  • Bridging: You can bridge two boards to create a BPA200 amplifier, but remember a) to lower the supply voltage to around +/-28VDC and b) that you need either a fully-balanced source/preamp or you need to invert the phase using a balanced line driver such as a DRV134/THAT1646 or or fully-differential amplifier of some sort.
  • Mechanics: The C-to-C spacing between the ICs is 1.5” (38 mm).

Download design files here

Related information:
Note: Always read the “intro post” for additional important information about my designs.

You can find additional information about the LM3886 amplifiers in the data sheet, the AN-1192 appnote linked above and several other resources – check them all out 🙂

The “Whammy” headphone amplifier…

Although I did my own version of the Pass/Colburn “Whammy” headphone amplifier before there were boards available for sale in the diyaudio store (and before it was officially called the Whammy), I have still considered getting an original all-in-one board as well.

The cost of shipping from the US originally deterred me enough to do my own version, but a couple of weeks ago a board popped up on diyaudio from a fellow hobbyist in Europe, so I was able to get one at a reasonable cost. Unlike the diyaudio board this one is green (which I massively approve of) and also 2mm thick and plated with gold (ENIG) so it looks and feels really great. Because the board was thicker than usual and I knew I had to mount it in a big case I decided to go “all out”, use tall caps and heatsinks and maybe experiment with turning up the current compared to normal (haven’t do that yet though).

The power supply is running at 20V courtesy of some 7×18-regulators and a pair of green LEDs. This limits my choice of opamps, more or less to either the original OPA2604 or the (now-discontinued) LME49860 which is supposed to be a 22V-tolerant LME49720. Not sure if that is true, but I did chose the latter and I have no complaints about the sound. I might try the OPA2604 at some point instead since I haven’t listened to that since the comparison was an OPA2134 – that’s been a while. The output FETs are the recommended 2SK2013/2SJ313 which I already had matched pairs of, but obviously plenty of other options available that are easier to source.

Just like my clone version this one worked immediately after being powered up, but that is probably more to Wayne’s credit than mine 🙂 I don’t have a case idea just yet, so for the moment it’s going in a box until I come up with a plan for what to do next – still sounds great though 😀

A parallel amplifier with the LM3886…

Gainclones or chipamps are a popular DIY-topic and I’ve done a couple of designs myself and assembled a few others as well. The only one of the “original” National semi amplifier IC’s that I haven’t really done anything with – and coincidentally the only one that’s still in production – is the LM3886.

But not any more, because I just finished a simple design with two LM3886s in parallel configuration. The circuit is built (mostly) according to the “PA100” design from the original National application note (AN-1192) and not the Jeff Rowland-derived PA150/BPA300 that has different configuration and of course a third IC per board.

The configuration with two parallel ICs gives full current output at +/-35V into 4 ohms where a single IC would otherwise be thermally limited, but of course the power is still modest. As I recently swapped my faithful Sonus Faber speakers for a set of Scansonic MB towers which have a fairly low impedance, that’s exactly what I needed though (not to mention that I had a 35V supply left over from another project 🙂 ). The two-chip configuration also means boards can be kept small (and cheap), and there’s still the option of using two boards per channel in bridge-mode to make a BPA200, although the supply voltage would have to be reduced – only the BPA300 will run at 35V rails in BTL-mode as well.

The boards worked first time on power-up and seem to be well-behaved (quick tests only though). I need to do a bit more testing and make some minor (mechanical) changes to the layout and then I’ll publish the project files 😀

Waiting for parts…

The summer weather still doesn’t show any signs of slowing down here – at least not significantly – and so building is a little on the backburner. However, I have been keeping up a steady flow of PCB-orders over the last weeks (partly my own designs, partly not) so that when I go on holiday in a couple of weeks the finished boards should be waiting for me. Assuming the weather is more suitable for indoor activities at that point, there should be a few interesting things coming up in the not-too-distant future then 😀

Already now though, I have started putting together a few things including another line-level buffer, an ebay tube-kit and a couple of headphone amplifiers but it’s stop-start traffic most of the way. A constant interruption to these builds are a lack of parts – not massively so, but a resistor here and a capacitor there is enough to slow everything down. Case in point is a buffer by Kevin Gilmore where I have the boards (and have had them for a while) and most of the assembly is done, except that I am missing four ceramic caps (odd value and specific form factor) and four RN60 resistors (a standard value that I simply ran out of).

For some odd reason this actually tends to delay overall progress by quite a lot because by the time I’ve accumulated enough volume for an order from a specific vendor and the missing parts show up, usually something else has caught my eye…  😀

Anyway, Mouser order just completed so the last parts for the buffer and a few other half-finished projects should be here by the end of the week. Maybe I should spend my holidays working out a queuing system for new builds of some sort instead? 😀

Project files: Line Attenuators

What is it?
If you are using a preamp with gain you may have problems with only using a fraction of the available range on your volume control which is very annoying. The problem is usually caused by too much gain and/or an incorrect gain structure. If it is not possible to reduce the gain of one or more of the amplifiers in the chain, a solution can be to use inline attenuators from e.g. Rothwell Audio instead. These are quite expensive though, and they only come in predefined attenuation levels so for testing purposes a DIY-option such as I am presenting here might be better.

The attenuator is built on a small board with RCA sockets for input and output, as well as an option for fitting two parallel resistors on the output side. The gives two (or even three) selectable attenuation values. The selection can be either by jumpers or even via a switch to make the boards suitable for testing etc.

How big are the boards?
The board measures 1.75″ by 0.9″ (app. 44 x 23 mm) – plus of course the off-board part of the connectors.

What is the status of the boards?
The board is in v1.0, meaning it has been tested and confirmed working.

Does it use any special/expensive/hard-to-find parts?
The RCA sockets are clones of Vampire RCAs. They are normally the best board-mounted RCAs I know of and available on ebay. If you don’t want to use connectors or can’t find them, just connect the signal via a 0.1” header (or a JST XH/Molex KK connector) instead.

Anything else I need to know?

  • Important: The reason that Rothwells are built into the RCA-plug is to keep the signal path short and especially the load capacitance on the output side as low as possible. Use the shortest possible cables on the output of these to avoid the cables inducing an RF-rolloff.
  • The resistor values are quite important and should ideally be matched to the source and load impedance. I’ve used this thread (post #6) as a starting point but it’s worth reading up on the theory behind the operation as there are a few trade-offs involved.
  • The center-to-center spacing of the RCAs is 1.1″ (28mm)

Download design files here

Related information:
Note: Always read the “intro post” for additional important information about my designs.

Inching forward…

Another long(ish) break from posting – this time mostly courtesy of some extremely nice late-spring weather and a couple of house-related DIY-projects. Just about the only thing that has moved forward (at least enough to notice) are my ICEpower 700ASC-based mono blocks (which I discussed here). A couple of weeks ago I got the mounting plates I designed for the modules + supporting circuitry which meant I could drill the chassis and start putting some mechanicals together at last.

Some of you may have guessed that this is where my BalBUF design is supposed to end up, but there was a piece missing. A matching power supply to drop the 700ASC’s 15V aux power supply to something more manageable for the OPA1632 (which gets very hot in operation). Because I was running out of space in the enclosure I wanted to use, a key design criteria was that the PSU should be “stackable” with the BalBUF board.

I quickly found what looks like the perfect device for this use – the TPS7A39 from TI – which is a dual pos/neg low-noise regulator with the right specs. Unfortunately, it is also only available in a 3×3 mm leadless package and as my odds of hand-soldering that are pretty much = 0 I dropped that pretty quickly. Instead I went for a bog-standard LM3x7-based design, but managed to squeeze it down to size because of the modest heat sinking requirements.

In a nod to “reusability”, which is something I always aim for where possible, the PSU board includes SMD resistors on the bottom in front of the caps, which means it can also be used with the unregulated supplies on the other ASX-boards such as the 50ASX and 125ASX. This means that you can use the BalBUF with any ASX-module without a separate offboard supply for the low-voltage circuitry, and because the BalBUF and the PSU stack on top of each other it should be very compact. Assuming everything works as expected with the 700ASC when I test it, I’m pretty sure that means I’ve just figured out what to do with my last remaining pair of 50ASX’es 😀

The sketch for the rear panels is also pretty much done, but given that Schaeffer/FPX panel work is getting more and more expensive I have decided not to order the rear panels “blind”, i.e. before I have tested that the monos work electrically. If this weather continues, that might be a while though 😀