JISBOS/Alpha20 buffer clone…

I’ve got a few projects that are now very nearly finished and also a couple of new PCBs in the mail that I am pretty excited about, but before we get to any of that I just want to show a bit more about the buffer design that I presented in my last post about the integrated ICEpower amp.

The Alpha20 buffer by AMB was originally called the “JISBOS” because of its “JFET input stage, bipolar output stage” and its original design pages are still online here. Sourcing the JFET inputs are of course a constant problem, but fortunately I bought some with my original boards from AMB. The first version of the design was a pure buffer, but since then AMB added the option to have gain as well – not something I need right now, but definitely adds to the versatility of the design.

Technically there is of course nothing wrong with AMB’s original board layout, but a couple of things were annoying me. One is that the original board is intended for very small resistors which I would have to buy, instead of just being able to use the RN55s that I have already (yes, I know it is possible to make the RN55s fit a 7.5mm lead spacing, but for me it always ends up looking like crap and a real 10mm LS is much easier to work with anyway).

Another niggle is that power and I/O connections are just holes in the board. That makes it pretty easy to solder up a permanent design, but it is a royal PITA for testing and also in case something ever goes wrong. Last but not least, I really like having LEDs to give some indication that the board is powered and operational. Of course this is not bullet proof in any way, but as a quick indication that everything is OK I find it works well.

The original plan was to run these boards without heat sinks (because they are only for line-level applications and not supposed to deliver a lot of power) but at the last minute I chickened out and put some small heatsinks on anyway. I’ve actually got another layout version with the output transistors turned 90 degrees. Then there is space for heatsinks to extend over the sides of the board, which for headphone use and other high-power applications would probably be better.

Now as I wrote in the previous post I don’t normally set out to make my boards twice as large as the original, but in this case I am willing to take that tradeoff for the improvements I have made – so let’s hope the finished amp will sound as good as I expect it to! :).

A blast from the past…

Recently I was rummaging around one of my (many) boxes of half-finished designs looking for something else when I found this – a Sijosae Gilmore board which I never put to any use.

For those of you that haven’t been doing DIY for as long as I have: This is a version of the original Kevin Gilmore class A headphone amplifier modified by Korean diy’er Sijosae to fit a much smaller board. Sijosae was an absolute artist who made miniaturised versions of pretty much all the popular headphone-amp designs of the day while also experimenting with different topologies for buffers, rail splitters and similar circuit components. Even if he is no longer actively posting you can still see his characteristic schematics pop up in google searches and being referenced in new designs as well.

Sijosae’s version of the Gilmore amp could (theoretically at least) be squeezed into an Altoids tin like a CMoy-amp. In reality there would be no space for batteries and the battery life would be very short because this amp runs in class A, but at least mechanically it would fit. He also made a simplified “EZ-gilmore” version of the Gilmore circuit which I cloned as well (but also never used, now I come to think of it…)

The Gilmore design is back from the headwize-days and the final PCB layout was done by an american user called Subsonic who subsequently offered it as a “group buy” on Head-fi in 2003. As I recall, this was the first group buy I ever participated in and one of the first headphone amp PCBs I bought internationally – if not the first. To say this started a tradition for me is something of an understatement (“avalanche” is more like it 😉 )

The board has been in storage for so long I don’t remember exactly why it was put away in the first place, but now that I have dug it out I am actually going to test it. I seem to remember it had offset-issues that I found very puzzling at the time, but I am thinking that the 15+ years of diy-experience I have added since might help me solve them this time… 😀

Starting the 250ASX-int…

For a while I have been thinking about doing an “all-in-one” integrated amplifier and therefore I’ve been looking for a suitable class D amplifier module as the “centrepiece”. A few weeks ago the perfect candidate showed up in a local classifieds ad and so I picked up a single unused ICEpower250ASX2 module at a fair price. Conceptually this build is quite simple – two switched single-ended inputs and a buffered ICEpower module with a volume control inside. However, just doing that would have been a little bit boring, so I added a some complexity to make it interesting 🙂

Part of my reasoning to build this at all was that I wanted try out an ESP8266-based amplifier monitoring & control board I developed based on my IoT-T design. This control board was actually intended for ICEpower-modules so that I was lucky enough to pick up a 250ASX was really good. The original inspiration for the control board wasn’t even the ASX-modules but rather the Pascal-module which has the ability to output quite a lot of monitoring and diagnostics signals. However, as I only have one Pascal module and no reliable way to get more I decided to build a first version to suit the ICEpower ASX modules instead.

I don’t really have working software yet, but when completed the finished amplifier should have the option of basic web control and monitoring via the ESP’s Wifi connection as well as driving a “local” front panel LCD display via I2C. I haven’t fully decided if I want to use this feature for this particular build, but at least the option is there. A potential problem is that the ESP8266 is going to be enclosed in a aluminium and steel box and the Wifi-connction might not like that. Obvious solution #1 is to use a ESP pro module that can be fitted with an external antenna on the back but my mechanical layout is suboptimal for this purpose to say the least. Obvious solution #2 is to ditch the steel lid in favour of acrylic or something else – we’ll see where I end up with that.

Another goal of this design was to try using a discrete buffer such as the JISBOS/Alpha20 with the ICEpower amp as I’ve never really done that. However, once I started looking at the A20 boards from AMB that I already have I decided I preferred to do my own version instead. Normally my goal with clones is to make stuff smaller but in this case I ended up making it about twice as large as the original… Still, I think it was worthwhile to do and I’ll probably do a separate writeup on this design later. For input selection I have a basic design that works (I only need two inputs), but once I got the boards I have out of storage I couldn’t resist messing with them a little, so I can’t finalise this until the new boards show up (which may take a while if all the factories in China stay closed due to Corona-virus…)

For volume control I ended up with a very difficult requirement, namely that it had to be controllable by I2C from the control board. That’s a surprisingly difficult one since the “usual suspects” for high-quality audio (e.g. a PGA23xx or similar chip) requires SPI, so my solution ended up being something else – we’ll see if that works 🙂