Bridging amplifiers…

In response to one specific (and several non-specific) requests and questions, here is a little overview of what to be aware of when bridging single-ended amplifiers. This could be either two channels on certain ICEpower-modules (with some particular considerations mentioned at the end) or e.g. two PA100/PA150 chipamps. Note: The following assumes you are bridging two existing amplifiers, either at board-level or two channels on a finished amplifier. If you are designing a BTL amplifier from scratch or connecting an amplifier IC that supports BTL mode, I suggest you look elsewhere (such as the data sheet for the IC) or this Infineon application note if you are designing from scratch.

On a normal SE-amplifier, both the input and output signals are referenced to a common “ground” (“0V” is probably more accurate, but never mind that for the moment). When you connect two amplifiers together for BTL you couple their grounds together and drive the signal inputs separately. That way, one amplifier “pushes” when another “pulls” the loudspeaker and in theory you can get four times the power (= twice the voltage swing) this way, but there are a few caveats to note.

Firstly, in BTL-mode each amplifier sees half the effective load impedance. Since most normal amplifiers don’t double their voltage swing into a half-impedance load, the max. power you really get is frequently less than four times but still higher than the original.

Secondly, the lower load impedance also means the amplifier has to be able to deliver more current, so the “price” of more power is typically increased heat and a higher risk of tripping an amplifier’s protection circuitry. This is also why many commercial amplifiers that can be bridged state the they should only be connected to loads of 8 ohms or more in BTL-mode, because the amplifiers are not able to handle the 2-3 ohm loads that they would otherwise see.

Lastly, on the signal side a prerequisite for more power in BTL-mode is that you feed one amplifier channel an input signal that is out of phase to the other. Only this way will one amplifier “pull” when then other “pushes”, which is how you double the output voltage swing. That means you need a “balanced” signal, meaning a signal with separate hot and cold pins with respect to a mid-point (ground). You can’t cheat and connect the normal SE input signal as hot and cold because that is also referenced to ground and you would have grounding issues (hum/noise).

To get a balanced signal you will have to either use a balanced source or include some sort of SE-to-BAL converter in the signal chain. This converter can be passive (= a balancing transformer) or it can be an active circuit either discrete or simply built from two opamps. Last but not least it can be a purpose-built integrated circuit in the form of something like a DRV134/THAT1646 line driver or a fully-differential opamp like the OPA1632 (examples of both of which are on this site 😉 ).

I honestly don’t know the textbook definition of “fully-differential”, but whereas a BTL amplifier is two separated channels connected together, a fully-differential amplifier at least has a feedback loop which is shared between the two sides. A fully-differential amplifier will usually perform any combination of SE<->BAL conversion you can think of. If you want an unbalanced input you simply ground the “cold” side of the signal, and if you want an unbalanced output you can take it from the “hot” output and ground.

Specifically for the ICEpower amps that allow bridging such as the 125ASX2 and 250ASX2: The principles are exactly the same as decribed above and as showed in the datasheet, but there is a “BTL sync” pin that should be pulled low as well. The reason is that the ICE-amps have load-dependent switching frequencies and the BTL sync pin ensures that the switcing frequencies track each other (with a fixed offset between them), no matter how you load the two channels. This prevents any risk of the two channels interfering with each other, even under adverse load conditions. Would it work without this connection – yes it should, but why risk problems that are so easily mitigated 🙂

I know this post should have come with a drawing or two, but I can’t be bothered to do that now – plenty of “googlable” examples anyway 😀

Elektor softstart…

Over the summer, Elektor magazine has been giving away a free article every week to subscribers to their newsletter. Most of them have been interesting, if not especially audio relevant. However this week’s article is a new softstart based on an ATtiny microprocessor with IR functionality, ability to learn remote commands, audio signal detection etc. (so actually more of a control center than purely a softstart)

I haven’t looked through the details of the project and the software yet, but the article includes some softstart-math which is nice to have, and the combined project might well serve as inspiration for how to implement the features you want into your own project (because at least I honestly don’t see myself needing everything on offer here). You can find the project information here but be quick to grab it and save it to your computer because I expect it will be behind Elektor’s paywall again soon 🙂

Image (C) Elektor

The hybrid that wasn’t…

The Borbely Hybrid amplifier continues to be one of the most popular projects I’ve posted based on site stats. What I didn’t realise until very recently was there is also a “non-hybrid” version of the hybrid published by audioXpress, with the glass triode replaced with a dual JFET.

I don’t really need any more headphone amps, but since the basic circuits are completely identical it was too easy to spend a couple of hours converting my original hybrid layout to an all solid-state PCB version and the finished article just showed up.

The boards are smaller at 2.9” by 3.6” but otherwise it’s pretty much the same. I kept the dual footprint for the input JFETs because I still have some stock that I wanted to use and because it gives the greatest versatility in my opinion, but a more “future-proof” solution would be to use the LSK489 which is current production and available in both a metal can TO-71 and a standard SO-8 IC packages.

The original 2SK389 dual JFETs are of course nearly completely unobtanium, as are the single 2SK170s that can be used instead (although fakes still abound). The recommended output devices are the (equally unobtainium) 2SK2013/2SJ313, but it should work with IRF(9)510/IRF(9)610 pairs as well – despite the much higher Vgs of the IRFs.

Another test will be to see if this version is more well-behaved with respect to DC-offset than it’s tubed brother. If not, I guess there is always the option of using a delay-circuit but it would be nice if that wasn’t required.

Hope to have time to put one together within a week or so and see if it works :).

ESP/Arduino accessories…

As I’ve spent more time working on my “IoT-T” design – I really should publish the files for that soon – I’ve found myself making a few additions to the main board. They are small extension boards that add functionality without cluttering up the mainboard – in no particular order:

– Different breakout boards to convert the I/O pins on the board to screw clamps for prototyping use or for easier connections.
– Board to add opto-isolation to a pair of digital out pins.
– A smaller watchdog timer to reset an IoT-device that is located remotely in case it freezes up for some reason.
– A breakout board for DS18B20 “One-Wire” temperature sensors.

Alone these boards are not really very impressive, but as extensions they really add to the versatility of the IoT-T main boards and they allow the mainboards to stay simple and universal. Of course these add-on boards can also be used for straightforward experimentation and prototyping, so I’ll be building a few extras to keep on my desk as well.

Project files: Filtered IRM power supplies (part 1)

As promised a while ago, here are my designs for the “filtered” power supplies based on IRM AC/DC modules. These are excellent for adding compact and powerful single and dual supplies which still have a reasonably good performance to any small preamp/headamp amplifier.

Read more of this post

A new flavour…

I finally managed to get an opportunity for a taste of a type of class D that I haven’t tried before – a Pascal Audio S-Pro2 module. I unfortunately missed buying a small lot of these a couple of months ago at a price which were at least the deal of the year if not the decade, but here was a single module at a reasonable price on a local classifieds side and so I could not resist buying that instead.

Pascal Audio is a(nother) Danish amplifier module company. It was actually founded by a group of ex-ICEpower people and although there are some clear similarities in product portfolio and product thinking, most of the Pascal products are focused on professional, PA- and musical instrument applications. This does show in things like power levels, channel configurations and module features. However, Pascal have also managed to creep into several hifi brands including Gato Audio, Jeff Rowland Design and many others – even edging out ICEpower from some of them. Irrespective of that, the quality reviews I’ve seen range from app. “massively better than ICEpower and Hypex” to “horrible sound and very poorly engineered”. As usual, the truth is probably somewhere in the middle but I guess we’ll see about that.

The S-Pro2 is a two-channel amp with onboard PSU that will do around 500W/channel in stereo and 1000W in BTL. This means it’s providing roughly twice the power of an ICEpower250ASX2 in almost the same form factor, making it (supposedly) the smallest 1000W amp on the market. As you can see from the pictures this version is an OEM-version without the usual aluminium base plate, which actually doesn’t bother me since it should make mounting the module to a “proper” heatsink much easier (and yes, even at 90% total efficiency a 1000W amplifier & PSU combo is still going to need pretty serious heat sinking if you want to get close to full power!).

While I do (sort of) have a specific project idea in mind for this module it’s going to take a while. First step is to (attempt to) develop a proper adapter PCB for the 26-pin signal connector to break the various module connections out to something that is easier to work with. Once that is done I’ll move on from there to some real testing, but that will definitely take some time 🙂

Project files: Electronic Load files (untested)

My post on Kerry Wong’s electronic load continues to be one of the posts that consistently drives referrals and traffic to the site, even four years on from publishing it. I guess this is more of a testatement to Kerry Wongs’s original idea and design than to my efforts, but it also means that regularly someone will email me and ask what happened to the project. For the last few years I’ve only really be able to answer “not much more than what you can see” and “no, I am not expecting to do anymore about it in the foreseeable future”.

However, I have been thinking that I should get a grip on myself and at least package up my design files so that someone else might be able to carry the project forward forward. Last week another email from a reader managed to prod me just enough that I finally relented and started doing just that, so here I am publishing something that I generally do not like – a completely untested design. However, i do hope that someone can continue with the design and get it tested and running because I’m sure it is worth it (and if at some point I need a load myself, I will be grateful that someone else has done the work 😉 )

Read more of this post

More MeanWell IRM PSUs…

As regular readers will know I have been using the Mean Well IRM-series AC/DC PSU modules quite a lot. There are many positives to these modules (size, cost, standby consumption etc.) and only a few downsides, mostly the limited output voltage options and the rather noisy output. The noise doesn’t matter so much when used for auxiliary supplies, but for things like preamps, headphone amps etc. I’m sure it’s possible to do better.

However, because the IRMs are a switching design with a relatively high switching frequency (app. 100kHz), cleaning up the output on the IRMs should be relatively easy with just a passive “Pi”-filter (CRC or CLC). Because of the high ripple/noise frequency even a low series resistance/inductance and a little bit of extra capacitance should have a dramatic effect. Some time ago I started making a single-rail PSU based on an IRM-module for a clean and compact 5V supply for my “Music Box” project, but I realised that actually these would be applicable in many projects so I ended up with several different versions (single/dual, different power levels, integrated/separate filtering etc.)

As I’ve just received what I think is the last variant, an integrated dual version with space for two 15/20W IRMs, now seems to be a good time to collect everything and release the project files so they will be up as soon as I am back from my summer holiday in a week or so. Until then, I hope you enjoy the summer 🙂

Experimenting with ESPs… (part 2)

One of the few projects that has moved a little lately is my ESP-based IoT-experiments (which started here). As mentioned then, I had just managed to crack how to do the mains-powered PCB layout I originally wanted to make so that’s what I have been spending time on building and refining since. Having a mains-powered board makes more sense when you need mains power for a relay anyway, otherwise a plug-in USB supply is just as good (or actually better/safer). The board is shown here in full prototyping mode, it is going into a case – of some sort – very soon.

Apart from adding mains power to the board I also removed the original DHT22 sensor and replaced it with an off-board BME280 instead. That was super smooth and it works even better than the DHT, not to mention that it also measures barometric pressure. I’ve been looking at other sensors as well (UV, air quality, light intensity etc.) but they don’t really make a lot of sense for my immediate application (which is remote monitoring of temperature and humidity in my basement).

Since I finished my original version I’ve made a few enhancements to the software and so now I’ve got the code for both LCD and web-UI mostly finished and especially the web part was a great learning experience. As mentioned in the previous post, it’s also a learning experience I am not sure I would have been able to complete without the help of the excellent ESP- and Arduino tutorials by Rui and Sara at, so obviously very grateful for those.

Now I can still do more improvements to the software but instead of picking at it for another six months I think I’ll try and package it up shortly and then publish it here so that someone else can hopefully have a go at it as well. Stay tuned! (but as usual, don’t hold your breath while you wait…)

Back to the future – of DACs?

For pretty much as long as I can remember, there has been a DAC “arms race” going on where manufacturers competed to give the biggest numbers – 16, 18, 20, 24, 32 bit resolution and 48, 96, 192, 384, 768 kHz sampling rates. Irrespective of whether you had the source material to make use of these massive numbers – or even whether the laws of physics made them obviously pointless in the real world, the “bigger number is better” philosophy was still adhered to.

For a few years though, some people have been going “back to basics” with DAC chips such as the TDA154x and others, as wells as discrete “R2R” dac designs such as the Soekris boards and the Schiit Multibit converters. These designs often don’t offer the full resolution of high-res material, but if you don’t care about that – or if you are still listening primarily to 16/44.1 material – then they offer something else, namely a different (many would say “better”) sound. I’ve previously written that I find the ESS ES90xx dacs to often be very impressive at first, but after a while I get a bit of listening fatigue. It could be me or my imagination, but it’s happened enough times that I start to see a pattern (at least when system matching doesn’t mask it) and so I have been trying to stay clear of ESS-dacs and see what else is on the market.

Now, as usual for this I don’t really need another DAC, but I am still curious 🙂 Not quite curious enough to splash out on a Soekris board to play with to be honest, but still curious enough to clicking the “buy it now” button on ebay for this board. It’s based on the AD1865 IC in dual-mono configuration and once again what pushed me over the edge was that I could get a half-finished board with SMD components soldered, so I had some influence over the design of the board.

Assembly was quite easy because I only had to pick and mount the through-hole parts, but even that was enough to make me remember how much I hate black PCBs. Not only because the seller showed that the board was green in the picture, but also because this is a matte black finish that I haven’t seen before and which makes it 99% impossible to see any traces on the board. Fortunately everything worked the first time, but if it hadn’t I am not sure I would have bothered with that much troubleshooting before giving up. Manufacturers: I know green PCBs are considered “boring”, but they WORK! or if you desperately want something else then blue or red still allows you to see the traces, so please stick to those colours. (OK, rant over 🙂 )

The board has one coax input. I opted for a BNC here because that was the socket I had on hand and that might end up being a mistake, but it works now (with an adapter). There’s also an I2S input meaning it’s possible to connect a second source such as a USB card or similar. As usual, I’ve only done basic listening testing with “baseline” I/V opamps (the OPA2134 and the LME49710) but it’s definitely not making a bad first impression 😀

Not quite sure what to do with this one yet, but possibly a “Music box” version 2? Anyway, more listening impressions to be added later I guess.