Having built a pretty big/powerful boombox (the Mark I), I found I don't use it very often as it's too big and heavy, and having the battery external to the box is a bit of a pain. I want a Mark II which is smaller, lighter, fully integrated, but still powerful enough to 'fill' a room.
The goal is to come up with something like this:
- roughly 300 x 150 x 120mm - <5kg approx. - 15+15W or more? - good bass! - minimal controls - internal LiPo/Li-ion - low quiescent consumption - solar panel on one face (to extend run time when away from civilisation)
Total parts cost is about £50, not including plywood, screws, etc for the enclosure which I had lying around, plus various machine tools and cutters I found I needed.
TPA3116-based bluetooth amplifier module (£10)
coaxial 4" car speakers (£20 for two)
(optional) solar MPPT 3S Li-ion charger module
CN3277-based board from eBay
22/8: Just a few snags to sort on the enclosure, and some kind of grille to fit to the little magnets on the front, but we're nearly there...
24/8: ...but! I've just realised I made a mistake in the first op-amp gain stage (what was I smoking?). When I tried to make the gain variable so I could add a volume control, it didn't work as expected because I'd forgotten the op-amp is a summing and differencing stage. Need to find a way to make the gain variable with only a single-gang potentiometer...
30/8: Now with 'gold' cloth grille to keep small fingers out of the speaker cones. Held on with magnets so it's easy to remove.
6/8: Having spent a couple of weeks trying to source a Li-ion pack of any kind, it's time to write up some ideas. I decided to aim for 11.1V (3S) based on the peak power I'd like the amp to deliver. The next challenge was to source either a 3S pre-made pack, or individual Li-ion cells (e.g. the ubiquitous 18650s) and make up a pack. Here is a summary of my findings:
pre-made 3S packs (made from 18650s) appear to be pretty rare, unless you are prepared to pay £50+ for a top brand
the cheap Chinese pack I bought on eBay never arrived
...which may be because UK postal carriers don't carry Li-ion cells unless they are contained within a product of some kind
making a pack requires connecting the cells either by a) soldering, b) spot-welding or c) using spring terminals/magnets
I didn't want to waste time making a spot-welder only to find that I cooked the cell internals (see here), and likewise soldering directly to the cells. Spring terminals is an extra faff too. The best option IMO is soldering tabbed cells.
tabbed 18650s on sale in the UK from reputable suppliers are also pretty rare, unless you're prepared to pay £5-6 or more per cell
so the remaining option is to break open products containing 18650s and harvest the cells (with tabs already fitted of course!). The best sources for my needs were either Dyson vacuum cleaner packs (six cells, 6S1P) at ~£20 each, power banks with 8 cells (e.g. this one for £23), and aftermarket laptop batteries (6 cells, 3S2P) at £9 each.
I bought a couple of used Dyson packs, only to find they were mostly junk (new ones would have been fine as they use Sony 1900mAh cells), and a new laptop battery with a claimed 5400mAh capacity which so far looks pretty good. Will report back with measured capacity.
7/8: How disappointing: measured at only 3400mAh. Think I'll try another charge/discharge cycle and see if it improves. But at least the pack works and stays nice and cool even during quite fast charging (0.8xC). The MPPT module, on the other hand, got pretty hot even in free air. Once it's in an enclosure it'll be verging on too hot. I might change the max. charge rate from 2.8A down to 2A if I can.
9/8: After a second charge-discharge cycle the pack managed 3800mAh. Probably good enough for my needs, but I'll steer clear of Lenoge batteries in future!
25/8: Ha ha ha! The "3800mAh" pack I bought from China finally turned up today, 3 weeks late and long after I'd had the money refunded. Marked on the packet as a "$3 phone case", that's about all it's good for. I measured the capacity at a measly 860mAh. What a joke!
7/7: My 'milling' rig is ready! A hilariously low quality setup for machining MDF and plywood.
15/8: Loads of progress on the enclosure due to a) finally deciding not to include solar panels (so I don't need to wait for them to arrive in the post) and b) a few days off work. Photos in no particular order...
17/6: Still searching for a 3S pack of 18650's that looks like it might actually have the claimed capacity, but there are so many lies on eBay...
But I have now chosen some solar panels (picked to stack side-by-side to completely fill the top panel of the box) and an MPPT charger module. I'm keen to see how long it takes to charge the battery on a sunny day...
19/6: Finally chosen a cheapo 18650 pack from eBay. It has a claimed capacity of 3800mAh (ha!) and protection circuit. We'll see what turns up.
25/6: Had the amp running all weekend indoors, at a comfortable 'background' music level, and measured the current consumption at ~50mA, i.e. 0.6W total power dissipation. If the solar cells manage the claimed 1W max each (x3), then I should get break-even charging/discharging even on a cloudy day.
11/7: Just a quick update on this. The solar panels arrived the other day, and I quickly cobbled them together with the MPPT charger module and three Li-ion cells I had kicking around. Long story short: the Voc of the panels (in series) is about 16V, so their Vmpp is about 0.8x16=12.8V. However, the module is set up for an 18V panel and its Vmpp is set (using resistors) a bit higher. I've added a trimmer pot to be able to vary the module Vmpp and tune it for maximum charging current. We'll see how well that works the next time it's sunny. For now, it's cloudy and the panels are only managing about 5-15mA. I'm not convinced they're going to be much better in the sun...
12/7: Had a bit of sun today, and thus the chance to measure Isc and peak charging current of the solar panels. Sadly, the most I managed was about 110mA into a Li-ion at 11.5V, in other words 1.3W. And that was with the charging module's Vmpp adjusted for best current. For a panel with a total area of 0.02m^2, and assuming say 1000W/m^2 solar irradiance, that's an efficiency of only 6.3%! Rubbish! Some of that will be the buck converter in the charging module, but that ought to be 70-80% I'd have thought. Back to eBay to find a better panel...
24/7: Just remeasured the solar panels in bright direct sun, without the MPPT board. Best I managed was Voc = 17.5V, Isc = 138mA. Peak power therefore likely to be around 1.5W for the string of three (they're meant to be 1W each!).
30/7: In a moment of (almost) madness I bought two used Dyson cordless vacuum cleaner batteries. Each has 6 li-ion cells in series, so the plan is to cut the packs in half, rewire them as 3S2P and add a new protection circuit. Sadly, of the 12 cells that arrived, only 3 are any good. Still, cheaper than buying new, and I've now got two plastic cages for holding the cells.
1/8: Finally had three 18650 cells, with their protection circuit, charging via the MPPT controller - all seems to be working fine! The MPPT controller module (based on the CN3722) is set to have a maximum charge current of 2.8A, which is a bit high for my rather small and sickly 3S1P pack (about 1500mAh), but will be ok when I have a bigger 3S2P pack of bigger capacity cells. For now I can use a power supply with current limit to avoid charging the pack too quickly.
2/8: Another observation: When the charging supply is <Vmpp, the charger module simply stops charging. This is a problem as I was hoping to set Vmpp to, say 14V, to get the most out of the solar panel, but still be able to charge from a vehicle cigarette lighter socket (12-13.5V ish) without any other circuitry. Now it looks like I'll need a separate boost converter to get up to 14-15V. The upside is that with a dedicated boost converter, I can also charge from 5V USB. Or anything really.
15/8: After a *lot* of agonising, I've finally decided to do away with the built-in solar panels. This is partly because I can't wait for the new ones to come in the post (I really wanna get this box built!), partly because the benefit is...
I chose the JBL CS742's as they are small (4"), cheap (£20/pair), have relatively high sensitivity (90dB/W at 1m, claimed) and low bottom-end response (claimed 75Hz).
This log documents the design of the enclosure to take the two speakers.
12/6: I've been playing with this bodged-together enclosure for a while now, and learned a lot. It's got a removable top and moveable back so I can change the volume easily. It's also got holes for two 20mm ports, and I've cut some 20mm tube to various lengths. The upshot is: completely sealed is best, ports do almost nothing at this size! Plus, my coaxial speakers have no bass response to speak of. Probably not surprising.
So, big 180 degree u-turn. The new plan is to use a single 4" coaxial mid-top and a 4" sub. I can use the two amplifier channel to drive each separately if I mix the left and right down to mono. A passive radiator or two will hopefully solve the problem of the port doing nothing. Parts on order, just have to wait...
19/6: Started thinking about final enclosure design. Solar panels stacked end-to-end:
...or if I turn the solar panels around to make it short and fat:
Think I prefer the latter.
6/7: Have now fitted the passive radiator and swapped out one of the coaxial speakers for a dedicated sub-mid. I've also partitioned the box into two spaces for each of the speakers. The amp is now wired as mono, with one channel driving the sub (with a low pass filter response) and the other the mid-top.
The box generally sounds pretty good. Bass is not bad, but let down by the crappy speaker I bought (it was very cheap!). I've seen one for a bit more cash with an xmax of 4mm and an Fs of 73Hz, so I might try that - plus wind the gain in the amplifier up a bit. The passive radiator works surprisingly well...
7/7: By way of some kind of semi-scientific method, I measured the response of my speakers plus a Gale Gold monitor for reference. The response of the microphone is not compensated - I reckon mostsome of the low-freq roll-off is due to the mic, not the speakers (sealed enclosure roll-off should be 40dB/dec and I've measured 60dB/dec). But I can at least compare them:
There's some peaking at ~200Hz (Fs of the speaker?) that I'm going to have to watch out for. Hopefully the new driver will be a bit flatter/have a lower Fs.
Just ordered the new speaker: it was a toss-up between the Peerless SLS-85S25CP04-04 and the Visaton KT 100V. I went with the Visaton after a bit of modelling in WinISD shows it's apparently better at <100Hz response in a small volume.
11/7: Got the new Visaton speaker today - so much better! Check the measurements (now normalised to the response at 500Hz)
It's about 10dB better than the JBLs and the cheapo 4" driver at low frequencies, and not far off the same response as the Gale monitor, at least down to 40-50Hz. Plus when driven hard it doesn't tear itself apart like the other one - lots of excursion.
14/7: I've cut down the prototype box to roughly the finished size, and am planning on taking it camping this weekend as a test run. Here it is:
Out of the box, the amplifier module has some minor flaws:
- there's no volume control (except from the device streaming the audio)
- it's a bit noisy at low volume levels (background hiss)
- there's a 12V linear reg and reverse-polarity protection diode on the board which do almost nothing AFAIK except waste power and voltage headroom
- the audio response is (probably) flat across the band and I'll be wanting some bass boost
So this log covers the mods I make to fix these problems.
First off, quiescent power consumption. Baseline: 14mA (after BT pairing) from 15V supply => 210mW. It looks like the 12V reg is only there to power the op-amp stage and the 3V6 reg which feeds the BT module. The TPA3116 seems to be connected directly to the main supply (via the diode), so it's not as bad as I thought. Still, the NE5532 op-amp will happily run from 22V (way more than I intend to use) so we might as well skip out the diode (saving 10mW) and 12V reg (11.5mA x voltage drop, in this case 3V so 35mW).
The BT module ("BT64X", datasheet here) claims 2.8-4.2V operation but doesn't specify current. However, when it's not actually streaming, the current is very low (per CSRA64215 chipset datasheet). The bulk of the rest of the 11.5mA I measured must be the NE5532 - spec'ced at 8-16mA no load.
Not really much scope for further power savings, except running the BT module directly from first LiPo cell and doing away with the 3V6 regulator.
10/6: Did some measurements today. The op amp stage combined with the gain in the 3116 is way too much, easily clipping the output long before the volume on my phone is anywhere near maximum. In fact about 10x too much. One solution is to change the opamp stage to 0.3x (from 3x). This should reduce the hiss and the very loud bing bong tones from the BT module on startup. Or get rid of the opamp altogether and save power?... Oh, and yes the response is essentially flat down to the roll-off below 50Hz due to the series caps (at the BT module o/p and at the 3116 input).
11/6: Bypassed the diode and 12V regulator. Changed op-amp gain from 3x to 0.3x - gain looking a lot better now, music at 'normal' levels only just clips at full volume when running from 3S LiPo (11.1V) supply. Quiescent current: 11mA paired but not playing, 49mA playing music at (almost) zero volume level. Will plug speakers in later and see if the hiss is better, and if it's still loud enough! (update: yes, it is just about. Hiss is almost acceptable for bed-side use, and fine for general use)
12/6: Had a quick look at noise levels. I get ~15mVrms across the load when music is playing but at zero volume, i.e. just noise. If I short across the BT module outputs, this value does not change, meaning the noise is not originating at the BT module. If I remove the NE5532 and short an output to GND, the noise in that channel drops to ~10mVrms, suggesting that some comes from the op-amp and upstream of it. The remainder (i.e. the majority) either comes from the 3116 itself (which I doubt very much) or poor layout of the board causing pickup of switching currents. There's not much I can do about that, short of carving up ground planes, so I might just leave it. Had the designers simply routed the BT module diff outputs directly to the 3116 diff inputs, it would probably have been a whole lot better! Changing to single-ended was unnecessary (no gain required) and made things more difficult...
13/6: A number of people who've reviewed this particular 3116 board have complained about the connection tones at startup. Mainly the...