The coolest Taobao SD card player yet - now further improved
Posted 10th December 2013 at 04:50 AM by abraxalito
Updated 21st September 2015 at 03:23 AM by abraxalito
Updated 21st September 2015 at 03:23 AM by abraxalito
This one rocks - it plays FLAC, WAV and mp3s. The DAC is WM8740 (not so hot) but the designer has taken the power supply very seriously. Just check out all those caps and TDK noise filter on the PCB on the right 
Power comes from a lithium battery behind, so its portable but there's no case - there are acrylic front and rear panels screwed on to the PCB spacers. Two 3.5mm jacks provide headphone and line out.
Its using STM32F407 to do the decoding which is a first for ARM in this sphere and the design is configured that there's an I2S link between the two PCBs. So in theory anyone could design a better sounding DAC and replace that bottom PCB whilst keeping the UI. Which I may just do.... Price is 568rmb (just under $100).
Update - I was a little curious about the PSU so I've pulled out all of those SMT 220uF caps to get a closer look. The block filter's a Murata (not TDK) BNX024 which looks like a wrong choice of part primarily because its rated for 15A. Hence the filtering's not going to be optimum for a relatively low powered DAC/headphone amp. There are also a couple of 100uH bobbin wound chokes in there as series elements. I put some of the 220uFs on my LCR meter - they show hideously poor ESR (about 0.5ohm at 1kHz). Compare this to some Sancon (cooking grade Chinese made) lytics of the same value, case size and voltage rating, just through-hole. They measure 0.15ohms, over 3X better. Moral of this story - steer well clear of SMT lytics! Next I'll remove the BNX and see if I can do better with discrete coils and caps....
Update 2 - the through-hole lytics are in place now. Block filter removed and replaced by a couple of strings of ferrite beads. 100uH bobbin inductors (0.5ohm) replaced by 22uH (0.05ohm). Significant SQ upwards step - some semblance of soundstage now, Pogorolich's piano has moved behind the speakers
Since I took the pic I've added ceramics (10uF and 228F) across the rails of the WM8740 and OPA2604 - this latter one doing the differencing of the DAC's outputs. More apparent soundstage now and HF has been cleaned up a little.
Update 3 - more serious RF filtering to the reg supplying the analog stages of the DAC chip. Piano pleasant enough, decent bloom so I decided to load up the TFcard with a few other selections. Gunter Wand's Mozart 40 - rather a disaster in rendering violin sound. Keith Jarrett's harpsichord (Bach French Suites)- sounds fine until he plays more than one note at a time, then it just falls apart. Rather fatiguing on the ears this one!
Upshot - abandoning the WM8740 and will work on a compatible board with a TDA1387 for this one in due course. I really love the convenience of the hand-held player.
Update 4 - the same guy has a newer version now, price has come down and its got a proper metal case - https://item.taobao.com/item.htm?spm...cket=13#detail
The Cortex M4 CPU has gone being replaced by (probably) a Rockchip SoC. The DAC part remains the same but its added on an S/PDIF out - more bang for the buck this time around. I checked the line out SQ - par for the course for an S-D, might try implementing a balanced output buffer on this if there's enough room inside.
I've checked out the implementation of the output filter now - it was using 2k4 resistors into 3n3 differential mode filtering. Followed by an LME bipolar opamp. The mods I've done to it have turned it into a really dynamic beast on the line output (nearest the volume control) and they're quite simple to do. Just remove all the passive filtering components (0603 resistors of 2k4 and 150R, 0603 capacitors of 680pF, wired 3n3 cap). Then replace them with 100X the original impedance values - so resistors go to 240k and 15k, capacitors to 33pF and 6.8pF. The bipolar opamp will be a tad noisy working with such high impedances so swap that to a JFET input one. I have fitted ISL28210 which is a nice low-noise dual JFET with a decent bandwidth and slew rate. I also slapped 4 10uF 1206s across the supplies of this opamp (which is in SO8, it needed bodging onto the DIL socket by filling the turned pin holes with solder then attaching short thin wires to the previously bent up pins of the SOIC). Dynamics are very satisfying now - they do get lost again on the two headphone outputs though - you'll need an external amp to get the best from this as a source.
I have added the vendor's own picture of the circuit board - the components to be swapped out are those adjacent to the two green box caps - most are directly behind them.
Doing some more listening on speakers rather than headphones reveals that the soundstage depth is good, at least in low frequencies and LF dynamics are awesome. However on massed strings, massed voices there's too much confusion, a little harshness. Impressive sounding but not ultimately about the music on complex stuff. This might be because the opamp loading at 240k is still a bit heavy. So higher impedance buffers prior to the filtering opamps is worth a try. Also better decoupling of the DAC's analog power stages and reference voltage might be explored.
Power comes from a lithium battery behind, so its portable but there's no case - there are acrylic front and rear panels screwed on to the PCB spacers. Two 3.5mm jacks provide headphone and line out.Its using STM32F407 to do the decoding which is a first for ARM in this sphere and the design is configured that there's an I2S link between the two PCBs. So in theory anyone could design a better sounding DAC and replace that bottom PCB whilst keeping the UI. Which I may just do.... Price is 568rmb (just under $100).
Update - I was a little curious about the PSU so I've pulled out all of those SMT 220uF caps to get a closer look. The block filter's a Murata (not TDK) BNX024 which looks like a wrong choice of part primarily because its rated for 15A. Hence the filtering's not going to be optimum for a relatively low powered DAC/headphone amp. There are also a couple of 100uH bobbin wound chokes in there as series elements. I put some of the 220uFs on my LCR meter - they show hideously poor ESR (about 0.5ohm at 1kHz). Compare this to some Sancon (cooking grade Chinese made) lytics of the same value, case size and voltage rating, just through-hole. They measure 0.15ohms, over 3X better. Moral of this story - steer well clear of SMT lytics! Next I'll remove the BNX and see if I can do better with discrete coils and caps....
Update 2 - the through-hole lytics are in place now. Block filter removed and replaced by a couple of strings of ferrite beads. 100uH bobbin inductors (0.5ohm) replaced by 22uH (0.05ohm). Significant SQ upwards step - some semblance of soundstage now, Pogorolich's piano has moved behind the speakers
Since I took the pic I've added ceramics (10uF and 228F) across the rails of the WM8740 and OPA2604 - this latter one doing the differencing of the DAC's outputs. More apparent soundstage now and HF has been cleaned up a little.Update 3 - more serious RF filtering to the reg supplying the analog stages of the DAC chip. Piano pleasant enough, decent bloom so I decided to load up the TFcard with a few other selections. Gunter Wand's Mozart 40 - rather a disaster in rendering violin sound. Keith Jarrett's harpsichord (Bach French Suites)- sounds fine until he plays more than one note at a time, then it just falls apart. Rather fatiguing on the ears this one!
Upshot - abandoning the WM8740 and will work on a compatible board with a TDA1387 for this one in due course. I really love the convenience of the hand-held player.
Update 4 - the same guy has a newer version now, price has come down and its got a proper metal case - https://item.taobao.com/item.htm?spm...cket=13#detail
The Cortex M4 CPU has gone being replaced by (probably) a Rockchip SoC. The DAC part remains the same but its added on an S/PDIF out - more bang for the buck this time around. I checked the line out SQ - par for the course for an S-D, might try implementing a balanced output buffer on this if there's enough room inside.
I've checked out the implementation of the output filter now - it was using 2k4 resistors into 3n3 differential mode filtering. Followed by an LME bipolar opamp. The mods I've done to it have turned it into a really dynamic beast on the line output (nearest the volume control) and they're quite simple to do. Just remove all the passive filtering components (0603 resistors of 2k4 and 150R, 0603 capacitors of 680pF, wired 3n3 cap). Then replace them with 100X the original impedance values - so resistors go to 240k and 15k, capacitors to 33pF and 6.8pF. The bipolar opamp will be a tad noisy working with such high impedances so swap that to a JFET input one. I have fitted ISL28210 which is a nice low-noise dual JFET with a decent bandwidth and slew rate. I also slapped 4 10uF 1206s across the supplies of this opamp (which is in SO8, it needed bodging onto the DIL socket by filling the turned pin holes with solder then attaching short thin wires to the previously bent up pins of the SOIC). Dynamics are very satisfying now - they do get lost again on the two headphone outputs though - you'll need an external amp to get the best from this as a source.
I have added the vendor's own picture of the circuit board - the components to be swapped out are those adjacent to the two green box caps - most are directly behind them.
Doing some more listening on speakers rather than headphones reveals that the soundstage depth is good, at least in low frequencies and LF dynamics are awesome. However on massed strings, massed voices there's too much confusion, a little harshness. Impressive sounding but not ultimately about the music on complex stuff. This might be because the opamp loading at 240k is still a bit heavy. So higher impedance buffers prior to the filtering opamps is worth a try. Also better decoupling of the DAC's analog power stages and reference voltage might be explored.
Total Comments 17
Comments
-
Not so hot, because it's S-D? How does it sound in raw form, what are its problems - does it improve after a solid workout?
FrankPosted 10th December 2013 at 10:41 AM by fas42
-
Guess because its S-D, opamps. Sound compared to Ozone is way forward of the speakers (Ozone is way back from them) - no soundstage. Not unpleasant, just not the recording, coloured. Only listened to one piano track though (Ivo Pogorelich, Scarlatti). Might give it a go on a wider selection of music. Also keen to see just how 'bad' mp3 is
This is most interesting to me as a digital source initially to fit my own back-end.Posted 10th December 2013 at 12:52 PM by abraxalito
-
I really feel MP3 has been given a bad rap, overall - a lot of the problem seems to be that decompressing done on the fly typically degrades the sound, for all the usual reasons. Would be interesting to compare a highest quality MP3, against an off-line decompressed WAV, and the latter resampled off-line to higher rates - and of course the original uncompressed file ...
FrankPosted 11th December 2013 at 05:16 AM by fas42
-
There does seem to be a powerful meme nowadays that 'bigger numbers = better'. Hence higher sampling rates, more bits is sought after, and conversely fewer bits (mp3) is despised.
Doing something on the fly degrading the sound to me implies RFI leakage from the CPU to the analog parts. Definitely poor design!
CPUs are getting less noisy for the same MIPS - I checked out a new ST part only last week - STM32F401 - looks very promising from the power consumption point of view. But so far nobody's selling them that I can see.Posted 11th December 2013 at 04:38 PM by abraxalito
-
Incidentally, I see Joshua's thread is generating plenty of heat and little light (except from J Neutron). Tickled to see that SY remembers an exchange with me for which I was banned for a couple of weeks :
https://www.diyaudio.com/forums/every...ml#post3735362
His recollection of it being entirely at odds with my ownPosted 12th December 2013 at 02:38 AM by abraxalito
-
Care to point to it, :)?
FrankPosted 12th December 2013 at 03:17 AM by fas42
-
You want the original discourse? It was erased of course as part of the 'tidying up' the mods do in such situations.
I was goading him for sure but the stupidity was entirely his. For an example in contemporary discourse watch the court scene in 'A few good men' where Tom Cruise wants to get at 'the truth' from Jack Nicholson. I scented 'emotional blood' and was curious to see if I could get it
Provocative Therapy: Amazon.co.uk: Frank Farrelly, Jeffrey M. Brandsma: Books
Update - had a brief chat with John Kenny - this comment of SY's could be referring to him, not me. The link John sent me is this one - https://www.diyaudio.com/forums/loung...ml#post2567653
Eliza use clearly evident therePosted 12th December 2013 at 03:26 AM by abraxalito
Updated 12th December 2013 at 01:53 PM by abraxalito -
hi Richard,
Can this player beat the ak100/dx50/ or may be tera player?Posted 12th December 2013 at 04:58 AM by jambul
-
Do you mean can the unmodified player beat the Tera? No, Tera is multibit (probably TDA1545) so will beat any S-D based player. But with my DAC/headamp design it would be able to beat Tera I reckon yes. Its much larger of course.Posted 12th December 2013 at 06:33 AM by abraxalito
-
what dac chip will you use?
tda1387?Posted 12th December 2013 at 07:53 AM by jambul
-
More likely TDA1545A as its lower power. But TDA1387 is cheaper and probably would work almost as well.
I have only got ideas about how to do a kick-*** headamp so far. I'll start out by using AD8016 as its specs look really good and I have a whole tube of them. I also want to use trafos but its hard to find small enough ones to fit on a PCB.Posted 12th December 2013 at 07:59 AM by abraxalito
Updated 12th December 2013 at 08:53 AM by abraxalito -
hi richard,
i'm looking forward for this project and considering to have this as well since i didn't have a portable music player for my commuter timePosted 12th December 2013 at 08:21 AM by jambul
-
I'm curious to ask you some questions about what you consider suitable specs for a portable player. Firstly, how big and heavy could you tolerate ? And secondly how long battery life would you need (in hours)? Thirdly what headphones do you use?Posted 12th December 2013 at 08:52 AM by abraxalito
-
for me portable enough to fit in my pocket. fiio x3 or colofly c3 would be ok but on size of ibasso dx100 will be ok too.
well as long the player easy to navigate i'll be fine(doesn't have to be fancy screen like ak100, monocrome like qls it's ok)..no need EQ..battery life min 4-5 hours
i used to take the 50-60ohm asg 1 iem on my commuting time, at home i used to listen 32ohm allesandro..sometime i listen my stax.Posted 12th December 2013 at 09:02 AM by jambul
-
Ah - thanks. Pocket sized will be a challenge but we shall see. I was thinking more like 16hours battery so a person could listen all day if necessary, but since you say 5hours then the battery could be quite a lot smaller. And 60ohms is the highest impedance you need to drive? How efficient are those 'phones - like how much power do they need? Remember this will be an extremely dynamic amp so it won't sound 'loud' like normal amps, even though it'll hit high SPL it will be without apparent effort, due to the low noise.Posted 12th December 2013 at 09:11 AM by abraxalito
-
Well,
5 hours is my maksimum hours i can listen to iem or headphone. more than that i got a headache :-p
well you can make a switch for low impedance and high impedance load maybe ?Posted 12th December 2013 at 09:23 AM by jambul
-
Sounds like the dac and amp you currently have give you listening fatigue. I don't get that with mine - you may well want to listen all day
Yes perhaps there can be multiple trafo taps, just like for a valve amp.Posted 12th December 2013 at 11:21 AM by abraxalito
