Taobao shopping for 'digital' amps

A number of interesting amps have caught my eye on Taobao of late, they (except for the last) are true digital amps in the sense that they're 'power DACs'. The first is using ST's STA333W which is a phenomenally cheap fully-integrated digital amp chip, hence the very low price of this board - https://item.taobao.com/item.htm?spm=a230r.1.14.19.4wjcvg&id=526018865372& ns=1&abbucket=13#detailhttps://item.taobao.com/item.htm?spm=a230r.1.14.19.4wjcvg&id=526018865372& ns=1&abbucket=13#detail USB input, speaker out but needs an external power brick of some sort, up to 20V.

The second one's a boxed product, albeit with an external power supply. Its using Intersil's D2-audio chipset (D2-41051), the first implementation I've noticed on Taobao. It uses discrete MOSFETs in the output stage. The enclosure and controls look to be in the PS Audio territory, a cut above the usual Taobao cheap stuff - https://item.taobao.com/item.htm?spm...cket=13#detail

The third item doesn't quite count as a digital amp as its using a normal classD amp driven by an on-chip BB DAC, its nevertheless an interesting proposition from SMSL using the TAS5766 - https://item.taobao.com/item.htm?id=...9828.17.c0bLil

Update - I spent a few hours yesterday with the first of these three boards, based as it is on STA333BW (datasheet here - https://www.st.com/content/ccc/resour...CD00166760.pdf)
Right out of the box I fitted some step-down transformers and connected my SuperLux HD688Bs for a quick listen. Very addictive, most particularly in the area of dynamics, there's an effortlessness about this sound which I haven't encountered before on any of my amps or DACs.

I figured this has potential so I ripped out the supplied filter components (22uH chokes, 470nF film caps, 6R2 damping resistors with 100nF in series) as those are all designed for a speaker load of 8ohms and I decided to optimize for a 600ohm load, taking care of all other lower impedance headphones via an output transformer. This required chokes of 1.6mH (1mH and 680uH in series, TDK SLF7045) and I used a paralleled pair of 6n8 NP0 ceramics as the main output filter. The damping resistors became 430ohm fed via 10nF.

One further mod I've made is to ditch the 470uF/25V main rail caps which look to be Rubycon and measure about 70mohms for ESR. Instead I've fitted some NCC 2200uF/25V which are taller and fatter.

The output transformers are the largest ferrite cored transformers I've wound so far, based on PQ50/50 cores and have 1660 turns of 0.25mm for the primary and the secondary is 700 turns (with several taps for varying impedance phones) of 0.38mm wire.

End result - audio nirvana I can't stop listening to my CD collection all over again...


[COLOR=Black]Update 2 - I wanted to hear how this board (configured as a DAC, not an amp) sounded on speakers so I've been playing with adding additional filtering. Turns out that with my Xindak amp (it has an opamp front-end) that simply the 2nd order LC filtering isn't sufficient to prevent quite audible IMD on complex material (massed violins, choirs). Hence I now have 4th order LC filtering - the first stage which is on the PCB I have modded to operate without snubbers and the loading comes from an off-board second LC stage. The transformers (now PQ4040) follow this extra stage of filtering, fed via customized common-mode chokes. The need for the chokes arises because my PC's a major source of CM noise - connecting it to my amp via the USB cable directly tends to eat up the dynamics and diminishes the soundstage depth.

Result so far - jaw-dropping dynamics but clearly the amp's internal opamp (despite being the best known to man HA5222) is a limitation (or its local power supplies are) so I desperately need a better amp. Might try my TDA8932 transformer paralleled creation....

Update 3 - part of the IMD problem is coming from the internal upsampling filter on the STA333. Its possible to bypass this by feeding in 88 or 96k material. So I upsampled some tracks using Audacity and played those - its like a veil being lifted, what I had thought was a deficiency in my analog filtering turns out to be a problem in digital filtering. In my own implementation of this chip I'll put in an ARM Cortex M-series CPU to act as controller and upsampler. Then I can have whatever filter I want to do the first stage upsampling.

Incidentally another interesting looking board on Taobao here, this one with display and remote too : https://item.taobao.com/item.htm?spm...cket=13#detail

Update 4 - I've now got quite a collection of these boards (six in total) and experimenting with various output filtering options (read - different inductors and filter orders). What I've found is I no longer need better headphones than the Superluxes - its apparent all the issues I had put down to these in terms of 'jangles' on dynamic piano material were down to the electronics, not the cans. With this board driving them the sound is effortless and without noticeable artifacts. I plan to reduce the size of the output trafos to make this a portable proposition next...
I also have on order the second link (the Intersil D2 based amp). Will write about that in another post I think when it arrives.
I hooked up one of the fresh new boards direct to my speakers to see how it would sound. In a word, dire - very 'shouty'. Less said about that, the better. The power supply I was using wasn't up to much mind. Longer term I would like to see what noise the amp generates on its own supply when connected to a low impedance load. With the headphones connected via transformers and my AC voltmeter monitoring the supply the audio bandwidth noise is under 1mV with music playing.

Update 5 - The STA333 chip itself tends to run fairly warm, not so hot I can't keep my finger on it long-term but its fairly uncomfortable to hold it on so I wondered what the reason for this is. First guess is that the output inductors I'm using are rather lossy. Those inductors which have plots for loss vs frequency (the SLF7045 is one of the very few that do) indicate substantially increased losses at the switching frequency (350kHz) used here. For example the 1mH in this range is 2.4ohm at DC which rises to 25ohm at 350kHz. Coilcraft's MSS1210 range provides no data like this. So I figured it was worth trying out some litz-wound inductors where in place of a single wire a bunch of thinner insulated wires is used instead. Making litz wire is very fiddly - not just because of the very small diameter of wire (0.1mm maximum) but because of its tendency to curl itself into knots when handled. Having played around with diameters of wire down to 0.06mm in bunches, using it does lower the subjectively perceived temperature of the chip so I reckon its worth the effort.

Update 6 - the Intersil D2 amp has arrived. Can't say as I like the sound, there's depth but only in the bass (which sounds like it goes pretty deep). The HF sounds like its coming out of the drivers themselves rather than set back in the soundstage. This is normally a sign of higher frequency artifacts. meaning the power supply impedance is too high at HF. I'll add some low ESR caps to the rails and see if that improves the HF...
(see my more recent blog post for how to fix up the SQ of this amp)

Update 7 - the link to the D2 (second one) now shows that vendor out of stock. There are plenty of others advertising the same product but in general their prices are higher, sometimes substantially so. Here's one which isn't too much higher - https://item.taobao.com/item.htm?spm=a230r.1.14.4.fwD6Ip&id=541453178890&n s=1&abbucket=13#detail.
With this amp less available I plan to try out this one, based on STA326 - https://item.taobao.com/item.htm?spm...cket=13#detail