Haven't heard it but the SMSL Sanskrit 6th anniversary looks like a nice DAC.
Amazon.com: SMSL Sanskrit 6th 32bit/192kHz USB/Optical/Coaxial to Analog Audio Decoder (Black): Electronics
SMSL M8a as well. It's $250. Has DSD capability, if that interests you.
Amazon.com: SMSL M8A 2nd XMOS ES9028Q2M 32Bit/768KHz DSD512 USB DAC Optical Coaxial Decoder: Home Audio & Theater
Amazon.com: SMSL Sanskrit 6th 32bit/192kHz USB/Optical/Coaxial to Analog Audio Decoder (Black): Electronics
SMSL M8a as well. It's $250. Has DSD capability, if that interests you.
Amazon.com: SMSL M8A 2nd XMOS ES9028Q2M 32Bit/768KHz DSD512 USB DAC Optical Coaxial Decoder: Home Audio & Theater
Another option comes to mind - one of the biggest weaknesses of most cheap USB DACs is the lack of isolation from the computer. You can fix this by getting a USB-Toslink converter then feeding the optical link into a separate DAC. Here's a DAC with Toslink input which I've bought a few of (and modded extensively), the Lite DAC-AH : 【北京安润】丽特LITE DAC-AH发烧音频HIFI解码器-淘宝网全球站
Its yet another 8*TDA1543 design
We're both suffering from the same bug which compels us to acquire DACs just for the hell of it
I've been corresponding with Alexandre who I think is one of the true experts on TDA1543 and he mentions he thinks there's a problem on one channel of this chip, perhaps a bug in the silicon? The 1543 has the advantage of more output current so with 16 devices you'll have something like 35mA to play with, enough to drive some headphones directly without any amp. Current mode is supposed to sound awesome but I've not figured out the best way to do it yet....As to whether passive I/V matters - yes, I've found it doesn't give the best bass, no matter how many chips are paralleled. Nowadays I'm only using active - my latest DAC project is up on Hackaday : Audiophile-sounding DAC for almost no money | Hackaday.io
Thanks for the generosity Richard!
The thing with the left channel of the tda1543 is noise, probably crosstalk from the digital inputs. Why is it noisier than the right channel? Layout I suppose.
I first noticed the left channel noise when I removed the preamp from my system, driving the amplifier directly from the TDA1543 (passive I/V) and using digital attenuation. In this config you are really listening to the low level of the chip, and the sound is quite good with a selected chip, but the noise on the left channel bothered me at night when ambient noise is low.
There is also a good deal of variability between chips and datecodes. Some are more musical to my ears but I haven´t yet measured the differences. I have tried two paralleled tda1543 on more than one ocasion, scaling the resistors accordingly, but went back to a single selected chip per channel. I´m not giving up yet, though; will try two paralleled chips again, with slight differences from my previous attempt. Maybe two selected chips in parallel, even though I don´t like to solder my favorite chips
I believe clock jitter makes a difference and I have always used my own clocking scheme, which I´m afraid is not very applicable to others (I use an ancient RME soundcard, toslink I/O, xilinx cpld and a reclocker board with master clock before the dac board).
One more thing about the 1543: the sample out is delayed by one bitclock on the right channel iirc (could be the other way around). Anyway, one bitclock shift.
The thing with the left channel of the tda1543 is noise, probably crosstalk from the digital inputs. Why is it noisier than the right channel? Layout I suppose.
I first noticed the left channel noise when I removed the preamp from my system, driving the amplifier directly from the TDA1543 (passive I/V) and using digital attenuation. In this config you are really listening to the low level of the chip, and the sound is quite good with a selected chip, but the noise on the left channel bothered me at night when ambient noise is low.
There is also a good deal of variability between chips and datecodes. Some are more musical to my ears but I haven´t yet measured the differences. I have tried two paralleled tda1543 on more than one ocasion, scaling the resistors accordingly, but went back to a single selected chip per channel. I´m not giving up yet, though; will try two paralleled chips again, with slight differences from my previous attempt. Maybe two selected chips in parallel, even though I don´t like to solder my favorite chips
I believe clock jitter makes a difference and I have always used my own clocking scheme, which I´m afraid is not very applicable to others (I use an ancient RME soundcard, toslink I/O, xilinx cpld and a reclocker board with master clock before the dac board).
One more thing about the 1543: the sample out is delayed by one bitclock on the right channel iirc (could be the other way around). Anyway, one bitclock shift.
I can vouch for the musicality of this one (only after some warming though, let's say half an hour on)
HiFi AK4118+AK4490+XMOS decoder Coaxial Optical USB DAC 192K 24BIT
I have tested it with lead acid batetries and it gains a lot more.
For AC power I would buy this (will eventually do) from the same firm
15W Ultra-Low Noise Linear Power Supply HIFI DC Regulated PSU 5V-24V Selection | eBay
I tested it with audiophile linux and deadbeef, sounds extremely well with that RT kernel, specially compiled for audio, coupled with a minimal fluxbox interface (almost all computing power is dedicated to audio, with audio getting absolute priority by scheduler, practically the system is transparent, cannot vouch for the hardware part but at least software is tightly tuned)
Also with a stock arch linux distro and deadbeef it works very well.
Audiophile linux is not well develeoped though, it easily gets problematic.
Arch is already very good, not bloated.
All my music is classical, with vocals tested with Christina Pluhar recordings (arpeggiata), very demanding for a system
HiFi AK4118+AK4490+XMOS decoder Coaxial Optical USB DAC 192K 24BIT
I have tested it with lead acid batetries and it gains a lot more.
For AC power I would buy this (will eventually do) from the same firm
15W Ultra-Low Noise Linear Power Supply HIFI DC Regulated PSU 5V-24V Selection | eBay
I tested it with audiophile linux and deadbeef, sounds extremely well with that RT kernel, specially compiled for audio, coupled with a minimal fluxbox interface (almost all computing power is dedicated to audio, with audio getting absolute priority by scheduler, practically the system is transparent, cannot vouch for the hardware part but at least software is tightly tuned)
Also with a stock arch linux distro and deadbeef it works very well.
Audiophile linux is not well develeoped though, it easily gets problematic.
Arch is already very good, not bloated.
All my music is classical, with vocals tested with Christina Pluhar recordings (arpeggiata), very demanding for a system
Is the left channel thing for a specific 1543 implementation? Or inherent in the chip itself?
Does this DAC allow you to select the digital filter (roll-off slope) that you want to use with the AK4490?
Also, do you notice any noise coming from the USB connection (ground noise)?
Last edited:
Agree. The USB receiver chip is as important as the DAC chip itself. M8A has XMOS and the current output ES9028Q2M, board design is top notch too. To be honest none of the DAC in your list is capable challenging it.
Last edited:
It´s nothing to worry about if you don´t hear it. You´ll only hear it if you ditch the analog volume and connect to power amp directly. In that case yes, the left channel is always noisier than the right. Not thermal noise like a transistor, but signal correlated noise, therefore the hunch about the chip layout.
The CTO from ESS gave an interesting talk which is on youtube... 5 years ago, but great content. Search for Martin Mallison. Funny how he describes some listening tests. Apparently his hired ears could hear the flaws with delta sigma even with source material such as New York New York by Frank Sinatra. Their proprietary modulator design fixes it, but he doesn´t claim superiority over true multibit dacs. He hints about state space and how their modulator transitions without the state variables going all over the place.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.