Chinese DAC-AH new modifications

[Karstenpg]

Hi

I am starting this as a new thread because the already existing ones on the DAC-AH appear to be rather old.
I have recently aquired two of these units from Ken Chang / Hongkong. A first review of the circuit diagram invites to some other modifications than those already discussed here. Two have come to my mind, which I could read nowhere about so far.

1. The Lite DAC-AH has separate voltage regulators for each group of four TDA1543 chips. Thus it lends itself to a complete separation of silicon for left and right channels. Simply cut the AOL and AOR rails in half and increase the I/V resistors accordingly. This should reduce crosstalk and maybe improve reproduction of sound details.

2. Completely eliminate the Vref rail, actually leaving each Pin7 open. Rewire output op amps as inverting I to V converters like in TDA1543 data sheet. Delete R35 and R36 and connect AOR and AOL rails directly to op amp -inputs. Feed positive bias current via separate 1 K pots from +15 V to op amp -inputs. Bias op amp +inputs to approx. 2 V, fixed. Adjust bias pots for 0 V quiescent output on each channel. Use non polarized output caps or even delete output caps as output will be DC free, if the following equipment permits.
Expected result: improved DAC performance. DAC output sees fixed current sink point and therefore comes closer to what the designers of the DAC probably had in mind, see TDA1543 data sheet. Moving bias current source from within DAC to op amp summing junction eliminates possible cross talk between audio and Iref currents. Leaving each Vref pin open instead of tieing them all together eliminates random compensating Vref currents between the DACs, caused by generic Vref differences per chip and temperature drift.

Has anyone already tried these mods?
DACs run on 6 V, output op amps are AD847.

Thanks for your comments, Karsten

[aparatusonitus]

Originally Posted by Karstenpg

Has anyone already tried these mods?

I have done it my Philips CD614 extensively used to try new thing like this, quick end dirty. Opamp is LM833 at +/-10VDC supply rails, one TDA1543 at 5VDC. Opamp inputs are at 2VDC, but I have to use 10k multiturn pot from +10VDC to bias inverting input at 2VDC and trim output to within +/-15mV DC offset.

I can only confirm your finding sound vice, great mod for sure.

[Karstenpg]

Originally Posted by aparatusonitus

but I have to use 10k multiturn pot from +10VDC to bias inverting input at 2VDC and trim output to within +/-15mV DC offset.

I can only confirm your finding sound vice, great mod for sure.

Thanks for the reply. It is encouraging. I will start with the mods and report my results. Btw: in case of four parallel DACs a 1k trimpot should be required for the increased bias current.

Regards, Karsten

[aparatusonitus]

There is one big problem with my setup...a big ~500mV offset from cold startup, so I have to keep it on permanently to stay within acceptable offset or design active DC correction with an opamp...more work is needed I guess. BTW, bas is fantastic

[Karstenpg]

Have ordered parts for the planned mods. Will start work next week, probably.
In the meantime I have done two things:
1. Started to read (never ending?) thread on Building the ultimate NOS DAC using TDA1541A . Very interesting. Has convinced me that non OS best adheres to Shannon and has most appeal to DIYers because easily understandable and easy to build. Besides, as Shannon IS right, I expect it to give best results.

Learned from one post, that AOR/L ac voltage compliance is only within +- 25 mV as per TDA1543 data sheet. Had overread this. Therefore, original circuit 274 ohms and even my 150 ohms resistor I/V is far out of specs.

2. Started to poke around in original circuit with oscilloscope. Created 22050 Hz 0 dB (16 bit full scale) square on left channel (2 samples per sqare), digital silence on right channel. I/V resistors 150 ohms each. Right channel shows decaying oscillation: 50 mV pp max, total duration approx. 6 us, frequ. approx. 800 kHz. Amplitude varies with excitation. Vice versa when changing channels --> -34 dB (!!) crosstalk problem. Will keep an eye on this.
Also tried to identify any crosstalk from BCK/WS/DATA lines to AOR/L but could'nt find anything significant. Small decaying 75 MHz oscillations are probably caused by suboptimal scope probe coupling (German readers: wer misst, misst Mist).
Unfortunately, I have no means for digitally recording oscilloscope traces, can only photograph screen. As there is nothing exciting yet, I will save this for later.

BTW: your offset problem 500 mV. Simple R voltage divider should work best to create the 2 V op amp bias. Also, only 500 mV and if you still use output caps, is'nt that more of a cosmetic problem?

[aparatusonitus]

Hi,

I have created voltage divider via trimmer, so opamp +in sits at 2VDC, no problems there, but output offset is 500mV from cold start, when fully warm I can adjust it to +/-15mV, which is almost tolerable since there is no output cap anymore. Also, offset significantly change if I press forward or play for example, so I think offset problem is PSU related - voltage apparently drops when motor have to do something I guess, have to check this to be sure.

[Karstenpg]

Following mods have been implemented:

• Cut Vref rail and connect individual Ibias resistors to each Vref pin. Result: AOL(R) noise reduction.
• Make existing op amp AD847 work as transimpedance amplifier followed by passive voltage divider giving approx. 1 Vss output at 0 dB fs digital input. Reason for passive divider: omitting it would require 100 Ohm feedback resistor which will degrade op amp performance (value too small). Result: DAC operates within specs: AOL(R) +- 25 mV AC compliance.
• Make op amp +Vbias adjustable for 0 Vdc on output.
• Connect 10 nF capacitor from AOL(R) to ground. Result: greatly reduced output noise. This size capacitor does not yet degrade output but only with this capacitor output remains inaudibly silent when other channel is excited with digital 0 dB fs.
• Attach 47 nF ceramics across both DAC voltage regulator output elcos.
• Change output decoupling elcos to 6.8 uF WIMA MKS2 type.

See circuit diagram for details.

Overall results in A/B comparison with second unmodified DAC: improved musicality, sounds silky, more "tube-like". Enhances piano and violin sounds, for example. Both outputs remain totally silent when CD player stops whereas unmodified DAC still emits faint but audible hiss.

Development work was carried out with LITE Audio PCM2704 based USB-SPDIF converter and Esser Audio TTG test tone generator here .

[DAC33]

Pcm2704?
Oh
i love

[Karstenpg]

After reading about current feedback op amps I tried an LME49713 as I/V converter. This amplifier has excellent reputation for its audio performance. It requires a 1.2 K feedback resistor, the total current from 8 parallel TDA1543 giving approx. 22 Vpp max output. To bring this high output voltage swing down to the common 2 Vpp range I use a 10:1 step down transformer and a servo amplifier to eliminate any DC component in the transformer primary. The servo circuit reduces the DC to less than 2 mV. This makes a coupling capacitor unnecessary.
When used in an inverting amplifier configuration, the LME49713 tends to ring. It can be easily "tamed" however with a 2.2 K resistor between current summing point and inverting input and a 470 pF shunt capacitor across the feedback resistor. With these measures impulse response becomes totally flat.
T1 is a cheap audio transformer made by German GT Elektronik. The Reichelt part number is NFU1-10, selling for 2.35 Euro. The 2.2 K resistor across the output suppresses any parasitic transformer oscillations caused by the typical high Z line load.
How does it sound: marvellous. Best non-oversampling DAC I ever heard. Also because English is not my native language I will not attempt to write any flowery or esoteric portrayals. Instead, go ahead and build this simple circuit and listen for yourself. Happy soldering!

[Karstenpg]

Circuit diagram had a 220 K resistor missing. Sorry for the fault. Here is the corrected version.
Karsten