16 x TDA1543 DAC directly drives headphones

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
After much reading of this forum and experimenting with DACs of the non-oversampling variety (why does it sound better? I don't know, it just does!) I found a way to exploit the extreme sensitivity of Sennheiser HD25 headphones with a DAC that directly drives them from 16 Phillips TDA1543 ICs in parallel.
:idea:

The starting points for this circuit were, how much voltage is needed to power a pair of headphones and how many TDA 1543 ICs (nice, cheap, plentiful) can be put in parallel? Could a not too ridiculous number of ICs provide a not too weak level into the phones?
:scratch2:

Opinions vary as to the maximum peak level that is suitable or sensible for domestic listening. The quoted figure of 125dB peak level for a live symphonic concert seems a good starting point. The Sennheiser HD25 headphones have a sensitivity of 120dB/mW. With an impedance of 70 ohms, this translates into just 264mV for 120dB. To reach 125dB with some headroom, an output voltage of 350mV was decided on.

Looking at the circuits developed on this forum for multiple TDA1543s in parallel, then experimenting with an initial 4 in parallel circuit (very promising, but a bit weak on some recordings) led to this 16 TDA1543 'monster'. The circuit is similar to those described in many previous postings by members of this forum, CS8412 and TDA1543 with a 47 ohm resistor for I/V conversion and 42 ohms for the Vref resistor. The output voltage at 0dB is 650mV, which means 379mV into 70ohms - nicely on target. A 5V supply is enough for this output voltage and the TDA1543s do not need heatsinks at this supply voltage.

I decided on a battery supply, which can also power the CS8412, as this gets rid of voltage regulators, essentially another form of amplifier. To adequately supply the 800mA+ needed by this circuit 4 5500mAH NiCd 'D' cells were used in series.

So, it all comes together nicely to make the headphone DAC with no filtering, no amplifier and no regulators. There are just two components in the audio signal path, the 47 ohm resistor and a 1,000 microfarad electrolytic capacitor. A 470 ohm variable resistor in series with the output acts as a volume control.
:)

After all this, how does it sound? I find it very good indeed (but then I might be prejudiced). Dynamics are very, er, dynamic, startling at times. Stereo imaging is excellent, with real 'depth'. Treble seems very clear and bass in control but weighty when needed.

The most important aspect of the sound of this DAC for me was the fact on recordings that I had heard many times before, it never failed to reveal musically important details which I had not previously discerned, even on the same headphones. At times, recordings seemed almost 'remixed', such was the difference. For example, the lyrics of Steely Dan's 'Aja' were now meaningful songs, with more insight into why they felt compelled to write and perform them in the first place. With Led Zeppelin's third album, not only were the lyrics clearer, bringing out the band's folk-rock roots, but I could visualise them playing together in a real venue, despite the obvious faults of the 1970 recording.

I'd be interested to know what others on this forum think of this idea and if it could be improved. Perhaps there are other DAC ICs with a higher output current, which could avoid the 'brute force' approach of using 16 in parallel. The above results were obtained using a CD-ROM drive as a digital source - could a 'better' transport make a difference?

I can post further details if needed, but this circuit is very much building on the previous work of many people on this forum and a few searches will find all the info needed to complete a prototype.
:hphones:
 
Driving possibilities of the CS8412

Hi Intermod,
very nice project.
Do you know the number of DAC the CS may drive without the need of a buffer ? I read somewhere else here that with an external NonOS DAC, or tree internal DAC directly fed with I2S, it should better driven with a buffer (74HC125 and a 2N4401 transistor ).
Best Regards,
Alain
 
Thanks for your replies and interest in this project.

I'm not sure how many TDA1543s can be driven by the CS8412 before errors occur. It seems to work fine with 16 and no one seems have tried more than that number.

I've considered how to make the DAC work with a wider range of headphones. One example is the Sennheiser HD650 with an impedance of 300 ohms and a sensitivity of 103dB at 1mW. To get 123dB 10mW will be needed. This means that the DAC will need to deliver 1.73V into 300 ohms.

The following methods could be used to get a higher output voltage:-

1) Increase the supply voltage. This will allow better matching to higher impedance headphones. However, each TDA1543 will need a heatsink, making assembly more complex. Also, the output impedance will need to be increased to get the higher output voltage. It should be possible to get about 1.8V from a Rout of 138 ohms and a supply voltage of 8.5V.

3) Use logic gates as buffers to allow more than one bank of TDA1543s to be driven. This could, combined with increasing the supply voltage, allow higher voltages from lower impedances. How about 32 TDA1543s at 8.5V giving 1.8V from 62 ohms? This could be continued almost indefinitely...

4) Drive the headphones differentially. This could double the output voltage by using logic to drive one bank of TDA1543s in opposite phase to the other. It would also double the output impedance as the headphones will be connected in series with two Routs. Most headphones have a common ground connection which would not allow this (but they could be modified).

Considering the HD650, a current favourite on various forums, this could be driven to 5mW by option (1) above. This equates to a peak level of 117dB, slightly less than the 125dB of the original version with HD25s. However this level is rarely needed in practice.

Hmm... (now saving up for HD650s) :) .

A further thought, that this is actually a digital amplifier, if a very low powered (4mW total) and inefficient (0.1%) one.

:hphones:
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.