lingDAC - cost effective RBCD multibit DAC design

In my estimation TDA1387 is more true to the source than TDA1543. The bipolar chip is very pleasant indeed to listen to with the lingDAC I/V and buffer stage after it, but somehow its just a bit romanticized when compared to the 1387 - a little softer and subtly coloured.
 
I sourced all my TDA1387 on Taobao, I have no documented chain of custody for them but I believe them to be 'pulls' from old 'Soundblaster' PC audio cards. They're obviously not new chips. Seeing they're so cheap (nowadays around 35 to the USD) I've never worried about whether they're 'legitimate' or not. Having used a few hundred of these parts during my prototyping years I've not found one yet which has sounded bad though I do note the DS shows the worst-case SNR can be just 86dB.

Gerbers I will make available after I've made the schematics more presentable (and corrected any errors) but I am not interested in organizing a group buy myself. Should anyone else wish to get involved in that (which seems to me to be rather a thankless task) I'll support them to the extent I am able. If anyone is interested in acquiring a finished DAC you're free to PM me.

I've attached a picture of one of the early incarnations of lingDAC - its a USB-fed DAC-AMP driving low impedance headphones direct up to 100mW or so. Three of the PCBs (DAC, I/V-filter, SE buffer) are in the plexiglass tower. In the wooden base are a USB-I2S interface (not my design, sourced from Taobao) and a power supply.
 

Attachments

  • laptopDAC.jpg
    laptopDAC.jpg
    315.7 KB · Views: 1,512
Last edited:
I sourced all my TDA1387 on Taobao, I have no documented chain of custody for them but I believe them to be 'pulls' from old 'Soundblaster' PC audio cards. They're obviously not new chips. Seeing they're so cheap (nowadays around 35 to the USD) I've never worried about whether they're 'legitimate' or not. Having used a few hundred of these parts during my prototyping years I've not found one yet which has sounded bad though I do note the DS shows the worst-case SNR can be just 86dB.

Gerbers I will make available after I've made the schematics more presentable (and corrected any errors) but I am not interested in organizing a group buy myself. Should anyone else wish to get involved in that (which seems to me to be rather a thankless task) I'll support them to the extent I am able. If anyone is interested in acquiring a finished DAC you're free to PM me.

I've attached a picture of one of the early incarnations of lingDAC - its a USB-fed DAC-AMP driving low impedance headphones direct up to 100mW or so. Three of the PCBs (DAC, I/V-filter, SE buffer) are in the plexiglass tower. In the wooden base are a USB-I2S interface (not my design, sourced from Taobao) and a power supply.

Hello Abraxalito,
the lingDAC project you are presenting here is for driving headphones if I understand correctly? How much modification (if any) would be required for interfacing with a preamplifier/amplifier ?
Thanks B.
 
Hi Bart

In reality there's no need to modify anything to drive a preamp or poweramp but in practice there may be a small improvement in SQ by replacing the electrolytic output capacitors for some film/foil ones of smaller value (say 6.8-10uF) and increasing the output cap biassing resistors by a factor of 20 (to 100k).
 
No - I have thought about that but not gone on to implement any ideas I had, I'd always assumed that any active driving of pin7 would potentially increase output noise. A few years ago I was advocating a few thousand uF to be put on pin7, these days I use nothing at all as I hear no improvement when I do add some amount of capacitance.
 
Yeah I originally preferred the 1545 to 1387 because it consumed about half the power and also was able to scale its output current up to 2mA. I built up a few DAC-AHs where I replaced the 1543s with 1545s and even increased the number of chips up to 32. However on careful listening I couldn't get rid of some lowish-level noise modulation on piano (a large proportion of the music I listen to has piano in it) which never has appeared on 1387.

TDA1311 has voltage output so does not lend itself to passive output filtering. LC passive filters have lowish impedance and at line levels draw too much current for comfortable drive.
 
I sourced all my TDA1387 on Taobao,...... they're so cheap .......


Thanks for sharing all your experience and expertise with the TDA1387. Given TDA1387 are so affordable is there any merit using multiple chip designs for a high performing DAC?


Are there three topologies for implementing multiple chips? Stacking them, analogue bit cascading them in time and segmenting MSB and LSB for eg 24bit processing? Do you think any of these approaches is useful for the TDA1387? Can any of them bump its performance up towards say a PCM1704? Thanks.
 
Thanks for sharing all your experience and expertise with the TDA1387.

You're welcome - I'm glad that there are people finding it useful.

Given TDA1387 are so affordable is there any merit using multiple chip designs for a high performing DAC?

I've been asking myself that exact same question for quite a few years now. I guess it all depends what's meant by 'high performing'. Certainly one of the TDA1387's weaknesses in today's digital market is that its only 16bits hence the noise floor is determined by that. I would guess that a lower noise floor can be achieved by paralleling outputs and driving each DAC independently from a DSP. Its not something I've tried so far, given there aren't so many DSPs to choose from which have at least 4 I2S ports on them (I'd say 4 would be the absolute minimum worthwhile to parallel).

Are there three topologies for implementing multiple chips? Stacking them, analogue bit cascading them in time and segmenting MSB and LSB for eg 24bit processing? Do you think any of these approaches is useful for the TDA1387?

I've done stacking and while its useful for getting more output current, it doesn't in my estimation result in better SQ. Its not like the 1543 where there are considerable linearity errors which, when averaged out result in a straighter line for INL.

I do have plans to build a spaced-in-time DAC for implementing NOS droop - effectively an analog FIR DAC-filter. I did build one with about 19 1387s a few years back, however that was in the days prior to my getting involved in LC filters and fancier power supplies.

Can any of them bump its performance up towards say a PCM1704? Thanks.

If by performance you mean noise, it would need a fair amount of complexity to get down to the 1704's noise floor (-120dB) - probably at least 16 chips. If you're interested in bettering the SQ of the 1704 then I'd say lingDAC already does that (on 16bit material) with only two chips - to my ears 1704 is a little bit 'soft' in comparison. About half-way to a 1543.
 
Member
Joined 2008
Paid Member
The aim there is to give the 'passive' DAC as close as possible the same voltage on its outputs as the active one. In the basic configuration only one DAC's outputs are taken to the I/V stage, the other DAC is there solely as a 'dummy' to balance out the current drawn by the active DAC (the current draw being signal dependent). I figured to get the best cancellation it might be best to give both DACs the same output voltage, but I've not actually gone and verified this.

Hi Richard! I only just now saw this thread, very much subscribed.

For those of us interested in balanced output: is it possible, instead of a dummy load for DAC #2, you could instead run it into a whole second filter+buffer, right?

I too have a v1 lingDAC that Abraxalito was kind enough to send me, but time constraints have prevented me from playing with it.
 
Hi Matt - welcome to the thread.

Yes, for balanced output its necessary to connect four twisted pairs to the DAC board - one pair for 'in phase' and the other for '180o phase'. These pairs then go to two filter-I/V boards and then on to two buffer boards, for a total of 5 boards to create the balanced lingDAC. If unbalanced outputs are required I recommend a trafo after the buffers.
 
Filter I/V board schematic

Here's the schematic for the 2nd board in the three-board set - it accepts the 0.6mA peak current output from the DAC board, filters it and then converts the current to a standard 2VRMS level.

Filtering is done by a 5th order LC which incorporates the NOS boost. Inductors (TDK SLF7045T range) do need pre-selecting to get an accurate frequency response - off the reel they're generally +/-20% tolerance.

The discrete transistor I/V stage is DC coupled and so there needs to be some tweaking to set the output DC offset to mid-way (3.8 - 3.9V). This is accomplished with some of the 'NF' resistors which adjust the bias currents coming from two-transistor current sources.
 

Attachments

  • Schematic_0DAC 2V filter.pdf
    55.8 KB · Views: 498
Here's the schematic for the 2nd board in the three-board set - it accepts the 0.6mA peak current output from the DAC board, filters it and then converts the current to a standard 2VRMS level.

Filtering is done by a 5th order LC which incorporates the NOS boost. Inductors (TDK SLF7045T range) do need pre-selecting to get an accurate frequency response - off the reel they're generally +/-20% tolerance.

The discrete transistor I/V stage is DC coupled and so there needs to be some tweaking to set the output DC offset to mid-way (3.8 - 3.9V). This is accomplished with some of the 'NF' resistors which adjust the bias currents coming from two-transistor current sources.

Very nice. :):)
I also see some Non Fitted capacitors (C14, C29) in the filter, what is their purpose?
Thanks B.
 
I also see some Non Fitted capacitors (C14, C29) in the filter, what is their purpose?

Good question - they're for tweaking the frequency response when the inductors aren't equal to their nominal values. As I mentioned above, the tolerances on inductors are pretty wide and at the extremes can make significant changes to the frequency response of the filter. Some variation in inductor values can be compensated for by adjustment of capacitor values, hence the NFs.

I do plan to do a more thorough analysis of the sensitivity to inductor values of my filter and I do have (in theory) a modification of this filter to accommodate a wider range of values. Still to be tested in practice though....

@kokoriantz - the absolute max supply is 6V but the recommended max is 5.5V. I've not had any problems so far running at 6V but if you'd like to run at 5.5V - the only penalty is a slightly lower maximum output - then you'll need to make changes to the LED string to reduce it by 0.5V. Currently there are two yellow and one red LED, I'd guess two green LEDs would have about the right voltage. The other LED locations will need bridging with 0R resistors.
 
Last edited: