lingDAC - cost effective RBCD multibit DAC design

Its taken about 3 days of soldering (not constant, mind) to put together this parallel 9 channel prototype. No 18 channel version is planned unless I do the layout on a PCB, the wiring is too time-consuming :cool:

Next is the filter stage which is going to need a large number of 104Js owing to some of the cap values being around 1uF hence 10 in parallel. I'm pondering whether it might be better to have a separate filter per pair of DACs - the number of caps will be about the same but the inductors multiply significantly....:eek:
 

Attachments

  • 9chanIVfilter.jpg
    9chanIVfilter.jpg
    576.4 KB · Views: 293
Having listened to this DAC for a few days on headphones and the last few hours on speakers (amp is modded TPA3220 micro) I'm not convinced doubling from 8 to 16 DAC chips makes a significant improvement. The sound is very well detached from the speakers at all frequencies - one of my most 'difficult' recordings doesn't create any 'smear' of HF to the drivers or any harshness, with instrumental placement remaining solid.


PCB layout of the PhiDAC Quad is well advanced into the routing stage so I think will be done by the end of the week. I'll post up the schematic soon.
 
First off schematic for PhiDAC Quad

Just for a sneak preview, here's the schematic of the first prototype PhiDAC Quad. Doubtless there will be changes to this once a few boards get built - the PCB is going off to manufacture within a day.

Significant changes vis-a-vis original PhiDAC are that the 4 DACs now are now sitting on an offset of ~3V to get their outputs to sit comfortably at mid-rail (~5V or ~2V relative to DAC GND). Neither the AD829s nor AD744s can run at such a low supply voltage as the AD8017s which necessitates this change. Level shifting is done by current sources and resistors. The opamps are augmented by darlington buffers running single-ended classA rather than relying on their internal classAB OPSs. The passive filter (on a daughter board) now is using custom wound P14 ferrite cores and has a steeper slope with rejection in the stop band better than 60dB (in theory at least).
 

Attachments

  • Schematic_PhiDACQuad-AD829_2020-03-20_.pdf
    295.4 KB · Views: 106
I am following your threads. Even though I haven't a clue what most of the words mean.
I am a complete noob and am so impressed that you can even consider designing stuff that, to me, appears to be the dark arts.
Might as well ask me to design the fuel to travel to another solar system.

I would love to build a cost effective DAC (or ten)
As a complete beginner, how far off being able to make this am I?
What stages of learning do I need. Or am I way off?
Obviously need to learn to SMD.
 
Hi Luke

Happy to talk you through the stages of the design, then if you have any detailed questions I'll have a go at answering them.

I2S inputs come in to U4 which buffers them so the outputs are 0-2.5V. The signals are then level shifted as the DAC chips 0V pins aren't at circuit ground. Level shifting is done by running constant current (from Q1-Q7 ex Q4) sources through resistors (R56,R65,R7) which act as floating voltage sources thus adding about 3V to each signal.

Output from the 4 paralleled DAC chips is paralleled and fed into the LC filter which is on a daughter board. This filter implements a sharp cut-off frequency around 18kHz and rejects by 60dB or so all frequencies above 27kHz. Its output is loaded by R29 (for L channel) and then into an I/V converter which is U2. External PNP transistors form its output stage, bypassing the internal one. The final stage of the DAC is a 3rd order MFB (multi-feedback) low-pass filter which gives a boost to higher frequencies (approx 3dB at 18kHz) to counteract the 'NOS droop'. That's built out of U3 (AD744) which also has a classA external OPS made from discrete PNP transistors. The idea of these discrete OPSs is they draw a constant current from the positive supply and hence don't create supply noise which is correlated with the signal as classAB OPSs do.

Power supplies for the DAC chips and opamps are created by the AD815 which uses as its reference the string of 3 red LEDs to give around 5V.

Overall the design is the same as the original PhiDAC in architecture, just some different implementation details.
 
Last edited:
@Grarea - this design might be a bit daunting for a total noobie as you'll need some equipment. At the very least a current-limited power supply, a DMM plus tools for SMD work like tweezers, fine tipped soldering iron, solder, cutters and a magnifier if your eyesight's as poor as mine. That's not an exhaustive list btw just what comes to mind. Perhaps complete beginners would be better off enlisting the help of more experienced builders to get them going.

Incidentally do you have a digital source which outputs I2S? I ask because this DAC isn't itself a complete solution for getting sound, you will need some kit to output I2S. The simplest solution would be an SDcard player - I think earlier in the thread I showed one which can work well. You'll then need at the very least a headphone amp to create the simplest system - I am working on a design for that or you could use the SEbuff from lingDAC which is earlier in the thread.
 
Thanks for that.
I shall continue watching your progress.
I am slowly learning what some of the words mean :)

Although from your response, I seem to be at least on the first rung of the ladder.

I have a half decent iron that has a range of tips available.
I intend to learn to do SMD anyway.
I have recently picked up some decent solder as well and discovered I am not as bad at soldering as I always thought. My iron was cheap but also faulty.
So I am playing and learning some basic soldering.
Have put together a keyboard for my pc and played with a couple of cheap kits and fixed a few things.

I am also playing with DACs and various cards with my Pis.
So I am moving along the right lines.
I have just bought some amp PCBs to get going along those lines as well.
I have a DMM (although had to look up what it was) Will look at what the power supply you mentioned is.

I have got some old phone boards that I am going to play with to practice removing an putting back on. I instantly lost the first component I removed which made me giggle and I learned from that. :)

I definitely need a magnifier.
I have bodged together a fume extractor.
Made an iron stand.
Need to make a 'helping hands' next.

I see that you have organised kits before.
When you offer the kits.
I can imagine that they are an awful lot of work to sort out.
Do you limit the number you are willing to sort (I for sure would) and limit the people you give them to.
I am very aware that my feedback would essentially be useless.
But I also assume that I could source the parts myself anyway eh?

With your designs, are all the difficulties (once designed) all about constructing them.
Or will i need more tools other than the DMM to measure things for example?

Sorry for all the questions, I am so intrigued and impressed.
 
I see that you have organised kits before.
When you offer the kits.
I can imagine that they are an awful lot of work to sort out.

Too right! Which is why last time we made up kits in batches of 10, not a lot more work to do that than make a single one. This time though I think 10 will be too many as the BOM cost is significantly higher - in the region of $10 rather than $3 as last time. So I'm wondering (not decided by any means) if batches of 3 or 4 might be appropriate.

Do you limit the number you are willing to sort (I for sure would) and limit the people you give them to.

For sure yes. Last time we took at stab at a number which was about 70 and created 7 lots of 10-batch kits. I don't think we have many left, if any. This time I dunno how many we'll do. The ideal way to distribute them would be to have a small number of people willing to do some of the kitting work to serve people in their geographical area. So then I (though its normally my wife) send out bags of components and the local kitter sub-divides and allocates these. That seems to me to be one way to share out the more boring work :)

I am very aware that my feedback would essentially be useless.

I can't think why that would be. Anyone's feedback is welcome.

But I also assume that I could source the parts myself anyway eh?

Most of them you can yes, though if you go through the usual channels for semis you'll pay a lot more (e.g. AD744 here is about 7 for $1, on Mouser over $5). That's because the semis I buy are pulls from old boards. On C0G ceramic caps, Mouser's prices are considerably higher and that's comparing new vs new. The inductors might be hard to source - Mouser does have some cores but I'm not sure they're the right ones.

With your designs, are all the difficulties (once designed) all about constructing them.

I've not come to any firm decision about the filter board yet - it might turn out that's too tricky to get right without a capacitance meter and tweaking cap values. In which case I might sell it as a finished board. But other than that, yes there's nothing too hair-raising.

Or will i need more tools other than the DMM to measure things for example?

You might need a DMM which can measure capacitance fairly accurately but that feature's pretty common these days on DMMs.

Sorry for all the questions, I am so intrigued and impressed.

I like the questions, apologies for them aren't necessary :)
 
Excellent.
Thanks for that reply.

I like the idea of a few for sure.
Might be able to get one right by the last one perhaps.

I would be willing to join in and do my bit if needed. (am in the UK) Well, if it is deemed that I am capable of making them anyway.

I guess my feedback would be useful in the way of 'how little amount of knowledge is required to put one together' type thing :)
I suspect I am right on the cusp. (well, somewhat below the line currently so take your time)
"looks up capacitance meter" - doesn't look like mine has that. Might be time to upgarde a tool :)
Sounds like a useful tool to have.

OK, seems like the steps are not too far away.
Just need a bit of time to practice mostly.
Thanks.
I shall continue to be bemused by your threads :)

If it needed five years of experience beforehand, I might have turned my attention elsewhere.
 
lol, yes, I can imagine you now.
Stuff is so amazingly cheap now.

Thank you for that.
Impossible for me to work out what is good and bad given all of the options. I have no idea what i am looking for.
Mine does more than I can presently but it doesn't measure capacitance. (I think.)
HAs done me fine for the limited amount I have needed it.
This is mine (nearly)
DIGITAL MULTI METER 987AA1770

I seem to recall that I can't buy from Taobao from the uk.
But this is the one right?
ZOYI ZT219 Digital Multimeter 19999 Counts True Rms Multimeter Transistor Tester Voltimetro Profesional Capacitance Meter|Multimeters| | - AliExpress

That would be an accurate enough capacitance meter for this would it?
 
I seem to recall that I can't buy from Taobao from the uk.

Yes I think you'd need to employ an agent to get it for you.


That is indeed the right one, but the price is a bit higher than on Taobao. I couldn't help but scroll down that page and found this one :

ANENG Handheld Digital Multimeter Voltmeter Ammeter Ohmmeter Voltage Multimeter Volt Tester Meter AC DC Tester Meter Portable|Multimeters| | - AliExpress

Looks to have identical specs just a different name at the top and its cheaper by about $5. Though it depends on a promotion.

That would be an accurate enough capacitance meter for this would it?

Good question - the spec says 2% (mid ranges) but I will compare mine with my dedicated LCR meter and see if its any better than that in practice.
 
Passive filter daughter board

Here's the PCB rendering of the filter board. The bulk of the capacitors making up the filter are on the reverse side.

Also here is the FR plot (from LTspice) for the NOS filter. If some end-users were going to use 2X oversampling to drive their PhiDAC Quad then a different filter design may well be more appropriate as with 2X OS (running at 88.2kHz) there's no need for such a steep slope as images won't begin until 68.2kHz. In giving up some steepness, deeper stop-band rejection can be achieved as in the second FR plot. 2XOS allows a 'textbook' response where rejection at 68.2kHz and beyond can be >96dB.
 

Attachments

  • LCdaughter.png
    LCdaughter.png
    65.5 KB · Views: 194
  • filter.jpg
    filter.jpg
    157 KB · Views: 190
  • 2Xfilter.jpg
    2Xfilter.jpg
    156.1 KB · Views: 189
Last edited:

TNT

Member
Joined 2003
Paid Member
Textbook is suppression of images by filtering at Fs/2. In your basic implementation, it's only 25-30 dB down and your second filter don't help here - right? And for Fs 88,2 the "textbook frequency" is not 68,2, its 44,1 - no? So ~70 dB down. There will be imaging distorsion in this DAC but the question is if it's less bad than the impact of a steep digital filter. I believe you think so or you wouldn't build it like this!?

Will you publish measurements?

//
 
The NOS implementation suppresses the lowest possible image (24.1kHz) by around 30dB and higher frequencies by more. So yes there's imaging distortion in this DAC, it will typically be around 0.05%. In the 'textbook' version with 2XOS the imaging distortion will be below the DAC's quantization noise.

I'm not following your math @TNT - to find the first (lowest) image frequency, subtract the highest audio frequency from the sample rate. At 44.1kHz its 44100-20000 = 24100. At 88.2kHz its 88200-20000 = 68200.

Happy to show measurements when I have a prototype up and running though they won't be comprehensive ones as I have no AP or equivalent equipment. Just a USB ADC.
 
I'm not one of those guys incidentally who thinks imaging distortion is preferable to a steep roll-off filter. I'm building it without a digital filter because I can't see a way to add one while keeping the design cheap and yet robust. An ARM chip can do the job but I'm not experienced enough yet with their foibles to put a design using one out for public consumption. Adding firmware to a hardware design increases the complexity considerably. Hopefully I'll get there eventually though.