tda1387 dac pcb "front end"

The 2X upsampling seems like a good compromise, since half the samples are still original, and it simplifies your analog filter. I have subjectively preferred upsampling Redbook to 88.2K versus higher sample rates in the limited listening tests I did using SoX algorighms. Have you ever tried upsampling in conjunction with a more gradual, linear phase Bessel slope?

Thanks for describing the differences between TDA chips and their response to heat. As for warming up caps, that is useful to know...I always figured it was only the actives that were impacted by warm-up.
 
Last edited:
I think that is correct. It is my understanding, that in a practical sense, upsampling to 88.2k essentially eliminates the need for filtering, and corrects the sin(x)/x effect. Hmm, maybe I should pull the filters from TeraDak. Leave just a ferrite bead. Play only upsampled 88.2k. On tda1387_rpi-hat_se_v1.6_schematic_20180903 remove C9, C10?
 
Last edited:
The particular digital filters I use don't preserve the original values - in my understanding to have that feature they'd need to be half-band filters which allow some imaging products through in the transition band. No, I've not tried any Bessel slope filters - I'm not sure but I think I could design a digital filter to give a linearized phase response for my current Chebyshev filter. This could in theory be combined with the upsampling function. Btw my comments about caps only apply to electrolytics, I'm not aware of other kinds being so temperature sensitive.

@hifiamps - 2XOS doesn't eliminate the filtering requirement, just relaxes the gradient of the slope that's called for.
 
Are there any available through hole subs for the SSM3J327R? Or is a SOT23 conversion board a better choice? Breadboarding a TDA1541 substitution module (TDA1387 x 8) with passive IV (~500R). I would like to try active IV with a MOSFET, as configured in Garman v1.6 (modded v1.6 sounds much better, thanks Matt and Kevin).
 
Last edited:
I think the Garman sounds better than my perfboard 1387 x 8 passive IV in my system, which consists of passive preamp (stepped attenuator) and Icepower 500ASP modules. Mods to the Garman. Power supply quality is extremely important. Keeping with the data sheet value of the Vref Pin 7 C1 cap at 1uF. A highest quality you can afford 10uF film coupling cap for C7/C8 (so far CRC M83421/01, yes, it won't fit on the board). Another thought I have yet to implement is eliminating MAX6071, C2, C3, C6, C13, C14 for the precision voltage reference, and simply use a voltage divider. There is no significant current draw from the J327 gate, so a voltage divider will work with lower noise. Could still use C15. Might need individual decoupled supplies for each channel, which complicates things if using the Garman pcb. It might also benefit from gate stoppers as well. But, a precision voltage reference seems completely unnecessary. Also, based on the MAX6071 data sheet, the max output cap value is 10uF. "Capacitive-Load Stability Range" is 0.1 to 10uF and C15 is 220uF. I'll try the voltage divider out with the perf board 1387 x 8 J327 active IV first. The 1387 is such a great sounding DAC. Extremely musical. My current 1387 x 8 passive IV setup is just stupidly simple. Yes, not the lowest SINAD, but dang Garman v1.6 sounds good.
 
Last edited:
The 1387 x 8 TDA1541 module with passive IV described above is much improved by adding a 1000uF cap across pin 28 (Vdd) and ground. I was previously powering the circuit directly from a super regulator to the module. Adding the cap reduces 3D, but increases volume, bass and dynamics. The 1387 x 8 module only uses pins 1,2,3 for I2S, pin 28 for Vdd, pin 5 is ground, pin 6 AOR, pin 25 AOL. The other pins have no connection.
 
The 1387 x 8 TDA1541 module or the 1387 x 4 are probably the easiest to implement perf board (0.100" centers and 7x pin jacks) testbed for the 1387. Two IV resistors (~500R for 8 or 1000R for 4), three series 470R I2S resistors, two coupling caps and the +/- 1000uF PSU decoupling cap. With or without low pass filters or active IV. Other ideas in the TDA1387 x 8 wiki.
 
Last edited:
Two IV resistors (~500R for 8 or 1000R for 4), ..[/URL].

I was wondering if you would need to up the supply voltage for these values. Correct me if im wrong, but tda1387 sheet shows 2.7k i/v resistor probably because 2.8v p-p is 1vrms. On a 5v supply it sheet says up to 3.5Vdcc, so maybe 3.5k. Actually 4V you suggest doesn't seem too bad, but i was wondering why Matt's front end v2 has 675R for 8, 5.6V (for 2VRMS i suppose), isn't this pushing the output compliance? I am sorry if i have grossly misunderstood something, was just wondering.
Matt i am bored, do you mind if i butcher your board? :D
 
I had a quick search back over the thread for the circuit details of v2 but didn't find it. I'm guessing in that version he isn't using a MOSFET for I/V so I agree with your math @Zbunjen. The max compliance (@5V supply) is 0 - 3.5V. 675ohm for 8 chips would give 5.4V p-p if the chip could handle that, but it can't so you'd end up with clipping at high signal levels.
 
675ohm for 8 chips would give 5.4V p-p if the chip could handle that, but it can't so you'd end up with clipping at high signal levels.

I guess this is true, but my apologies for misleading as i was going by memory and did a silly! It is for only one channel, so it's actually 675*4 for 2.7k. I found the schematic in post 196 now. But i guess it's good to still discuss these things, even if a mistake or two might be made by me :)
 
I did a tda1387 design with this sort of "quick release" system where you can have banks of tda1387 piggybacked, and just plug them in and out of sockets (if you want to experiment or change number of dacs or one goes bad of the bunch you can quickly get it out), so you're then working with dip and dont have to unsolder anything. Not sure about lead inductance or if you even gain that much in practice, since if you change the number of dacs other design parameters also change, this one has 30 converters. But i thought it was interesting anyway ;)
This dac is a pure test platform which is why i have not posted anything else yet. I am courious to compare with abraxalito's kubelik which is on the way.
(LCSC part numbers for the socket and pins are C9742 and C72129 if anyone is courious)
 

Attachments

  • tda_thing.jpg
    tda_thing.jpg
    356.3 KB · Views: 279
Those stacks look totally awesome! Have you run into any issues with decoupling so far?

How does it sound with 30 chips? I'm super-curious! Also - what's the I/V stage?

Excellent work, keep going.

Thank you :)

No issues, but the film caps i ended up using are physically a bit larger than the ones i planned to use when designing, but these ones have ESR under 0.1ohm, and i can just alternately solder them on the upside and bottom side of the PCB. The other film caps fit perfect but 1.1ohm ESR, the same as some original WIMA 100nf's i have.
Will test out all sorts of I/V's (passive, discrete, tube, opamp...) with it but still missing some components to complete everything. Building Jung/Didden super-reg as well to power everything so i guess im not done yet or come to some conclusion. Also built a TDA1541 dac, more stuff to compare to as i will test concurrently. More to come next week!
 
Zbunjen: looking forward to the updates.

I have mounted the 1387 x 8 module on an RPi HAT protoboard. I used DIP pin sockets for the module (two rows of six), in case I needed to remove the module and get board access in a design change. I also plan to use sockets for the IV resistors and the Vcc cap (from stellarelephant), which I think is an excellent idea for easily swapping out critical parts. I am currently using an RN70 metal film IV resistor, and I have acquired some metal foil resistors to try (Vishay RNC90). The only downside with sockets that I have encountered in the past, is that they can develop poor contact in time, and potentially generate noise. Coupling caps are 10uF CRC polycarbonate (definitely off board, too large for board mounting), but I would also like to try others that I have. I am still not using any output low pass filtering. I am using a China LT1963 7805 module mounted to the HAT on the 5V rail (with a filtered 6V SMPS), with added electrolytic caps, since the ceramic caps on the module are inadequate based on my read of the data sheet. I think the LT1963 sounds about as good as the slightly modded China Jung/Didden-esq board that I have tried.

Unfortunately, I haven't had much time for tweaking or even listening recently. But, this simple 1387 x 8 DAC HAT sounds better in my system and to my ear than other NOS DAC HATs I have tried, including the Matt TDA1387 v1 and v2 HATs (thanks stellar!), AD 1865 and a discrete NOS.
 
Last edited: