tda1387 dac pcb "front end"

Member
Joined 2008
Paid Member
Balanced RPI HAT Front-End v1.0.0

Here are the files for the Balanced RPI HAT Front-End v1.0.0, see post #236 for an overview and pictures of my prototype build.

In short, this attempts to merge the concepts of the "Front End" with a form factor suitable for use with a Raspberry Pi. It's not a true HAT, because it's slightly bigger than the RPI. It's also not as flexible as a true front end, due to space limitations.

Attached are the schematic, Gerber files, BOM and KiCad files. The BOM is a zip archive containing PDF, CSV and ODS formats. Note the ODS version has multiple tabs: basically, BOMs for several of my tda1387 designs.
 

Attachments

  • tda1387_rpi-hat_bal_fe_v1.0.0_gerber_20170728.zip
    129.3 KB · Views: 81
  • tda1387_rpi-hat_bal_fe_v1.0.0_schem_20181215.pdf
    81.5 KB · Views: 205
  • tda1387_rpi-hat_bal_frontend_v1.0.0_BOM_20181216.zip
    88.7 KB · Views: 88
  • tda1387_rpi-hat_bal_frontend_v1.0.0_kicad_20181215.zip
    110.9 KB · Views: 77
Member
Joined 2008
Paid Member
To be honest, I haven't listened to it in quite a while. It has a jumper that allows you to use RPI power or dedicated DAC power. My original tests were with shared RPI power. But I think that's unwise due to the amount of capacitance on the board: it will basically "starve" the RPI of power, at least initially. So to avoid that, it needs dedicated power. I don't have a "clean" dedicated 5v supply---what I mean is "clean" in the literal sense, my supplies require wires and terminal blocks and transformers hanging all over the place. This makes me nervous with young kids running around. So in short, I haven't had the opportunity to spend a lot of time with it with a dedicated supply.
 
To be honest, I haven't listened to it in quite a while. It has a jumper that allows you to use RPI power or dedicated DAC power. My original tests were with shared RPI power. But I think that's unwise due to the amount of capacitance on the board: it will basically "starve" the RPI of power, at least initially. So to avoid that, it needs dedicated power. I don't have a "clean" dedicated 5v supply---what I mean is "clean" in the literal sense, my supplies require wires and terminal blocks and transformers hanging all over the place. This makes me nervous with young kids running around. So in short, I haven't had the opportunity to spend a lot of time with it with a dedicated supply.

Noted, makes sense. I’m waiting for the Salas BiB v1.3, so hopefully that should be out of the equation for me xD
 
Hey guys,

The more one learns, the more questions one has. Please bear with me.

I finally bought a S/PDIF+I2S to I2S board, this will act as a source selector, and send the I2S to the TDA1387.

In parallel, I've been reading about gain structure and gain related to noise. The basic concept I am getting is that I should move a bit of the gain from the power amp into a "pre-amp" or IV stage.

If you move part of the voltage gain in front of the volume pot, noise at the power amplifier's input won't be amplified as much any more

My aim was to have a 2x or 4x to generate a differential output, then a passive output, but I might need to reconsider a bit, since many things come to mind now. Facts:

- I will get two sources into the DAC, thus I will need a hardware volume control (if I only had the RPi through I2S I could do w/ the software control).

- Need gain before the volume control in order to "reduce SNR" after the volume control. Thus, in search of an active I/V. This gain structure also seems to invalidate the alternative discussed previously, adjusting the voltage reference of the TDA1387.

- Some HF filter still needed after the DAC but before the I/V, right?

- Since I will have an active I/V I would probably just need a single DAC even though I can parallel many. What would be the pro/con of adding more DACs before an active I/V? (more current, less impedance, ...)

- Differential output circuit would be placed after the volume pot. I think I like the one suggested by Mark4 in post #356.

Well so far so good (or not), these are my new steps/doubts in this journey.
Thanks for reading
 
This gain structure also seems to invalidate the alternative discussed previously, adjusting the voltage reference of the TDA1387.
Sorry, to be more precise. Another reason to invalidate this method is that the min voltage seems to be around 1V. Below that, no sound, just a bit above that, it produces too much sound for my taste. Increasing the voltage also increases the volume quite smoothly, but apparently the initial raise in volume is to abrupt for my taste. I should probably test with a finer trimmer on the PSU though.
 
Also, and now a bit off-topic, has any of you tried the TDA1311 and/or the UDA1334 DACs? There's an adafruit/ebay board for the later that might be interesting to have a look at, for the price. Unless you guys that buy/try all the chips can already advice against for whatever reason.
 
Member
Joined 2008
Paid Member
tda1387 RPI HAT DAC v1.6

Finally! Here is v1.6 of the single-ended tda1387 RPI HAT DAC. I designed this back in the beginning of September, but due to part shortages and general lack of time, only now was I able to actually build up the prototype. Good news, it works!

Differences from previous version(s):
  • Use of MAX6071 precision voltage reference for I/V transistor base voltage ("BREF", 2.5v) instead of TL431. This is the most significant change.
  • Use bigger screw terminals for power and output (5mm pitch)
  • Jumper to select RPI power (closed), or dedicated DAC power (open)
  • 2x6 pin header for I2S to use as a standalone DAC instead of RPI hat

The pictures shown here are of the first build, which worked without any debugging! Listening to it now as I type this message.

As always, I have a few bare boards available. They are free, you just need to cover shipping.

Attached are the schematic, BOM, Gerbers and KiCad (v5.0) files. Please let me know if you have any questions!
 

Attachments

  • tda1387_rpi-hat_se_v1.6_gerber_20180902.zip
    93.3 KB · Views: 131
  • tda1387_rpi-hat_se_v1.6_schematic_20180903.pdf
    52.4 KB · Views: 340
  • tda1387_rpi-hat_se_v1.6_image-live_20181228.jpg
    tda1387_rpi-hat_se_v1.6_image-live_20181228.jpg
    231.8 KB · Views: 541
  • tda1387_rpi-hat_se_v1.6_image-top_20181228.jpg
    tda1387_rpi-hat_se_v1.6_image-top_20181228.jpg
    378 KB · Views: 512
  • tda1387_rpi-hat_se_v1.6_kicad_20180902.zip
    105.4 KB · Views: 92
  • tda1387_rpi-hat_se_v1.6_BOM_20181228.pdf
    62.6 KB · Views: 158
  • tda1387_all_dac_boms_20181228.zip
    26 KB · Views: 111
Finally! Here is v1.6 of the single-ended tda1387 RPI HAT DAC. I designed this back in the beginning of September, but due to part shortages and general lack of time, only now was I able to actually build up the prototype. Good news, it works!

Differences from previous version(s):
  • Use of MAX6071 precision voltage reference for I/V transistor base voltage ("BREF", 2.5v) instead of TL431. This is the most significant change.
  • Use bigger screw terminals for power and output (5mm pitch)
  • Jumper to select RPI power (closed), or dedicated DAC power (open)
  • 2x6 pin header for I2S to use as a standalone DAC instead of RPI hat

The pictures shown here are of the first build, which worked without any debugging! Listening to it now as I type this message.

As always, I have a few bare boards available. They are free, you just need to cover shipping.

Attached are the schematic, BOM, Gerbers and KiCad (v5.0) files. Please let me know if you have any questions!

Congrats on another success, Matt. :cheers:

Happy to see bigger terminal blocks! The old ones were a bit small...worth the upgrade for that alone :) Power jumper will be super handy, too. But yes, it seems that any audible differences will be due to the new precision voltage reference. Very cool. How's it sounding?
 
Member
Joined 2008
Paid Member
Also, what VRMS does it output?

I don't think the I/V transistor changes the output at all, so I believe the VRMS calculation is the same as for a passive I/V single-chip tda1387 DAC.


Also, trying to understand the filter. Looks very similar to the one Abraxalito suggested me in post 369 but I think one cap position is different. What are the key purposes of these two filter designs? I mean differences between one and the other. Still learning the basics I guess. TIA

What I'm using is about as simple as it gets: a ferrite bead between DAC out and I/V, and also a cap in parallel with the I/V resistor. The latter cap is taken straight from the reference design in the tda1387 datasheet.

Abraxalito is working on a much more sophisticated filter with the goal of removing all of the noise from the DAC chip itself. I just copied the datasheet and added the ferrite bead based on someone's suggestion along the way. :)


Happy to see bigger terminal blocks! The old ones were a bit small...worth the upgrade for that alone :) Power jumper will be super handy, too.

Yes! I can't tell you how many times I cursed myself for using those tiny terminal blocks on the earlier designs.


Listening impressions? Compared to the front v2 for instance?

But yes, it seems that any audible differences will be due to the new precision voltage reference. Very cool. How's it sounding?

I think it sounds good. It's been way too long since I've listened to the front-end, so can't compare to that. There's still a tiny audible hiss if I put my ear right on the tweeter: that's what I was hoping the precision voltage reference might eliminate, but no such luck. Otherwise, I think it's mostly on-par with the other designs. Though right now I'm running it from RPI power, I haven't yet listened with independent DAC power.
 
I don't think the I/V transistor changes the output at all, so I believe the VRMS calculation is the same as for a passive I/V single-chip tda1387 DAC.

Its a little bit more nuanced than that. The I/V transistor does constrain the output a little bit more in a single supply implementation.

First we chose the base voltage of the I/V transistor so that its emitter (which goes to the DAC output) is within the compliance range of the DAC. Normally one or other end of the compliance range is chosen (0V or 3.5V) - I've chosen the top end, 3.5V. For a PNP transistor this sets the base voltage somewhere close to 2.8V. A transistor can't operate with its collector voltage too close to its emitter - this area of operation is called 'saturation' and for the kind of transistors we use the saturation voltage is typically in the region of 0.2V. The transistor's saturation voltage limits the top end of the voltage output range - for the 'bare' DAC its 3.5V and when the transistor's used it'll be 0.2V lower, so 3.3V.

If you were to use a negative supply to feed the I/V resistor then the transistor's output compliance range can be much greater than for the unaided DAC, limited only by the transistor's voltage rating.

Abraxalito is working on a much more sophisticated filter with the goal of removing all of the noise from the DAC chip itself.

I'm playing around with many kinds of filters as I don't know that its really necessary to get rid of _all_ the noise (say down to the 16bit level). I'm curious to know how much noise elimination is going to be enough but I rather suspect it depends to a large degree on the downstream amp. On headphones, where I use a much simpler amp (actually just a buffer) I don't notice as much difference when including a filter as I do with speakers. But this could simply be because there's no depth with headphones, just 'spread'.
 
I don't think the I/V transistor changes the output at all, so I believe the VRMS calculation is the same as for a passive I/V single-chip tda1387 DAC.
Silly me, of course!


What I'm using is about as simple as it gets: a ferrite bead between DAC out and I/V, and also a cap in parallel with the I/V resistor. The latter cap is taken straight from the reference design in the tda1387 datasheet.

Abraxalito is working on a much more sophisticated filter with the goal of removing all of the noise from the DAC chip itself. I just copied the datasheet and added the ferrite bead based on someone's suggestion along the way. :)
Noted. I think it's pretty similar to the one Richard told me in that old post. just a cap changed position if I'm not mistaken.

I'm playing around with many kinds of filters as I don't know that its really necessary to get rid of _all_ the noise (say down to the 16bit level). I'm curious to know how much noise elimination is going to be enough but I rather suspect it depends to a large degree on the downstream amp. On headphones, where I use a much simpler amp (actually just a buffer) I don't notice as much difference when including a filter as I do with speakers. But this could simply be because there's no depth with headphones, just 'spread'.
Good to know.

Thank you all again
 
You can do differential out yourself. See attached diagram. Just a couple of dual opamp buffers, one for each channel to do make SE into diff. Resistors could be 5k or 10k, or whatever you think would be suitable, they just need to be the same for unity inverting gain. Vcm can be tied to ground.

If you want to read an article on another way, here you go: Versatile, Low-Power, Precision Single-Ended-to-Differential Converter | Analog Devices

Since your power amp will probably need something like this from time to time anyway, you may as well make something you can use when you need it. The first diagram I posted is commonly used in pro audio grear to produce balanced outputs from single ended circuitry. The version at the link above may be a little more fancy.

Personally, the I would just use some low distortion opamps well suited to audio whichever way you prefer. On the other hand, the two dac approach seems a bit much to me, but I guess its yet another choice.
Hi Mark,

I've been reading (a lot) lately. This by itself doesn't mean anything, but hopefully smth will settle :)

Just found another way to convert SE into (true) balanced without opamps, from Jensen AN-003 (highlights are mine).

attachment.php


What do you guys think about this method?
TIA
 
Right. I forget to mention transformers sometimes. :)
For good quality, they are the most expensive solution, and they still can't beat the low distortion of opamps and That Corp. differential drivers and receivers, but they are very good at galvanic isolation when it is truly needed (not usually).

But I see no transformer there, just a pair of resistors and a capacitor. Am I missing smth?
 
But I see no transformer there, just a pair of resistors and a capacitor. Am I missing smth?

Okay. I am already familiar with using Jensen transformers for balanced and unbalanced conversions, so I assumed the reference was to that sort of thing. There may be other approaches that can work in some situations, although they may lose some of the benefits having balanced connections in the first place. The solutions I would recommend for general use and that are most often used for interfacing professional audio equipment are the ones I have mentioned. In some cases I have used other methods when I knew the circuitry inside an electronics box would work okay with it.
 
Okay. I am already familiar with using Jensen transformers for balanced and unbalanced conversions, so I assumed the reference was to that sort of thing. There may be other approaches that can work in some situations, although they may lose some of the benefits having balanced connections in the first place. The solutions I would recommend for general use and that are most often used for interfacing professional audio equipment are the ones I have mentioned. In some cases I have used other methods when I knew the circuitry inside an electronics box would work okay with it.
Sure, that's what I thought :)

What surprises me a lot of this approach is that I haven't seen it commented anywhere else. According to Jensen scan I pasted above
The modified output will have balance as good or better than most current pro gear and, with the exception of the possible "gain reach" problem mentioned earlier, will produce excellent results in a professional environment.
(where "gain reach" refers to 1.23 V RMS two pages before that figure).

How come it's not used/cited more often?