ProtoDAC TDA1387 X8 project

[hifinet]

ProtoDAC is an I2S (slave), NOS (nonoversampling), multibit DAC HAT using eight parallel Philips TDA1387 chips, passive I/V and is capable of 16 bit 384kHz sample rate PCM conversion. It can play higher resolution files by truncating to 16 bits. The legendary Philips TDA1387 DAC in a simple circuit with premium components results in outstanding sound quality. The DIY build is easy. As a named I2S device in Moode 8.3.0 and later, configuration is easy.

Output level: 1.24Vrms for 0dB 1kHz sine wave (DC voltage compliance for the TDA1387 chip is 0 to 3.5V. Any higher output, and clipping will occur. 3.5V peak to peak is 1.24Vrms).

Distortion:
With 420R I/V resistor THD = 0.013%  H2 = 0.012% H3 = 0.0022%
With 100R I/V resistor THD = 0.0070% H2 = 0.0068% H3 = 0.00056% Note that output is 0.235Vrms.
Detailed info from this chart of I/V resistance and distortion.

For more information, please refer to the following links:

ProtoDAC on Moode

ProtoDAC TDA1387 X8 Builders Guide and Gerber zip file for PCBs Rev B. Updated Rev C with pads for 27.5mm lead space radial output coupling caps to accommodate polycarbonate MKC caps. More info here . Updated Rev D

Original post with some additional info.

Gabster video review

For those in the US interested in a ProtoDAC parts kit, order by email ProtoDAC.HAT@gmail.com . The complete kit is $59 including shipping (USA only). Payment by PayPal invoice. Please confirm in your email that the shipping address is in the US and your PayPal address for the invoice. 

The complete parts kit includes:
One Tested TDA1387 X8 module (eight parallel TDA1387). The modules are newly manufactured and contain TDA1387 chips which are used, as is the case with all of these modules from China. Selected for balanced AC signal at 0dB.
One high quality two layer PCB.
Two Texas Components TX2575 audio metal foil 420R I/V resistors (Vishay Z-foil).
Now with two Audiophiler MKP 6.8uF/250V metallized polypropylene film capacitors. Matched pair if available. 
Two Panasonic FR-A (EEU-FR1A102L) power supply electrolytic decoupling capacitors.
Two PCB mount RCA jacks. These are made by AEC Connectors, Taiwan and have teflon insulators.
One 40 pin GPIO header
Three metal film I2S resistors

International shipping of complete ProtoDAC parts kits is available through eBay. eBay provides international shipping services through eBay International Shipping (eIS).  Due to eBay selling fees, the kit is $69 USD (plus shipping and any VAT fees collected by eIS) . Search for ProtoDAC or item # 266402263899. Seller is hifiamps. The listing may not be active if I am out of parts.

Unfortunately, eBay International Shipping will not ship ProtoDAC kits to the UK and some of the EU countries (Germany). 

Messages from eBay support chat:
I have checked and it seems that UK is not eligible for eBay international shipping for the categories of the item that are listed where the option of UK is not there.
Please allow me to share that our shipping partners need to abide by all customs laws and shipping regulations with carriers. In addition to regulations and laws, we also reserve the right to restrict listings based on high defect rates/high value items at our discretion.
It is important to note that our shipping partners will always err on the side of caution. If there is any concern that an order may not be fit to ship, we will choose to restrict it rather than risk it not being deliverable at all. In the event the category, or item was previously supported, but is no longer.

DIY SUPPORT: If you want me to check your assembled ProtoDAC, just email pictures to ProtoDAC.HAT@gmail.com . Clear pictures of the bottom of PCB and top with a good look at the GPIO. I will email back with a constructive critique of your work.

Obtaining the parts for the kit yourself: 
1) Gerber files are provided. PCBs can be ordered from JLCPCB, or many others (see here ), at very low cost.

2) Be aware that some of the TDA1387 X8 modules sold on Ali are not compatible. I recommend using modules from HuaXinHuiDianZi Store. Buy more than one. I have no affiliation with the store. Make a simple module test rig with one of your spare PCBs with GPIO soldered, three I2S resistors (R1-R3), a socket for the module (14 pin DIP socket OK), two metal film 430R I/V resistors and 2000uF power supply decoupling cap (C3/C4). There is no need for any other parts on the test rig. Download this 60s_Sine_50Hz_0dB.wav file and play on Moode with repeat track on. Measure AC voltage at the two test points, which are the output coupling cap pads at the I2S resistor end of the PCB. Measure between test point and ground. Voltage should read 1.2-1.3 Vrms, and the two channels should read very closely. I reject modules for the kit that can't pass this test. I do think modules that read 1.10-1.19 are good, but I only include 1.2-1.3 Vrms modules with the kit. I hot swap modules and wear a grounding strap. Most DMMs are accurate at 50 Hz, and accuracy suffers as the frequency increases.

Update 12/20/2024: Post # 707

3) You can use inexpensive metal film I/V resistors if you want to reduce the cost of your first build. The expensive TX2575 metal foil resistors do sound substantially better, but at significantly higher cost (about $13 USD each). Forum members are investigating tantalum nitride SMD resistors [Stackpole RTAN, Vishay PTN and Yageo NT] (@eculcea) and Caddock TF020 (@Onionhead) as alternatives to metal foil. Rhopoint 8G16D Econistor 100R I/V resistor.

[hifinet]

While the TDA1387 x8 is a drop-in replacement for the TDA1541 from the limited data I have the reverse might not be possible.

Is this assumption correct?

You are correct. I am definitely not an expert, but looking at the TDA1541 datasheet, the requirements are much more complicated. It needs 17 decoupling capacitors. It needs +5VDC, -5VDC and -15VDC supplies, and the current draw is significantly higher. I see quite a few DAC boards for the TDA1541 on Ali, most variants of the Philips HiFi Forum 10th Anniversary design. I have actually never heard a TDA1541 based DAC or CD player. From the few reviews that I have read, the TDA1387 X8 module as a replacement for the TDA1541 apparently sounds very comparable. Comparing the data sheets for THD+N for the TDA1387 is 88dB for the single chip, which is likely much better with 8 chips in parallel. The signal to noise plus THD for the TDA1541 is listed at 95dB. The TDA1387 was the last of the multibit DACs by Philips, and it is quite advanced and greatly simplified compared to the TDA1541.  The TDA1387 X8 module has only 8 active pins (and pin 4 is a duplicate BCK), and needs only a single +5VDC power supply.

[Skip Pack]

Yesterday I ordered four of the X8 1541 replacement boards from two ebay vendors and pcb boards from JLCPCB. That comes to a cost right around $50. While I really only want one well worked out DAC, the minimum on the boards is 5, so I thought it would make sense to make one base level 'test' board and then decide how many finished/optimised ones I'd complete after that.

For the test board, I thought it would make sense to put in a socket for the multi 1387 board. I could then check out each of the four I expect to receive without soldering them. Is it crucial to solder that board directly to the hat board or is this socketing a practical way to test?

I noticed that you, @hifinet, highlight soldering only a limited number of specified pins to the hat board. I'm inclined to also solder a couple of pins at the other end of the 1387 board just to limit any (unlikely) vibrations. Can I indulge my structural obession here without making a real potentially harmfull connection?

Thanks to @hifinet and others who have helped with the design, development, testing, and writeup of this project.

Skip

[hifinet]

Yesterday I ordered four of the X8 1541 replacement boards from two ebay vendors and pcb boards from JLCPCB. That comes to a cost right around $50. While I really only want one well worked out DAC, the minimum on the boards is 5, so I thought it would make sense to make one base level 'test' board and then decide how many finished/optimised ones I'd complete after that.

For the test board, I thought it would make sense to put in a socket for the multi 1387 board. I could then check out each of the four I expect to receive without soldering them. Is it crucial to solder that board directly to the hat board or is this socketing a practical way to test?

I noticed that you, @hifinet, highlight soldering only a limited number of specified pins to the hat board. I'm inclined to also solder a couple of pins at the other end of the 1387 board just to limit any (unlikely) vibrations. Can I indulge my structural obession here without making a real potentially harmfull connection?

Thanks to @hifinet and others who have helped with the design, development, testing, and writeup of this project.

Skip

You can usually get an introductory first order discount with the PCB manufacturers. I got my first 5 PCBs from JLCPCB for $1.24 shipped with a coupon. Check PCBShopper to compare.

A socket makes a lot of sense for a test DAC.  I use an old proto board setup with sockets to test modules. The pins of the module are round.

With Moode, play a 50 or 60 Hz 0dB sine wave file and measure DC and AC voltage at the input to the output coupling caps (module pins 6 and 25), with a DMM. Typically, DC voltage is 1.9V and AC is about 1.3VAC RMS with the 420R I/V resistor.

Another thing that you might consider with the test DAC is soldering the C3 and C4 decoupling capacitors on the end of the leads so they can be easily exchanged. The sound character can be affected quite a bit with variations in these capacitors. Mostly in terms of dynamics and clarity. These caps need about 60 hours of burn in.

Soldering additional pins is not a problem. The other pins have no electrical connection.

[hifinet]

I am exploring options for selling parts kits and pretested modules, Texas Components resistors, etc. for people here in the US. The TX2575 (Vishay Z-foil) resistors are expensive at $13 each plus shipping ($11 since they ship them FedEx) when purchased from the Texas Components webstore. They make a big difference in sound quality compared to other resistors. I think they are definitely worth the extra cost. I am also exploring the TX2352 (Vishay S-foil), which are less expensive at $9 each. I haven't listened to them yet. I have ordered several of the TX2575, and will sell them for my cost plus first class postage to get the ball rolling on these DACs. I also have some WIMA MKP 4 6.8uF output coupling caps (do not attract a magnet), which sound excellent.

[Skip Pack]

This morning I bought 4 ERO MKC 6.8uf caps on ebay. They are in the 'red box' configuration with the short leads, like the Wima caps you see on many, many builds. I've used them on other projects with success and they have a very good reputation. Generally considered more smooth/even than metalized Polypropylene. I bias my builds toward making the most music enjoyable regardless of the recording quality. The MKC (polycarbonate) caps have not been produced since 2002, but there are fair supplies around. I'll have to extend at least one lead.

When you get the Z-foils that you are willing to sell, I'll be interested.

[hifinet]

This morning I bought 4 ERO MKC 6.8uf caps on ebay. They are in the 'red box' configuration with the short leads, like the Wima caps you see on many, many builds. I've used them on other projects with success and they have a very good reputation. Generally considered more smooth/even than metalized Polypropylene. I bias my builds toward making the most music enjoyable regardless of the recording quality. The MKC (polycarbonate) caps have not been produced since 2002, but there are fair supplies around. I'll have to extend at least one lead.

When you get the Z-foils that you are willing to sell, I'll be interested.

I use Component Research 10uF 30V polycarbonate caps (in the ProtoDAC photo) for personal use. They are hard to find, and I only have a limited quantity on hand. I agree that PC sound a bit more revealing and smooth/even than PP. The polycarbonate is also physically small for capacitance.  Thanks for the tip on the ERO MKC. In my opinion, these WIMA MKP 4 PP sound 98% of the way to the CRC PC. I am listening to them now, and they are very nice. They are red boxes with 37.5mm lead spacing 41mm L x 15mm W x 25mm H. They fit well on the ProtoDAC PCB. The right channel cap needs to sit above the inside strip of the GPIO pins, since there is an overlap. They are much less expensive than the PC caps.

I will post here when I have Z-foils for sale.

[Tim Curtis]

Just some info from my setup :-)

- moOde 8.3.2 64-bit
- Pi-3B+
- Allo Kali Reclocker
- Two power supplies: One for Pi and the other for Kali
- CamillaDSP "Polarity Inversion with VC"
- Chunk size set to 4096

I've also used the Polarity inversion option in the MPD section of Audio Config with MPD Software volume and I can't really tell the diff between this and the CamillaDSP option but either way it's good to have choices.

Also, I'm not having any issues playing 24 bit audio through Kali with this DAC.

[DRONE7]

Also, I'm not having any issues playing 24 bit audio through Kali with this DAC.

Interesting !  and now I am even more eager to start my build, just a few more components to arrive.

Thanks for that gem Tim.

[Tim Curtis]

Just remember that the ProtoDAC TDA1387 chips will truncate any input to 16 bits.

My point was that there is no Kali "while noise" or other issue with this particular DAC and > 16 bit input.