ProtoDAC TDA1387 X8 project

[thomaschan]

https://www.youtube.com/watch?v=5Rxz75QFQig

A FET output stage is added to the 1387 chip for improved performance. Better dynamic range and musicality with a warm touch.

Three power supplies for optimal performance. 4A for Pi 4, low noise 5V for1387 DAC chip, linear power supply for FET output. Internal shielding to minimize noise.

Beautiful work. Active (like a Pass D1 w/o buffer) or passive I/V? Very inspiring.

FET output is active powered by DC15V original from founder of Lampizator.

[hifinet]

The "FET-ishizator"? Quite a sense of humor.

That is passive I/V with a simple FET inverting gain stage. The 90 ohm resistor is the passive I/V resistor. The FET VAS gets you probably the 12-13 dB of gain for line level. This will also get you polarity inversion, like the other Lampizator circuits. If you are using SPDIF input, this will invert polarity to normal. The BF245B is discontinued. I was just looking at the BF245A with another DAC I/V, and that is discontinued also. I see they have them on Ali.

Have you tried it without the 220uF cap on the 820R source resistor? That will give some local feedback and lower distortion. It may reduce the gain too much.

It sounds like TDA1387 X8 plus passive I/V plus FET VAS is a winner. Thank you for sharing and posting your findings, and the detailed picture of your streamer.

For true active I/V conversion, the gate of the FET, grid of the tube, or noninverting input of the op amp need to be grounded or at a voltage referenced level.

One interesting active I/V circuit I just found is the D1NOB1 in this diyA thread. It is balanced, but just use half of the circuit for single ended. It is a mod of the Pass D1 I/V stage. There is a version with a buffer output (like the Pass B1 preamp) called the D1B1 in the same thread, but performance is worse, so they dumped the buffer. I think this could be easily adapted to the TDA1387 X8 as is. 

If you reduce it to a single-ended version, it is only slightly more complex than the FET circuit above. They later upped supply voltage to +45V/-45V to get ridiculously low THD numbers (-130dB). Power dissipation was quite high though , requiring significant heat sinks. Then in the later posts, OPC realized you could do away with the -45V supply and still have excellent performance with reduced power dissipation.

The TDA1387 X8 would see very low resistance with the D1. Much lower than passive I/V. I am not shure if the TDA1387 X8 really cares. Once you get below 100 ohms, it may not matter.

[hifinet]

I was just informed by a ProtoDAC kit customer that ProtoDAC would not work with the PCM2706 USB to I2S card (GY-PCM2706). 

Changed the three I2S resistors (R1-R3) from 430R to 100R and it works.

[thomaschan]

The "FET-ishizator"? Quite a sense of humor.

That is passive I/V with a simple FET inverting gain stage. The 90 ohm resistor is the passive I/V resistor. The FET VAS gets you probably the 12-13 dB of gain for line level. This will also get you polarity inversion, like the other Lampizator circuits. If you are using SPDIF input, this will invert polarity to normal. The BF245B is discontinued. I was just looking at the BF245A with another DAC I/V, and that is discontinued also. I see they have them on Ali.

Have you tried it without the 220uF cap on the 820R source resistor? That will give some local feedback and lower distortion. It may reduce the gain too much.

It sounds like TDA1387 X8 plus passive I/V plus FET VAS is a winner. Thank you for sharing and posting your findings, and the detailed picture of your streamer.

For true active I/V conversion, the gate of the FET, grid of the tube, or noninverting input of the op amp need to be grounded or at a voltage referenced level.

One interesting active I/V circuit I just found is the D1NOB1 in this diyA thread. It is balanced, but just use half of the circuit for single ended. It is a mod of the Pass D1 I/V stage. There is a version with a buffer output (like the Pass B1 preamp) called the D1B1 in the same thread, but performance is worse, so they dumped the buffer. I think this could be easily adapted to the TDA1387 X8 as is. 

If you reduce it to a single-ended version, it is only slightly more complex than the FET circuit above. They later upped supply voltage to +45V/-45V to get ridiculously low THD numbers (-130dB). Power dissipation was quite high though , requiring significant heat sinks. Then in the later posts, OPC realized you could do away with the -45V supply and still have excellent performance with reduced power dissipation.

The TDA1387 X8 would see very low resistance with the D1. Much lower than passive I/V. I am not shure if the TDA1387 X8 really cares. Once you get below 100 ohms, it may not matter.

I did not put in any cap in parallel to the 820 ohm resistor. The sound quality and gain is quite OK to me. I choose this FET design as opposed to tube because I want to put everything together in a not too big chassis. The FET output only requires DC15V which takes up very small space. 

I built this for an old friend together with a preamp (based on MBL6010) and small mono block LM1875 power amp. He will be using B&W 606S2 speaker. Should be quite decent setup for sweet vocal.

[speedy210]

Hello, is it possible to use a Gipo like this? The background to this is that I have installed a controller that is also connected via Gipo.

[hifinet]

Hello, is it possible to use a Gipo like this? The background to this is that I have installed a controller that is also connected via Gipo.

Yes, as long as GPIO pins 2,4,6,12,35,40 pass through.

[speedy210]

Hello, is it possible to use a Gipo like this? The background to this is that I have installed a controller that is also connected via Gipo.

Yes, as long as GPIO pins 2,4,6,12,35,40 pass through.

Yes, that shouldn't be a problem.  I only need 17,27,22 and one GND.

[hifinet]

@thomaschan This reminds me of a favorite FET preamp from Audio Amateur in the 70's. A single stage FET common source amplifier is essentially a single ended triode, without the heater supply. THD will be in the 0.1-0.2% range, and with the proper part, mostly H2. The best sounding tubes have low THD, almost all H2 and virtually no H3. THD with the FET circuit can be lowered by increasing the supply voltage with excellent voltage regulation.

The output of TDA1387 X8 is up to 800mV with a 100 ohm I/V resistor. I wonder if that is high enough to be fed directly to the MBL6010 based preamp. Does it have enough sensitivity, gain and is it low noise?

In the D1 thread, the consensus was that the best sounding circuits have the lowest parts count with the shortest signal path. In that circuit, they used high output MOSFETs in a simple circuit and cranked the voltage up to +45/-45V to get lowest THD.

[hifinet]

@thomaschan The BF245B circuit with the 820 ohm source resistor is biased for a TDA1541A. The TDA1387 X8 will have a different bias point. I think the TDA1541 sinks about 4mA current at idle. The TDA1387 X8 produces about 4.5mA current at idle (based on 1.9V/420 ohm I/V resistor). After the 1541, the voltage on the gate with 100 ohm I/V is -0.4V. With the 1387 it is +0.45V (you can confirm and measure this voltage at the gate with a DMM). With the 820 ohm source resistor, I think the BF245B is under biased. To achieve the same bias point as the 1541, the value of the source resistor should be increased to 1000 ohms. You could try it and see if it sounds better.

[hifinet]

@eculcea kindly sent me the tantalum resistors discussed in post #712. After 20 hour burn-in, I think they sound excellent. Their best quality is openness of sound. They initially lacked the clarity of the TX2575, but now after burn-in that has improved significantly. I can recommend these as an inexpensive and readily available alternative to the TX2575.

Part numbers: Vishay PTN1206E1000BST1 100 ohm 1206 size, Stackpole RTAN1206BKE100R 100 ohm 1206, Yageo NT1206BRD07100RL 100 ohm 1206.

The SMD mounting PCB is from Ali.