I run custom filters on an old DAC with PCM1704 but are not going to extend that to needed 8 ch.
I can check when i get home, in Thailand on a business trip right now. What cores do you use?
Those in the pic are P14s gapped at AL=160. Seeing as you mention your current DAC is PCM1704, you do realize this is only a 16bit DAC? The limitation on number of bits in the TDA1545s is the main reason for wanting to incorporate an analog volume control, with more bits to play with digital volume may be a more attractive solution.
Last edited:
IIRC i got my cores from Mouser, TDK RM10 or 12 gapped to ? It is 10 years ago but i guess they still keep a decent stock.
I am still on the Phidac quad. My previous setup was usb isolator -> usb es9018 -> spidif -> AKM4418 ->I2s -> Phidac. Very nice, musical and better than anything else I have ever heard. So after years of humming and hawing over getting the Singxer F-1 to replace the es9018 I pulled the trigger.
This setup was usb Singxer F-1 -> AKM4418 -> I2s -> Phidac. Immediate improvement. Dynamics galore. Bass !! Full and thick and glorious. It moves air in my headphones now. Presentation is slightly muddled on complex songs. Some music is too intense because the top end feels sharp and muddled at the same time. It is too hard to follow and so becomes fatiguing.
So today I had time to pull out the soldering iron. Now it is just usb Singxer F-1 -> I2s -> Phidac. It didn't make sense to have the F-1 decode usb just to encode to spdif and then have the AKM4418 decode spdif to I2s.
Now there is a solid presentation. Things don't get muddled anymore. But this is not the thing I wanted to mention. The Singxer F-1 is an enormous improvement in dynamics, bass, pratt, full thick and detailed sound. It is just so much more and the interesting part is that the little Phidac stepped up. I hadn't even known what it was capable of before. It also prefers 24bit to 16bit input. 16bit sounds flat compared to 24bits.
So now I will be moving on to Celibidache because I have to know how could it possibly get better than this.
Let me put it this way, if I was told the Phidac quad I have right now would be the last dac I could ever have, I would be ok with it.
So a question to you all who have moved through the various iterations. In what ways is the Celibidache better than the Phidac?
This setup was usb Singxer F-1 -> AKM4418 -> I2s -> Phidac. Immediate improvement. Dynamics galore. Bass !! Full and thick and glorious. It moves air in my headphones now. Presentation is slightly muddled on complex songs. Some music is too intense because the top end feels sharp and muddled at the same time. It is too hard to follow and so becomes fatiguing.
So today I had time to pull out the soldering iron. Now it is just usb Singxer F-1 -> I2s -> Phidac. It didn't make sense to have the F-1 decode usb just to encode to spdif and then have the AKM4418 decode spdif to I2s.
Now there is a solid presentation. Things don't get muddled anymore. But this is not the thing I wanted to mention. The Singxer F-1 is an enormous improvement in dynamics, bass, pratt, full thick and detailed sound. It is just so much more and the interesting part is that the little Phidac stepped up. I hadn't even known what it was capable of before. It also prefers 24bit to 16bit input. 16bit sounds flat compared to 24bits.
So now I will be moving on to Celibidache because I have to know how could it possibly get better than this.
Let me put it this way, if I was told the Phidac quad I have right now would be the last dac I could ever have, I would be ok with it.
So a question to you all who have moved through the various iterations. In what ways is the Celibidache better than the Phidac?
H
HAYK
I want to consult you about biasing the TDA1387.
The TDA1545 is very similar, the DS says if you bias 2ma instead of1ma they Ifs, the -60db THD+N falls to 1.4% from 1.7%.
The TDA1387 has also the same 11k in Iref input with Vref of 1/6Vcc. If I sink the current to devide 1ma Ifs, I get 13.2 which is exactly the Afs of TDA1545. I conclude that if I ground the Vref pin with another 11k, I will get 2ma Ifs= I bias on TDA1387 and maybe 1.4%.
The question is, with Ifs doubled, did I also doubled the gain to 2ma p-p output?
The TDA1545 is very similar, the DS says if you bias 2ma instead of1ma they Ifs, the -60db THD+N falls to 1.4% from 1.7%.
The TDA1387 has also the same 11k in Iref input with Vref of 1/6Vcc. If I sink the current to devide 1ma Ifs, I get 13.2 which is exactly the Afs of TDA1545. I conclude that if I ground the Vref pin with another 11k, I will get 2ma Ifs= I bias on TDA1387 and maybe 1.4%.
The question is, with Ifs doubled, did I also doubled the gain to 2ma p-p output?
Hi Hayk - the TDA1387 pin7 is a voltage input, not current. The way to get more current out is to bias pin7 with a resistor to pin5 (VDD). I don't think 2mA is possible but I get ~1.5mA with 110k to VDD when VDD=5.5V. You might find my blog useful : https://www.diyaudio.com/community/threads/tda1387-and-tda1545a.384224/post-7030615
H
HAYK
H
HAYK
This is the good one. It is identical to 1545, the 22k+33k are inside. The Iref is the current through the 11k. But the Rref is not 11k as the DS says. Or else your 110k added will give 1.75ma or 1.7ma if the pull up resistor is a CCS.
What I found in my experiments was the Iout didn't increase linearly with pin7 voltage. As you keep on increasing the pin7 volts, the Iout doesn't increase in proportion. So you end up needing a higher pin7 voltage than you would if it was a linear relationship.
H
HAYK
I don't think Philips troubled themselves to make the Iref generator much different. From the figure 1 of TDA1545, the voltage across the Rref is already 1/6Vdd. As they could bring the compliance to 0, the Vref was needless so as the outside resistors.
If the 1545 can out 2ma Ifs with 22k from Vdd, I don't see why the 1387 can't. In your case that would be too much heat because multiple// but I am investigating to use only one pair loading each positive and negative of each channel in the same chip for differential output.
The minimum possible THD+N at -60db is -42db, with 0.14% only 4db is away from best.
If the 1545 can out 2ma Ifs with 22k from Vdd, I don't see why the 1387 can't. In your case that would be too much heat because multiple// but I am investigating to use only one pair loading each positive and negative of each channel in the same chip for differential output.
The minimum possible THD+N at -60db is -42db, with 0.14% only 4db is away from best.
Heat certainly was a problem with biassing up the 1387s to higher current, I swapped to 1545 as it was a more efficient (slightly lower power, fewer chips) way to achieve the same output current. I'd be interested to hear how you get on coaxing your 1387s to reach 2mA.
H
HAYK
Probably the Iout at 0v is too low for 2ma, I will see what potential gives the lowest THD, as you found that it is not linear.
I suspect (though haven't verified) that the THD doesn't much change, the THD+N is lowered because the noise lowers.
H
HAYK
George, a customer of you, has posted very interesting measurements.
https://www.diyaudio.com/community/threads/16-bit-dac-at-60db.406935/page-2#post-7547967.
https://www.diyaudio.com/community/threads/16-bit-dac-at-60db.406935/page-2#post-7547967.
H
HAYK
Tda1387, tda1387t, tda1387/n1. What is the difference?
I see the philips labeled ones used in your projects, how about NXP ones?
I see the philips labeled ones used in your projects, how about NXP ones?
I haven't noticed any differences between those labels in my experience. I don't think I have any NXP labelled chips, all mine are older, recycled ones from 'Soundblaster' cards and the like.
What I have noticed in testing a few hundred chips is the ones with 'THAILAND' printed vertically like this :
TL
HA
AN
ID
have on average a lower output current level than those with 'THAILAND' printed horizontally. The difference is small and unlikely to be audible.
What I have noticed in testing a few hundred chips is the ones with 'THAILAND' printed vertically like this :
TL
HA
AN
ID
have on average a lower output current level than those with 'THAILAND' printed horizontally. The difference is small and unlikely to be audible.