Melodious MX-U8 USB Interface 8 Core XMOS chip

[b0bb]

The LT 3042 is a new class of regulator designed to work at very high/radio  (>1MHz) frequencies. 
The article below compares it to the LT1962, a conventional high performance regulator, the 3042 is operational @10MHz with 60dB of PSRR, the LT1962 is effectively out of the picture at the range.
 
http://powerelectronics.com/regulators/quiet-ldo-employs-unique-architecture-cut-noise-and-boost-psrr
 
I am using the 3042 on another project, USB-I2S DDC . The 3042 is good enough that separate supplies to the XO/clocks are not required, hence the  supply inputs are jumpered together on the DDC  board. Top right on picture.
 
Top board is the DDC, bottom board is the 3042 regulator board, it drives and external transistor to provide > 200mA.
 

 
There are couple of disadvantages.
 
The 3042 is a very small SMD flatpack
RF frequencies require special handling and several problems with the PS board shown here has to be corrected, picture was taken before I rebuilt it.

 

[rb2013]

   
Hi and i think i have some issues with the english ... it is only my 5th language ... 

Question ... what stops us to mod the Melodious to get a Hydra like unit ... what ?
  Thanks again, gino 

The Hydra Z is based on a completely different processor - not the XMOS.  So that is impossible.
 
It would be difficult to make the Melodious adapted to a 5VDC external feed.

 

[rb2013]

   
Because the post is from 2011, LT3042 is from 2015

I see thanks!  I see the TI LM 2941 is not on the chart as well - so I assume it's also newer the 'New Breed of Ultra Low Noise Regulators' as well.
 

  The LT 3042 is a new class of regulator designed to work at very high/radio  (>1MHz) frequencies. 
The article below compares it to the LT1962, a conventional high performance regulator, the 3042 is operational @10MHz with 60dB of PSRR, the LT1962 is effectively out of the picture at the range.
 
http://powerelectronics.com/regulators/quiet-ldo-employs-unique-architecture-cut-noise-and-boost-psrr
 
I am using the 3042 on another project, USB-I2S DDC . The 3042 is good enough that separate supplies to the XO/clocks are not required, hence the  supply inputs are jumpered together on the DDC  board. Top right on picture.
 
Top board is the DDC, bottom board is the 3042 regulator board, it drives and external transistor to provide > 200mA.
 

 
There are couple of disadvantages.
 
The 3042 is a very small SMD flatpack
RF frequencies require special handling and several problems with the PS board shown here has to be corrected, picture was taken before I rebuilt it.

The 3042 PSRR is really excellent at these extreme frequencies.  What about the LM2941 the PSRR looks very good @ 1MHz 68dB.  TI does show beyond that on their PDF spec sheet, but it looks flat from approx .5Mhz to 1Mhz
 
Good luck on that project

 

[rb2013]

Thanks for this great link - so looking at the specs and comparing the Linear LT3042 ad LT1962 to the TI LM2941  I see some interesting numbers - but don't know if they are directly comparable (different mauf).
 
So SSRP at 1 Mhz:
LT1962 20dB, LT3042 79dB, LM2941 68dB
 
Output noise (uVRMS 10Hz to 100kHz)
LT1962 20, LT3042 0.8, LM2941 .003
 
http://www.ti.com/lit/ds/symlink/lm2941.pdf

 RMS output noise, % of VOUT 10 Hz to 100 kHz, IOUT = 5 mA 0.003%

 
http://powerelectronics.com/regulators/quiet-ldo-employs-unique-architecture-cut-noise-and-boost-psrr

 

[b0bb]

  Thanks for this great link - so looking at the specs and comparing the Linear LT3042 ad LT1962 to the TI LM2941  I see some interesting numbers - but don't know if they are directly comparable (different mauf).
 
So SSRP at 1 Mhz:
LT1962 20dB, LT3042 79dB, LM2941 68dB
 
Output noise (uVRMS 10Hz to 100kHz)
LT1962 20, LT3042 0.8, LM2941 .003
 
http://www.ti.com/lit/ds/symlink/lm2941.pdf
 
http://powerelectronics.com/regulators/quiet-ldo-employs-unique-architecture-cut-noise-and-boost-psrr

The PSRR at the high frequencies is the important parameter
LT3042 79dB, LM2941 68dB
 
If you look at the graph in the article, the 3042 maintains 60dB PSRR @10MHz. The LM2941 is pretty decent at @1MHz, it is a pity TI did not characterize it any higher. PSRR >50dB is sufficient
 
0.003%@5V is 150uV, quite tolerable as this is a 5V pre-regulator and this is a digital system. If this was a DAC it is another matter.
 
My rebuild of the regulator board is almost done.
 
Before:

 
After:

 
The LT3042 needs both low ESR and low ESL (Equivalent series inductance). It can go unstable with the wrong choice of components.
 
The brown cap (C4) was marginal, ESR measured at 60mOhm, 3042 recommended ESR for the output cap is 50mOhm and the inductance was quite high given it wound foil construction.
 
The replacement is the black SMD Kemet tantalum polymer cap with an ESR of 4mOhm.
 
R4 (SMD resistor in the center) sets the output voltage and C6 is the AC bypass for R4.
C6 uses a stacked construction to reduce the ESL, it is a 10uF TDK XR7 MLCC ceramic cap sitting ontop of  a 1uF Rubycon Acrylic Stacked film cap. The film cap is to help cushion the big MLCC to reduce microphonics and smoothtout the performance in the mid freqs (100kHz-1MHz)
 
As comparison, the big red WIMA and C6 are both 10uF, C6 holds a lot of capacitance in a very compact package.
 
C5 is the bypass for the output cap, same stacked construction of MLCC + plastic film.
 
Power supply filter caps are 1000uF OSCON SEPF replacing the Panasonic FCs, the diodes were replaced with fully insulated low leakage Schottky diodes, this helps reduce  the potential of an unwanted short with the exposed metal tabs of the original.
 
The only thing I have left to do is to replace R4 with a Vishay VSMP bulk foil SMD resistor to reduce inductance even further.

 

[rb2013]

The PSRR at the high frequencies is the important parameter
LT3042 79dB, LM2941 68dB

If you look at the graph in the article, the 3042 maintains 60dB PSRR @10MHz. The LM2941 is pretty decent at @1MHz, it is a pity TI did not characterize it any higher. PSRR >50dB is sufficient

0.003%@5V is 150uV, quite tolerable as this is a 5V pre-regulator and this is a digital system. If this was a DAC it is another matter.

My rebuild of the regulator board is almost done.

Before:



After:



The LT3042 needs both low ESR and low ESL (Equivalent series inductance). It can go unstable with the wrong choice of components.

The brown cap (C4) was marginal, ESR measured at 60mOhm, 3042 recommended ESR for the output cap is 50mOhm and the inductance was quite high given it wound foil construction.

The replacement is the black SMD Kemet tantalum polymer cap with an ESR of 4mOhm.

R4 (SMD resistor in the center) sets the output voltage and C6 is the AC bypass for R4.
C6 uses a stacked construction to reduce the ESL, it is a 10uF TDK XR7 MLCC ceramic cap sitting ontop of  a 1uF Rubycon Acrylic Stacked film cap. The film cap is to help cushion the big MLCC to reduce microphonics and smoothtout the performance in the mid freqs (100kHz-1MHz)

As comparison, the big red WIMA and C6 are both 10uF, C6 holds a lot of capacitance in a very compact package.

C5 is the bypass for the output cap, same stacked construction of MLCC + plastic film.

Power supply filter caps are 1000uF OSCON SEPF replacing the Panasonic FCs, the diodes were replaced with fully insulated low leakage Schottky diodes, this helps reduce  the potential of an unwanted short with the exposed metal tabs of the original.

The only thing I have left to do is to replace R4 with a Vishay VSMP bulk foil SMD resistor to reduce inductance even further.

Sweet! You're light years ahead of me on this level of mod/build.

 

[b0bb]

Sweet! You're light years ahead of me on this level of mod/build.

Thanks. if you have any plans to use the DiyINHK 3042 boards, suggest checking them over, I see they are also using regular low ESR aluminium caps  for the output.
The reference  and bypass (C6/R4 equivalent on their board) might benefit from closer inspection.

 

[ccschua]

hi B0bb, can I know if I can just remove the LT 1963 and 220uF in the U8 and replaced with a 3.3V TPS7A4700 set at 3.3V ? 

 

[b0bb]

  hi B0bb, can I know if I can just remove the LT 1963 and 220uF in the U8 and replaced with a 3.3V TPS7A4700 set at 3.3V ? 

Yes but there a few problems to solve.
 
The TekDevice board you mentioned earlier has a diffrent pinout. You need rewire the TekDevice board to patch the PCB lands on the Melodious

 
In my case I was translating from 78xx pinout on the TekDevice to the LM315 pinout on the board.
The TekDevice board is not decoupled properly and the regulator can go unstable, you will need to add a10uF  tantalum cap to the DC input.
 
The next problem is to figure how to mount the TekDevice board onto the Melodious, this is a mechanical issue.
 
The last thing is to figure out how to remove the LT1963 from the PC board, it is a surface mount DDPak with an integrated heatsink.
You will need a lot of heat  applied in a carefully controlled fashion to avoid warping the PCB and causing the copper foil to peel off.
 
 
The tool for that job is a hot air wand and it will take between 15 and 30mins of slow careful heating.

(I do not recommend cutting the leads and leaving the thing on the board, care needs to be taken to avoid overstressing the PCB connection points (lands) if you do ).

 

[rb2013]

Sweet! You're light years ahead of me on this level of mod/build.

Thanks. if you have any plans to use the DiyINHK 3042 boards, suggest checking them over, I see they are also using regular low ESR aluminium caps  for the output.
The reference  and bypass (C6/R4 equivalent on their board) might benefit from closer inspection.

Looking at getting on of their DXIO PRO 3A with the NDK's, they mention ultra low noise regulators, but the photo is too grainy to make out which. So when I get one, I'll keep that in mind.

The Breeze Audio is blowing me away! Getting close to the Hydra Z in SQ. This is by far the best of the low cost units. Thinking about how good it might be with CCHD957's. Will have to get one of those sweet heat wands, so I can donthe DIP14 mounting easier.

 

[ccschua]

sorry.
 
I am using the following board which can accept 6 to 9V input.  
 
http://www.ebay.com/itm/TPS7A4700-Ultra-Low-Noise-Adjustable-1-4V-20-5V-1A-Power-Supply-/391133464079?hash=item5b115f060f
 

 

[b0bb]

  sorry.
 
I am using the following board which can accept 6 to 9V input.  
 
http://www.ebay.com/itm/TPS7A4700-Ultra-Low-Noise-Adjustable-1-4V-20-5V-1A-Power-Supply-/391133464079?hash=item5b115f060f
 

This is not suitable as it has a built in diode rectifier and AC input.
In order to get the best results with the high RF frequencies, the replacement regulator as to be as close as possible to the connections of the LT1963
 
The TPS7A4700 is also an RF LDO, so care has to take to keep the connections as shorts as possible or it will oscillate and go unstable.
 
The one in my photo is what I would recommend
http://tekdevice.com/chapter2/index.php?route=product/product&product_id=131

 

[b0bb]

The Breeze Audio is blowing me away! Getting close to the Hydra Z in SQ. This is by far the best of the low cost units. Thinking about how good it might be with CCHD957's. Will have to get one of those sweet heat wands, so I can donthe DIP14 mounting easier.

 
I am glad you like the Breeze Audio. The differences between it and the Hydra Z is boiling down to the Crystek XOs, quite impressive given it is 1/4 the cost of the Hydra.

 

[ginetto61]

Originally Posted by rb2013 /img/forum/go_quote.gif
The Hydra Z is based on a completely different processor - not the XMOS.  So that is impossible.
It would be difficult to make the Melodious adapted to a 5VDC external feed.

Hi and thanks a lot for the very helpful advice.
Please let me rephrased a little.
 
I did not measure the V out of the LT1963 ... let's say that is 7 V just as an example.
I understand that:
1)  the LT1963 is not the end of the world for performance
2)  the apd150 after it are decent but not spectacular
On this basis
 
could providing the needed 7V with an external very low noise power supply be beneficial ? 
 
I am sure there is some noise at the LT1963 out pins ...
A higher quality external DC power supply could generate less noise and less EMI and vibrations than with the transformer inside the unit and close to the board.
 
However ...
I have looked at your mods ... very very excellent.  Congratulations. They look extremely neat.
Questions:
1)  do you select the new parts (i.e. caps, diodes)  looking at their specifications ?
2)  after mods do you check the result with a scope ?
Thanks a lot again
gino

 

[Sergey83149]

Сообщение от  abartels 

 
For those who wonder why I 16V of a cap instead of share 25V caps:
The rail a warehouse of food of food provides max. 9.5V, before regulation (when using the transformer with an exit 7B)
It also counts on noise food Ultra Low.
 
Hi Abartels. If to replace only caps with NICHICON FG, it is better to choose 16v or 25v? I plan to change nothing more.
These it is better?

http://priceinchina.ru/buy_elna_cerafine_roa_220uf_16v_audio_aluminum_electrolytic_capacitors_aliexpress_42ONCCA