Open-source USB interface: Audio Widget

Didn't mean to intimidate anyone!

I'm happy to share my layout knowledge. Plus, there are so many ways to do things. Perhaps you belong to the figurative 4-layer school while I'm more of a 2-layer impressionist :)

And if you buy into what Ott writes we have the same starting point.

Børge
 
Now that we are in the knowledge sharing mood, any thoughts on using TL431 shunts as shunts for the supplies? I have seen onoclocks interesting studies of the using the TL431 in a series regulator, but on diyaudio's analog forums, shunt regulators seem to have more street-credit. :cool:

In the attached schem, I have included the option for a npn SOT-23 device in cases where more current needs to be shunted.
 

Attachments

  • ab_cv_supply.png
    ab_cv_supply.png
    71.3 KB · Views: 253
I finally got my camera working.

These really belong over in the oscillator thread but since I mentioned here...
It is mixed signal though.

These are development boards. I panelized them to save setup charges.
There is one big booboo, see if you can find it. These are from Make PCB in China. I'm not 100% satisfied but these will do. They seem to have trouble getting the holes centered with the pads.

Cheers,
 

Attachments

  • Top.jpg
    Top.jpg
    295.8 KB · Views: 247
  • Bottom.jpg
    Bottom.jpg
    255.3 KB · Views: 236
Didn't mean to intimidate anyone!

I'm happy to share my layout knowledge. Plus, there are so many ways to do things. Perhaps you belong to the figurative 4-layer school while I'm more of a 2-layer impressionist :)

And if you buy into what Ott writes we have the same starting point.

Børge

When something just makes sense it's difficult to ignore.
What Ott writes just.....

Cheers,
 
I have used Make PCB a lot, and you are right that putting the holes in the middle is not their strongest skill. They also have the most hilarious order status system, but very reasonable priced.

Why do your two layout panels share the same gnd plane? Isn't that taking Ott's points too far? ;)
 
Member
Joined 2004
Paid Member
Now that we are in the knowledge sharing mood, any thoughts on using TL431 shunts as shunts for the supplies? I have seen onoclocks interesting studies of the using the TL431 in a series regulator, but on diyaudio's analog forums, shunt regulators seem to have more street-credit. :cool:

In the attached schem, I have included the option for a npn SOT-23 device in cases where more current needs to be shunted.

There is nothing intrinsically better about a shunt regulator. The single clear benefit is unambiguous clear indication of where they are connected. Audiophile cred often has no basis in reality and fails miserably when really put to test.

For a USB powered device the lack of efficiency is a real issue.

The TL431 is a great device but its not very low noise. The circuit I posted early in this thread is about as low noise as is possible (1 nV/rt Hz) with an active regulator and optimized for oscillators. It also has much higher line noise rejection than a shunt regulator, really important for oscillators.

I have a variation that is a shunt regulator, attached.
 

Attachments

  • Shunt Reg.JPG
    Shunt Reg.JPG
    38.4 KB · Views: 224
  • shunt noise filter.asc
    2.6 KB · Views: 42
There is nothing intrinsically better about a shunt regulator. The single clear benefit is unambiguous clear indication of where they are connected. Audiophile cred often has no basis in reality and fails miserably when really put to test.

For a USB powered device the lack of efficiency is a real issue.

The TL431 is a great device but its not very low noise. The circuit I posted early in this thread is about as low noise as is possible (1 nV/rt Hz) with an active regulator and optimized for oscillators. It also has much higher line noise rejection than a shunt regulator, really important for oscillators.

I have a variation that is a shunt regulator, attached.

Thanks for your input. I am just toying around with different ideas, but ADP151 being the baseline it seems like there are many good options out there. I believe I have found your other regulator here: http://www.diyaudio.com/forums/digi...sb-interface-audio-widget-52.html#post2810970 - is LT1431 better than Onsemi TL431?

BTW: You are right about the Audiophiles. One of the best one I have heard so far is about the importance of which way you place resistors - can't remember if tolerance band needs to be in or against the direction of the signal. But not to worry, you can hear the difference - especially if you have a high end fuse in your circuit. ;)

EDIT: And regarding USB, I plan to use another supply, so that is not really a concern for me.
 
Last edited:
I have used Make PCB a lot, and you are right that putting the holes in the middle is not their strongest skill. They also have the most hilarious order status system, but very reasonable priced.

Why do your two layout panels share the same gnd plane? Isn't that taking Ott's points too far? ;)

Because they are two boards. I.m cutting them down the middle.
 
There is nothing intrinsically better about a shunt regulator. The single clear benefit is unambiguous clear indication of where they are connected. Audiophile cred often has no basis in reality and fails miserably when really put to test.

For a USB powered device the lack of efficiency is a real issue.

The TL431 is a great device but its not very low noise. The circuit I posted early in this thread is about as low noise as is possible (1 nV/rt Hz) with an active regulator and optimized for oscillators. It also has much higher line noise rejection than a shunt regulator, really important for oscillators.

I have a variation that is a shunt regulator, attached.

This is very impressive. Is there a way of improving the efficiency.
With no load it's drawing 218mA.
 
Member
Joined 2004
Paid Member
This is very impressive. Is there a way of improving the efficiency.
With no load it's drawing 218mA.

The shunt will draw whatever is necessary to pull the supply down to the target voltage. Efficiency be damned if you are going this route. Its the least efficient regulator possible. . .

R9 is a simulated load. That's where most of the current is going. Remove it and add your circuit there.

I have not looked at or run the simulation in a long time. The series resistor from the 5V supply is what controls the current.

I may cook up a two quadrant variation on the concept and see what it does.

I had to build a very special test fixture to measure the noise of this supply. If there is interest I'll post the details.
 
Last edited:
Because they are two boards. I.m cutting them down the middle.

David, one thing you might consider in designs like that is to add a straight V-cut line down the middle. The the boards are going to come apart nicely. The other alternative is to use a routing tool >2.0mm between breakaway tabs.

Some board manufacturers want to charge extra because they see that there are now two boards in one production panel. But it is easy to find ones which don't charge more. I've used PCBCART - Professional China Prototype PCB Manufacturer quite a bit.

Børge
 
And a bit of promotion: The AB-1.2 has two unpopulated layouts of Demian's series LDO regulator on the PCB. The various loads on the board can be easily strapped to these regulators. Or the regulators on the AB-1.2 PCB can be populated and strapped into whatever you want. I did populate one, and the output voltage was dead-on 3.3V. Adding a pair of XOs as a load didn't change it at all. Follow my signature to the AB-1.2 schematics to learn more.

Picture of regulators: http://www.qnktc.com/AB_12_01.jpg

OK, enough promotion for now :)

Børge
 
The shunt will draw whatever is necessary to pull the supply down to the target voltage. Efficiency be damned if you are going this route. Its the least efficient regulator possible. . .

R9 is a simulated load. That's where most of the current is going. Remove it and add your circuit there.

I have not looked at or run the simulation in a long time. The series resistor from the 5V supply is what controls the current.

I may cook up a two quadrant variation on the concept and see what it does.

I had to build a very special test fixture to measure the noise of this supply. If there is interest I'll post the details.


I haven't done power supply for my OSC so I'm interested. We can discuss this over in the OSC thread if it doesn't fit here.
 
David, one thing you might consider in designs like that is to add a straight V-cut line down the middle. The the boards are going to come apart nicely. The other alternative is to use a routing tool >2.0mm between breakaway tabs.

Some board manufacturers want to charge extra because they see that there are now two boards in one production panel. But it is easy to find ones which don't charge more. I've used PCBCART - Professional China Prototype PCB Manufacturer quite a bit.

Børge

Hi Borge. I use a 1mm or 1.5mm fishtail chipper router bit for PCB cutting. I built a nice table with a moveable fence and a Proxxen drill press. The neat thing about cutting this way is it doesn't matter what angle the bit is approached from I always get a straight clean cut. I took a chance and tried to fool then manufacture to save money. Make PCB didn't seem to care.

I'll try PCBcart next time.

Thanks,
 
Member
Joined 2004
Paid Member
Low noise measurement preamp

Here is the circuit. Its based on a surplus Geoformer I found on eBay. The transformer has a 35:1 ratio. I used a low noise preamp on the output to boost the overall gain to 80 dB. The equivalent input noise resistance is about 3 Ohms or .2 nV/rt Hz. Which means you can measure below the .7 nV/rtHz of a typical low noise transistor.

Calibrating this is a pain. I made a resistive divider 1000-1 50 Ohms in .05 Ohm out. Its pretty load insensitive. And to measure the noise floor I used a decade resistor and increased it until I got a 3 dB increase in noise, between power line harmonics. My FFT analyzer has a uV/rt Hz mode that makes measuring noise easier and more rational. Used with this preamp the whole process is pretty easy.
 

Attachments

  • LNP.PNG
    LNP.PNG
    12 KB · Views: 468
  • ultra low noise measurement preamp.zip
    9.8 KB · Views: 56
What is in the pipeline?

1. Isolated i2s AW's:

An externally hosted image should be here but it was not working when we last tested it.


http://luta.7u.cz/imgLinks/audioweb/sdrwidget/DAC2/Left48k-4dB.png

2. Using TI C5535 dsp as the engine:

C5535 eZdsp USB Stick Development Kit - TMDX5535EZDSP - TI Tool Folder

The list price is $99 but you can get promotional price for half of that sometimes.

The stock ezdsp5535 comes with uac1 firmware allowing 16bit 48khz audio in adaptive mode.

We have developed firmware for the ezdsp5535 with its on board AIC3204 codec:

a. AC2 stereo playback at 32 bit (of which 24 significant bits passed to AIC3204) at sampling rates of 44.1/48/88.2/96 khz (limited by the AIC3204).

b. Current version of firmware customized for Linux ONLY for development and testing. Another version of firmware will be available for MAC OSX in future. Another version with an ASIO driver will be available for Windows in future.

Join the audio-widget google group if you are interested in downloading an experimental boot image file for your ezdsp5535 to play with. The software is still under active development and is open source. The source code is in github:

git://github.com/alexlee188/C5535.git

The ezdsp5535 will be the engine for a line of AW with 5.1 and even 7.1 playback capabilities, as the C5535 DSP has 4 I2S ports. Plans are underway to use the expansion port of the ezdsp5535 to connect to external DACs such as PCM51xx to confirm 32/384 capabilities and the SQ.

3. Capture Widgets:

This is in the conceptual stage.

Alex