Hi All.
I am designing an 8ch DAC with a active XO. The system will consist of 4 DAC boards using 1xTDA1387x8 and 2xNE5534 as I/V each, a Raspberry Pi 5 running CamillaDSP, and an Arduino Uno as MCU (to control trigger in/out, mute relays, etc). Pi+Uno will be powererd by a 5V 5A SMPS power supply, and this circuit will be completely isolated from the DAC boards. I2S from the PI to the DACs will go through a SI8660BA 6 channel isolator.
I am designing a linear PSU (LM317/LM337 based) to provide +12, -12 and 5 VDC. The +/- 12V will power the I/V stage, and the 5V will power the DAC chips + isolator. My initial idea is to add a denoiser (the simplest version, as I am not aiming for ultra high end) to each rail. I came up with the attached design (transformer, rectifier and filter caps ommited for simplicity) for the 12V rail. My concerns are as follows:
- It seems that more complex designs, like the dienoiser e nonoiser, requires specific component values depending on the output voltage. However, for the denoizer, AFAIK, the values remain the same across the entire operating range. Am I correct? If so, are the values in the design adequate? (I got these values from another design in this thread)
- The cabinet layout has become quite a challenge. The cables from the PSU to all 4 DAC boards will probably be 15/20 cm long. My concern is the noise that will be picked up by the cables. Is this a valid concern? If so, would it be a good idea to place the denoiser circuit in each one of the DAC boards instead? Ddon't know if there will be enough room for it, but the parts are cheap so why not.
Thanks a lot, specially Elvee and Trileru for the wealth of information and different designs in this thread.
I am designing an 8ch DAC with a active XO. The system will consist of 4 DAC boards using 1xTDA1387x8 and 2xNE5534 as I/V each, a Raspberry Pi 5 running CamillaDSP, and an Arduino Uno as MCU (to control trigger in/out, mute relays, etc). Pi+Uno will be powererd by a 5V 5A SMPS power supply, and this circuit will be completely isolated from the DAC boards. I2S from the PI to the DACs will go through a SI8660BA 6 channel isolator.
I am designing a linear PSU (LM317/LM337 based) to provide +12, -12 and 5 VDC. The +/- 12V will power the I/V stage, and the 5V will power the DAC chips + isolator. My initial idea is to add a denoiser (the simplest version, as I am not aiming for ultra high end) to each rail. I came up with the attached design (transformer, rectifier and filter caps ommited for simplicity) for the 12V rail. My concerns are as follows:
- It seems that more complex designs, like the dienoiser e nonoiser, requires specific component values depending on the output voltage. However, for the denoizer, AFAIK, the values remain the same across the entire operating range. Am I correct? If so, are the values in the design adequate? (I got these values from another design in this thread)
- The cabinet layout has become quite a challenge. The cables from the PSU to all 4 DAC boards will probably be 15/20 cm long. My concern is the noise that will be picked up by the cables. Is this a valid concern? If so, would it be a good idea to place the denoiser circuit in each one of the DAC boards instead? Ddon't know if there will be enough room for it, but the parts are cheap so why not.
Thanks a lot, specially Elvee and Trileru for the wealth of information and different designs in this thread.
Don't worry too much about absolute cable lenghths:their path and layout matters more. Identify the most likely return path for each supply, and run the return path close to supply wire.
If you need an ultralow impédance at the POL, place the denoiser locally
If you need an ultralow impédance at the POL, place the denoiser locally
+12 and +5 rails use LM317s and -12 rail uses a LM337. Quite standard setup I guess.
I finished the complete schematic late last night and still need to double and triple check everything. I will go with a single transformer and rectifier for the sake of simplicity (as I mentioned, not aiming for ultra-high-end).
Thanks Elvee! I'll go with the denoisers in the PSU PCB then.
What about the component values? Are they adequate for the 3 rails? I tried to read all of the thread but I am quite sure that I was not able to process all the information here. 😂
Thanks a lot!
the values for +/-12 are OK and they will also work for 5V, but for 5V 1K/100K for the collector and base resistor will gain you 1~2dB.
Note that the 337 can be fussy, it depends on the maker. You might have to tweak R16.
Read the VRDN thread for more detail, in case you have oscillations
Note that the 337 can be fussy, it depends on the maker. You might have to tweak R16.
Read the VRDN thread for more detail, in case you have oscillations
I would not use a 337 but build two identical positive regultors powerd by two independant secondary windings ( instead of a center tap transformer winding ).
I do not take any chance about stability and like a shorter BOM list.
I would use a much smaller output capacitor, 1 or 10 uF, using the 100uF at the DAC boards decoupling.
I do not take any chance about stability and like a shorter BOM list.
I would use a much smaller output capacitor, 1 or 10 uF, using the 100uF at the DAC boards decoupling.
I don't remember who was the PCB designer or what was the name of the project, but a few years ago I saw a really cute idea for positive and negative voltage regulator boards:
A single PCB held three voltage regulator circuits -- arranged as three adjacent rectangles which can be scored or sawed or Dremel-tool-wheel-cutoffed apart if so desired.
In this way, both types of "customer" can be accommodated. If someone wants a positive voltage regulator and a mirror-image-circuit negative voltage regulator, they simply cut off and discard rectangle #1. Leaving #2 for the positive regulator, its exact mirror image circuit #3 for the negative regulator.
Equally delightfully, if another person wants two identical voltage regulator circuits, which they wish to configure as positive and negative, they simply cut off and discard rectangle #3. Leaving #1 and #2 which are exactly identical.
Everybody's happy.
A single PCB held three voltage regulator circuits -- arranged as three adjacent rectangles which can be scored or sawed or Dremel-tool-wheel-cutoffed apart if so desired.
- PCB rectangle #1 (on the left) held a positive voltage regulator using circuit-design-X
- PCB rectangle #2 (in the middle) held a second exact copy of the positive regulator using circuit-design-X
- PCB rectangle #3 (on the right) held a negative voltage regulator using circuit-X-with-Pdevices-and-Ndevices-swapped
Equally delightfully, if another person wants two identical voltage regulator circuits, which they wish to configure as positive and negative, they simply cut off and discard rectangle #3. Leaving #1 and #2 which are exactly identical.
Everybody's happy.
That sounds practical but with a dual supply you can have shared ground. With a triple version each has to have its own ground path. Higher chances of ground loops?
Is that so? As I went with a single transformer/rectifier and only two filter cap banks, my intention is to go with a single ground plane for all 3 supply circuits. As the transformer has a 12-0-12 output and the center wire is the ground one, I thought a single ground plane would work just fine fine. At least it works on a protoboard.
The alfa version of the board looks like this (double check pending):
Board would look like this (still have to make room for the heatsinks).
The top circuit is -12V, center is +12V and the bottom one 5V. All paths are in the back layer. Front layer is the ground plane.
My PCB design experience is equivalent to my ancient Greek cousine knowledge 😂, i.e., nothing more than a few Youtube videos. But it is quite fun to connect the dots and get to something that might work!
Hello, can using an external pnp transistor to expand the current 3A ensure stability and low impedance?
As I said many times, the denoiser is not supposed to be used together with other "non-regular" tweaks or additions. It works safely and effectively with a "naked" 317, but any addition risks compromising the stability. You are free to try it for yourself, but there is no guarantee.
A safer option is to use a LM338 instead
A safer option is to use a LM338 instead
The parts finally arrivend and I built (in a protoboard) the negative rail (-12VDC) of the PSU in post #3247. As many others, I am facing oscilation under load. With a 680R load, the oscilation has about 70mV peak to peak and a frequency of 2.42Mhz. The oscilation is proportional to the load (a 330R load aproximately doubles the peak to peak voltage):
The circuit is (a snippet of the complete circuit in the post above):
If I remove R16 completely (keeping C19), it stops oscilating. However, I am not experienced enough to understand R16 function and the side effects of removing it.
Is it OK to go without R16 (I'll keep its place in the PCB and will put a jumper instead)? Is there anything else that you guys recommend me to try/check?
Thanks!
The circuit is (a snippet of the complete circuit in the post above):
If I remove R16 completely (keeping C19), it stops oscilating. However, I am not experienced enough to understand R16 function and the side effects of removing it.
Is it OK to go without R16 (I'll keep its place in the PCB and will put a jumper instead)? Is there anything else that you guys recommend me to try/check?
Thanks!
Last edited:
The 337 is known to be temperamental, even in its simple datasheet configuration, and combined with a denoiser, two sub-varieties exist, depending on the manufacturer and production date. One requires some damping of the compensation cap, thus R16 (although 22 ohm seems somewhat excessive), the other is happy with no resistor at all.
You seem to belong to the latter category.
Note that it is possible to optimize the value, by applying a chopped load and monitoring the collector of the correction transistor to look for the minimal overshoot.
The chopped load can be applied through a contraption like this one:
You can of course content yourself with the stability of the circuit with zero-ohm series resistance
You seem to belong to the latter category.
Note that it is possible to optimize the value, by applying a chopped load and monitoring the collector of the correction transistor to look for the minimal overshoot.
The chopped load can be applied through a contraption like this one:
You can of course content yourself with the stability of the circuit with zero-ohm series resistance
Thanks Elvee! Without the resistor it seems that everything is working just fine.
I built the +12V circuit now. And I am getting another "strange" behavior. When I power up the PSU, the voltage goes do 12.8V, then goes down to 11.5V, then up to 12.6V, down again to 11.8V, back up to 12.4V.... You got the idea. After about 25 to 30 seconds, output voltage stabilizes at 12.2V (which is just fine). It is like an VERY slow oscilation that keeps decreasing and then stops.
I replaced every single component in the circuit to be sure that it is not something defective. Nothing changed. Tha's the circuit:
If I add a 1k resistor as load and, when I turn on the PSU, the oscilation is smaller/quicker, but still happend for about 10 to 15 seconds.
The same behavior happens in the 5V circuit (which uses 1k/100k in R2 and R3 respectively, as per Elvee recommendation). Of course I tested without the denoiser and the oscilation doesn't happen.
I am assuming this behavior is not normal. I understant that it might happen if there is no load, but even with a 1k load, it is still there.
Any tips on what to probe or what to adjust to fix this (if it is indeed a problem)?
Thanks a lot you all!