Thanks for the quick reply Salas.
Since I'm going to buy a new MOSFET anyway, would you have a recommendation between the STP55NF06L or the MTP3055VL?
Since I'm going to buy a new MOSFET anyway, would you have a recommendation between the STP55NF06L or the MTP3055VL?
Since strong 1/F noise filtering C2 high value is not your subjective priority with your DAC and you liked lower values maybe you can find many types 100-220uF of small size like special Low Z HF electrolytic, solid polymer, or bipolar for audio to test too. I mean its OK you like your now components but they look big and difficult in the photo. Compromising for size when you want to mount securely. Also some Wima like MKS2 50-63V are not too bad and they give some μf in still smallish size. LM329 will change things, its much better for technical spec than any normal Zener.
So now call for Salas then to adjust Ref-D boards for Mundorf 1000uf Ag+ (4 pin) in C2 position got leds filtering. It really outperforms other caps.
I got also same experiences with Mundorfs - you need to give them some time to burn in to fully bloom.
Didn't you like 16v polymer here too?
Right now, the mini is serving lots of needs in that it can go to higher output current with, especially with the IRF9530 as CCS.
I suppose Ref-D could be done with this, and J2 in it, making it a more tweakable board.
Polymers do the job yes. Although they are very clear, especially on heights, they have very bad tone. Due to that I only use them for powering DVDDs. For powering AVDDs/Vref in voltage out DACs like ak4490 or Soekris I use Mundorf 1000uf AG+ 4 pin. Among many other caps I tried here this one is the best sounding. Heights are clear, bass is tight and have very good resolution.
Hi Salas, I already use 100+ 47uf. small size, in C2. I use Nichicon KZ, small and cheap. Panasonic pureism PX is also good, but beware sanyo oscon and elna silmic II, I think these best only in signal pass. In PSupply others still worse.
The problem is how place a large 4,7uF PIO. In analog supplies, provide overall midrange and TONE to complement the bass coming from 147uF electrolytic. For sure 4,7 Mundorf Evo Oil not as large as mine.
Replacing it for an small MKP can make sense for low volts digital supplies, I try soon in next supply project 3,3volts USB board.
On analog supplies I would upgrade instead to Mundorf M-lytic if space limitations.
I together with Tiofrancotirador, strongly believe yes yes... the board deserves keeping place M-Lytic as C2. Eyes closed.
The board R-D may look simple, but rewards in spades any effort done. CONGRATS Salas - Teabag anyway for achievement.
My Zener bought straight in local store, also can't reach 8,2V. I must deleted order from Spain provider RSComponents and hope soon receive purchase from DigiKey.
I will see till some next gb IF such a 35mm x 35mm 160V four pin cap can be ALTERNATIVELY fitted for C2 with some jumpers still converting to normal C2 option. Because I wouldn't force a boutique part as only way. Maybe turning the Big Ref-D to more of a "connoisseur" PCB including even more decisions than already since there is the straightforward Mini also available anyway.
Fine going upgrade Big REF-D.
My vote also to defeat extra components related to parallel output.
My vote also to defeat extra components related to parallel output.
Why you don't use remote sensing? Near to load and finding no difference? Use short thick wires when in normal mode (two wire) at least. I will see in practice because big parts both.
not compare myself Salas, but somewhere this thread, I read two wire has better sound.
I like simplicity, not only for sound, also, having to much troubleshot to deal with in the past, related to complex circuiteries.
I like simplicity, not only for sound, also, having to much troubleshot to deal with in the past, related to complex circuiteries.
Salas,
Polymer of 1000uf capacitance has around 6A of ripple current (electrolytes achieves that at 10000uf)
So, would it make sense for powering digital circuits (like DVDDs, digital parts of the dac chips in general) to use 1000uf polymer caps as main filtering cap instead of big 6800/10000uf electrolytes?
Polymer of 1000uf capacitance has around 6A of ripple current (electrolytes achieves that at 10000uf)
So, would it make sense for powering digital circuits (like DVDDs, digital parts of the dac chips in general) to use 1000uf polymer caps as main filtering cap instead of big 6800/10000uf electrolytes?
Its not a high current switching power supply operating at tens or hundreds of kHz so to prioritize how much ripple current the cap takes before heating much and wearing fast. In such application it has to do with its HF ESR basically. There its punished with rapid high current triangular shots by a coil charge, whatever. What we mainly need make sure in our case of filtering is there is enough capacitance in the reservoir for the max expected load current it sees (here fixed by CCS) so the voltage is not to drop below an acceptable level at the regulator's input.
Say you have a 9VAC transformer and you set 500mA CCS in a REF-D of 5V output. 9*1.41=12.7V. Minus about 1.4V from two conventional bridge diodes conducting at any one time. 12.7V-1.4V=11.3V. They convert 50Hz to 100HZ (10mS) also. Reservoir capacitor must keep 5V higher Vin than Vout for our MOSFET CCS to be working very well. Even more is better. The lower that voltage difference the less rigid our CCS becomes until under 2V that it will even let pass 100HZ ripple at a point. 11.3V-10Vin=1.3V max charge wave is our goal (V).
C=I*t/V. C=500mA*10mS/1.3V or C=(0.5A*0.01sec)/1.3V=0.003846 Farad=3846uF. Thus with 3846uF the 100HZ ripple will be 1.3V and the Vin will not drop below 10V. That's borderline, we don't normally like Volt+ level ripple trying our circuit and polluting around with its charge cycle harmonics if we can do better. So we use chunkier value reservoir capacitor to have few tens or low hundreds mV level ripple. The standard value was chosen to cover well the max CCS spec in the guide with 9VAC Tx. Since we try keep Tx VAC low enough not to have too much heat in the CCS sink. I hope I gave you the main idea of how things play out. You can manipulate the parameters and parts each time by taking application specific decisions given the settings etc. If CCS was 100mA in the same example then the minimum reservoir capacitor fulfilling the set criteria could be 770uF for instance. Or 1500uF for 200mA. And so on and so forth.
Say you have a 9VAC transformer and you set 500mA CCS in a REF-D of 5V output. 9*1.41=12.7V. Minus about 1.4V from two conventional bridge diodes conducting at any one time. 12.7V-1.4V=11.3V. They convert 50Hz to 100HZ (10mS) also. Reservoir capacitor must keep 5V higher Vin than Vout for our MOSFET CCS to be working very well. Even more is better. The lower that voltage difference the less rigid our CCS becomes until under 2V that it will even let pass 100HZ ripple at a point. 11.3V-10Vin=1.3V max charge wave is our goal (V).
C=I*t/V. C=500mA*10mS/1.3V or C=(0.5A*0.01sec)/1.3V=0.003846 Farad=3846uF. Thus with 3846uF the 100HZ ripple will be 1.3V and the Vin will not drop below 10V. That's borderline, we don't normally like Volt+ level ripple trying our circuit and polluting around with its charge cycle harmonics if we can do better. So we use chunkier value reservoir capacitor to have few tens or low hundreds mV level ripple. The standard value was chosen to cover well the max CCS spec in the guide with 9VAC Tx. Since we try keep Tx VAC low enough not to have too much heat in the CCS sink. I hope I gave you the main idea of how things play out. You can manipulate the parameters and parts each time by taking application specific decisions given the settings etc. If CCS was 100mA in the same example then the minimum reservoir capacitor fulfilling the set criteria could be 770uF for instance. Or 1500uF for 200mA. And so on and so forth.
Thanks Salas for nice explanation.
If I got your point right then:
For my specific needs: 3.3V @ 600mA
Specific environment: 50Hz line.
Components used: 9VAC trafo, Schottky diodes (0.7V drop)
I would need the following capacitance:
Voltage on reservoir capacitor: 9V*1.4=12.6V, 12.6V - 0.7V = 11.9V
Vin - Vout difference: 11.9v-3.3V = 8.6V
5V margin (voltage left to ripple 50Hz) : 8.6V-5V = 3.6V
Capacitance: C=(0.6*0.02sec(50Hz))/3.6 = 0.003333 Farad=3333uF, assuming +/-20% tolerances on capacitance of caps then it gives ~4000uf.
Which means I would need to use at least 4x1000uf/16V polymer capacitors array. It is doable.
Now I got following consideration:
Do 16V rated caps have enough margin to be used in 12V environment.
will it bring any improvement anyway 🙂, but since it is reported that the quality of voltage drop resistor R1 has impact (some replaces it with Caddock 930) then I assume main reservoir should have impact as well on overall quality.
Assuming the following:
- Worst case (50Hz line) - needs the biggest capacitance for ripple.
- using Schottkies diodes on rectification to have less voltage drop. They are not that expensive.
- DVDD - almost every chip is powered with 3.3V, so are most clocks
- AVDDs/Vref - here tonality is important on voltage out DACs (different caps gives different tone, where polymers are the worst from my experiences) and voltage can go as high as 7.2V (e.g.: ak4490)
then maybe following could be considered:
Instead of having one universal ref-D having two:
- one dedicated for DVDDs and clocks running @3.3V with array of 4 1000uf/16V polymers for rectification (or maybe even 6, since they are very small and still can fit without loosing footprint of ref-D) and one 1000uf/16V polymer for LEDs filtering
- the other one dedicated for AVDDs/Vref (still universal also) with option for 4 pin cap on rectification and with option for 4 pin cap for LEDs filtering definitely.
If I got your point right then:
For my specific needs: 3.3V @ 600mA
Specific environment: 50Hz line.
Components used: 9VAC trafo, Schottky diodes (0.7V drop)
I would need the following capacitance:
Voltage on reservoir capacitor: 9V*1.4=12.6V, 12.6V - 0.7V = 11.9V
Vin - Vout difference: 11.9v-3.3V = 8.6V
5V margin (voltage left to ripple 50Hz) : 8.6V-5V = 3.6V
Capacitance: C=(0.6*0.02sec(50Hz))/3.6 = 0.003333 Farad=3333uF, assuming +/-20% tolerances on capacitance of caps then it gives ~4000uf.
Which means I would need to use at least 4x1000uf/16V polymer capacitors array. It is doable.
Now I got following consideration:
Do 16V rated caps have enough margin to be used in 12V environment.
will it bring any improvement anyway 🙂, but since it is reported that the quality of voltage drop resistor R1 has impact (some replaces it with Caddock 930) then I assume main reservoir should have impact as well on overall quality.
Assuming the following:
- Worst case (50Hz line) - needs the biggest capacitance for ripple.
- using Schottkies diodes on rectification to have less voltage drop. They are not that expensive.
- DVDD - almost every chip is powered with 3.3V, so are most clocks
- AVDDs/Vref - here tonality is important on voltage out DACs (different caps gives different tone, where polymers are the worst from my experiences) and voltage can go as high as 7.2V (e.g.: ak4490)
then maybe following could be considered:
Instead of having one universal ref-D having two:
- one dedicated for DVDDs and clocks running @3.3V with array of 4 1000uf/16V polymers for rectification (or maybe even 6, since they are very small and still can fit without loosing footprint of ref-D) and one 1000uf/16V polymer for LEDs filtering
- the other one dedicated for AVDDs/Vref (still universal also) with option for 4 pin cap on rectification and with option for 4 pin cap for LEDs filtering definitely.
Your minimum capacitance is half what you calculated because a bridge rectifier makes 50HZ to 100Hz so plug 0.01 sec time in your calculations
Any chemical capacitor used to no higher than about 70% its nominal voltage is considered good practice for its longevity
Schottky diodes are a good thing for your voltage envelope. Must watch their peak reverse voltage spec is high enough though. 30-50V reverse ones should be plenty in this case
Any chemical capacitor used to no higher than about 70% its nominal voltage is considered good practice for its longevity
Schottky diodes are a good thing for your voltage envelope. Must watch their peak reverse voltage spec is high enough though. 30-50V reverse ones should be plenty in this case
P.S. Find any candidate diode's forward drop on its published curve taking into account the current to use in your application. Nominal drop is quoted for a relatively high current usually. Two diodes are dropping simultaneously in a full wave bridge rectifier at any moment. So two times the found forward drop is their voltage loss always
Thanks Salas for further tips. To be specific I am using SiC Schottky Cree C3D04060A. I took a look at their Forward Voltage characteristics specs and it is hard to read from graph exact voltage drop for so low current. It seems like I will have to experiment and measure real voltage across rectification with different VAC applied and CSS setting and then plan how much capacitance do I really need.