Hello,
I love this forum, many talented people help each other to create very nice projects. Big thanks to all contributors here.
After a long time only reading i began creating my own diy projects and now i also have a question or two.
I created a dac based on the pcm5102 because of its simplicity (my first project, so i didnt want to shoot too high
).
Anything works really fine, but now i would like to create a linear psu for this dac. I bought a toroid transformer of nice quality ( audio grade by toroidy.pl ) and i use 2 tps7a4700 ldo regulators behind it. My question now is : do i need some special rectifier diodes ? What sounds good ? Are fast switching better for a dac psu or slower ones ?
My transformer is 2x7V with 3,5VAC. The Dac itself needs only little power, so i think it should do fine.
Second question is : the DAC has a digital 3.3V input, an analoge one and a chargepump input ( it still requires only positive voltage, negative is created internally). I have only two rails. Should i use one rail for the digital input and the other for chargepump AND analoge or should i use the two rails differntly ?
Thanks in advance for your help.
Greetings
Tom
I love this forum, many talented people help each other to create very nice projects. Big thanks to all contributors here.
After a long time only reading i began creating my own diy projects and now i also have a question or two.
I created a dac based on the pcm5102 because of its simplicity (my first project, so i didnt want to shoot too high
). Anything works really fine, but now i would like to create a linear psu for this dac. I bought a toroid transformer of nice quality ( audio grade by toroidy.pl ) and i use 2 tps7a4700 ldo regulators behind it. My question now is : do i need some special rectifier diodes ? What sounds good ? Are fast switching better for a dac psu or slower ones ?
My transformer is 2x7V with 3,5VAC. The Dac itself needs only little power, so i think it should do fine.
Second question is : the DAC has a digital 3.3V input, an analoge one and a chargepump input ( it still requires only positive voltage, negative is created internally). I have only two rails. Should i use one rail for the digital input and the other for chargepump AND analoge or should i use the two rails differntly ?
Thanks in advance for your help.
Greetings
Tom
Last edited:
Hi Tom and welcome to diyAudio
Soft recovery diodes would probably be a recommended choice if you believe they would make an audible difference. You might also find standard devices like the 1N4004 used with a snubber (0.01uf and say 2.2 ohm in series across each) are just as effective.
You would have to search for soft recovery types, there are hundreds to choose from.
It makes sense to separate the digital and analogue supplies if possible, not just the supplies but the grounds as well. I would keep the analogue supply totally separate if possible.
Soft recovery diodes would probably be a recommended choice if you believe they would make an audible difference. You might also find standard devices like the 1N4004 used with a snubber (0.01uf and say 2.2 ohm in series across each) are just as effective.
You would have to search for soft recovery types, there are hundreds to choose from.
It makes sense to separate the digital and analogue supplies if possible, not just the supplies but the grounds as well. I would keep the analogue supply totally separate if possible.
For a <40V DC output with <0.5 ampere load, I suggest you include these rectifier diodes among your pool of candidates
- SBYV28-100 (100V, 3.5A, axial)
- SBYV27-200 (200V, 2.0A, axial)
- UF4004 (400V, 1.0A, axial)
- MUR460 (600V, 4.0A, axial)
- MBR1100 (100V, 1.0A, axial)
I believe people started to call these xformer-rectifier contraptions 'linear' AFTER the introduction of SMPSs,to distinguish them. Only explanation I can think of.
an
In my opinion the "low hanging fruit" is to prevent oscillation of the transformer secondary's leakage inductance. There are other things you can do but this is the easiest and offers the greatest benefit for the least effort and lowest parts count.
In my opinion the easiest way to do this is to install a snubber which overdamps the transformer secondary's resonant circuit. For those who prefer to consider the resonant "Q" instead of the "damping", I remind you that Q = (1/(2*damping_ratio)), so my suggestion is to install a snubber which reduces the resonant Q to 0.5 or less.
How?
I recommend searching right here on diyAudio for the Quasimodo test jig. It includes a 20 page "no math" discussion.
After reading how to do this with "no math", if you decide that you prefer "with math" the Quasimodo paper gives References which point you to (1) Jim Hagerman's math-filled treatise; (2) Cornell-Dubilier's math-filled treatise; (3) Morgan Jones's math-filled treatise.
Disclaimer: I sleep with the person who coined the name "Quasimodo" so I am not a disinterested third party approaching this topic with insouciant indifference.
In my opinion the easiest way to do this is to install a snubber which overdamps the transformer secondary's resonant circuit. For those who prefer to consider the resonant "Q" instead of the "damping", I remind you that Q = (1/(2*damping_ratio)), so my suggestion is to install a snubber which reduces the resonant Q to 0.5 or less.
How?
I recommend searching right here on diyAudio for the Quasimodo test jig. It includes a 20 page "no math" discussion.
After reading how to do this with "no math", if you decide that you prefer "with math" the Quasimodo paper gives References which point you to (1) Jim Hagerman's math-filled treatise; (2) Cornell-Dubilier's math-filled treatise; (3) Morgan Jones's math-filled treatise.
Disclaimer: I sleep with the person who coined the name "Quasimodo" so I am not a disinterested third party approaching this topic with insouciant indifference.
Last edited:
No, the word "linear" doesn't refer to the rectification, but to the regulation part of a PSU. In a linear PSU, the regulating transistor is used in the linear part of its curve (on or off and any state in between). In an SMPS the regulating transistor is only used in on or off states, hence the term "switching".
Last edited:
Last edited:
When the diode drops out of conduction on each cycle, it can cause ringing and a burst of high frequency interference to be produced. Its very real, I've encountered the effect twice on commercial products, a radio and a tuner where the effect was a 'mains buzz' type interference on the audio (but caused by the tuner stages picking it up). The snubber damps it all down and slows the rise time of any sudden interference produced.
Just thought I'd throw in another option the BYV-32E http://www.nxp.com/documents/data_sheet/BYV32E-200.pdf
I've been using them for quite a while. Hugh Dean put me onto them. I've not done any tests to determine if they make a difference or not, but they are not terribly expensive, so I just use them as a matter of course
There are two diodes per package, the first time I only used one, but now I use the two in parallel. There doesn't appear to be any issue with the current sharing.
Tony.
I've been using them for quite a while. Hugh Dean put me onto them. I've not done any tests to determine if they make a difference or not, but they are not terribly expensive, so I just use them as a matter of course
There are two diodes per package, the first time I only used one, but now I use the two in parallel. There doesn't appear to be any issue with the current sharing.
Tony.
They look a useful device. I remember the first time I came across the problem of diodes and 'hi speed' diodes. It was one of the EW (east west) modulator diodes in a TV, and one was short circuit. I replaced it with what would probably be a 1N5408 or similar and I couldn't understand that while the fault symptom changed drastically, it wasn't fixed.
A lot of water has passed under the bridge since then
A lot of water has passed under the bridge since then
Unfortunately NXP does not state the junction capacitance, either in the datasheet or in any of the application notes or product literature. This could ruin the effectiveness of a single-R, single-C snubber if the calculations assumed one value of Cj while the real diode provided quite another.
One way to skin this cat is to assume a monstrous value of Cj; "pessimists only receive pleasant surprises." Then calculate your required 1R,1C snubber, and feel comfortable: when the real Cj turns out to be smaller than assumed, it decreases the denominator of the math expression for zeta. Smaller denominator means bigger zeta. Bigger zeta (lower Q) means greater comfort. Just set Cseries >= (30 X Cj_monstrous_assumption)
Another way to go is to squander money like a drunken sailor: install a second capacitor "Cx" (giving a 1R,2C snubber) which swamps out any and all possible Cj values. Quasimodo suggests Cx=10nF which is conservatively enormous; I'm shipping product today with Cx=3.3nF which is conservative but not quite so enormous. Admittedly this is with a non-NXP rectifier whose Cj IS plotted on the datasheet: 500pF. If I didn't know Cj I would probably have used a bigger and more conservative Cx. Beware this second capacitor Cx might cost as much as ten cents AND you have to lay it out on your board.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.