Oh... Go on.
The OP is a HiFiNut^2 so a bit of fiddling can only lead to pain and, as long as it is reversible, no pain no gain. Of course if it means hacking tracks to bodge something on then it might be worth thinking twice.
Not sure what sort of noise they might be thinking about but perhaps they do want things to work at RF...
Words of warning might/would be that three terminal regulators are prone to 'hooting' if you do not use the 'right capacitors' in the right places. The 7809/7909 have low value decoupling at the input.
More importantly look at what is going on at the inputs and outputs of U7 and U8. It is hard to see from the picture but the original designer has added a pair of low value capacitors in parallel.
One is labelled PEF which might indicate a Polyester film. 0.018uf or 180nf. The others seem to be 1uF Y5V ceramics. Similar loss to X7R but dirt tolerance and therefore likely to be cheaper... You really need to look at the impedance curves.
The values and/or selection of dielectric type do not seem to be arbitrary.
Someone appears to have thought about this but of course if the OP knows better then they should get their hacksaw out and go for it. Do not under any circumstances look at the data sheets for the regulators concerned.
Last edited:
My thought was that due to finite impedance at regulator output the digital rails are far from DC and rail noise corresponds to the load from the digital circuit at higher frequencies. The current draw from the 3V regulator supply would then generate noise at the 5V supply output, and so on.
In other words, the LCRs / ferrites are not there to help getting rid of the ripples from the power source but to reduce the noise coming from the load affecting the upper stream circuits. The noise may come from spiky current draws or from capacitive coupling. Since the rail impedance is provided by the regulator and the capacitors, the LCRs / ferrites would not increase the rail impedance anyway so cause no harm.
Of course, I can be wrong because I have not done any simulations / studies / measurements, and my electronic knowledge is at beginner's level. My source of information and/or misinformation could have come from reading posts in this forum. 🙂
It is just a thought. If you can give me a good lesson then I can learn something from you.
I had a dear friend of mine who inspired me to start audio DIY. Because of my craziness/passion about sound and audio, he called me HiFiNut and sometimes simply Nut Nut. He later passed away. I joined this forum at about that time so I called myself HiFiNutNut to pay tribute to my friend as if he were still there calling me that name, whether it were a silly name or not.
I never assume that I know better than this (reputable) manufacturer. I question because I don't understand and I want to learn.
But in this particular case, I sometimes doubt if parts of the circuits of this player were contracted out to some graduates to do. There are dozens of parts that are not connected, indicating that they might use the same prototype boards for final production. For an example, the 0.1uF 25V Y5V ceramic after the 7809, the AZ1117_5V and the CW1117_3V regulators are not connected.
The manufacturer's datasheet for AZ1117 recommends 10uF tantalum at the input and 22uF tantalum at the output. These caps give a very smooth, flat impedance that don't get below 0.3R. They may be required for stability of the regulator. So why the schematic has those ceramic caps but not installed?
CE14, the input cap for the 3V regulator, is placed possibly 80cm away from the input of the regulator! Is that OK? I would put it right in front of the regulator. or am I wrong? why do we want inductance in front of a regulator input?
I do read the datasheets. Amazingly, while the datasheets indicate tantalum caps at the outputs, the player has 0.018uF polyester caps at the output as well as 100uF low ESR/impedance caps nearby, as well as 0.1uF X7R MLLC SMD at the chip on the same rails! and the player is working!
I also noted that the designer loads the IV opamp OP275 with 780R DC load and only 490R AC load... I think the designer likes them "HOT".
Here is my question again.
I think, theoretically, low value decoupling at the regulator input may be needed, seeing it from the regulator point of view. But in practice, the power supply cable originated from a SMTP supply on the other side of the player to the regulator on this side of the player goes through 300mm to 400mm length. I guess a low value decoupling cap 0.1uF/25V/Y5V at the input here being a high Q capacitor could potentially form some LC resonant circuit with the cable inductance, introducing some impedance/noise peak.
Would a modern low impedance capacitor 100uF 25V Rubycon ZL at the regulator input be good enough to ditch the 0.1uF/25V/Y5V?
A resonant tank with high Q must be avoided, the low ESR electrolytic cap thus is indeed the better solution.
I have just compared the datasheets of the LM1117 and AZ1117 and found the functional block diagrams to be identical and most parameters are near identical. I guess I can safely assume that they are basically the same thing.
The AZ1117 has an example circuit with 10uF tantalum input cap and 22uF tantalum output cap. The LM1117 specifies output capacitance to be 10uF at the minimum but larger the better - improve the loop stability and transient response with the ESR of the output capacitor should range between 0.3 Ω to 22 Ω.
But this player has a low ESR (possibly less than 0.1R) 100uF 50V cap and 0.1uF X7R SMD at the output of the AZ1117. Perhaps the AZ1117/LM1117 can handle much lower ESR than its datasheet specifies?
The AZ1117 has an example circuit with 10uF tantalum input cap and 22uF tantalum output cap. The LM1117 specifies output capacitance to be 10uF at the minimum but larger the better - improve the loop stability and transient response with the ESR of the output capacitor should range between 0.3 Ω to 22 Ω.
But this player has a low ESR (possibly less than 0.1R) 100uF 50V cap and 0.1uF X7R SMD at the output of the AZ1117. Perhaps the AZ1117/LM1117 can handle much lower ESR than its datasheet specifies?
The output impedance of the regulator is already inductive, and probably more inductive than a ferrite bead.
The regulator provides DC. Dealing with digital signal currents is the task of local rail decoupling, not the PSU regulators. If you have a problem (and you don't know that you have a problem?) then you are looking in the wrong part of the circuit to solve it.
The regulator provides DC. Dealing with digital signal currents is the task of local rail decoupling, not the PSU regulators. If you have a problem (and you don't know that you have a problem?) then you are looking in the wrong part of the circuit to solve it.
LT1117 recommend ESR lower than 0.5 ohm as output cap. it will be easier for people if manufactures provides a chart of capacitance , ESR stable region chart.
I think the sure way to do it, is through measurement.
Thank you for your insights. I am now convinced that there is no benefit (or perhaps there is harm) to change the circuit.
I like to play with things. Below are the two images of the SMPS power supply before and after my modifications on the Blu Ray same player. This is the stage before the cascaded linear supplies to the DAC. I did the modifications a couple of years ago. I am now doing it again on my second player and I hope I can do better this time.
Attachments
Ooooh.... RC snubber on the primary of what looks likely to be a Flyback Converter. I wrote some embarrassing babble about similar on a Forward Converter...
Snubbing The Forward Converter
Your first ring at turn off is leakage inductance in combination with primary/switch capacitance. The second ring is primary inductance with combination with primary/switch capacitance.
You can also see ringing when the switch is on. That will be leakage inductance with secondary/diode capacitance when it is off... That's why I guessed Flyback. You might want to put an RC snubber across that one as well.
This is the sort of thing where you can make meaningful improvements without too much stress.
Hi MobidFractal,
You are spot on.
Would you be happy with the results of the damping of the first 2 ringings? Any further improvements could be made?
As for the 3rd ringing, there are multiple of secondaries. So I need to get all of the snubbers right before I can see the resonance(s) gone. I didn't know the exact capacitance of the diodes and didn't measure the leakage inductance at higher frequencies so I was unable to calculate the snubbers correctly.
You are spot on.
Would you be happy with the results of the damping of the first 2 ringings? Any further improvements could be made?
As for the 3rd ringing, there are multiple of secondaries. So I need to get all of the snubbers right before I can see the resonance(s) gone. I didn't know the exact capacitance of the diodes and didn't measure the leakage inductance at higher frequencies so I was unable to calculate the snubbers correctly.
Looks like you have nailed the important primary side one. The result does modify the lower frequency ring out but as suggested since they are on the same node you cannot really do much more.
In respect of values there is a way of getting quite good answers. I did derive the sums for this but that was some time ago and my brane has been addled with too much alcohol to try again.
The premise is that you measure the frequency of ringing, F1. Then you add some known capacitance in parallel with the Diode, C2, and measure the frequency again, F2. C1 is the raw diode junction capacitance.
You end up with two equations....
F1 = 1/2PI.SQRT(L.C1)
F2 = 1/2PI.SQRT(L.(C1+C2))
Where L is the leakage inductance. You rearrange those two in some way, I forget how, and that gives you a value for C1. Then you go back with that into F1 and the value of the Leakage inductance drops out.
Your snubber capacitor value is set to 3 times that discovered for C1 and the snubber resistor is...
Rsn = SQRT(L/C1)
Yes... with multiple outputs it can be tricky to isolate things but many times assuming the secondary windings are tightly coupled it turns out that hitting one by and large tames the other ones.
Yes, I agree it is hard to do the primary side damping. I found in my previous note:
Dampping Lm Coss: 770kHz, 3971R, 326pF
Dampping Ll Coss: 3.5MHz, 870R, 326pF
I cannot remember how I calculated it at the time (2 years ago). The result is that it requires 3971R to damp the Lm Coss, and 870R to damp the Ll Coss. I chose a resistor 2 x 470R (6W) and the above picture was the result.
For the leakage inductance, I measured each of them while shorting all other windings of the transformer. But I used only a cheap LCR meter and I presume it was done at relatively lower frequencies.
How different it is to measure leakage inductance at lower frequency and higher frequency? Could my LCR give me readings that are completely useless?
Dampping Lm Coss: 770kHz, 3971R, 326pF
Dampping Ll Coss: 3.5MHz, 870R, 326pF
I cannot remember how I calculated it at the time (2 years ago). The result is that it requires 3971R to damp the Lm Coss, and 870R to damp the Ll Coss. I chose a resistor 2 x 470R (6W) and the above picture was the result.
For the leakage inductance, I measured each of them while shorting all other windings of the transformer. But I used only a cheap LCR meter and I presume it was done at relatively lower frequencies.
How different it is to measure leakage inductance at lower frequency and higher frequency? Could my LCR give me readings that are completely useless?
Do you think that in the image for the modified PSU, the damping on the primary side is slightly under-damped vs over-damped? Perhaps I can change the resistor value 20% on either side and see if it improves?
Last edited:
- Status
- Not open for further replies.