Now let me do some maths. There seems to be no need to know the leakage inductance of the secondary winding.
Since we know the C, if we can observe the ringing frequency F via the probe-in-the-air, then we have the answer.
Since F = 1 / (2 * pi * sqrt(L * C)), so we can work out L = 1 / (4 * Pi * Pi * F * F * C). We can select R = sqrt(L/C). So R = sqrt( 1 / (2 * pi * F * C) ^ 2), or R = 1 / (2 * pi * F * C). Oops, I just realised that I turned a big circle to get back to a very familiar formula, excuse me.
So for C=1.2nF, if f = 10MHz, R = 13R, and if f = 100MHz, R = 1.3R. A 20R trimpot is suitable for the test.
C can be 4.7nF up to 10nF.
The RC can be installed at the secondary windings before the diode.
What do you think?
OOOOOops! I can be wrong - there are 3 secondary windings so which winding is causing which resonance?
Since we know the C, if we can observe the ringing frequency F via the probe-in-the-air, then we have the answer.
Since F = 1 / (2 * pi * sqrt(L * C)), so we can work out L = 1 / (4 * Pi * Pi * F * F * C). We can select R = sqrt(L/C). So R = sqrt( 1 / (2 * pi * F * C) ^ 2), or R = 1 / (2 * pi * F * C). Oops, I just realised that I turned a big circle to get back to a very familiar formula, excuse me.
So for C=1.2nF, if f = 10MHz, R = 13R, and if f = 100MHz, R = 1.3R. A 20R trimpot is suitable for the test.
C can be 4.7nF up to 10nF.
The RC can be installed at the secondary windings before the diode.
What do you think?
OOOOOops! I can be wrong - there are 3 secondary windings so which winding is causing which resonance?
Last edited:
I have previously simulated a bit for snubbers. I know that a wide range of R can still work reasonbly well.
So perhaps I could try 10nF + 5.6R and solder them underneath the PCB and see if I have any luck.
So perhaps I could try 10nF + 5.6R and solder them underneath the PCB and see if I have any luck.
Before I do the snubbers, I think I may try replacing the diodes first, and this will change the resonant frequency.
Most if not all of the diodes surrounding the transformer are Schottky diodes.
But not all Schottky diodes have soft recovery characteristics written in their datasheets. This is something I don't know the answer. Looking up Wiki or something I remember there is a paper stating all Schottky diodes have soft recovery characteristics.
I have just observed the datasheet of one of the Shottky diodes which states "fast recovery" and "low switching noise" while having trr = 500nS. 500nS is not fast! MBR40250 has trr < 35nS and claims to have soft recovery characteristics. MBR40250 is my favourite diode for audio, though it can be 10 times more expensive than the other Schottky diodes of lower power. MBR40250 is rated for 250V 40A and it still has the lowest forward voltage drop. It is over kill but if all the parameters are good, not not use it?
Most if not all of the diodes surrounding the transformer are Schottky diodes.
But not all Schottky diodes have soft recovery characteristics written in their datasheets. This is something I don't know the answer. Looking up Wiki or something I remember there is a paper stating all Schottky diodes have soft recovery characteristics.
I have just observed the datasheet of one of the Shottky diodes which states "fast recovery" and "low switching noise" while having trr = 500nS. 500nS is not fast! MBR40250 has trr < 35nS and claims to have soft recovery characteristics. MBR40250 is my favourite diode for audio, though it can be 10 times more expensive than the other Schottky diodes of lower power. MBR40250 is rated for 250V 40A and it still has the lowest forward voltage drop. It is over kill but if all the parameters are good, not not use it?
The link I posted makes it very easy. Here is the procedure, copied and pasted from there:
Determining Optimal Snubber Components:
1. Measure the frequency, f, of the resonance or ringing.
Bill's (HiFiNutNut's) trick of connecting a scope probe's ground clip and then holding the probe near the circuit but not touching it is a good one.
2. Add a shunt capacitor and adjust the value of this temporary capacitor until the frequency of the ringing is reduced by a factor of two. I've left out the math but the value of this resulting capacitor will be three times (3X) the value of the parasitic capacitance, C, that is creating the resonance. Remove the temporary shunt capacitor.
3. Because the parasitic capacitance is now known, the parasitic inductance can be determined using the formula:
L = 1 / [ (2πf)² · C ]
where f = (the original) resonant frequency from step 1, and C = the parasitic capacitance that was determined in step 2.
4. Now that both the parasitic capacitance and inductance are known, the
Characteristic Impedance, Z, of the resonant circuit can be determined using the following formula:
Z = √(L / C)
i.e. the square root of the quantity L divided by C,
where L = parasitic inductance and C = parasitic capacitance.
5. The resistor value used for the for the RC snubber network should be equal to Z, the value of the characteristic impedance, and the capacitor, if used, should be sized between four and ten times the parasitic capacitance. The use of larger (than 4X) capacitors slightly reduces the voltage overshoot at the expense of greater power dissipation in the resistor.
Last edited:
gootee,
I did read it a little while ago and I checked the link the other day. I just did it myself which matched 100% of yours. It is always good to go through it once more to make sure I understand it.
Regards,
Bill
I did read it a little while ago and I checked the link the other day. I just did it myself which matched 100% of yours. It is always good to go through it once more to make sure I understand it.
Regards,
Bill
MBR40250 won't fit.
I have found these babies: http://www.vishay.com/docs/88736/sbyv27.pdf
These are fast and soft recovery diodes made for SMPS. Trr=15nS so they can't be bettered. However, they are not Schottky diodes and they don't have the extremely low forward voltage drop.
If somebody thinks they are problematic please let me know, as I want to order them from Farnell tomorrow.
I have found these babies: http://www.vishay.com/docs/88736/sbyv27.pdf
These are fast and soft recovery diodes made for SMPS. Trr=15nS so they can't be bettered. However, they are not Schottky diodes and they don't have the extremely low forward voltage drop.
If somebody thinks they are problematic please let me know, as I want to order them from Farnell tomorrow.
Changing the smps diode type would be a different can of worms.
Faster edge times might make the smps better with less noise or it might break it or create worse noise.
It would be interesting to see what would happen. But "interesting" is probably not your primary goal.
I think it would be safer to first try the option that would have fewer unknowns, which would be snubbering the existing diode(s). That would be easy to remove if it was done underneath the board (if this is a two-layer pcb).
But I can't see the circuit so if the diodes are easy to change, and the change would be easy to reverse, and if the diodes you want to try are a type that is used in smps supplies a lot, then it probably couldn't hurt permanently, at least. And maybe it will push the noise to higher frequencies where it might be easier to get rid of it.
It's your call.
Faster edge times might make the smps better with less noise or it might break it or create worse noise.
It would be interesting to see what would happen. But "interesting" is probably not your primary goal.
I think it would be safer to first try the option that would have fewer unknowns, which would be snubbering the existing diode(s). That would be easy to remove if it was done underneath the board (if this is a two-layer pcb).
But I can't see the circuit so if the diodes are easy to change, and the change would be easy to reverse, and if the diodes you want to try are a type that is used in smps supplies a lot, then it probably couldn't hurt permanently, at least. And maybe it will push the noise to higher frequencies where it might be easier to get rid of it.
It's your call.
The diodes are quite easy to disolder and solder.
I think most fast recovery type diodes are suitable for SMPS. The soft recovery diodes should produce least noise.
The only issue here is that Schottky diodes always have the lowest forward voltage drop than the non-Schottky types. If there is only a small margin of voltage built in then I can get into trouble by using non Schottky types of diodes.
If this fails, I can always put Schottky diodes back in place. At this point in time, I don't expect it fails. The circuit design should have sufficient margin built-in.
I think most fast recovery type diodes are suitable for SMPS. The soft recovery diodes should produce least noise.
The only issue here is that Schottky diodes always have the lowest forward voltage drop than the non-Schottky types. If there is only a small margin of voltage built in then I can get into trouble by using non Schottky types of diodes.
If this fails, I can always put Schottky diodes back in place. At this point in time, I don't expect it fails. The circuit design should have sufficient margin built-in.
Here is a progress update.
Unfortunately, the soft recovery sbyv27 which has trr=15nS is rated only 200V. The two diodes on the primary side are rated 1000V so the sbyv27 is not useful.
My favourite diode MBR40250 (trr=35nS) are radial type so they don't fit.
The best one that can do the job I could find is UF5408 (trr=75nS), 3A 1000V that has soft recovery characteristics. They should still be way better than the onboard Schottky diodes which have trr=500nS. 500nS happens to be at the point separating fast recovery and slow recovery diodes. The UF5408 has a higher forward voltage drop than Schottky diodes, but still reasonably low.
So the two diodes on the primary side as well as two on the secondary side for the -12VA and +5V have been replaced with UF5408. The +12V diode was not changed as it is an isolated diode with a much higher reverse current rating, so I had better not to touch it.
I pulled out the +12VA and +12VM 2nd order LP filter and put the 1st order LP filter back, but added two ferrite beads in series. I kept the -12VA 2nd order filter.
My scope showed that after the diode upgrade the amplitude of the ringing is reduced to about half. When "zoomed out", the switching spikes now look thinner too.
Subjectively, the sound is the best so far. The treble harshness is further reduced, now to an almost accpetable level. I use Soprano voices for the test. Violin tones are more accurate than before.
Video quality is good. However, there is still a dark line 2 inches from the right edge, which faded away after some minutes. I really can't think of what I changed could have caused this. Perhaps the connection of the HDMI cable is not as good as before?
Next, I will add on some snubbers.
Unfortunately, the soft recovery sbyv27 which has trr=15nS is rated only 200V. The two diodes on the primary side are rated 1000V so the sbyv27 is not useful.
My favourite diode MBR40250 (trr=35nS) are radial type so they don't fit.
The best one that can do the job I could find is UF5408 (trr=75nS), 3A 1000V that has soft recovery characteristics. They should still be way better than the onboard Schottky diodes which have trr=500nS. 500nS happens to be at the point separating fast recovery and slow recovery diodes. The UF5408 has a higher forward voltage drop than Schottky diodes, but still reasonably low.
So the two diodes on the primary side as well as two on the secondary side for the -12VA and +5V have been replaced with UF5408. The +12V diode was not changed as it is an isolated diode with a much higher reverse current rating, so I had better not to touch it.
I pulled out the +12VA and +12VM 2nd order LP filter and put the 1st order LP filter back, but added two ferrite beads in series. I kept the -12VA 2nd order filter.
My scope showed that after the diode upgrade the amplitude of the ringing is reduced to about half. When "zoomed out", the switching spikes now look thinner too.
Subjectively, the sound is the best so far. The treble harshness is further reduced, now to an almost accpetable level. I use Soprano voices for the test. Violin tones are more accurate than before.
Video quality is good. However, there is still a dark line 2 inches from the right edge, which faded away after some minutes. I really can't think of what I changed could have caused this. Perhaps the connection of the HDMI cable is not as good as before?
Next, I will add on some snubbers.
I am trying to calculate the RC snubbers.
There are 3 obvious resonances.
(1) Resonance at somewhere between 500kH and 600kHz.
I suspect this comes from the mains input inductance of approximately 0.75uH and the X2 capacitor of 0.1uF. When L=0.75uH and C=0.1uF then F=581kHz and R=2.7R.
The problem is that I need a 0.47uF X2 cap which may be too bulky to fit beneath the PCB.
(2) Resonance at around 4MHz.
I suspect this could come from the -12V winding. The diode capacitance is around 80pF. There is 22R + 1nF (Marantz snubber) in parallel with the diode.
So how do we calculate the snubber for this one? If there is no 22R then it is easy. In that case we take C=1.08nF, L=1.47uH and we have R=36. But there is 22R in series with the 1nF (Marantz' snubber)! Should this 22R be ignored? or the 1nF be ignored?
(2) Resonance at around 8MHz.
I suspect this could come from the +12V winding because this resonance shows the strongest in that winding. The diode capacitance is around 80pF. There is 22R + 2.2nF (Marantz snubber) in parallel with the diode.
So how do we calculate the snubber for this one?
gootee, what is your take on these?
There are 3 obvious resonances.
(1) Resonance at somewhere between 500kH and 600kHz.
I suspect this comes from the mains input inductance of approximately 0.75uH and the X2 capacitor of 0.1uF. When L=0.75uH and C=0.1uF then F=581kHz and R=2.7R.
The problem is that I need a 0.47uF X2 cap which may be too bulky to fit beneath the PCB.
(2) Resonance at around 4MHz.
I suspect this could come from the -12V winding. The diode capacitance is around 80pF. There is 22R + 1nF (Marantz snubber) in parallel with the diode.
So how do we calculate the snubber for this one? If there is no 22R then it is easy. In that case we take C=1.08nF, L=1.47uH and we have R=36. But there is 22R in series with the 1nF (Marantz' snubber)! Should this 22R be ignored? or the 1nF be ignored?
(2) Resonance at around 8MHz.
I suspect this could come from the +12V winding because this resonance shows the strongest in that winding. The diode capacitance is around 80pF. There is 22R + 2.2nF (Marantz snubber) in parallel with the diode.
So how do we calculate the snubber for this one?
gootee, what is your take on these?
I have a major difficulty - there is now only a limited number of tests I can do.
It is not the first time I mod a CD player. It is my 4th time. The cable connectors and sockets are the biggest issue, especially those with ribbon cables. After plugging and unplugging a number of times, the contacts become loose and the player can die for this reason. It is not easy to order the cables or sockets fit for the player, even from the manufacturers' repair department. I have a couple of Marantz SA11 Reference Series SACD players that are currently offline pending for repair for this reason.
I have only unplugged the ribbon cable from this player once for the analogue board upgrade, and I can perhaps do it once more but I would not dare to do it more than a couple of times.
However, for the power supply board (I am glad that there is no ribbon cable here), I have plugged and unplugged for about 10 times now. I can feel that the connection has become more and more loose each time. I am worried it will die if I do this a few more times.
The snubbers must be soldered underneath the board and there is no other way. So I must get the snubbers right in one or two attempts, or I can risk killing the machine.
It is not the first time I mod a CD player. It is my 4th time. The cable connectors and sockets are the biggest issue, especially those with ribbon cables. After plugging and unplugging a number of times, the contacts become loose and the player can die for this reason. It is not easy to order the cables or sockets fit for the player, even from the manufacturers' repair department. I have a couple of Marantz SA11 Reference Series SACD players that are currently offline pending for repair for this reason.
I have only unplugged the ribbon cable from this player once for the analogue board upgrade, and I can perhaps do it once more but I would not dare to do it more than a couple of times.
However, for the power supply board (I am glad that there is no ribbon cable here), I have plugged and unplugged for about 10 times now. I can feel that the connection has become more and more loose each time. I am worried it will die if I do this a few more times.
The snubbers must be soldered underneath the board and there is no other way. So I must get the snubbers right in one or two attempts, or I can risk killing the machine.
Here are some measurements of the RF transformer. I expect that the numbers are indicative only and may not be exact. The measurements were made using a relatively cheap LCR metre with inductance display resolution of 0.01uH. However, shorting the two probes showed 2.2uF, which is deducted from the measurement results. Leakage inductance is obtained by shorting the primary windings but not the other secondary windings.
Primary: 0.84R, 840uH
Controller power: 0.13R, 22uH, leakage 1.5uH
+12V: 0.07R, 15uH, leakage 0.8uH
-12V: 0.23R, 17uH, leakage 1.1uH
+5V : 0.01R, 3uH, leakage 0.2uH
As an indication, the schematic indicates the primary inductance to be 0.82mH, and it was measured 0.84mH.
Primary: 0.84R, 840uH
Controller power: 0.13R, 22uH, leakage 1.5uH
+12V: 0.07R, 15uH, leakage 0.8uH
-12V: 0.23R, 17uH, leakage 1.1uH
+5V : 0.01R, 3uH, leakage 0.2uH
As an indication, the schematic indicates the primary inductance to be 0.82mH, and it was measured 0.84mH.
You need to snub at the right place. It seems like the only way to tell where that is will be to try the procedure I gave. i.e. Try temporarily connecting a capacitance across the secondary, for example, and see if that changes a resonant frequency. You cannot usually go by specs, or measurements, of a single known L or C that are involved! You don't know what they ALL are, and especially what the resultant L and C are, from all sources. The only practical way I know of is to "measure" by following the procedure I gave.
Sometimes in switching circuits snubbing can make noise worse. For instance a transformer may have a high source impedance when off but a low source impedance when being driven. In these two situations, the snubbing values will be totally different. In something like a transmission line, it would be the difference between series resonance and parallel resonance. I've had a lot of thought about how best to snub rectifiers for this reason. There are many different resonances involved and you have to either find a solution that works in both cases or you have to isolate the stage that's switching somehow.
I redid my LC metre measurements carefully and believe that the new parameters I got for the transformers are fairly accurate.
I also calculated the resonant frequencies based on those transformer parameters and Marantz RC snubbers, and in 2 out of 3 cases my calculations matched the measured results with the scope.
The resonances formed by the transformer and didoes would have been between 50MHz and 100MHz. The Marantz snubbers, while did not damp the resonances well, have brought the resonances down to 3.5MHz to 8MHz. Whether I could damp the resonances successfully or not, the new C in the RC snubber should bring the resonances down to hundreds of kHz at the highest. There is a good chance that low ESL capacitors become effective in these frequencies, so the harm is significantly reduced.
So I feel I can give it a try.
With regards to changing source impedance, would the LCR filter's characteristic impedance be much higher than the transformer / source impedance to make it insignificant?
I also calculated the resonant frequencies based on those transformer parameters and Marantz RC snubbers, and in 2 out of 3 cases my calculations matched the measured results with the scope.
The resonances formed by the transformer and didoes would have been between 50MHz and 100MHz. The Marantz snubbers, while did not damp the resonances well, have brought the resonances down to 3.5MHz to 8MHz. Whether I could damp the resonances successfully or not, the new C in the RC snubber should bring the resonances down to hundreds of kHz at the highest. There is a good chance that low ESL capacitors become effective in these frequencies, so the harm is significantly reduced.
So I feel I can give it a try.
With regards to changing source impedance, would the LCR filter's characteristic impedance be much higher than the transformer / source impedance to make it insignificant?
I do hope the snubbers will work. If so, they are the ones that ultimately address the most significant issue of this player - RF noise.
When I first started this thread, I was thinking about boosting up the output filtering to suppress the RF noise. I am happy with the progress that I have now realised that I need to tackle the problem at its source. The RF noise is created during diode switching, so we really need snubbers.
Last night, I realised that not only do I need snubbers at the secondaries but also at the primary as well. The primary switch is the father of all other secondary switches. So I will need to deal with not only low voltages, but also high voltage (+360VDC) as well.
I have to be very careful here with 360VDC voltage and if you have the knowledge please reply to confirm with me that X2 rated cap would work fine here. If I am looking at the datasheet of some of the X2 caps I have bought: http://www.kemet.com/kemet/web/homepage/kechome.nsf/vapubfiles/F3301_R46S.pdf/$file/F3301_R46S.pdf
It shows that the rated voltage is 275Vac (50/60Hz) / 560 Vdc. I presume that this can be used as the snubber C. As for the Resistor power rating, I will check closely with P=C*f*V^2.
Two days ago, I put a 470R 5W ceramic resistor at the output of the +12.4V regulator to do some testing. After only 1 minute, the resistor was so hot that it was not touchable. That was only dissipating 0.33W power in a fairly large surface.
When I first started this thread, I was thinking about boosting up the output filtering to suppress the RF noise. I am happy with the progress that I have now realised that I need to tackle the problem at its source. The RF noise is created during diode switching, so we really need snubbers.
Last night, I realised that not only do I need snubbers at the secondaries but also at the primary as well. The primary switch is the father of all other secondary switches. So I will need to deal with not only low voltages, but also high voltage (+360VDC) as well.
I have to be very careful here with 360VDC voltage and if you have the knowledge please reply to confirm with me that X2 rated cap would work fine here. If I am looking at the datasheet of some of the X2 caps I have bought: http://www.kemet.com/kemet/web/homepage/kechome.nsf/vapubfiles/F3301_R46S.pdf/$file/F3301_R46S.pdf
It shows that the rated voltage is 275Vac (50/60Hz) / 560 Vdc. I presume that this can be used as the snubber C. As for the Resistor power rating, I will check closely with P=C*f*V^2.
Two days ago, I put a 470R 5W ceramic resistor at the output of the +12.4V regulator to do some testing. After only 1 minute, the resistor was so hot that it was not touchable. That was only dissipating 0.33W power in a fairly large surface.
You also need to look at the resistor datasheets, and only use resistors that are rated for the voltages and frequencies involved, for the HV part at least. Many common resistors are not suitable for more than 300 Volts.
That is the hard part. With a lot of common resistors there are no datasheets provided.
I would not order a single 1W carbon composite resistor from Digikey for $30 shipment to ship it to Sydney.
I am thinking about buying some no brand 2W metal film. The good thing about snubber is that if the resistor gets burnt open or short there is still a capacitor there that blocks all DC.
I would not order a single 1W carbon composite resistor from Digikey for $30 shipment to ship it to Sydney.
I am thinking about buying some no brand 2W metal film. The good thing about snubber is that if the resistor gets burnt open or short there is still a capacitor there that blocks all DC.
- Status
- Not open for further replies.