Snubber network and the values

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello everyone.

I read the article about snubber network which's been reffered here so often a couple years ago.
And now I'm wondering about the value with simple simmulations.

I coppied the schematic from the artivle and set these values.
Vi = 20v [60Hz]
Rt = 500mohm
Lt = 1mH
Ct = 560pF
R_load = 10ohm
C_load = 1nF

Diodes are MUR860, and I got the model and used it to simmulate the circuit.
And I checked the spike current between the C_load.

case1
no snubber
case 2
Rs = 0ohm
Cs = 1nF
case 3
Rs = 0ohm
Cs = 10nF
case 4
Rs = 0ohm
Cs = 100nF
case 5
Rs = 0ohm
Cs = 1000nF
case 6
Rs = 1ohm
Cs = 100nF
case 7
Rs = 10ohm
Cs = 100nF
case 8
Rs = 100ohm
Cs = 100nF

The results were below, and case 4 & 5 were the best snubber network at this simmulation.
case4 = case5 > case3 > case6 = case7 > case8 > case2 > case1

in case4 and case5, I couldn't obseve any spike current.
 

Attachments

  • num1.png
    num1.png
    77.9 KB · Views: 814
Here is the schematic.

And now, I'm not sure if my approach was correct or not.
Besides, I used the spike current to compare the snubbers.
If it was wrong approach, please give me some hints to simmulation the snubber network.

Thank you.
Okina
 

Attachments

  • num4.png
    num4.png
    80.5 KB · Views: 835
Then, I tested at Lt = 10mH
The circuit's response was fully changed and spike currents were pretty larger than at Lt = 1mH.

The results were below.
case5[3uA] > case4[3uA] > case8[4uA] > case7[6uA] > case6[8uA] > case3[14uA] > case2[241uA] > case1[333uA]

This time, I read the values of AC RMS as the amount of spike current.
The case1 & 2 were so bad.
Cs = 100nF was enough.

Then, I tested at Lt = 100mH
[Though I think it was meaningless]

And the results were almost the same as at Lt = 10mH
case5[1.3uA] = case4[1.3uA] > case6[1.5uA] > case8[1.5uA] > case7[1.7uA] > case3[4uA] > case2[59uA] > case1[147uA]

Cs = 100nF was enough, again.
 
Coud anyone give some suggestions ?
I actually have no idea if it was wrong, so I'm wondering if anyone give me hints.

And I tested some other sommulations, and I got some results.
If the Lt is lower than 0.1mH, you don't have to add any snubber, but I got slightly better results with only a cap.
And then, I added ESR to 100nF cap to simmulate the circuit more real at Lt 1mH - 10mH.
The results were that Rs around 100m was bad for snubbing.
and that resistors around 1-100ohom didn't always work well.
Sometimes pretty bad, and the other times worked well.
1mohm ESR and no ESR were ideal snubber.
Then, I added ESR to 1000nF cap.
The results were that 1000nF + ESR[around 1m - 100mohm] worked pretty better than each snubbers with 100nF cap.
And that you don't have to add any resistor, because they made spike current larger.

After all, if my approach wasn't wrong, I can say that 1000nF cap without a resistor is reasonable and that the lower the cap has ESR, the better the snubber worked.

Thank you.
Okina
 
Konichiwa,
Remember that the purpose of fitting snubbers across the rectifier diodes if to suppress the noise generated when they turn off. This noise only occurs in a short burst lasting only a few hundred microseconds, and the millisecond scale used in your simulation is probably hiding it.

Try re-simulating with a much lower timebase. Figure 8 in the Hagtech article shows where the burst occurs in the cycle:
http://www.hagtech.com/pdf/snubber.pdf#search="snubber"

I’d also increase the load capacitance in your simulation to something more representative, say, at least 1000uF.

Nice one,
David.
 
okina

Come on! Swith on your 'scope!
Why do you go on with those useless sims?
The real world is different ..parasitics are there, not in your spice model!

I experimented with a 1n4007 x4 bridge and a SBVY27 x4 one.

With the test transformer the first bridge is best smoothed with 22nF+1R // 100nF on secondary side. The last with 22nF across each diode.

The transformer and the reverse switch characteristics of the diode dictate the conditions ..is your model so reliable?
 

Attachments

  • rec_1n4007.jpg
    rec_1n4007.jpg
    69.3 KB · Views: 683
daatkins said:
Konichiwa,
Remember that the purpose of fitting snubbers across the rectifier diodes if to suppress the noise generated when they turn off. This noise only occurs in a short burst lasting only a few hundred microseconds, and the millisecond scale used in your simulation is probably hiding it.

Try re-simulating with a much lower timebase. Figure 8 in the Hagtech article shows where the burst occurs in the cycle:
http://www.hagtech.com/pdf/snubber.pdf#search="snubber"

I’d also increase the load capacitance in your simulation to something more representative, say, at least 1000uF.

Nice one,
David.


Hello, daatkins.
About the scale of the burst, please see the second post on this thread.
I measured the burst as 300kHz oscillation.
Though it seemed to keep oscillating.

And now, I changed the value of C_load to 1000uF as you mentioned.
But I have no idea where I should put the probe to observe the burst voltage.
I'm just seeing the point just before the bledge with no reason.
And I notived that the oscillation Freqs are depend on Cs value the same as in the article and its theory we're looking on.

So, should I calculate the reasonable snubber to remove this oscilattion ?

mrjam said:
okina

Come on! Swith on your 'scope!
Why do you go on with those useless sims?
The real world is different ..parasitics are there, not in your spice model!

I experimented with a 1n4007 x4 bridge and a SBVY27 x4 one.

With the test transformer the first bridge is best smoothed with 22nF+1R // 100nF on secondary side. The last with 22nF across each diode.

The transformer and the reverse switch characteristics of the diode dictate the conditions ..is your model so reliable?


Thank you for nice example.
But just now I'm a bit comfused and actually I can't say if my model is reliable one.
Could anyone give me some more suggestions to get me out from this mud? ;)

Thank you for reading.
Okina
 
daatkins said:
This noise only occurs in a short burst lasting only a few hundred microseconds, and the millisecond scale used in your simulation is probably hiding it.

The switching transient is very high energy and very short duration -- so it will propogate like a radio wave and cause "rectification effects" in analog circuitry. This becomes problematic if you are in the instrumentation business and are trying to measure microvolts and nano-volts -- or if you are building a moving coil amplifier (or other transducer). It can be problematic in the high-gain circuitry of a power amplifier as well if you're sloppy about how you dress the leads.

If you want to see it in the simulation eliminate the load on the power supply. Also -- put a few nano-henries of inductance and some hundreds of milli-ohms of resistance (ESL and ESR) in the simulated power supply caps --

The leakage inductance you are using is quite high -- at least if you are using a decent toroid transformer -- mine measure in the microHenries, not milliHenries. Don't assume, measure.

Usually the calculated values of C are in the low nano-Farads, not hundreds of nano-Farads.

The "Optimal Snubbers" article will get you started, but in the real world there are more moving pieces -- for one, rectifiers junction capacitance is a function of voltage.

There is a real tradeoff between snubbing all the rectifiers, or just across the transformer. You have to measure the effects in the system to see which works best.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.