Input current spikes in capacitive rectifiers

[AB2EZ]

Hi!

While using my new, home-built linear, I noticed that the HV transformer was buzzing at a frequency of about 2 kHz. I could measure this easily by using my graphic equalizer to remove the buzz from the output of my audio chain when picking it up with my microphone. Getting rid of the buzz was easy to do by placing the transformer in a box with some soundproofing material... but this got me thinking about why such high harmonics of 120 Hz would be present in the current passing through the transformer windings (thus exciting an apparent 2 kHz resonance in the transformer's vibrational characteristics).

A few quick, rough calculations explained the cause. Since the transformer is feeding a bridge rectifier with a 112uF capacitive load... I could quickly calculate the following. During each half cycle, the voltage across the capacitor drops linearly from its maximum level (approximately 3000 volts) to about 37 volts below that level. I.e., 500 ma of load current x 1/120 second x 1/112uF = 37.2 volts. During a brief portion of each half cycle, the transformer output voltage rises above the voltage on the capacitor, and the capacitor charges back up to its peak voltage.

A little math shows that for an ideal transfromer (which is clearly not the case in reality), the current starts to flow out of the transformer (and into the capacitor) about 8.5 degrees (out of 180 degrees in each half cycle) before the transformer secondary voltage reaches its peak value, and the peak secondary (not primary) current at the instant the current begins is 21 amps! The current decreases linearly from 21 amps to 0 amps over the 8.5 degrees (.0004 seconds) that it takes for the secondary voltage to reach its peak value.

Obviously, the real transformer doesn't deliver 21 amps of secondary current, i.e., there is some resistance and inductance that has to be factored in. Nevertheless, the peak current drawn from the secondary is very high, and flows for a small portion of each half cycle. In this case, probably at least 5 amps at the peak, and flowing for 20% of each half cycle.

Question, does it make sense to add a relatively small series choke, perhaps 1H between the transformer secondary and the bridge rectifier, to limit the current peaks... without going so far as to create a traditional choke-input rectifier?

Stu

[WA1GFZ]

Stu, I would think as long as you stay below critical inductance you will limit current by making the pulse wider. At some point your peak charge voltage will be effected.
How about a cap input filter and break up the large C on 2 sides of an inductor.

[w3jn]

Stay away from putting chokes in solid state supplies... unless the diodes are considerably over-rated the L di/dt when the field collapses can kill 'em.

Just put a series resistor in if you're worried about this.

[AB2EZ]

John

I'm not understanding why the inductor (assume it is between the diodes and the capacitor, and about 1/10th of the the "critical inductance") would create an L di/dt problem. The current in the inductor is always flowing in the direction that is allowed by the diodes of the bridge rectifier. If, for example, the primary of the transformer is suddenly opened, the inductor will not experience a sudden decrease in current. The inductor will continue to pump current into the capacitor (returning to the inductor via the forward biased diodes) to charge it up to a slightly higher voltage. The current in the inductor will gradually decrease to zero, and the capacitor will discharge into the bleeder resistor and any other load that is on the output of the HV supply.

Best regards
Stu

[K1DEU]

Dear Stu; I have reduced annoying noise like this before by placing a series resistor and cap (tuned snubber) across each rectifier diode. And please consider the MOV+Transorb  across the primary. http://amfone.net/Amforum/index.php?topic=11418.0
73 John, K1DEU

http://www.spiritualpeaceproject.com/  and  http://hamelectronics.com/k1deu/

[WA1GFZ]

usually a 1 amp diode can handle 40 or so amps for a short time without failurs. di/dt will be limited be winding resistance.

[k4kyv]

Stay away from putting chokes in solid state supplies... unless the diodes are considerably over-rated the L di/dt when the field collapses can kill 'em.

Just put a series resistor in if you're worried about this.

A resistor of 50 ohms or less should be enough.  Use just enough resistance to limit the peak current without lowering the output voltage under load more than 1-2%.  This may eliminate the need for step-starting the power supply as well.

The input capacitor shorts the voltage induced by the spike to ground.  So a capacitor input filter with a smoothing choke is OK.  But never use a choke input filter with solid state diodes as John said.

By the same reasoning, MV rectifiers are usually not satisfactory for a capacitor input filter.  The current spike will exceed the peak current rating and shorten the life of the tubes, if they don't flash over immediately.  This is not a problem with high-vacuum rectifier tubes, since they have enough internal resistance to limit the peak current.

That's why nearly all the old transmitters with 866A or 872A rectifiers used choke input, and why modern slopbucket leen-yars with solid state diode rectifiers use a large filter capacitor and no choke.

MV tubes are more forgiving of voltage spikes, but don't like current peaks, while solid state diodes will withstand spikes of up to several times the rated current, but are absolutely intolerant of voltage spikes that exceed their PIV by even a fraction of a volt.

[WA1GFZ]

I solid stated my viking 2 PDM rig in 1983 and the diodes are hanging in. My dual CG310 supply also choke input has diodes.
could I be lucky or crazy?

[AB2EZ]

I think the folks who have responded have pointed me in the right direction on this. I measured the DC resistance of the secondary (rated at 2400 VAC @ 1.5 A CCS) of the new transformer I am using... and it is only 8.2 ohms. On the other hand, the secondary of my Peter Dahl 1500 VAC @ 0.5A CCS transformer, which I was previously using (temporarily) with this same power supply, has a DC resistance of 94 ohms. The combination of the much lower DC resistance and the higher voltage yields a much "peakier" current waveform at the input to the capacitor bank (compared to what I had with the Peter Dahl transformer). I'm going to see what happens when I put a 50 ohm resistor in series with (the grounded side of) the capacitor bank. Assuming that I am shooting for a peak current of around 5 amps, and a current pulse that lasts 20% (or less) of each half cycle, the effect on the DC output voltage will be to reduce it by no more than around 250 volts (under load), and probably significantly less than that. We'll see what happens. 

Stu

[WA1GFZ]

Yup some resistance will be good unless you are using 6 amp diodes.
8 ohms is some real current. Not a square wave drive so the di/dt will have a slow rise time. I would think turn on will be the big issue with that monster cap value. I hope you have a step start.

[AB2EZ]

Update:

I decided to perform an experiment by placing a 150 watt light bulb in series with one of the leads on the secondary of the transformer. [Note: this approach is not approved for indoor lighting under the National Electrical Code]

It worked like a charm.

Without the light bulb in series: transformer "hum" centered at ~ 2kHz was present. DC output under load (~400 mA of average plate current + 37.5 mA of bleeder current) was 2700 volts

With the light bulb in series: no "hum" noticed, and the DC output under load was 2600 volts.

Thus, the presence of an additional 96 ohms (150 watt bulb) of series resistance in the secondary of the transformer reduced the DC output of the supply, under the load intended for this application, by about 100 volts (3.7%)... which I can easily make up for with the large Variac I am using to feed the transformer's primary.

Naturally, without load (only bleeder current) the DC output voltage of the power supply was about the same with or without the light bulb in series with the transformer secondary (~3000 volts).

The light bulb was not lit when there was no load on the power supply (as expected, since there is not very much current going through it when there is no load on the power supply). The light bulb was close to normal brightness when the power supply was under load.

Stu

[WA1GFZ]

so now you have a modulation monitor ....

[Bacon, WA3WDR]

I was getting strange ambient noise with a cheap power supply kit that I bought - and finally got to work after many mods - and it turned out to be diode switching transients working with the leakage inductance of the power transformer.  The solution is capacitors or preferably snubber networks across the diodes in the bridge rectifier.

The sudden onset and cutoff of current during the AC waveform causes spikes and pings on the leakage inductance, and this was making RFI.

I thought it had to be coming from active circuitry (the voltage/current regulator), but no.

[WA1GFZ]

Bacon this is a common problem with diodes snapping on and off.
Most diodes today are quite fast Trr so make a lot of noise.

[AB2EZ]

I used SWCAD III to simulate the power supply. The waveforms are here:


http://mysite.verizon.net/sdp2/id20.html

Note that the filter capacitance is 112.5 uF (900uF/8); the transformer secondary resistance is 45 ohms (9 ohms intrinsic to the transformer + 36 ohms added in series with the secondary using power resistors). The load is drawing 400 mA of average current.
 
The red trace is the voltage across the filter capacitor. It varies between approximately 3220 volts and 3240 volts... corresponding to a 20 volt triangular-shaped ripple at 120Hz frequency. I.e., the 120Hz ripple is about 0.6% of the average voltage... which corresponds to 44dB of ripple attenuation.
 
The light blue trace is the 0.4A load current.
 
The medium blue and light green traces correspond to the current through the diodes of the bridge rectifier on alternating half cycles. Note that the peak current (for this value of secondary resistance) is roughly 3 amps, and the current waveform is roughly trianglular in shape (the base of this triangle is roughly 27% of the width of a half cycle),

[WA1GFZ]

I often wonder how the meter on the side of the house handles short pulses.??