You actually don't have that data because it doesn't register on the graph. There's no way the leakage current is constant with temperature. That's fundamental semiconductor physics.
I looked at Digikey's selection of 2 W resistors. It looks like they top out at about 350-500 V, so you may need to connect two resistors in series across each diode to meet the voltage spec. Welcome to high voltage design.
You mention 1200 V. Why? I thought you had a string of four diodes across 2500 V. That's "only" 625 V per diode. Or am I missing something? With 625 V, you'll need two resistors in series. At 1200 V, you'll need three or four depending on the resistor of choice.
~Tom
Last edited:
These 2M, 1500V, 1W MOX HV resistors will probably suffice if needed.
RNX0382M00FKEE Vishay Dale | RNX2.00MBCT-ND | DigiKey
RNX0382M00FKEE Vishay Dale | RNX2.00MBCT-ND | DigiKey
It's actually a string of 5. My diodes are 1200V, my caps are 1250V, why would I want to put a resistor across them rated less? Seems like a step backward to me, when I could use a single MOX resistor rated at 1500V. Fits nicely in the space, too.
By the way, this quote is classic. I've been researching the topic for weeks, building the circuit, experimenting with it, measuring the results, reporting the findings, and seeking explanations for them. Hardly handwaiving and plug-n-play, is it?
Give me some credit already.
Yes, I'm aware that the graph runs into the abscissa and that Ir is temperature dependant. My point here is that even at much higher voltages than my diodes will ever see, where the data is visible, the difference in Ir between 25C and 75C is only 2 or 3X, and the lines are converging and the slope decreasing as the voltage decreases. Is that much of a difference really something to worry about when I have 700V of headroom per diode?
I'm trying to understand the fundamentals here, not just follow a cookbook.
Following up on this a little further, if you look again at the datasheet plot (post #120), at a temp of 175C (clearly a worse case than I will ever see, but useful because it's easy to see on the graph), the Ir increases from ~10uA to ~25uA from 500V to 1000V. Assuming the highest voltage drop diode in my worst stack stayed unchanged and the lowest changed by the full 2.5X, my calculated 27V difference in the worst stack at 600V (calculated assuming the diodes all changed linearly at the same rate) would become a 67.5V difference. I have 700V headroom per diode, or 10X that amount.
Where is the flaw in this logic? In this case it looks to me like adding 20 resistors per bridge would be an unnecessary expense, complication, and potential additional source of failure. Perhaps that's why the more recent editions of the ARRL handbook for ham radio amateurs recommend against adding resistors (or caps for that matter) to diode strings?
Again, this is not an attempt to be obstinate or argumentative, it's an attempt to understand why I should add complication and potential problems for what seems like little to no benefit.
It seems to me the next experiment is to run the bridge up through several voltage levels, measure the voltage differences between diodes, plot the data and see what it tells us about the change in voltage drop difference per unit applied voltage. Or as one of my grad school profs used to always insist "plot the data, plot the data, plot the data!".
Where is the flaw in this logic? In this case it looks to me like adding 20 resistors per bridge would be an unnecessary expense, complication, and potential additional source of failure. Perhaps that's why the more recent editions of the ARRL handbook for ham radio amateurs recommend against adding resistors (or caps for that matter) to diode strings?
Again, this is not an attempt to be obstinate or argumentative, it's an attempt to understand why I should add complication and potential problems for what seems like little to no benefit.
It seems to me the next experiment is to run the bridge up through several voltage levels, measure the voltage differences between diodes, plot the data and see what it tells us about the change in voltage drop difference per unit applied voltage. Or as one of my grad school profs used to always insist "plot the data, plot the data, plot the data!".
Last edited:
Magz:
Found a great application note from SGS-Thompson on using fast diodes in series and explaining resistors and diodes. Don't know if you have seen this:
http://www.datasheetcatalog.org/datasheet/SGSThomsonMicroelectronics/mXyzyuu.pdf
Found a great application note from SGS-Thompson on using fast diodes in series and explaining resistors and diodes. Don't know if you have seen this:
http://www.datasheetcatalog.org/datasheet/SGSThomsonMicroelectronics/mXyzyuu.pdf
Yes, I ran into that in my digging, and thanks for sharing it anyway!
I don't argue with the basic principles here, diodes can vary a lot from batch to batch, and there's no guarantee a customer will use diodes that are all from the same batch so they need to make recommendations based on the worst case scenario - for example, the datasheet for the diodes I'm using show Vf as 1.4V typical, 1.8V max. I'm pretty sure a 1.4Vf diode and a 1.8Vf diode might have some huge differences in reverse current and voltage drop.
The 40 diodes I have all were within .007V for Vf - do they still need to be treated with the worst case scenario? That's what I'm trying to find out. If you surf the Ham Radio forums the consensus now seems to be that if the diodes are from the same batch, then resistors aren't needed; the caveat is that they are not talking specifically about SiC Schottky diodes, rather they focus on standard Si diodes.
OK, enough theorizing. In the next few days I'll hook up my 2500V PT to a variac and run the HT output to the board. Then I'll run the voltage up as high as I dare and check the drops across my "worst" bridge leg (luckily I noted which one that is). That ought to give us enough data to answer this question. Stay tuned; first I want to clean up the board a bit to ensure the voltages stay where they should (remove adjacent solder pads, round sharp points, maybe some high dielectric conformal coating).
PS: Yes I have a 50kV high voltage probe if I get too high up there ;-)
PPS: I'm having fun!
PS: Yes I have a 50kV high voltage probe if I get too high up there ;-)
PPS: I'm having fun!
You're missing the point. If you read the SGS app note, you'll find the following in the first paragraph of the conclusion:
(Emphasis mine). Please read through that app note. It states specifically under which circumstances that you can eliminate the sharing networks. Do the math for your scenario rather than relying solely on experimentation and blog opinions.
You are trying to balance the REVERSE voltage, VR. **NOT** the FORWARD voltage, VF. It doesn't matter that the FORWARD voltage of the diodes in your batch are within a few mV of each other. It's the REVERSE voltage that causes the problem. If you exceed the REVERSE voltage across any of the diodes at any point in time, even briefly, it will break down catastrophically. When it does, it will take the remaining diodes with it. With the amounts of energy you are dealing with, this will cause the diodes to EXPLODE. Seriously.
No offense to any HAM out there. I used to be one myself, as a matter of fact. But how many HAMs, honestly, understand the semiconductor physics behind the reverse breakdown of a diode? Do they understand it better than the manufacturer does? I doubt it. Please read that app note.
If you're curious about the semiconductor physics, I suggest reading Ben Streetman, "Solid State Electronic Devices".
I cannot believe you are willing to risk a $10k amp build to save a few bucks on resistors. Never mind risking your life...
~Tom
Last edited:
Too much drama. Experimental evidence can't hurt, besides I'm enjoying myself, after all this is a hobby. Who knows, maybe it will help someone else down the road who has this same question come to a faster decision.
PS: Don't anyone else try this, EVER. It's too dangerous. There, that's my legal disclaimer.
PS: Don't anyone else try this, EVER. It's too dangerous. There, that's my legal disclaimer.
If you read in current arrl amateur radio hand book they say this is not necessary with modern diodes.
I'm not saying don't put snubbers on them. Just food for though is all.
Nick
I'm not saying don't put snubbers on them. Just food for though is all.
Nick
Maybe you can get away without, but balancing resistors can help reliability - and suitable parts can be procured to make it easy.
For example, I use the Welwyn MH37 series, which are rated for 3500V, and are almost as cheap as ordinary resistors:
MH37-470KJI - WELWYN - RESISTOR, H/V 0.5W 470K | Farnell United Kingdom
One more subtle reason for using them: Before the diodes break down destructively, you can sometimes get transient breakdown/leakage effects, that may not destroy the diode, but cause a lot of noise.
Especially if you choose to avoid balancing resistors, I would recommend a fuse in the PT secondary winding circuit - rated to be able to manage the HT voltage. A microwave cooker fuse is made for 5kV, and may be useful, depending on the VA rating of the PT:
Microwave Oven High Voltage Fuse 800mA 0.8A 5kV | eBay
A fuse in the secondary gives closer control, since the power-ON surge needs no accounting for.
For example, I use the Welwyn MH37 series, which are rated for 3500V, and are almost as cheap as ordinary resistors:
MH37-470KJI - WELWYN - RESISTOR, H/V 0.5W 470K | Farnell United Kingdom
One more subtle reason for using them: Before the diodes break down destructively, you can sometimes get transient breakdown/leakage effects, that may not destroy the diode, but cause a lot of noise.
Especially if you choose to avoid balancing resistors, I would recommend a fuse in the PT secondary winding circuit - rated to be able to manage the HT voltage. A microwave cooker fuse is made for 5kV, and may be useful, depending on the VA rating of the PT:
Microwave Oven High Voltage Fuse 800mA 0.8A 5kV | eBay
A fuse in the secondary gives closer control, since the power-ON surge needs no accounting for.