• The move to the new server is done. There are some software and database maintenance updates in process. This has us passing the hat around to help out. We appreciate any donations. Seriously, even a dollar helps. The payment page may be found here - https://www.audiokarma.org/support.html

470k resistors in parallel with B+ diodes -- why?

Recently, I've seen two schematics of modded Dynaco amps that had SS rectification and used 470k resistors in parallel with the silicon diodes in the B+ supply. One is the Paoli Model 60M, a MkIII mod.

Now, I've seen a few designs for snubbers across Si diodes involving capacitors and resistors, but never a 470k resistor by itself. What is the purpose supposed to be? Why do you want a diode to have reverse leakage?
 
Register to hide this ad
If these amps have diodes in series it has merit because it will distrubute the reverse voltage equally between those diodes. Otherwise it is just silly. If the designer wanted to bleed reservoir capacitors place them there.
 
This issue of R or RC "sharing networks" with series diodes — both R and C are actually required for proper functioning, the R for balancing the C for diode-destroying spike suppression, but many older designs only used R — comes up enough that it is worth my time to assemble the postings I've previously made on the subject. Here they are, somewhat edited together for content. It should explain more than you care to know about series-string diodes and why most of the time it devolves into the famed meme of "UR Diode String, UR Doin' It WRONG!".

Series diodes are used because a single diode is unavailable with a higher-voltage rating. The string works because each individual diode in the string only sees (V / n) where V is the overall voltage rating for the entire string and n is the number of diodes. (Similar tricks are used to increase the voltage ratings of resistors and capacitors.) The old books routinely suggested series diodes as a solution for higher voltage, but often failed to explain the issues, or include the sharing networks, because the author generally did not know or was repeating lore and dogma. I will explain those issues.

The attached excerpts from the Radio Amateur's Handbook by the American Radio Relay League (ARRL) (1981) elaborate on the generally accepted electrical engineering practice for series diodes:
The Radio Amateur’s Handbook
by The Headquarters Staff of the American Radio Relay League (ARRL)
(1981)

Page 5-12:


Diodes in Series

Where the PRV rating of a single diode is not sufficient for the application, similar diodes may be used in series. (Two 500-PRV diodes in series will withstand 1000 PRV, and so on.) When this is done, a resistor and a capacitor should be placed across each diode in the string to equalize the PRV drops and to guard against transient voltage spikes, as shown in Fig. 22A. Even though the diodes are of the same type and have the same PRV rating, they may have widely different back resistances when they are cut off. The reverse voltage divides according to Ohm’s Law, and the diode with the higher back resistance will have the higher voltage developed across it. The diode may break down.

If, we put a swamping resistor across each diode, R as shown in Fig. 22A, the resultant resistance across each diode will be almost the same, and the back voltage will divide almost equally. A good rule of thumb for resistor size is this: Multiply the PRV rating of the diode by 500 ohms. For example, a 500-PRV diode should be shunted by 500 X 500, or 250,000 ohms.

The shift from forward conduction to high back resistance does not take place instantly in a silicon diode. Some diodes take longer than others to develop high back resistance. To protect the “fast” diodes in a series string until all the diodes are properly cut off, a 0.01-uF capacitor should be placed across each diode. Fig. 22A shows the complete series-diode circuit. The capacitors should be non-inductive, ceramic disk, for example, and should be well matched. Use 10-percent-tolerance capacitors if possible.

Diodes in Parallel


Diodes can be placed in parallel to increase current-handling capability. Equalizing resistors should be added as shown in Fig. 22B. Without the resistors, one diode may take most of the current. The resistors should be selected to have about a 1-volt drop at the expected peak current.​

Fig_22_Diode_Series_and_Parallel_ARRL_Handbook_1981.png

Figure 22


Fig 20 - Diode Series in Typical Power Supplyl - ARRL Handbook (1981).png
Figure 20

​
Pages 5-12 to 5-12:

Transient Problems

A common cause of trouble is transient voltages on the ac power line. These are short spikes, mostly, that can temporarily increase the voltage seen by the rectifier to values much higher than the normal transformer voltage. They come from distant lightning strokes, electric motors turning on and off, and so on. Transients cause unexpected, and often unexplained, loss of silicon rectifiers.

It’s always wise to suppress line transients, and it can be easily done. Fig. 21 A shows one way. Cl looks like 280,000 ohms at 60 Hz, but to a sharp transient (which has only high-frequency components), it is an effective bypass. C2 provides additional protection on the secondary side of the transformer. It should be 0.01 uF for transformer voltages of 100 or less, and 0.001 wF for high-voltage transformers.

Fig. 21B shows another transient-suppression method using selenium suppressor diodes. The diodes do not conduct unless the peak voltage becomes abnormally high. Then they clip the transient peaks. General Electric sells protective diodes under the trade name, “Thyrector.” Sarkes-Tarzian uses the descriptive name, “Klipvolt.”

Transient voltages can go as high as twice the normal line voltage before the suppressor diodes clip the peaks. Capacitors cannot give perfect suppression either. Thus, it is a good idea to use power-supply rectifiers rated at about twice the expected PRV.
​
Fig 21 - Transient Suppression - ARRL Handbook (1981).png

Figure 21
​

So, in summary, placing solid-state diodes in series without a sharing network creates serious problems which can result in device failure:
(1) Diodes in series add Vfdrop together. This doesn't matter at all for most applications. Sometimes it is even highly desirable.

(2) Diodes in series require load-balancing resistors, as above explained, or the diodes may fail.

(3) Diodes in series must have a peak rating high enough to prevent failure of an individual diode from a voltage spike above the nominal voltage rating. Once a diode fails as a short, the entire string will be overloaded and fail.

(4) Diodes in series cause the diode with the lowest Qrr to turn off prior to the others, which means that diode is hammered with the brunt of PIV blocking every cycle, which can damage or destroy that individual diode. (Remember, first diode turning off takes the full voltage at either end of the string because the other diodes are conducting, like wires.) Load-balancing resistors are thus required to avoid exceeding the steady-state voltage rating of any individual diode. That takes care of steady state. Now consider transient response, which must be similarly dealt with to avoid having one diode conduct the entirety of the voltage spike, for much the same reason. Removing the spike requires load-balancing capacitors. Suddenly series diodes becomes a lot more complex than simply using a single diode with a higher voltage rating, even if that diode is significantly more expensive.

(5) Adding more diodes increases Qrr as the sum for all diodes. This amplifies the noise broadcast throughout the amplifier (Qrr discharge spikes act like a HF carrier wave that is modulated by mains harmonics) and increases ringing by stimulating the power transformer to ring at whatever resonant frequency it has (parasitic inductance and parasitic capacitance form a tank circuit damped by its Q). Each diode needs an RC snubber to dump Qrr. This snubber was, of course, needed for a single diode but now all of the new diodes need them.​

You can see (4) and (5) on a scope if you look for it. The issue of Qrr has been elsewhere been explained, and the solution of snubbers set forth in detail.

Strings of diodes thus make existing problems worse, and adds new problems which did not previously exist. The only time diode strings are appropriate is when a diode at the requisite voltage rating is not available. I always suggest purchasing (low-noise) diodes with the proper voltage rating and not adding extra diodes "just in case".

Attached is some engineering data (emphasis added):
Application Note AN443
Series Operation of Fast Rectifiers
STMicroelectronics
2004

INTRODUCTION
The use of several rectifiers connected in series is necessary to obtain voltage ratings beyond the capabilities of single diodes and also when some special requirement, such as very low switching losses, requires the implementation of several low voltage ultra fast diodes. Rectifiers connected in series tend to unequally share the voltage across the string in blocking conditions because of the variations in reverse characteristics: leakage currents and turn-off switching parameters. To ensure that each diode operates within its voltage rating, it is generally necessary to add a voltage sharing network.


...

CONCLUSION

When using several fast rectifiers in series it is necessary to make sure that no diode will be subjected to continuous or transient voltages in excess of their ratings.

In most cases, this is achieved by using sharing networks across each diode. It is important to optimize this circuit in order to reduce power consumption and to save space
.
​
 

Attachments

Resistors across the series diodes force them to share the reverse voltage. Without the resistors, the diode with the lowest reverse leakage will drop most of the reverse voltage, exceed its PIV rating and then proceed to take its last hot & steamy one on you. If the diodes were matched in reverse leakage, then the resistors would be superfluous. Of course there's no guarantee that the leakages would stay matched over time. Use the resistors.
 
Resistors across the series diodes force them to share the reverse voltage. Without the resistors, the diode with the lowest reverse leakage will drop most of the reverse voltage, exceed its PIV rating and then proceed to take its last hot & steamy one on you. If the diodes were matched in reverse leakage, then the resistors would be superfluous. Of course there's no guarantee that the leakages would stay matched over time. Use the resistors.

One issue: a single diode doesn't "drop most of the reverse voltage", as no voltage is ever "dropped" during the polarity inversion of the next half-cycle. (Vf is not the issue here.) The reverse voltage (from the perspective of the diode), instead, is blocked by the rectifier exactly like a switch.

The first diode to earlier (relative to the others, not to the AC cycle per se) switch off consequently must act as the sole switch in the line, providing the entirety of the blocking. As I above explained, the other diodes remain on because the Qrr has not yet depleted, and thus remain conducting exactly like wires. Only one diode is early to the shutoff and it will experience the entirety of the PIV. In a larger string, such as the high-voltage rectifier stack used in a color television, if several diodes with matched Qrr nearly identically shut off, the PIV still exceeds the diode's individual PIV rating, which is why a string was used in the first place instead of a single device. The peculiar characteristics of selenium make it particularly suitable for such stacks.

Plus the issue of transient response is not addressed via a purely resistive sharing network, which is why an RC network is required to snub the fast-moving spikes that can also destroy rectifiers.

The Qrr problem is worse with ordinary silicon diodes, particularly so with older diodes of that vintage. I have elsewhere explained how to obtain low-Qrr diodes which also reduce noise, and even identified candidate devices inexpensively available from Mouser.

I really did take a lot of time to initially create, and then now gather and collate, those explanations to demystify the care and feeding of diode strings.
 
CVElectronic, that Mod is quite old and used 1N4006 diodes. There was likely no need for balancing resistors back when that mod was proposed, but people just did it. Nowadays there is no need for balancing resistors and capacitors in mains frequency amp applications when using series diodes like 1N4007 or UF4007, with the caveat that the diodes are from the same batch and are installed so they operate at the same temperature. This type of application operates the diode in a very 'slow' manner such that dynamic capacitor balancing is not needed, and diodes from the same batch and at the same temp have sufficiently low leakage current and a consistent leakage current characteristic that no one diode ever experiences stressful conditions.
 
Wow. Where do you get those spherical diodes of uniform density? The devices were long-ago discontinued and the Japanese bought up all of the stock. Even the sockets are impossible to find.

Since we live in the real world with non-sperical diodes of non-uniform density, sharing networks are mandatory for best practices. That's just black-letter electrical engineering (or physics, take your pick).

The diodes never have identical characteristics, even among a batch (CPUs from the same batch are commonly graded to quite profitable effect), so the shut-off point is always slightly different. Long enough for the voltage rating to be temporarily exceeded. The lack of failure from such misuse is not accidental; the junction in a rectifier is generally engineered to take a great deal of abuse relative to the ratings, and likely has a greater safe-operating-area (SOA) to avoid failures (and customer complaints!) arising from transients or poor engineering practices. Not all semiconductor devices are so forgiving when abused. The issue of transient response, i.e. a fast-moving temporary spike exceeding the PIV rating even with sharing resistors, similarly must be addressed.

The "Series Operation of Fast Rectifiers", Application Note AN443, STMicroelectronics (2004), which I above posted as a PDF, clearly and unequivocally states that sharing networks are required and explains why. So does the other material I included. All of that material is somehow incorrect? Why? What? How? The OP saw sharing resistors for very good reason. It was not just something just done, as per your claim "There was likely no need for balancing resistors back when that mod was proposed, but people just did it." Why? Massive ignorance? Nope. Sound engineering practice.

Cutting corners and series-stringing devices is a poor practice which any electrical engineer would call out in a design review as a first-semester design mistake, and then deliver a long lecture about rectifier behavior and best practices. A power engineer would never make such a mistake. Rules for engineering practice dictate following the limits presented in datasheets. While failure analysis might calculate the tail risks on the distribution curve as relatively rare events, willfully violating specifications never has a good outcome. That's why buildings and bridges do not generally fall down, tunnels and pipes do not generally collapse, and the electric grid and communications grid do not generally fail. When simple devices, like threaded bolts, have load or shear-force limits violated the failures can kill people. Electronics just tend to go up in smoke and be replaced.

Selenium rectifiers fail in electronics not because selenium rectifiers are inherently short-lived, but for one simple reason: the specifications — voltage, current, and temperature — were routinely violated as a cost-savings measure, shifting the reduced manufacturing costs onto the backs of consumers because the device would last just long enough to evade the warranty replacement. Selenium rectifiers were widely used in welders and DC locomotive applications, for decades on end operating with high reliability. Why? Because the specifications were not violated. Gresham's Law applies to electronics.

It does a grave disservice to hobbyists to not explain designs and, furthermore, to promulgate poor engineering practices that tend to work in practice out of blind luck, but which actually violates widely accepted design goals and widely accepted best practices. When a learning opportunity presents, best to use it and advance practice. That is what I did. I explained all of the facts and circumstances. Knowledge and education.

A superior solution, of course, is upgrading to a single modern diode at the specified rating with low Qrr, as such devices may be trivially and inexpensively obtained as I have elsewhere explained and even suggested specific products.

But the issue at hand is explaining what the sharing network does and why is present, which I certainly did, using primary source material which may be trivially verified as correct if anyone so chooses.

Edit: fixed word omission, fixed a typo, clarified a sentence.
 
Last edited:
I've found empirically that adding a high value resistor in parallel with a rectifier diode, can soften the switching transients, and thus reduce "buzz". They probably did this mod because they also added a bunch of filter capacitance, with the undesired side effect of magnifying the switching transient. IMO this is kind of a band aid, and its presence is an indicator to me that the power supply isn't designed well.
 
AN443 is appropriate for applications where static or dynamic conditions require it. Mains frequency valve amps is not one of those applications, given the caveats I described. A discussion on this topic is in the link in the second section:
https://www.dalmura.com.au/static/Power supply issues for tube amps.pdf

Measured leakage current behaviour as rated PIV is exceeded can be appreciated by reading through the linked forum thread:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/

For DIY applications, including restorations of vintage equipment, my view is that the risk is high of people adding in equalising parts that are not adequately voltage rated, so even for 'peace of mind' I would not recommend this type of mod.

It is not uncommon for vintage equipment with ss diodes to have diodes that show some leakage (for the various amps I have restored), but that type of testing requires at least an insulation resistance tester with 500V to 1kVdc capability. So like replacing vintage e-caps, imho it is well worth replacing vintage ss diodes (unless they can be adequately tested).
 
Last edited:
AN443 is appropriate for applications where static or dynamic conditions require it. Mains frequency valve amps is not one of those applications, given the caveats I described. A discussion on this topic is in the link in the second section:
https://www.dalmura.com.au/static/Power supply issues for tube amps.pdf

Measured leakage current behaviour as rated PIV is exceeded can be appreciated by reading through the linked forum thread:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/

For DIY applications, including restorations of vintage equipment, my view is that the risk is high of people adding in equalising parts that are not adequately voltage rated, so even for 'peace of mind' I would not recommend this type of mod.

It is not uncommon for vintage equipment with ss diodes to have diodes that show some leakage (for the various amps I have restored), but that type of testing requires at least an insulation resistance tester with 500V to 1kVdc capability. So like replacing vintage e-caps, imho it is well worth replacing vintage ss diodes (unless they can be adequately tested).

Wow.

I will analyze that bit of nonsense to demonstrate that you, again, endlessly call me out because of your well-demonstrated lack of understanding.

You must stop doing this because you're not adding to the discussion, you're taking away from it by promulgating pernicious nonsense which is outright dangerous and will result in a failure and possible fire. That is unconscionable behavior for any engineer. The standard of responsibility is very high and I know the kids no longer grasp this concept, but it once was the case that engineering was a profession with the highest ethical standards.

Your position appears to be summarized as:
(1) The learned tomes I cited, none of which were written by me, are inferior to you merely quoting your own writings as a "definitive source".

(2) Because a diode from one manufacturer was once tested and found to have a breakdown voltage greater than the published specifications, all diodes from all manufacturers may similarly be abused without fear of failure.

(3) Load-sharing resistors should not be used because hobbyists might select resistors at the wrong rating.​

Point (1) requires no commentary other than to note that many years ago as an undergraduate I learned this type of argument as a Fallacy Loquitur ex Rectum. (I knew all those philosophy classes would someday be useful!)

Find a respected and definitive source, namely one you did not write, which contradicts the published learned tomes I cited
.

As far as Point (2) it does not refute any of my arguments. It actually proves my point, as the OP for the cited thread later explains the diode was damaged by the test, by way of not limiting the current through it to the microamps used when grading to avoid heating the junction and damaging or destroying it. So the diode was NOT perfectly fine with the excess voltage as you claim, it was permanently damaged.

I specifically addressed the issue of higher rating in Number 7:
The diodes never have identical characteristics, even among a batch (CPUs from the same batch are commonly graded to quite profitable effect), so the shut-off point is always slightly different. Long enough for the voltage rating to be temporarily exceeded. The lack of failure from such misuse is not accidental; the junction in a rectifier is generally engineered to take a great deal of abuse relative to the ratings, and likely has a greater safe-operating-area (SOA) to avoid failures (and customer complaints!) arising from transients or poor engineering practices. Not all semiconductor devices are so forgiving when abused. ​

But minor overload wasn't the case here. The device was permanently damaged by the breakdown test.

To continue with Point (2), even worse yet for you, it appears that you DID NOT EVEN READ YOUR OWN LINK and are, furthermore, cherry-picking data. The scenario related is idle testing and the diode was not put into service. Nor are the conditions real-world. It is simply a laboratory curiosity.

In Number 8 on that page, the poster wrote:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/msg2763198/#msg2763198
Putting the diodes in parallel would just divide the fixed 3 mA current. But the diodes would still degrade. If you look at the voltage current graph, even after 1200V, there is a substantial increase in current. So even at that voltage, the diode probably won't be the same after the voltage is removed.

The diodes in series, you would add the breakdown voltage of each diode. The voltage drop across each diode won't be the same because they will have different insulation resistances. The one to reach break down first will just be in standby because current is limited by the other diode. Then that other diode will reach break down and then both will get damaged.
​

Number 9:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/msg2763204/#msg2763204
And another thing to keep in mind: This type of failure would only occur in the real world where the voltage source is rather low energy. Maybe getting continuously hammered by ESD, or maybe the diode is clamping a small inductor (small relay coil?) and gets continuously hammered with back EMF ...

But many times in the real world, the voltage source is high energy, so the diode will probably fail short circuit, and if the current continues, it will melt away and become open.
​

Number 10:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/msg2763286/#msg2763286
Failed components like these are always real pleasure to try to find. Oldschool CRT monitors often had components that would measure ok with DMM but leak too much in real circuit.
​

Number 17 and 18:
https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/msg2765060/#msg2765060
So can we conclude that limiting the current prevents the junction from sustaining permanent damage under avalanche?

https://www.eevblog.com/forum/projects/testing-1n4007-with-a-megger/msg2765066/#msg2765066

Nope, we can't ...
I picked 3 new 1N4007 from the bin and measured their insulation resistance at 1 kV. They are 65 GOhm, 87 GOhm, and 102 GOhm.
Then I measured the diode that just underwent the ramp test. It has an insulation resistance of 260 MOhms ... So it has permanently degraded.
​

Notice a decline in resistance from an average value of 87 Giga Ω has declined to 260 Mega Ω.

Let's convert the units to be the same, noting that Giga (billion) is 1,000 Mega (million).

So that would be a decline from 87,000 Mega Ω to 260 Mega Ω, a reduction of a factor of 335.

The device clearly is damaged and failing after the excessive PIV test.

At this point I've utterly lost interest in further debunking for Point (2).

As far as point (3), then you should simply have stated, "The load-sharing resistor must be of sufficient voltage rating to function at the rated voltage." That is the caveat, not to deprecate my analysis because a hobbyist might select the wrong resistor! The very same issue exists with any other resistor in the existing circuit which is fed by that voltage!

Quod Erat Demonstrandum

Your approach of deliberately exceeding published design specifications because the device tends to not blow up is never a best practice. It violates every accepted doctrine and is gaming the specifications. Your designs are dangerous and do not meet UL, CE, PSE, PSC or any other accepted safety rating.

In conclusion, you have failed the assignment. I suggest studying harder for the final exam, because if this poor level of scholarship continues you will similarly fail the class.

I have only explained the error in detail at this level to keep someone from being led astray by your nonsensical claims that sharing networks are not needed. My citations and explanations clearly explain the issues.
 
Call it what you want, but a DC voltage source across 2 series diodes in reverse (nonconducting) is going to divide between the diodes according to their reverse leakage currents , with the least leaky diode getting stuck with more of the voltage across it than the less leaky one has. This is a measurable quantity. This current is given as a specified parameter in the data sheets for the diodes. An ideal diode would have no leakage current at all. Real ones, alas, do. Use the resistors to force even sharing of the reverse voltage.
 
Don't forget contaminants on the board and other environmental hazards. Those will screw up everything. Place resistors. Diodes remain OK.
 
Call it what you want, but a DC voltage source across 2 series diodes in reverse (nonconducting) is going to divide between the diodes according to their reverse leakage currents , with the least leaky diode getting stuck with more of the voltage across it than the less leaky one has. This is a measurable quantity. This current is given as a specified parameter in the data sheets for the diodes. An ideal diode would have no leakage current at all. Real ones, alas, do. Use the resistors to force even sharing of the reverse voltage.

That is correct. Because we can no longer purchase spherical diodes of uniform density and instantaneous shutoff. The diode cartel pulled them all off the market! I demand government investigate this cabal of evil for restricting our diode choices!

The key issue to bear in mind is that the string of diodes cannot instantly and fully shut off because the devices are ever so slightly different, even from the same batch. Some diode(s) will be earlier while some will be later. Which, as a limit condition, means that one diode will be slightly earlier than all of the others which are not shut off, acting like wires, thereby elevating the voltage across the one that is shut off. Very little current is required to puncture a junction at higher voltage.

Static electricity is an example of very high voltage at ultra-low current, and it readily punches holes in junctions, destroying them. I remember using a wrist strap for exactly this reason when handling pricey MOS/CMOS chips and even pricier power MOSFET transistors. Very expensive mistake to make in those days! I once blew up a very expensive varactor diode from being charged with static electricity in the winter. Forgot to ground myself. Humans are very bad at remembering which is why we have checklists.

Metal rectifiers, particularly selenium or mercury vapor, are much better at temporarily absorbing such transient conditions than silicon because the resistive characteristics vary with applied voltage, and that tends to provide a greater SOA. Up to a limit, of course, which is why selenium rectifiers often burn up. It's all over once an insulation breakdown occurs near the connection studs.

But not merely parallel sharing resistors are required! The √2 peak is bad enough, but noise — typically designed in at ≈ 5 to 10% overage — creates fast-moving transient peaks, be they from noise or coupling in of discharge (typically flyback) from other sources (inductors in power and output transformers, power-line noise from same and motors), must also be eliminated. Removing the transient peaks requires parallel capacitors to absorb the fast-moving transient and then gradually release it. Simple and inexpensive to add parallel ceramics to the sharing resistors, as I above explained.

The worst part of this is that a single high-voltage diode with low Qrr is readily and inexpensively available, so these sharing networks are only needed for supplies delivering above a few kiloVolt operating points. We truly live in amazing times, and can now throw away all those older kluges because devices were, quite frankly, awful compared to what we may today purchase.

Can you imaging going back in a time machine and explaining modern diodes to engineers? The look on a 1960's designer's face upon hearing that for under 10% of the cost of the diode string, sharing network, stocking, and assembly time that a single device, with vanishingly small Qrr thereby removing a noise source even without a snubber (best practice is to use one), can be delivered either next day or two days? Then tell them about how much we pay, relatively speaking, for polypropylene film capacitors, ultra-low noise metal film resistors, and fancy PTFE tube sockets. Plus glass-fiber phenolic boards (epoxy is a phenol, just a much better one) instead of the awful cardboard formaldehyde-phenolic boards. Of course, their cost for transformers and chokes are a fraction of ours, because copper and steel were far less expensive because of lower demand for metals and greater supplies of easily mined deposits.
 
Perhaps Figure 1 in AN443 is not being appreciated by some. No diode in a string can experience the onset of avalanche until all diodes enter their particular breakdown knee region.

The dispersion of diode leakage for modern common diodes is effectively negligible in the considered application. Perhaps if we get enough diyers to test their own batches then we could accumulate a 5 sigma acceptance - I have tested 2 batches of 10 diodes, and the characteristic is consistent (as would be expected) through each batch.

The insightful post in the linked eevblog thread is near the end by rodpp as it shows a plotted example of the typical ramp up of leakage current before the avalanche knee. That characteristic is what I have measured for 2 batches, of which one batch of 10 was 1N4004 which all exceeded 1.4kV prior to onset of avalanche, and had leakage increasing from 1kV (resolution was 0.005uA) to no more than 0.02uA at 1.4kV. A batch of UF4007 shows a similar dispersion and ramp characteristic, but in comparison had a higher leakage current, and breakdown knee onset around 1.2 to 1.3kV.

It is quite likely that diyers don't keep resistor parts in their original packaging with identification of part/datasheet. There would be no other resistor part in an amplifier that needs a voltage rating anywhere near that needed as an equalising part across for example a 1N4007.

Transient voltages to the level that would cause concern are not readily achievable on a power transformer secondary for this amp application. I've tried to force transient voltages on a few types of circuitry and transformer windings - the latest effort was related to output transformer over-voltage protection using MOVs - perhaps some readers can set up an example amplifier application and show what type of disturbance can push a practical diode string in to avalanche and relate the disturbance to a common cause.

Maybe more than 40 yrs of power electronic engineering experience allows me some appreciation of how to interpret datasheets and implement parts in particular applications.
 
Nobody, least of all me, is interested in my further fruitless attempts to educate you about the difference between bogeys and real devices. To the best of my knowledge and ability I always follow the documented and well-established best practices. You do what you want, I'm glad you're not designing any product I have in my home.

Neither, I fear, is anyone interested in a discussion about diode back-resistance, a parameter, BTW, is generally not directly specified in data sheets and must be inferred from the leakage current. Those, being bogey values, may not be mirrored by actual devices. Neither is anyone interested in how back resistance varies with temperature, and how otherwise identical devices may be at different temperatures because of vagaries in the enclosure.

Trying to avoid the trivial expense of a low-cost resistor and low-cost capacitor because the devices tend to work, particularly in overload situations, is a terrible design practice. My explanation of best practices is fully accurate and includes actual citations from learned tomes written by others, without the need to resort to hermeneutics to decipher the hidden meanings in a linked discussion. My analysis of the posted information was sufficient.

As I above noted:
Your approach of deliberately exceeding published design specifications because the device tends to not blow up is never a best practice. It violates every accepted doctrine and is gaming the specifications. Your designs are dangerous and do not meet UL, CE, PSE, PSC or any other accepted safety rating.​

This remains true. Your design practice simply does not meet the requirements for regulatory approval which requires, at a minimum, the swamping resistor to balance. So, by all means, keep designing products to be made in China and slapped with a decal falsely claiming the requisite safety certifications.

Finally, the MOV you reference is a well-known fire hazard, and with alarming regularity short out and heat to incendiary levels, often causing fires. Terrible technology which is no longer necessary as replacements exist. I have elsewhere explained the virtues of the TVS diode or, indeed, ordinary flyback diodes in the standard configuration. Exotic solutions are hardly required to protect output transformers. You should know better than to use the MOV for any new design, and to not remove it from any equipment you have. But that is outside the scope of this discussion.
 
Nobody, least of all me, is interested in my further fruitless attempts to educate you about the difference between bogeys and real devices. To the best of my knowledge and ability I always follow the documented and well-established best practices..........<snip>


It would be instructive if you were to run thru a practical example utilizing the info in the STM app note AN443 you linked for say 3xUF1007 in series subjected to 2.5KV PIV. Resulting resistor and cap values and voltage ratings would provide a useful point of reference.
 
It would be instructive if you were to run thru a practical example utilizing the info in the STM app note AN443 you linked for say 3xUF1007 in series subjected to 2.5KV PIV. Resulting resistor and cap values and voltage ratings would provide a useful point of reference.

Wow.

Ummm, did you not read my explanation? Oh, riiiiight, this is a continuation of the hassling you dished out to me in "Mystery Device found in Heathkit W4-AM", isn't it? Yeah, that.

The above tutorial, which I notably laboriously constructed, clearly provides the answers you above castigate me for "omitting". Only basic reading comprehension was required to find the details. I constructed the tutorial to be self-sufficient for the purpose of explaining sharing networks. Even if I had omitted it, which I did not, a google search would have identified the standard guidelines for sharing networks.

Okaaay. So that nobody thinks I shirked on the writing effort, when I obviously did not...

In Number 3 posting, I clearly quoted the AARL guidelines which sets forth, in part:
A good rule of thumb for resistor size is this: Multiply the PRV rating of the diode by 500 ohms. For example, a 500-PRV diode should be shunted by 500 X 500, or 250,000 ohms.

...

To protect the “fast” diodes in a series string until all the diodes are properly cut off, a 0.01-uF capacitor should be placed across each diode. Fig. 22A shows the complete series-diode circuit. The capacitors should be non-inductive, ceramic disk, for example, and should be well matched. Use 10-percent-tolerance capacitors if possible.
​

How could that explanation be any clearer? Oh, right, you didn't want to actually read (or want to read) what I wrote. Was it as Butthead says, "words, words, words" or was this just an opportunity to further hassle me?

As far as the voltage rating, well, I would have thought that would be obvious, as it would be for any other selection of a power-supply component, because the voltage rating depends upon the peak voltage seen by the rectifier! N'est ce pas? Which, as per all the peaks for RMS would generally be √2 × V ≈ 1.4 × V. That is simply how component voltage ratings are sized in a power supply, plus a little slop factor for transients. I then specifically stated in Number 12, that the peak transients were "typically designed in at ≈ 5 to 10% overage". That, too, is a rule of thumb for power supply design, and I would have thought it unnecessary and therefore beyond the scope of my answer.

Electrical Engineering frequently requires foundational knowledge, as well as a willingness to read texts, study, and cogitation about the meaning of the text. I cannot fix that, nor may the foundational knowledge always be reduced to a few simple sentences or a short YouTube video. If that were the case, an EE education would be a mere few weeks of reading forum postings and watching YouTube videos. Alas, not to be. Reading comprehension and the building of knowledge brick-by-brick is sometimes required.

The fact that you did not take the time to read my writings, which notably included extensive material from learned tomes, is not, and cannot be, my fault or error.
 
Don't forget the voltage rating slop factor due to tolerances of equalising parts.

Any pointers to a practical resistor and capacitor from say Mouser or RS etc?
 
Selection of specific components from specific supply houses was clearly left as an exercise for the reader, as it is for any project.

Are you saying you are unable to select R and C values for components to construct an equalization network in which you do not believe, and, thus, will never actually purchase?

That strikes me as asking for advice to purchase insurance against attacks by bandersnatch when you don't even believe in bandersnatch.

You clearly are up to something. That's obvious. Not playing.

I have clearly answered the OP's question and explained why best practices and safety certification requires the equalization networks.

Quod Erat Demonstrandum
 
Back
Top Bottom