| rtarbell |
I've seen several posts about the best audio quality diodes to use, but I get a little bit of contradiction between didoe types:
Schottky diodes seem to be more audio quality than standard, but do we want to use fast recovery or soft recovery?
==> Similar note, has anyone tried Fairchild's stealth diodes? |
|
|
| BWRX |
| Try both and see if you can tell a difference. You won't need fast recovery for dealing with 60Hz line frequencies. Diodes used in a switching supply are a different story... |
|
|
| Tweeker |
| Its not the speed of fast diodes that matters at 60hz, its thier better recovery giving less HF noise when they are shutting off. |
|
|
| pwillard |
I wish I could convince myself that fast recovery is an issue at 60 hertz...
This is one of those things I guess I'm just going to have to try myself and compare. |
|
|
| vectorplane |
This diode story is one of those placebo effects in audio.
You can use chicken fence wire to connect speakers to an amp. Then tell a guy it's gold plated solid silver $300 mega-monster interconnects, and he'll swear he can hear profound presence, soubtle overtones, rich full-bodied sound that envelops you like the flavor of fine cognanc...blah, blah, blah, ... you know the drill...
Then tell him it's chicken wire, and he can suddenly hear the rough, jagged, scratchy granularity caused by the myriad diodes formed by the cheap iron molecules in the cable...
There's never been any evidence, measured objectively with instruments, to suggest that any power supply rectifier diodes used in place of the 1N400x will provide any significant improvement in sound parameters.
The drive to substitute more expensive alternatives for all components in a power amp stems from a desire to improve the design without an actual knowledge of how to improve the design.
It demonstrates complete unawareness of a fundamental principle of electronic design:
The topology must factor out component characteristics,
so that overall quality of result depends on circuit arrangement, not on individual component properties and tolerances.
The sound of an amplifier should not depend on the quality of components, but on the arrangement of those components. With few exceptions, this design goal is to a large extent achieved in most designs today.
So you're trying to improve something which has already been rendered irrelevant by the design.
This in turn causes an individual to convince themselves they can "hear" an improvement, in order to achieve self-justification. Communicating with others who have done the same modification provides reassurance through cross-confirmation. So now you have a club of believers. Something like this:
http://www.alaska.net/~clund/e_djub...arthsociety.htm |
|
|
| lndm |
A spike (as can be seen across the diodes when they turn off), is energy with a uniform frequency distribution. That is it may set off resonances throughout the power supply if they exist.
Fairly easily fixed though, put some snubbers across the diodes, and place small resistances between stages in the ps.
I believe Nelson Pass has a few good things to say on his web site on this subject too. I seem to recall him recommending which diodes to use. |
|
|
| richie00boy |
Schottky diodes have the smoothest recovery of them all, but they are not as robust nor available in as high PIV rating as other types.
My own opinion is that with massive cap banks etc the conduction angle of the diodes becomes smaller so any recovery time will become more important. I use Schottky's, I like to over engineer. |
|
|
| Nordic |
| I recently started using UF5408 diodes, they are good for 1000V 3A, and are wicked fast for the price, they are also quite large, wich I like. |
|
|
| richie00boy |
| They may be fast, but their recovery - which is the most important parameter on 50/60Hz mains - is poor. I would not use these diodes. |
|
|
| Upupa Epops |
| Sound of diodes....Jesuschrist !.... :eek: |
|
|
| richie00boy |
| quote: | Originally posted by Upupa Epops
Sound of diodes....Jesuschrist !.... :eek: |
Some people claim to notice a difference. That aside, there are also EMC/RFI issues to take into account. |
|
|
| Upupa Epops |
| It is confusing of reason and result - low PSRR, amp on border of stability, wrong design of PCB, sensitivity for HF disturbing at input etc... Isn't problem to get with normal diodes SNR over 120-125 dB, with any special low noise devices ( at line levels... ), if all is correct designed .... ;) |
|
|
| jotape |
| We should always think in the audio systems as a WHOLE. |
|
|
| lndm |
Diode commutation noise is not like common thermal/resistance noise which is usually relatively quiet and resembles white noise.
Diode noise is short spikes that on their own are largely inaudible, but they are very high in level. It takes this wideband high voltage spike to ring the power supply. |
|
|
| powerbecker |
Hello,
there is no current recovery problem, but in some cases hf-generation can occur if one uses diodes with higher capacities together with a transformer with some leakage inductance.
In this case there is one (not four!) small snubber the problem solver!
The "funny" thing is that usually schottkey´s have bigger capacities!
You can start read about from here:
http://www.diyaudio.com/forums/show...t=&pagenumber=5
Regards
Heinz! |
|
|
| pwillard |
I'm sure I read somewhere that people were using Switching Mode Power Supplies and Microcontrollers in their Hi-Fi designs... all of which, in my mind, do WAYYY more to induce HF noise than a Power Diode like a 1N4004 could ever hope to do.
I realize that a lot of designs, especially Gainclones, like to use those really cool looking fast diodes in TO-220 cases... but WELL FILTERED DC power (both HF and RIPPLE filtered) would seem to be able to completely overcome anything a "standard" diode could do while changing state.
I'd be more worried about what a local FM radio tower could induce into my amp than what the diode might do...
So... I guess I remain unconvinced it makes a difference... still haven't lab tested though. |
|
|
| lndm |
| quote: | Originally posted by pwillard
Switching Mode Power Supplies and Microcontrollers in their Hi-Fi designs |
Not me...
| quote: | | but WELL FILTERED DC power (both HF and RIPPLE filtered) would seem to be able to completely overcome anything a "standard" diode could do while changing state. |
Yes, but the trouble is getting there. Filtering requires resonant circuits, and whilst we try to have them behave themselves and to act at infrasonic frequencies, there are always parasitics.
Electrolytics in particular usually have an inductive component due to their plates being rolled, that when combined with their capacitance, resonates in the upper midrange. |
|
|
| peranders |
| quote: | Originally posted by lndm
Diode commutation noise is not like common thermal/resistance noise which is usually relatively quiet and resembles white noise.
Diode noise is short spikes that on their own are largely inaudible, but they are very high in level. It takes this wideband high voltage spike to ring the power supply. |
When you mention "very high in level" you must put this into perspective, compared to what? Bear in mind that 99.99% of all high-end gear have regular diodes. |
|
|
| lndm |
| quote: | Originally posted by peranders
When you mention "very high in level" you must put this into perspective |
No I don't. I understand the problem conceptually as I'm sure you do too. I'm explaining it conceptually. I build according to guidelines, common sense, and my training. If it's broken I fix it and I enjoy much success. Therefore I speak from experience.
Not interested in specifics here.
http://www.passlabs.com/downloads/a...powersupply.pdf |
|
|
| AndrewT |
Hi Upupa,
tell us how you switch off the Forums's auto delete to let through "JC"?
Although I'm joking it would be nice to know. |
|
|
| Upupa Epops |
| No Andrew, it is mistake - JC is for me great ideal... his knowledges are great and he is man of worth... But two times I was asking him for listenig results and two times I was ignored.... I will never try it again.... In all hifi are now many superstitions and I can to know " uncolored " thruth, 'cos I like it very much.... ;) |
|
|
| pwillard |
| quote: | | Bear in mind that 99.99% of all high-end gear have regular diodes. |
What I use as the basis of my own conclusions. I too agree that the humble power diode designed years ago is still in use.
---
Over-engineering with parts that were designed for other purposes does not always make something better. I don't think faster means better. The only thing that I can imagine that could come into play here is what has already been mentioned.
A properly designed "system" would be less susceptible to power diode issues. If your AMP or PREAMP is affected by the diodes... maybe it's not so good? |
|
|
| Upupa Epops |
| To pw : Full agree... ;) |
|
|
| jackinnj |
when are we going to kill this over-flogged horse -- you can see the effect of a diode's junction capacitance when you use an EI core transformer (with large leakage inductance) -- I said you can see itwith a scope but I don't know if you can hear it in an application which draws a few amps. If you are an analog circuit designer trying to squeeze out acuracy in the LSB's then it's germaine.
the diode capacitance forms a "tank" circuit with the transformer and will oscillate -- the cure is inexpensive -- an RC network in parallel with the diode is the simplest cure.
folks like OnSemi, Fairchild, ST Micro, Cornell Dublier, International Rectifier pay a lot of attention to diode noise issues.
For me, I use MUR860's, MBR20100's, 1N4007's. |
|
|
| Elso Kwak |
| Put a low valued resistor in the ground return to the transformer's center tap, something like 3 Ohm. Then put your scope's probe across the resisitor and you will get an idea of the charging current spikes.......:idea: |
|
|
| pwillard |
| No doubt it's a regularly debated issue, but a circuit like this using standard power diodes makes for a good supply. |
|
|
| jackinnj |
| quote: | Originally posted by pwillard
No doubt it's a regularly debated issue, but a circuit like this using standard power diodes makes for a good supply. |
putting a 10nF cap across a diode just lowers the resonant frequency of the "tank" circuit -- |
|
|
| Eva |
| quote: | Originally posted by Upupa Epops
No Andrew, it is mistake - JC is for me great ideal... his knowledges are great and he is man of worth... But two times I was asking him for listenig results and two times I was ignored.... I will never try it again.... In all hifi are now many superstitions and I can to know " uncolored " thruth, 'cos I like it very much.... ;) |
Everyone involved in serious design and research employing scientific methods is going to avoid discussing listening tests. That's because we all have a long experience in listening tests and we know very well that these tests never ever give the same result twice. On the other hand, salesmen will always use these tests as a tool to persuade customers because they know quite well that the result is highly self-biased and random.
Trying to find serious repeatable criteria to "improve" audio gear by listening to it is like trying to repair a broken amplifier employing a "speacial" multimeter that has a 50% error in its measurements and the display never *ever* reads the same twice.
You will never find two people doing "listening research" reaching the same results, but you will see them always arguing, as opposed to people employing scientific methods that will usually end up agreeing. Then again, you can repeat any scientific test that I have previously done and you will obtain exactly the same results, but you won't be able to repeat a listening test carried out by anyone else.
In fact, that kind of people expend their whole life switching equipment and trying new exotic components, as they become tired of everything quite quickly without finding any solid criteria. They have even invented the "component burn-in" concept in a desperate attempt to explain the randomness and the lack of repeatability of our own hearing.
Every human being should learn and understand the limitations of our own bodies and senses, instead of blaming other stuff for it. |
|
|
| Elso Kwak |
Tell that to Stradivarius and Johannes Ruckers Eva!
:rolleyes: |
|
|
| Eva |
| quote: | Originally posted by Elso Kwak
Tell that to Stradivarius and Johannes Ruckers Eva!
:rolleyes: |
They wouldn't have gained any reputation if they hadn't been able to get repeatable results through a very scientific mind, that allowed them to understand very well what they were doing through experimentation.
Are you trying to compare that with a bunch of audiophiles throwing fashionable capacitors at their gear without knowing how to bias a transistor (or a tube for that matter)? It's ridiculous. They neither know what they are doing nor they will ever know or reach any solid criteria. Not to mention that after three months they will have already thrown away their current equipment (with all the non-sense mods inside) and bought new more fashionable stuff.
Remember that for any Stradivarius-like genius we got, there were also thousands of gossip people that were never able to get repeatable results, and though they probably made a lot of money by selling snake oil and telling nice stories, nobody remembers them now.
Do I have to mention that Beethoven composed nice music, just by mind, while he was *deaf*? In fact, a very powerful mind is required in order to filter out the remarkable amounts of false information that out hearing system gathers. It's sad but we have to accept that our hearing lacks a lot of precision and is very poor in comparison with other species (that can precisely locate and perceive things happening thousands of meters away). By far, our most precise sense is sight (and of course we have a lot of thinking power, altough we seldom use it). |
|
|
| vectorplane |
Lesson to learn: If you venture into this forum, you need to be ready for Eva's scrutiny.
In all fairness to those trying to "improve" their gear by replacing each component with a more expensive version, you need to notice that there are a lot of commercial interests out there perpetuating this cult.
And the number of these commercial parties is growing. They grow by inventing a new pseudo-science, and then nourishing the ego of their cult members by reassuring them that they have understood this science.
It's a way for someone to feel they excel at something which they would not be able to understand if objectivity were enforced.
It's all ego for the cult member. And it's all money for the cult leader. Two very strong market driving forces.
You'll often see some of those cult leaders blurt out some samples of subjective logic: "...I like a diode at this position...", or "...I don't like high-value capacitors here..."
If they were ever in the surgical intensive care ward of a hospital, I hope they will accept a doctor who throws out all the monitoring equipment, and starts using his nose and his ear to decide how to treat the dying patient.
Andy |
|
|
| poobah |
Interesting that someone brings up Stradivarius. Many people search and search for his "secret".
It is generally agreed that the violins he built later in life were the best. Years and years of perfecting a craft and making steady and methodic changes.
Those who have achieved similar results in the modern day have said that Stradivarious had no secret (no magic capacitor). They say that his "secret" was merely doing everything right... absolutely everything. He knew and understood every aspect of his craft thoroughly.
Sounds like science...
Keep in mind... you never fully know the sound of a violin until it is assembled... and it is not easy to reverse assembly and correct prior mistakes. I imagine he was highly motivated to make violins and not firewood.
:) |
|
|
| forr |
Poobah,
---Those who have achieved similar results in the modern day have said that Stradivarious had no secret (no magic capacitor). They say that his "secret" was merely doing everything right... absolutely everything. He knew and understood every aspect of his craft thoroughly.---
It reminds me Andrei Tarkovski's film "Andrei Roublev".
Its last part shows the making of a huge bell under the orders of a boy. He pretends to possess the "know how" from his father who just died. I won't tell the end of the story.
Hi Poobah,
In an other thread, Quasi suggested you needed a hair-cut. I've seen you have been to the hair dresser. But not for the intended aim.
regards ~~~~ Forr
§§§ |
|
|
| poobah |
Thank you for noticing Forr. I had my hair "styled" by a professional.
:D |
|
|
| Elso Kwak |
There is nothing wrong with voicing parts in your equipment. Even when these parts are boutique quality.......
I do it all the time with transistors, FETs, caps, resistors.......
:cool:
PS. the secret of Stadivarius, Johannes Ruckers and other ancient instrument builders.......? Good listening ability! Seldom found with electronic engineers; hahaha now Jocko, Charly Hansen and others are rolling over me!
:clown: |
|
|
| vectorplane |
| quote: | Originally posted by Elso Kwak
... the secret of Stadivarius, Johannes Ruckers and other ancient instrument builders.......? Good listening ability! Seldom found with electronic engineers; hahaha now Jocko, Charly Hansen and others are rolling over me!
:clown: [/B] |
It is not clear that Stradivarius' product was only a result of listening ability, or that today's subjectivist afficionados share any skills with the violin maker.
The fact that one bases decisions on what they hear (or think they hear) does not demonstrate good listening ability. Not any more than using your head to hammer a nail into the wall would demonstrate good thinking ability.
Many may have had good listening ability in Stradivarius' days, and did not produce violins.
Having good listening ability does not, unfortunately allow us to manufacture a speaker, or a violin, or a trombone for that matter.
You are attempting to attribute subjectivity to someone who may have used a high level of objectivity in implementing his craft. Stradivarius may have liked to use some today's sophisticated equipment, if he were given the chance. Since he is not available for comment, none of us can really say for sure.
The Stradivarius analogy is probably not very relevant.
Everyone has a baggage of well-rehearsed analogies to support their own opinions and beliefs.
As for engineers lacking listening skills, it is not clear where this fact comes from, of if it is even a fact.
When you hear someone praise fast-recovery rectifiers because they (the TO220 case) "are cool looking", the form-over-function credo stares you in the face. Noone should care how a diode looks, *unless* they intend to show it to friends...
( EGO !!! )
Andy |
|
|
| jackinnj |
| quote: | Originally posted by jackinnj
putting a 10nF cap across a diode just lowers the resonant frequency of the "tank" circuit -- |
I should add (to my own post) that when you lower the resonant frequency of the tank circuit you also DECREASE the voltage spike that the diode will see as a result of the transformer's leakage inductance -- |
|
|
| tinitus |
Stradivarius - not two pieces off wood are alike and "he" was good at selecting the right wood and to judge how thick or thin it should be - and probably wood was better at that time
true art and craftmanship |
|
|
| AndrewT |
| What happened to the topic? |
|
|
| peranders |
| I'll guess we don't come any further.... |
|
|
| Elso Kwak |
There was a thread in the past about high speed diodes, started by Fred dieckmann.......
Most was said there.
:cool: |
|
|
| Eva |
Can't anybody build a nice test setup allowing to measure reverse-recovery current and voltage spikes for various transformer and diode types???
I did it for conventional diodes and published the measurements and real oscilloscope waveforms. I also showed how to damp the spikes associated to these diodes until they were barely measurable. However, I'm busy now with an urgent project.
Anyone? |
|
|
| forr |
Hi EVA
--- did it for conventional diodes and published the measurements and real oscilloscope waveforms. I also showed how to damp the spikes associated to these diodes until they were barely measurable.---
Is there a link, please ?
~~~ Forr
§§§ |
|
|
| Eva |
Check this thread. I start posting in page 6 (I think) and the interesting stuff starts from page 7. Other people also attached waveforms, but as theirs came from simulations and simulators doesn't model diode reverse recovery properly at all, they had little to do with reality.
http://www.diyaudio.com/forums/show...&threadid=66542
Note that diodes suffer an abrupt impedance change just when they stop conducting, and when they are connected in series with an inductor in which some current is flowing, a voltage spike is produced that excites any possible HF resonant modes of the entire system. Placing a low HF impedance in paralell with the diode damps the voltage spike, and may also damp system resonances if the right component values are chosen. Don't expect any simulator to model that properly, it has to be measured with oscilloscope and real components (that's the dirty work nobody wants to do).
|
|
|
| jackinnj |
| quote: | Originally posted by Eva
Placing a low HF impedance in paralell with the diode damps the voltage spike, and may also damp system resonances if the right component values are chosen. Don't expect any simulator to model that properly, it has to be measured with oscilloscope and real components (that's the dirty work nobody wants to do).
|
Doesn't the junction capacitance of a diode changes through the conduction cycle?
It is a lot of work, tedium, and it must be painstakingly acurate. |
|
|
| Eva |
| Capacitance of a diode changes with everything, including temperature, current and voltage (forward or reverse). However, I think that capacitance is of little interest when the diode is forward biased and conducting, as it's effectively shunted by the own diode and all the possible HF resonance excitation energy would be dissipated in the first cycle of HF excitation. |
|
|
| poobah |
Actually Jack, the cap value is not super critical... just get the decimal point in the right place and you can make a vast improvement...
;) |
|
|
| jackinnj |
| On-semi has some nice pictures: |
|
|
| Eva |
| Cap value is not super critical for spike supression, but this added capacitance is the one that will resonate with transformer leakage inductance. This makes the value a bit more critical. Also, a series resistor will help to avoid resonance but will reduce spike supression performance, that's why I employed a capacitor and a RC in paralell with it. |
|
|
| Eva |
This graph has no useful information for us. Note that the current downslope is 1A per 10 nanoseconds (100A/us standard test for ultrafast diodes) and the thing rings at several dozen Mhz.
Reverse recovery behaviour is entirely dependent on this current downslope (as charge is removed from the junction slower or faster and the diode stops conducting only some time after all charge has been removed), that for 50/60Hz rectification applications is more like 1A per 100 microseconds. I doubt that the own manufacturer has even measured their diodes in these working conditions. I will do it and post the results when I find some spare time to prepare the experiment.
Note that several graphs from diode datasheets show that reverse recovery time increases substantially as current downslope is reduced, so we may easily find out that ultrafast diodes acually behave much like standard diodes when current downslopes are as low as 1A per 100us. |
|
|
| Joseph K |
Eva,
RANT MODE ON
| quote: | | Can't anybody build a nice test setup allowing to measure reverse-recovery current and voltage spikes for various transformer and diode types??? |
here:
First part
continued here
I think one could have found almost everything - measurements, explanations, formulas and graphs - if one would have made that small effort of searching the forum for previous art.. :D
And yes, diode capacity counts, and it is visible from the measurements. What is also visible is the clear superiority of the MUR type fast & smooth recovery diodes, over the classic rectifier bridges, in this particular reference [reverse recovery spikes].
And yes, this was a sidenote responding to You fom that other thread.
And also the formulas and methods were shown if someone wants to use C // RC damping of those spikes, instead of completely avoiding them.
RANT MODE OFF
Otherways I think we agree in most of the points.. :)
Ciao, George |
|
|
| jackinnj |
JosephK -- yeah, that was me -- but instead of using the real world transformer (for the moment) I would now change the procedure to just use several inductors 10uH, 33uH, 50uH, 100uH etc. in series with the various diodes. For the moment "assume away" the coupling coefficient inefficiency of the transformer and interwinding capacitance, resistances.
Like any other experiment, at first you want to separate out the independent variables one at a time, then see if you can predict the interdependencies from first principles. |
|
|
| Eva |
Good work Joseph.
But your oscilloscope pictures show that ultrafast and schottky diodes take much longer to stop conducting when they are driven with very low di/dt and also produce a turn-off spike (smaller) and some ringing (of higher frequency so more likely to be radiated, and that should be damped with a suitable RC anyway).
On the other hand, my measurements contrasted to yours show that a classic KBPC25 diode bridge with just a suitable C and a RC across the AC terminals does not produce any spike or ringing and behaves better than MUR and MBR diodes when they are used without anything else.
Finally, you also found out that placing 100nF capacitors across electrolytics actually worsens any supply ringing problem because dissimilar capacitors resonate mutually when paralelled. That's something that I'm always emphasizing, because most people believe they are doing the right thing when the paralell dissimilar capacitors (sometimes three or four!!).
p.s.: If you do similar measurements again, include a small shunt resistor in order to measure diode current. Here is where the spike will be best observed. |
|
|
| jneutron |
I'm afraid I haven't dealt with reverse recovery speeds and fixtures that are as slow as that graph, but I believe that ringing is simply fixture inadequacies.
Get the fixture to settle in 200 or so picoseconds, then you'll get a clearer picture of what's going on.
Cheers, John
If you wish to monitor the current, you have to balance the CVR symmetrically, to eliminate I dot errors. An easy task with some simple matched carbons. (don't use precision metal films, they're spirally trimmed.. |
|
|
| Eva |
| He should be kidding :) |
|
|
| jneutron |
| quote: | Originally posted by Eva
He should be kidding :) |
Should be, but am not.
As a test engineer for the largest manufacturer of diodes on the planet (4 million units per day at the time), One of my tasks was measurement of TRR, both for existing product, and for newer higher speed R and D. In that capacity, I developed a TRR fixture capable of measuring the snap recovery of a diode with sufficient bandwidth to show a square wave with 250 picosecond rise times. This of course, required the use of microwave power resistors (BeO) and the layout from "heck".
I discovered then (1981) a signal generator called a Tek type 109 (at the time it was over 20 years old) which produced 250 pico rise/350 pico fall times. A little known attribute of mercury wetted reeds is their amazing rate of current change during make or break.
My fixture allowed the diffusion Phd's to see exactly what the diode was doing in the sub nano regime, and gave them feedback for diffusion profile modifications.
Seeing the waveforms being presented on these posts remind me of the horribly speed challenged TRR fixtures I have seen in the intervening years, and the incorrect assumptions resulting from the inadequacies of the test methods and materials. Also, the fundamental errors which we (I include myself) have committed in trying to measure these speeds in low Z circuitry.
One of the easier things to remove, is the inductance of the current viewing resistor. It is actually rather trivial to make one which theoretically goes below 60 picohenries. I made one, pic attached. I must note that I have tried to measure the inductance of this beast, but have not been able to acheive a reliable measure at the 60 pHenry level, I am only able to get the measurement down to 250 picohenry maximum. (this inability stems from the current path taken as it spreads out to the ring of resistors, I cannot physically remove that.)
Anyone interested in making a simpler one that will be good to the single digit nanohenry level, I'll describe it. It would take only 2 or 3 matched 1/4 or 1/2 watt carbons to make a simple one which outperforms even the caddock ceramic types by an order of magnitude.
(I have much experience in this area, having been bit in the hindmost on many occasions)
As the waveforms are such a confusing mess, it it very difficult to figure out which is due to the diode, which the surrounding circuitry, and which the result of measurement error as a result of inductive coupling and inductive reactance. This is why I said get the stuff to settle in 200 pico or less, that way it is much easier to understand the waveforms without wondering how much of it is related to the measurement technique. I would be happy to evaluate and put my two cents worth in if pictures of the setup including scope probe were presented. There is a very small possibility that my experience could be of some use. Ok, maybe a small possibility, but I can dream..;)
Cheers, John
Edit: yah, the right side is not soldered, this is a pic of the incomplete one. I did not have a good pic of the finished which shows the detail of the resistors.
This configuration is that of a 36 pole magnet, and the external field drops off as 1/R37, internal as1/R35 . Another way of looking at it is paralleling of 36 inductors of small value, in an antiparallel configuration.
Oh, forgot..I don't work at that diode company now. That was back in '81 to '84. |
|
|
| poobah |
That look's an aweful lot like a space modulator. I have had to build IGBT snubbers that were as messy.
:D |
|
|
| jneutron |
| quote: | Originally posted by poobah
That look's an aweful lot like a space modulator. I have had to build IGBT snubbers that were as messy.
:D |
What is wrong with you...??:D
Given your Avatar, I was sure you'd say it looked like a tie fighter:confused::confused: You're under your game there..
Actually, I figgered it looks more like a space heater, what with the fins and all.
It's amazin what ya can do with a section of slantfin from Home depot and some resistors from Digikey. (my first test load resistor got overheated and melted the solder, it was on a 1/2 inch diameter tubeset, with no heat sink fins...using the slantfin allows more dissipation.
My next config will be a symmetrical block style, and I'll epoxy encapsulate it with a CPU cooler as the base of the structure. That'll get me into the several hundred watt RMS regime without melting the solder again.
Cheers, John |
|
|
| Joseph K |
Eva, all -
Bah, yesterday I've just lost a 500 page answer to Eva bacuse of my stupid gprs home connection....
John, are You talking about my graphs or about that one shown by Jackinnj in post49?
Because I can only talk about my things. Yes, You are right in that I was not conducting a checkup for my test fixture - I can still do that. But: the observed resonance in my case, at around 400 kHz, is way too low a frequency to be a result of a parasitic resonance due to fixture / probe lead inductance.. while Your observation is totally in place as regards the pushed to the edge test environment in your case, in this simple case of ours with ~100 uH leakage inductance of the home transformer's secondary, and the several nF stray capacitances in play, the measured resonance seems to be on spot. Then, I could shift it by will by adding extra capacitance, and the results were in accordance with the calculations. Also I was able to note the resonance shift due to diode reverse junction capacitance, [in case of different diodes] and the measured offset was in accordance with the datasheet values controlled later..
Eva,
I think we are in a gross agreement in almost all but this:
| quote: | | that ultrafast and schottky diodes... also produce a turn-off spike (smaller) |
I would like to point out again, that in accordance with what was written here, in post 545, the fast diodes does not produce at all a spike, but rather a "step" response, with a step level in accordance with their forward voltage drop! Now, this step response, of about .5 V magnitude in it's really fast part, is riding on the whale-back of the 50 Hz sinusoidal. So the already much smaller high frequency energy content [with respect to the ten volts order real spike in case of a slow bridge rectifier] is subject of an even further LP filter effect because of the "hold"-like function with 50Hz fundamental...
Note also that this "step" exists the same way even in case of normal diodes, only is hardly visible because of the high level of reverse snap and ringing!!
Then, while the reverse snap with slow diodes is proportional to the original forward current, so provides for a nice AM modulation with the audio signal, with spectral components up to the MHz range, with the fast diodes this mechanism is totally absent.
Then, while it's true that one can slow down & dampen the spike and ringing with the appropriate values of snubber components,
how many real applications have You seen done that properly?
Finally, yes, I know about the recharge current spikes refilling te reservoir caps, have played with measurements and simulations, and got some [at least for me] unexpected results..:bawling:
http://www.diyaudio.com/forums/show...3560#post603560
ciao, George |
|
|
| Joseph K |
Uhm,
In the same time would like to emphasise: I don't want to be a defensor of fast diodes at all costs, only wanted to show that this kind of argumentation provided by Eva, in my opinion, is not sufficient for showing the superiority of the conventional type bridges.
Now, if somebody could point out other aspects clearly in favor of the classical bridges, that would move really forward the debate!
Ciao, George |
|
|
| Eva |
jneutron:
I said that you were kidding because we are not trying to isolate diode behaviour from test setup influence, what we want to measure is how the diodes behave when they are fed by the distributed inductance/capacitance of the secondary winding of a typical 50/60Hz EI or tororid mains transformer (with the corresponding low di/dt slopes) and when they are feeding electrolytic capacitors through PCB tracks or wires of some lenght and prone to ringing. We were trying to reproducing the real working conditions (sometimes lousy) in which people employs rectifiers in their DIY amplifier projects, and this has not much to do with 250 ps rise times :D |
|
|
| jneutron |
| quote: | Originally posted by Eva
jneutron:
I said that you were kidding because we are not trying to isolate diode behaviour from test setup influence, what we want to measure is how the diodes behave when they are fed by the distributed inductance/capacitance of the secondary winding of a typical 50/60Hz EI or tororid mains transformer (with the corresponding low di/dt slopes) and when they are feeding electrolytic capacitors through PCB tracks or wires of some lenght and prone to ringing. We were trying to reproducing the real working conditions (sometimes lousy) in which people employs rectifiers in their DIY amplifier projects, and this has not much to do with 250 ps rise times :D |
I took no offense.
What I am pointing out is the problems measuring what you guys are talking about, with concurrent errors.
As an example, a 24 guage wire pair exhibits about 1 uH per foot when it's anticurrent mate is 60 inches away.(This spacing chosen to get a reasonable per wire estimate of inductive reactance). For a single current carrying conductor of 24 guage, it is reaonable to assume half a microhenry of reactance.
A diode which will shutoff in 1 usec, and 1.5 amperes (the old jedec spec numbers), that is 1.5 amps per uSec. Using V =L di/dt, the induced voltage across .5 uh is .75 volts, across one inch of 24 guage wire, that voltage error will be 63 millivolts.
If the CVR (as you recommended try) is low enough in value to not disturb the series circuit, say 100 milliohms, the IR drop you are discussing will be on the order of .15 volts give or take.
At 1 amp per microsecond, the error will be 63 / 150, or 42 percent.
A 1 uSec device is a general purpose rectifier.
At 10 amps per uSec, the error becomes 420%.
At 100 amps per uSec (this is typical of a 10 nSec device), the measurement error is 42 times the entity you are trying to measure. Attempts to view this device using a simple resistor, or even using a scope probe with the alligator clip type ground lead, will result in wildly erroneous waveforms of the type that diode manufacturer (onsemi I believe) has provided.
I know how to remove that error very easily. But it is an error that most are unaware of. Discussion of wild transients must be tempered by the fact that they certainly do not exist at the levels that the measurement equipment is reporting.
Cheers, John
edit:math errors, sorry ran outta fingers.. |
|
|
| Eva |
| We are working at approx. 1A per 100 microseconds. |
|
|
| Joseph K |
John,
I appreciate very much Your intention in pointing out possible caveats!
But I am afraid that momentarily there is some confusion here.
In my diagrams, I have not measured current at all. We were looking for possible voltage spikes induced in the transformer secondary leakage inductance by the diode r.r.current snap. My sniffer is a small 10nF - 2kohm highpass filter, I've been measuring the voltage drop across that 2kohm carbon resistor. So I think the fixture is well damped, no way for MHz resonances.
Then, in the case of measuring current, Your precautions could applie - but:
I was using, as CVR, an ~ 8mm long 50 mOhm sense resistor, from resistor wire. So no way 500nH there, a couple of nH maximum rate.
Then, te change of rate. You talk about 1.5 A / usec rate, while in my actual measurements, because of the slow LRC tank exited, in case of a real setup with ~100uH secondary leakage inductance, 10000 uF filter cap, ~.1 ohm DCR, I got some 8A / 1.5 millisec rate of change, that is, several hundred times lower than that in your example!
Then, while it's true that the alligator ground clips can be a problem, though no current is flowing accross them, I was using a differential setup: two channels with symmetrical probe setup, subtracted.
But the main point here really is the constant presence of the transformer leakage inductance in the order of hundred uH-es, dominating everything in the loop, as pointed out by EVA, correctly?
Ciao, George |
|
|
| AndrewT |
Hi,
are we getting close to a conclusion that us DIYers can use?
How do we measure that leakage inductance of a toroid?
Then, how do we calculate the component values needed for a useful (not useless) snubber? |
|
|
| jneutron |
| quote: | Originally posted by Eva
We are working at approx. 1A per 100 microseconds. |
While the surrounding circuitry and drive currents are indeed doing such, the diode is under no such constraint.
When a diode is forced to carry current in the reverse direction, when the carrier supply collapses either by sweep or recombination with dislocation sites (gold or platinum), the current will collapse at the diode's desire, not the circuit.
If the circuit has high inductance, as is the case for the supplies we are speaking of, it's reaction will be to very rapidly sweep the depletion zone of the silicon, charging the internal capacitance very swiftly. Note that the capacitance continues to fall as the reverse voltage increases.
What I am pointing out is that while you believe you are working in the 1 amp per microsecond realm, the diode is not.
Cheers, John |
|
|
| jneutron |
| quote: | Originally posted by Joseph K
John,
I appreciate very much Your intention in pointing out possible caveats!
But I am afraid that momentarily there is some confusion here.
In my diagrams, I have not measured current at all. We were looking for possible voltage spikes induced in the transformer secondary leakage inductance by the diode r.r.current snap. My sniffer is a small 10nF - 2kohm highpass filter, I've been measuring the voltage drop across that 2kohm carbon resistor. So I think the fixture is well damped, no way for MHz resonances.
Then, in the case of measuring current, Your precautions could applie - but:
I was using, as CVR, an ~ 8mm long 50 mOhm sense resistor, from resistor wire. So no way 500nH there, a couple of nH maximum rate.
Then, te change of rate. You talk about 1.5 A / usec rate, while in my actual measurements, because of the slow LRC tank exited, in case of a real setup with ~100uH secondary leakage inductance, 10000 uF filter cap, ~.1 ohm DCR, I got some 8A / 1.5 millisec rate of change, that is, several hundred times lower than that in your example!
Then, while it's true that the alligator ground clips can be a problem, though no current is flowing accross them, I was using a differential setup: two channels with symmetrical probe setup, subtracted.
But the main point here really is the constant presence of the transformer leakage inductance in the order of hundred uH-es, dominating everything in the loop, as pointed out by EVA, correctly?
Ciao, George |
Ah, last math error of mine in post..forgot to convert from 500 nH of a foot, to 1 inch CVR. 42 nH for one inch. At 8 mm, that'd be 13 nH.
So let's have an abrupt I dot of 1.5 amperes/1 uSec.
.013 uH times 1.5 is 19.5 millivolts.
That is for a standard recovery device. BTW, that is the timeframe for the device to arrive at zero conduction, and does not represent the rate that it "zones out". Abrupt recovery devices are very bad in this respect. So I use the device spec'd number just for reference, with the caveat the real devices will transit from reverse conduction to zero at approximately 10 times the rate. My estimates are lowballs.
100 Nsec devices...195 millivolts
10 nSec, 2 volts.
IF the 10 nano device actually recovers in 1 nSec, 20 volts?? That is the voltage response on your 8mm long resistor, 50 milliohm resistor.. Your resistor design would not be acceptable for use in a TRR fixture as a result of it's inductive storage. Yes, it is low, no it cannot accurately distinguish the current change in a fast or medium fast diode.
The one or half amp signal in reverse direction is being viewed as a 50 millivolt signal, but yet the resistor will generate 4 to 400 times that depending on the snap characteristics of the silicon.
As I said, how does one distinguish the characteristics of the silicon independently of the cvr?
Personally, I would be less apt to recommend the selection of a diode based on it's recovery characteristics if all I had to go on was waveforms and writeups based on those waveforms where the test conditions were incapable of discerning real differences. If you really want to identify the issue of diode selection vs sound, this is not the best place to look. I note that both of you are also looking at the other circuit element waveforms, and are doing so quite well. I have been very happy to read both your's and Eve's posts.
Using a differential probe setup does not remove the pickup inductance. You must consider the physical signal path loop.
I have found that those little probe tip females, the kind that come with the hp or tek probes, are the best, but the physical size of your resistor does not allow taking advantage of that.
This is why I said I can detail a resistor which does not suffer the issue I speak of.
Cheers, John
Edit: Joseph, I just remembered that you were the one I had the braid over capacitor discussion with...
Your resistor suffers the exact same thing..the external inductance..
If you take 10, 1 ohm resistors and parallel them to form a cylinder..have the circuit current flow through those. To measure the current through this resistor, run a wire through it on axis, use the scope from this wire to the tie point opposite the wire connection. That removes the scope wire from all collapsing magnetic fields. This is exactly what I had mentioned as possibly sending to you for measurement on the analyzer.. |
|
|
| Joseph K |
John,
You are putting me under the torque...:bawling:
I'm afraid I have to retire now, and will not be able to continue with that damn GPRS connection, but one point, just to underline my confusion:
Here, in the scope shot below, I was using that 50 mOhm CVR, and one can see the ~2A current pulses[100mV peak in the pic], or the ~8A current pulses[400mV peak], the 2A ones having a half width of ~ 1.5mSec, the 8A pulses are having ~3msec half width.
The current is changing, because I was testing the system response to a "step" in the load current.
Evrything is inverted because it came like this in the process off subtracting, the zero line is obviously on top.
What, in fact, I was NOT able to see, never, are the "they should be there" fast current glitches at the diode turn off moment.. The rectifiers used here were actually MBR 760s, but I saw the same with MUR860, which are ~ 60nsec Trr devices, so should have shown ~ 400mV peak?
These glitches were only visible while looking at the tension of transformer secondary, and only with conventional rectifiers, nothing with the fast rectifiers...
Ciao, George |
|
|
| jneutron |
| quote: | Originally posted by Joseph K
What, in fact, I was NOT able to see, never, are the "they should be there" fast current glitches at the diode turn off moment.. The rectifiers used here were actually MBR 760s, but I saw the same with MUR860, which are ~ 60nsec Trr devices, so should have shown ~ 400mV peak?
These glitches were only visible while looking at the tension of transformer secondary, and only with conventional rectifiers, nothing with the fast rectifiers...
Ciao, George |
At that time base, I don't think you can see any recovery transients. Actually, I believe a lot of the high end talk of transient stuff is just, um, interesting creative dialogue.
Looking at the voltages is really the way to go in the supply.
For audio sound, don't look in the supply. Look at the induction of currents elsewhere, like ground.
Cheers, John |
|
|
| Joseph K |
John,
| quote: | | For audio sound, don't look in the supply. Look at the induction of currents elsewhere, like ground. |
That is a very valid point!
Ciao, George |
|
|
| jackinnj |
| quote: | Originally posted by AndrewT
How do we measure that leakage inductance of a toroid?
Then, how do we calculate the component values needed for a useful (not useless) snubber? |
Short the primary leads together and measure the inductance of the secondary -- you can use the simple (and acurate) bridge method discussed by Steber: http://www.arrl.org/qex/2005/Steber.pdf
It is also helpful to know the winding coupling capacitance, and DC Resistance of the primary and secondary -- you can also measure this with a bridge.
There are a number of papers on calculation of snubber values -- and different types of snubbers for the application -- one often cited paper is by Hagerman "Optimum Snubbers" -- easy to find on the web. |
|
|
| AndrewT |
Hi Jackinnj,
thanks for that link.
I have Hagerman saved. |
|
|
| percy |
Jackinnj,
should the primary or the secondary be shorted ? Not arguing, just thought I had heard/read otherwise somewhere so I thought I'd confirm.
I used Speaker Workshop to measure the inductance. |
|
|
| Joseph K |
Elso,
It's good that You have reported that thread here, thanks!
Though if You intended to say "we are going in circles; everthing had been told before" then I say yes, we are going in circles -- though just the post 8 of that thread [by SY] justifies the new efforts here.. :D
Ciao, George |
|
|
| quasi |
| quote: | Originally posted by forr
Hi Poobah,
In an other thread, Quasi suggested you needed a hair-cut. I've seen you have been to the hair dresser. But not for the intended aim.
regards ~~~~ Forr
§§§ |
....and it seems you've taken up smoking ! |
|
|
| jneutron |
Hmm..
The links in the first post are nice. It's good to see that what I recall from previous work experience is still consistent with what they write. It is bad to see that they have not advanced a bit in TRR testing over the last 20 years.
Gensemi link was dead, though.
Any manu's actually write about using fast diodes in line supplies, or are all the manu tech papers out there still all about switching stuff and transient V-I losses (losses during switching)?
Cheers, John |
|
|
|
