Nelson thanks for this. Just last week I succeded to almost halve the standing current in the Vas stage by applying your active current source idea (hope you don't mind) so there you bailed me out there. 🙂
With this paper you sort of undo that!😱
I find it curious that three different IXYS support engineers all failed to come up with this app note which is quite to the point when I asked them about the issue.
Anyways. I think in the Vas stage with 10mA DC bias at 2.1kV I am not going to worry about this, there's just not enough energy in this low dissipation of just 21W in a TO-247-equivalent package on a heatsink. With a heatsink temp of say 40C and a 1.1 Tjc + 0.3 Tcs the Tj will be about 30C above heatsink. That is very much safe for a max Tj of 150C. I am assuming here that the hot spot issue is less likely to develop at low temp, which may or may not be warranted.
The class AB output stage is another story. My design Iout max is 150mA RMS with 2 output devices in // so each will face a max Id of just above 100mA per half cycle. I will go and look at the max Pds depending on output voltage with this. I know there's a rule of thumb which I just don't remember, but Google surely will. Later.
Jan
Last edited:
One question (and I do not know if it needs answering), if the MosFet is created from many (analog of) small MosFet's (as the application note [seems to] indicates) and they do have different sizes and ballast arrangements, then, at 10mA's (this is 1% of the design limit), how many of these MosFet(imals) do carry the bulk of the current (in your application)?
Frans.
Frans.
Last edited:
Good point. My reasoning is that the unequal cell sizes and ballasts across the die counteract the high voltage effect of bunching, and that it spreads the current such that each cell carries an equal amount (or equal amount per area). I would expect that to be independent of the total current level.
BTW Found the equation for max class B power dissipation. It occurs at 50% efficiency with Pd being the same as Pload, and is (1/pi)*(Vcc^2/Rl). In my application that is about 70W per device.
Jan
BTW Found the equation for max class B power dissipation. It occurs at 50% efficiency with Pd being the same as Pload, and is (1/pi)*(Vcc^2/Rl). In my application that is about 70W per device.
Jan
I actually owned 4 or 5 of the previous Fluke HV calibrators, the ones with half a dozen of 4X150 tubes. I needed help to lift one up 😉
But they were full range, no output transformers, to 1100V RMS out to 100kHz at 100mA RMS with 0.01% THD. Awesome!
At one point I thought about re-configuring them for ESL drive use, but they really were too monstrous. They all came from a dismantled US Army calibration lab in Germany. New price in the 80-ies was $ 18k each.
For my Vas I use a single device with a bootstrapped active current source, works like a charm. The output stage is something I'd like to keep for myself at this point...
Jan
I'm not so sure expectant; switching devices ara designed (and characterized) for high loads (and the 1% load is real small).
[edit]
but then again, the 1%-load could be carried by 1% of MosFet(imals) [maybe]
Frans.
Last edited:
Jan, I find it comforting to stay far below the maximum junction temperature of military "MIL-SPEC" submicron integrated circuits. I believe that the military is especially conservative because their operating environments are so harsh, and discrete devices are more rugged than submicron ICs with 9 layers of metal. So that gives me two extra servings of safety margin. MIL-SPEC is 125C maximum TJ and I try to make sure I'm never above 98C even in the worst case "ridiculous but theoretically not impossible" scenario.
I have no problem to stay far below Tjmax with this. I am just going to go ahead and do extensive test for this.
I will spend the next few weeks (officially 'vacation') to lay out a 'channel board' PCB integrating the various parts that have been proven - the Vas, the output stage and the Tj estimator. Next up is the control board and the input/protection board.
No guts no glory, right? ;-)
Jan
I will spend the next few weeks (officially 'vacation') to lay out a 'channel board' PCB integrating the various parts that have been proven - the Vas, the output stage and the Tj estimator. Next up is the control board and the input/protection board.
No guts no glory, right? ;-)
Jan
I honestly can't tell a thing from those graphs. First, why characterize a 4500 volt device out to 140 and 160 volts? And I hope somebody flogged the individual who decided to present two graphs of related data with totally different axis scales.
Leakage is never a good thing to use for anything other than leakage. I assume they used a guard ring structure around the periphery, and maybe burying the edge junctions. Leakage is primarily used only to check the health of the passivation and edge integrity, never as a reliable temperature measurement.
John
I thought that was stupid too.
OK, I was just thinking, if the leakage stays very low that would be a Good Sign. Is there any way that I can test for this potential current hogging without having to rent the LHC?
Jan
Thanks for the link, an interesting read.
They say threshold voltage has negative tempco. I suspect that is incorrect as a blanket statement, as the IR fig I provided indicated, as well as the IR app note I mention in passing. It is possible IXYS figured out how to diffuse out the dependency and I never heard about it, or they simply neglect it because it's an article on linear operation.
In figure 2, they show an electro-instability boundary. Below the DC powerline, that slope isn't making sense. When time is a parametric, as in the pulse region of the graph, it's necessary to include the silicon heat capacity for pulses faster than 10 mSec, as well as lateral thermal transfer limited by die thickness and attach integrity. So they could draw any curve from 10 mSec and faster and nobody in the world could disagree.
As to transient thermal response testing, that was being done back in the late 80's, mainly as a die attach process control. Back then, moto was doing BVceo(sus) testing to cull out bad ones, but that test is unfortunately a destruct test.
But the concept of anisotropic cell structure is an interesting one.
John
Long time ago, I built an active load exactly like the one in figure 4. If you take the Id control line in the schematic, and pulse it half sine, watch the op amp drive signal to the mosfet. When the device is behaving, the drive will mimic the control signal. If the device is starting to do a runaway thing, the opamp output signal will drop. This because less drive is needed for the ID being forced into the source resistor.
If you are careful, you can see the onset of thermal runaway in the depressing of the gate drive required.
Mind you, I did this with BJT's, not mosfets. I do not know if the overall gate voltage will drop if the transfer gain goes anisotropic, as the current may simply shift within the real estate without requiring less drive. In fact, it might even be possible that when it hogs it may need more gate drive. Never tried one of these new types.
For 2 to 5 dollar devices, I'd go for that. For 40L devices, maybe not..😱
John
Last edited:
Actually what I should use is something like the IXYF30N450 IGBT. Almost square SOA, absolutely safe.
It's just that they come in at almost 100 bucks a pop...
But 2 // HV MOSFETs are close to that anyway.
Where's my lottery ticket?
Jan
It's just that they come in at almost 100 bucks a pop...
But 2 // HV MOSFETs are close to that anyway.
Where's my lottery ticket?
Jan
Hi,
Just for once, I would like to see someone do modifications with parts that fit properly! And replacing the 1A bridge with other diodes? A disaster for the PCB. If the power supply was any good at all, those changes wouldn't be audible.
Sometimes the designer and manufacturer did build it exactly right, and any differences heard are the result of confirmation bias. Changes to the actual circuit components can be audible, and fixing problems with a design power supply can be audible. Changing power cords and rectifiers shouldn't be in properly designed equipment.
If you are contemplating making changes to a piece of equipment, you should be damn sure that the internet modification information comes from someone smarter than the designer of that gear first. Otherwise you're probably damaging the equipment. Why are some "Mid-fi" pieces of equipment better than most "high end" equipment? Because people like Marantz, Namamichi and Denon have real labs with real good test equipment and more than one engineer who specializes in the art. It's not romantic, but it is reality. It just takes some good technicians (with real good test equipment and an ear) to eek out that last bit of performance that allows the design to operate at it's best. It's very rare for a bunch of folks with a soldering iron to actually improve anything. They do lay waste to mountains of equipment chasing "discoveries". Once reality (or the wife) sets in, good technicians can sometimes salvage some of it.
Just saying ...
-Chris
Just for once, I would like to see someone do modifications with parts that fit properly! And replacing the 1A bridge with other diodes? A disaster for the PCB. If the power supply was any good at all, those changes wouldn't be audible.
Sometimes the designer and manufacturer did build it exactly right, and any differences heard are the result of confirmation bias. Changes to the actual circuit components can be audible, and fixing problems with a design power supply can be audible. Changing power cords and rectifiers shouldn't be in properly designed equipment.
If you are contemplating making changes to a piece of equipment, you should be damn sure that the internet modification information comes from someone smarter than the designer of that gear first. Otherwise you're probably damaging the equipment. Why are some "Mid-fi" pieces of equipment better than most "high end" equipment? Because people like Marantz, Namamichi and Denon have real labs with real good test equipment and more than one engineer who specializes in the art. It's not romantic, but it is reality. It just takes some good technicians (with real good test equipment and an ear) to eek out that last bit of performance that allows the design to operate at it's best. It's very rare for a bunch of folks with a soldering iron to actually improve anything. They do lay waste to mountains of equipment chasing "discoveries". Once reality (or the wife) sets in, good technicians can sometimes salvage some of it.
Just saying ...
-Chris
Anatch, in these neoliberal days of laissez faire economy. our problem is often down to parts supply. <<few manufacturers of almost anything can supply a lot of its products in regular batches just as a manufavturer might want them, so almost every part with which it's possible has an alternative source or two. Unfortunately, many alternatives have different dimensions and/or pinouts. Hence the fitting problem.
I get around it by making the PCBs with alternative added pinout holes, starting from smaller (say 7.5 mm) and on to larger (10mm, 12.5 mm and 15 mm). the smaler one for say Wima caps, and the larger ones for say Plessey caps, while the smallest is reserved for Siemens caps, which are good but also smaller than almost everybody else's. More holes mena more money for PCB manufacture, but that's a small price to pay for my peace of mind and a proper fit.
I get around it by making the PCBs with alternative added pinout holes, starting from smaller (say 7.5 mm) and on to larger (10mm, 12.5 mm and 15 mm). the smaler one for say Wima caps, and the larger ones for say Plessey caps, while the smallest is reserved for Siemens caps, which are good but also smaller than almost everybody else's. More holes mena more money for PCB manufacture, but that's a small price to pay for my peace of mind and a proper fit.
Hi dvv,
I'm not complaining about the manufacturer at all. What I am talking about are the self-appointed experts that feel they must modify all equipment. These folks are the ones that need to either obtain parts that fit properly, or call the whole thing off.
Changing types of parts isn't a trivial action. You have to understand parts and where they are best suited to be used. Without this knowledge, equipment most often underperforms, or has very real problems up to and including circuit failure.
-Chris
I'm not complaining about the manufacturer at all. What I am talking about are the self-appointed experts that feel they must modify all equipment. These folks are the ones that need to either obtain parts that fit properly, or call the whole thing off.
Changing types of parts isn't a trivial action. You have to understand parts and where they are best suited to be used. Without this knowledge, equipment most often underperforms, or has very real problems up to and including circuit failure.
-Chris
Today, we are finishing a mod on a Vendetta Research SCP-2 preamp by taking out the 1A full wave bridges made of conventional diodes and making our own bridge with high speed, soft-recovery diodes.
It is a hassle! It costs time and takes care, but it will be worth it for those who can hear the difference. I would not bother if it did not.
Like so many subtle mods, I did not realize that this could be important at first, and that is why I still have many hundreds of full wave 1A bridges that I will never use. My departed CTC business partner, Bob Crump, did the first mod on his own Vendetta and got me to do it on mine. Once you look at what the diodes are doing, it should be obvious that it could be an audible improvement.
Now should every amateur do this 'upgrade'? Not if you are not good at fabrication and at least an electronics technician in knowledge. I, too, have had to repair 'mods' from amateurs, but I don't want to discourage those who do the mods to improve the sound quality of a piece of equipment.
It is a hassle! It costs time and takes care, but it will be worth it for those who can hear the difference. I would not bother if it did not.
Like so many subtle mods, I did not realize that this could be important at first, and that is why I still have many hundreds of full wave 1A bridges that I will never use. My departed CTC business partner, Bob Crump, did the first mod on his own Vendetta and got me to do it on mine. Once you look at what the diodes are doing, it should be obvious that it could be an audible improvement.
Now should every amateur do this 'upgrade'? Not if you are not good at fabrication and at least an electronics technician in knowledge. I, too, have had to repair 'mods' from amateurs, but I don't want to discourage those who do the mods to improve the sound quality of a piece of equipment.
- Status
- Not open for further replies.