How do you know what the pickup level is? And why is it a bad thing to have more rejection?
It seems pretty unlikely that even a single FPGA is going to pull less than that. A board I just sent off to the client uses a CPLD that pulls
80 mA quiescent all by itself.
3) any kind of high frequency disturbance on any of his
That's what bypassing is for. And it will in general be better than what you'd get with completely separate 1117s, because when you stack them they're all running with a gain of 1.0, i.e. maximum loop gain and maximum BW.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
The 5 volts is from a VME crate, and the loads are an FPGA and an ARM processor and a bunch of miscellaneous logic. The tail current from the 3.3 regulator divider resistors will try to pull up the 1.2 volt supply, but the FPGA and CPU core static power dissipations will easily eat that up.
Of course the logic supplies are bypassed all over the place, so there won't be any noise problems.
Looks OK to me. I've done this before, stacking LM1117s or 78XX regulators, and it works fine. Yes, the minimum loads need to be considered.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Pickup almost certainly exceeds the nV of line frequency remaining after the regulator cascade rejection.
It's generally bad practice to rely on the load, the minimum of which isn't specified on the dropbox schematic, to prevent possibly damaging overvoltage mal-operation of his power supply, he should incorporate protection.
Ummm, they're not all running at gain of 1.0, and the additional bandwidth, which is about an octave, won't be all that apparent with the outputs swamped with the dominant pole compensation e-caps.
That's a really crap way to configure your regulators, especially
If there's any great possibility of a FPGA supply running at less than 5 mA under any conditions, sure. You do have to design stuff like that.
You keep tossing around numbers like 60 dB (based apparently on zilch), chopping and changing as though a factor of 2 in bandwidth doesn't affect regulation, and generally making up stuff as you go along.
You're just throwing rocks.
How would you suggest designing it, and why?
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Uh-huh, in the case of a regulator, the main advantage of BW is a broader spectrum of low output impedance, and that's just not a problem in this application.
About the only reason I can see for crosscouplng the regulator operation would for load sharing. But he can do that directly off the 5V too, except it might interfere with his quad pack requirement.
You didn't do away with kitting, you just moved it to the CA. Now you pay for it as part of your assembly invoice. Something like Kanban is only useful in a volume production setting to optimize production costs. In John's situation I expect they are assembling their own boards to optimize some aspect other than cost. It is hard to manage costs for runs of 3.
You can do a poor job of kitting just like you can do a poor job of Kanban.
BINGO! There is the difference. You are judging a low volume assembly process by your high volume needs.
I have to be careful to get the 317/1117 type dividers right. I keep wanting to flip the upper and lower Rs on these types, where ADJ is referenced to Vout, not ground. We have a ritual: I do it wrong, and the guys in the design review fix it. Can't mess with tradition.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
It's June 1st at 8AM. It's 52F here and 48 in Truckee. The heat is on in both places. I'll wear a parka to work today.
The 1.2 volts (FPGA and ARM core) power will be low, and the 2.5 (PLLs and a few line terminators) will be lower.
I guess I could drop some shunts into the rails, to measure the currents. I should remember to always do that.
I like the 1117. It's an "MDO", less drop than a 317 but still cheap and simple. I wish it was ceramic-cap stable, but nobody's perfect. I use tantalums on the output to maintain low-temperature performance.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Only if you consider rigging a P&P machine kitting. Usually it isn't. Also, since they are larger they have lots of jelly beam parts constantly loaded as long as you ECO-release their parts along with yours. Then there's machines that can auto-load from a storage facility. At the end of the day you have all but eliminated the labor costs of kitting.
We successfully used it on system production runs at the rate of several dozen per month. Not exactly volume production.
Companies like Toyota have this down to a science. The way to get there with such a large array of products that include low-volume ones is to standardize as much as possible. Use the same starter on nearly all cars, and so on.
True, but the P&P machine essentially does its own Kanban. It eats 128
10k resistors per board off the reel so you'd know when a new reel is going to be needed.
Kitting usually increases labor costs in a product significantly.
[...]
Again, we also did it on low volume runs. Long term numbers do not lie because they are vetted over and over again by the accountants, we scored a substantial cost reduction after going Kanban.
and every single one of the cars leaving the line may be different with every option ticked off when bought arriving at the line at just the right time to be mounted on the specific car that needs it
considering how complicated a car is it is mind boggling how they build so many so fast so cheap
It's probably also a mixed calculation. The US car market is large and very different from yours. Here, they have lots of cars at the dealer and they are usually equipped quite well. Poeple pick a car fairly quickly, then sit down with the dealer for a serious haggling session with bluffing and the whole nine yards, then drive off in their new car right afterwards. So the car manufacturer just makes 50,000 of these and another 70,000 of those and ships them. In 5-6 of the usual (bland ...) colors maybe.
This results in being able to produce cheaply in a very repetitive mode. For other parts of the world they have to individualize almost every car because people pre-order and then wait many weeks. Then the cars are more expensive.
Kanban works well also in a custom-build environment. On our medical machines it could happen that they assemble two dozen without and one dozen with video page printers. It didn't matter, as long as the sales forecast was somewhat accurate.
They do have cars in stock here though not many I think, if they it'll often be some sort of campaign with model with some fixed set of options or "demo" cars
Sometimes there's a discount on popular models if you can wait for delivery
I heard a story that Fords CEO once visited and couldn't understand why Denmark had more than one Ford dealer, a single US dealer sold more than all of Denmarks combined
the 180% tax + 25% vat on cars might have something do to with that
Building a fixed configuration also keeps the line going when there is less orders. I don't think it makes a difference to them if they have to put option x in car number 12356 and option y in car number 123457 everything is just in time and all the systems are in place
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.