XC7K325T Low cost Power management

[ali_asadzadeh]

Hi,
I want to design a new Board with this puppy I'm searching for preferably chines power solution chips for this FPGA, do you recommend any thing special or good?
I was searching LCSC and find this part JW5068A it's from JoulWatt  and can deliver 8A and it's only 0.5$, do we have some parts from a chinese manufacturer with multiple outputs or higher current with similar prices? also in this JW5068A  data sheet there is no info regarding the L and other passives calculations.

[OwO]

Above 10A you probably want to look into designing your own multiphase buck converter using a CPLD, gate drivers, and discrete FETs. It's not going to be any more expensive than using an overpriced all-in-one solution and potentially performs better (if you did your design work well).

[ali_asadzadeh]

Thanks, the space is limited and I do not have time for Designing it right now, I'm looking into cheap solutions,
Also I have found this part
RT6242A, it's a 12A puppy, I have no price info, But maybe it's under 1$

[tszaboo]

Thanks, the space is limited and I do not have time for Designing it right now, I'm looking into cheap solutions,
Also I have found this part
RT6242A, it's a 12A puppy, I have no price info, But maybe it's under 1$

Are you that desperate to design with a 1 dollar chip for a 800 dollar digikey price FPGA? What if you blow up one chip, because poorly documented transients on that JW5068A or just having the wrong capacitor in the loop filter? Are we really having this race to the bottom? Instead of placing a 5 dollar module from TI on it, and concentrating the engineering where it actually matters.

[OwO]

xc7k325t is around $100 on LCSC and cheaper on the market. TI power management ICs are known to be poorly designed and have an above average tendency to blow up: https://www.eevblog.com/forum/projects/psa-do-not-use-the-tps61099-boost-reg-in-your-designs/

[asmi]

That "puppy" can consume up to 12 Amps of power on the Vccint rail, and that rail is also required to be extremely precise (3%) and extremely low ripple (<10 mVpp over 0-20 MHz). No way your cheap Chinese crap is going to deliver this. Nor do you unless you have an extensive experience in DC-DC converter designs. I will just say one word - thermals, that thing is almost never properly designed in by typical wannabe designers.
Also K325T requires paid Vivado license to build designs - do you have that? Or you're going to do it the usual Chinese way - i.e. use pirated version?

[ali_asadzadeh]

Guys thanks for sharing,

Also K325T requires paid Vivado license to build designs - do you have that? Or you're going to do it the usual Chinese way - i.e. use pirated version

I have the license , Please do not worry.

Are you that desperate to design with a 1 dollar chip for a 800 dollar digikey price FPGA? What if you blow up one chip, because poorly documented transients on that JW5068A or just having the wrong capacitor in the loop filter? Are we really having this race to the bottom? Instead of placing a 5 dollar module from TI on it, and concentrating the engineering where it actually matters.

This project is for mass production and we can get it under 100$ from china, so the winner is the one that can do the job @ lower price.

[asmi]

Above 10A you probably want to look into designing your own multiphase buck converter using a CPLD, gate drivers, and discrete FETs. It's not going to be any more expensive than using an overpriced all-in-one solution and potentially performs better (if you did your design work well).

It won't be unless you have a lot of experience in that area. Infact it will probably work much better as a broadband jammer than as a DC-DC converter
So, yeah - do yourself a favor and get one pre-designed by people who know what they are doing.

[OwO]

I've designed current mode DC-DC converters before, it's really not that hard but yes you need to know what you are doing. EMI is my expertise and I have no issues designing circuits with 100dB isolation on tiny 4 layer boards. I also dispute that the designers at TI really know what they are doing since they can not even get a power supervisor working correctly, seem to have trouble designing load switches that don't randomly blow up, and battery management ICs that don't get into a state where it drains the battery.

[asmi]

EMI is my expertise and I have no issues designing circuits with 100dB isolation on tiny 4 layer boards.

We've all seen your "expertise in EMI" on that Zynq board which would radiate like there is no tomorrow Nobody who claims to care about EMI would ever place high-speed traces on external layers.

I also dispute that the designers at TI really know what they are doing since they can not even get a power supervisor working correctly, seem to have trouble designing load switches that don't randomly blow up, and battery management ICs that don't get into a state where it drains the battery.

I use their power parts almost exclusively for years and never has a single issue with them. So I have my doubts that you will ever do a better job at this than they would, especially given what I said above.

[asmi]

Every cheap consumer device (which has to pass the most strict class B EMI) is designed with high speed signals on outer layers. Many motherboard running DDR3 signal at up to 4266MT/s (2133MHz) are only 4L, with 2 center power and ground layers and 2 outer layers, running 128/144 bits of 2133MHz signal.

Do you mean the ones which are certified only when working inside a Faraday cage (a.k.a. computer case)?

[daqq]

Analog has a few very compact modules with the coil included. While they are pricier per chip, you just add a few capacitors and you are done. You save on space and the inductor and some passives.

[OwO]

Those DDR traces are on the same board as a AD9363 transceiver. If there was a noise problem I would know by now. Open up any WiFi router and you will see DDR traces on outer layers. Microstrips on 0.1mm prepreg doesn't radiate much *if* you actually keep ground planes solid and well stitched at layer transitions.

If you do decide to go with an off the shelf module stick with LTC/ADI, they actually have analog design expertise which seems to be absent at TI these days.

[ali_asadzadeh]

Are you that desperate to design with a 1 dollar chip for a 800 dollar digikey price FPGA? What if you blow up one chip, because poorly documented transients on that JW5068A or just having the wrong capacitor in the loop filter? Are we really having this race to the bottom? Instead of placing a 5 dollar module from TI on it, and concentrating the engineering where it actually matters.

If some company can make a chip, they have at least spent a lot on resources, like time and money! EEVBLOG members told me in this thread that making even samples of chips cost a lot
https://www.eevblog.com/forum/projects/fabrication-your-own-chips/

Laptops and other high volume dirt cheap devices use something called a DrMOS, with gate drivers, current sensors and high/low side FETs packaged together. You can buy some AO Semiconductor ones for a few tens of cents per 35A phase block.

Do you have any part numbers, It seems very interesting!

[ali_asadzadeh]

Visit AO's website, and find whatever the part tat you can buy on LCSC and fits your requirements.

you mean https://ao.com/? it's a washing machine website, do you have their website address.
Thanks

[ali_asadzadeh]

http://www.aosmd.com/products/power-ics/drmos

Thanks for the info.

DO they need a controller too? I checked some parts data sheet, and They only contain MOSFET and driver, what do you suggest for the controller?

[TimCambridge]

xc7k325t is around $100 on LCSC and cheaper on the market.

LCSC prices for Xilinx parts are a mystery to me. The XC6SLX9-2TQG144C (a terrible package) is available on LCSC at around $5 (31 in stock), it's $17 on Digikey.

Is that because sourcing this sort of component via LCSC is a bit of a gamble on availability?

[ali_asadzadeh]

I got price info on RT6242BHGQUF from it's company, it's around 0.35USD and it can deliver 12A, I need around 15A for the Vcore of the FPGA, Can I use two of this RT6242BHGQUF  and connect half of the Core pines to one IC and the next half to the another one!

[ali_asadzadeh]

LCSC prices for Xilinx parts are a mystery to me. The XC6SLX9-2TQG144C (a terrible package) is available on LCSC at around $5 (31 in stock), it's $17 on Digikey.

Is that because sourcing this sort of component via LCSC is a bit of a gamble on availability?

you should not buy from Digikey then!

[OwO]

Before designing in any part always check taobao. If there are many sellers selling it it's usually safe even if it's long discontinued. I designed the cmy210 RF mixer into a new design when it was already discontinued at all major distributors, but right now many years later I can still get it on taobao below 1CNY/pc.

No you can not feed different Vccint pins with separate buck converters. The pins are all internally shorted in the FPGA package.
RT6242 and JW5068A are only good up to 6A. The efficiency is 90% and you simply can not dissipate more than about 0.6W in that package.

[OwO]

Have you designed Artix/Kintex boards before? If not, I highly recommend practicing with a Artix XC7A100T. The power distribution layout design is going to be the hard part for most people. See also https://www.eevblog.com/forum/repair/how-to-test-salvageable-xilinx-ultrascale-board-from-ebay/

I will be playing with these soon (please don't go and buy a bunch of them and drive the price up):

[ali_asadzadeh]

Have you designed Artix/Kintex boards before? If not, I highly recommend practicing with a Artix XC7A100T. The power distribution layout design is going to be the hard part for most people. See also https://www.eevblog.com/forum/repair/how-to-test-salvageable-xilinx-ultrascale-board-from-ebay/

I will be playing with these soon (please don't go and buy a bunch of them and drive the price up):

Yes, I have done it before for ARTIX (XC7A100T) and I have used a TI part LMZ31710, So for this Kinetis, I want to use low priced parts.

[ali_asadzadeh]

Unfortunately, Taobao does not open to other people, it needs registration and needs a cell phone for registration, which I do not have a chinese phone number

[tszaboo]

xc7k325t is around $100 on LCSC and cheaper on the market.

LCSC prices for Xilinx parts are a mystery to me. The XC6SLX9-2TQG144C (a terrible package) is available on LCSC at around $5 (31 in stock), it's $17 on Digikey.

Is that because sourcing this sort of component via LCSC is a bit of a gamble on availability?

Or they bought a partial reel remaining from someone, and have 0 guarantee to have the parts again.
Or they are selling it below their cost, because they want to get rid of a stock.
Or they found a reel of it in a corner of a warehouse with unknown origin and want to sell it.
You dont build a business on LCSC. Or even digikey for that matter.

[TimCambridge]

xc7k325t is around $100 on LCSC and cheaper on the market.

LCSC prices for Xilinx parts are a mystery to me. The XC6SLX9-2TQG144C (a terrible package) is available on LCSC at around $5 (31 in stock), it's $17 on Digikey. Is that because sourcing this sort of component via LCSC is a bit of a gamble on availability?

Or they bought a partial reel remaining from someone, and have 0 guarantee to have the parts again.
Or they are selling it below their cost, because they want to get rid of a stock.
Or they found a reel of it in a corner of a warehouse with unknown origin and want to sell it.
You dont build a business on LCSC. Or even digikey for that matter.

Digikey gives you a starting point, then you negotiate - a skill that we all have

I've just looked up the LCSC prices for the old Cypress FX2 ( CY7C68013A, 8051 based! ). There's a price spread of 4:1 between the package options, and all cheaper than list price ! Has this part been cloned?

More seriously, is there a broker service that navigates taobao for foreigners?

[ehughes]

Sorry, putting a $0.50 power supply on an expensive FPGA is the dumbest idea ever.

That is like building a house on a foundation of toothpicks.

Don't give me this is "production/volume" nonsense.    There is no project that Uses a Kintex 7 that is a make/break because you need to save a couple dollars on a power supply.    Someone needs to reassess the economics of the project if you really need to use a shitty DC-DC converter to run a high end FPGA.     

[asmi]

Sorry, putting a $0.50 power supply on an expensive FPGA is the dumbest idea ever.

That is like building a house on a foundation of toothpicks.

I guess that "price is everything" mentality used for cheap crap from China is finding it's way into more expensive products. There is a number of chinese FPGA boards on Aliexpress with underpowered PDS. They can't seem to grasp a simple truth that "saving" a couple of bucks on PDS can end up ruining the whole project. And pretty much on all of those boards (that I've seen or read about) PDS has insane amount of switching noise that is way beyond the spec (the spec is <50 mVpp at 20 MHz BW limit for A7/S7, <30 mVpp for K7, <10 mVpp for MGT power rails). In my experience S7/A7 are not very picky about the noise (except for MGTs), but K7s are.
Another factor that is very commonly neglected is thermal regime. Take any of A100T "modules" that are aplenty on Aliexpress - vast majority of them have zero provisions for any kind of thermal control. The thing they don't realize (or just don't know) - when FPGA heats up, it consumes MORE current, exacerbating the heating problem. This can lead to a thermal runaway in come cases.

[SiliconWizard]

Agreed. Anyway, those big FPGAs are very power hungry and dissipate a lot of heat (at least if you use them as it makes sense to use them), and they are pretty expensive already. I don't really see the point of shaving off a dollar or two for what should be seen as the already upper-end of the series.

Underpowered - and without any proper thermal management - boards with big FPGAs are just deceptive items, close to scam IMO, as you won't be able to reliably use them for anything except the most trivial things (for which again it wouldn't make sense to use such FPGAs anyway.)

[OwO]

Yes but putting a $10 DC-DC module on without any thought is also careless design. If you do that here and there, it all adds up and you'll double your BOM cost. Not to mention those modules aren't always well designed and you will have to qualify them anyway. Experience gained in designing and troubleshooting your own DC-DC solution carry over to all future projects, and can save huge amounts in the long run. You don't want to be the designer that can't build anything and have to rely on pre-engineered solutions for everything.

I'm not going to go engineer my own 5A DC-DC solution because off-the-shelf chips do that well enough at reasonable price, but >10A and the value-add is large enough to justify in-housing it. It will be a project on its own, but if you design a lot of FPGA boards it's a worthwhile investment.

[OwO]

And uhh.. XC7A100T is a cheapie $10 FPGA. Those QMTech boards are overpriced for what they are, and I can tell they are designed hastily (there is no other reason you would use an expensive TI buck converter for 3A output). I will absolutely use a $0.5 power solution for a XC7A100T. I have no reason to believe that the JW5068A is any bit worse than what TI puts out, although I'd only trust it up to 6A because of power dissipation.

[ali_asadzadeh]

I think, I would first evaluate these chips to see what they can do, then If they pass the test, I will add them to the design.

[tszaboo]

                Δ
                |-------------------------------------------¬
Please read this part number somewhere here -----˩

And uhh.. XC7A100T is a cheapie $10 FPGA. Those QMTech boards are overpriced for what they are, and I can tell they are designed hastily (there is no other reason you would use an expensive TI buck converter for 3A output). I will absolutely use a $0.5 power solution for a XC7A100T. I have no reason to believe that the JW5068A is any bit worse than what TI puts out, although I'd only trust it up to 6A because of power dissipation.

[OwO]

I'm replying to the comment about A100T boards being inadequately designed, which are because it's a toy FPGA for playing around.

[SiliconWizard]

And uhh.. XC7A100T is a cheapie $10 FPGA.

I know you guys have incredible prices for Xilinx parts, but the current Digikey price is about 140 euros. I'm sure you can get better prices with high volumes and another distributor, but there's no way us poor europeans can get this for $10.

Oh, and for the XC7K325T that the OP was talking about, it's somewhere between $1000 and $2000 depending on package and quantities. For how much can you get them in China?

[olkipukki]

And uhh.. XC7A100T is a cheapie $10 FPGA.

10$ for XC7A100T ?? 
How much would XC7A35T cost then?

Where I can buy with a proper paperwork make sure these are not be nicked and okay to be (re)exported?

[olkipukki]

On the end of 2014, I bought first Artix-7 chips (XC7A35T-2FTG256C) for ~$36 that available only on Avnet. Eventually, the price has increased bit-by-bit and today a few bucks more in same quantity.

https://www.avnet.com/shop/us/products/xilinx/xc7a35t-2ftg256c-3074457345626042585/

You can buy these from LCSC for $17 (assuming they can supply required qty!)

https://lcsc.com/product-detail/CPLD-FPGA_XILINX_XC7A35T-2FTG256C_XC7A35T-2FTG256C_C181716.html

That's mean - 2 times (!!) cheaper, but not a magic 10+ times... 


Am I missing something? 

[olkipukki]

Buying from LCSC guarantees no fake or bad parts. It doesn't guarantee if it was not illegally imported or exported, nor guarantee if it was acquired from grey market. LCSC only guarantees the part is new, working and genuine, unless LCSC specifically identifies it as authorized distributor of brand X.

OK, understood, thanks

I can find XC7A100T cost $50 on LCSC, still better than 3 times compare to Avnet, but nothing close to $10 

https://lcsc.com/product-detail/CPLD-FPGA_XILINX-XC7A100T-2FGG484I_C410276.html

[TimCambridge]

I can find XC7A100T cost $50 on LCSC, still better than 3 times compare to Avnet, but nothing close to $10 

https://lcsc.com/product-detail/CPLD-FPGA_XILINX-XC7A100T-2FGG484I_C410276.html

And out of stock. LCSC have over 100 Xilinx parts, almost all with tiny inventories - only five parts with over 100 pieces in stock.

[OwO]

There are many stores selling the A100T around $10:

I've bought zynqs from one of these sellers before; they didn't have signs of reballing but are old stock and not in its original packaging. The date codes ranged from 2013 to 2018. The 2018 ones are unlikely to be salvaged at least, I haven't heard of devices with FPGAs get retired after a year. However you will definitely want to bake these before soldering. Salvaged A100T can be had for 40rmb if you can deal with the desoldering yourself.

[olkipukki]

And out of stock. LCSC have over 100 Xilinx parts, almost all with tiny inventories - only five parts with over 100 pieces in stock.

Not wonder, the most distributors have low FPGAs stock, and most stuff RFQ and 3+ months wait time.

Maybe LCSC can restock FPGA, never asked, but they definitely can do this for another components...

[OwO]

XC7K325T on LCSC shows 0 stocked but 140pcs sold in the past month. Most likely people are backordering. Similarly most "correctly priced" SMD capacitors are unstocked half the time, but still show lots of orders flowing. I've backordered chips from LCSC before (no FPGAs yet though) and was usually quoted 1 month lead time.

[ali_asadzadeh]

There are many stores selling the A100T around $10:

I've bought zynqs from one of these sellers before; they didn't have signs of reballing but are old stock and not in its original packaging. The date codes ranged from 2013 to 2018. The 2018 ones are unlikely to be salvaged at least, I haven't heard of devices with FPGAs get retired after a year. However you will definitely want to bake these before soldering. Salvaged A100T can be had for 40rmb if you can deal with the desoldering yourself.

Please share your good FPGA sellers in here(thier Taobao or other sites links), I have Chinese friends that will buy them for me directly from them