Microcontroller suggestions?

Are you asking for 4 actual ADCs or 4 channels to a single ADC? I haven't ever noticed a micro with four separate ADCs.

If you are looking for channels to an ADC, then I think you have some options. Even one I know of with a 16-bit that operates at 1MHz.

Jon

I don't know of any microcontroller that would have more then two independent ADCs. However, TMS C28xx has two ADCs each capable of sampling at 12MHz with automatic sequencing of the inputs; the only problem is lousy accuracy.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Neither did I until today:

Pricey though... and only 1KSPS.

Hang on a quad SPI ADC . Processor choice is almost academic then....

Not really. 1M * 2 bytes *4=8MB/s = 64MBit/s.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

On a sunny day (Fri, 1 Oct 2010 17:05:39 +0100) it happened "TTman" wrote in :

The only ADC type I could imagine that would need more then one, as no sampling is needed, is the so called flash-converter (basically a lot of comparators, 256 comparators for 8 bits). So putting 8 of those at 10 bit on chip would take up a lot of silicon :-) In any other case sampling is used by means of CMOS switches. If one MUST sample at the exact same time, you could use some external CMOS switches, store in an external cap, and then use a normal ADC to sample those caps in a sequential way. Maybe use buffers too.

Sink0 had written this in response to

: The processor is not that good but DSPIC got 4 S/H for 10 bits ADC, and according to its datasheet it can sample 4 channels simultaneously...

##-----------------------------------------------## Delivered via

Electronics Enthusiasts' Community of the Net Web and RSS access to your favorite newsgroup - sci.electronics.design - 427146 messages and counting! ##-----------------------------------------------##

Offhand I can't think of any processor that could do much useful work on

4Msps, even if it could get it in the door with on-board hardware.

I think you need to have an FPGA in the mix, connected to an appropriate ADC.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

Some processors include a DMA capability and some applications do NOT require a sustained data rate.

I require a 1MHz to 1.5MHz ADC for a particular application. In this case, I'm using a 16bit. The processor the ADC is on is a Cygnal (sorry, SiLabs) 8051 core -- which is downright lousy at memory to memory copies in software. And the DMA unit won't do that, either. Just ADC to memory, which is all that is seriously needed for my application.

Slow CPU -- certainly no DSP. Yet I need to do high speed bursts of about 5 millisecond durations and do them up to 10 times a second.

Which I'm quite able to do with a basically lousy at many things, 8-bit ALU, register-starved, poorly designed for DSP applications, 8051 core.

(In another application, I need simultaneous sampling on 2 channels as a slight skew affects the resulting measurement. But not two ADCs.)

Your point is well-taken, too. I just wanted to point out that there are cases where 1MHz, perhaps even with 4 channels, might be reasonable with a low-brow micro and no FPGA and FIFO queues in the mix.

Jon

On the contrary, it is not uncommon to sample at that kind of rate to get average PWM current or something like that. 4Msps is within the capabilities of the low end DSCs.

1. FPGAs are expensive.

1. Every other programmable device on the board is a pain for development and manufacturing.

I use FPGAs only upon careful consideration.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

But not with much math behind it, at least if it's sustained.

I get your point, though -- if all you're doing is running a motor, then it may make sense to expend that much processor time on doing an average. And in this case that's certainly "useful work".

FPGAs are expensive and much harder to program than processors. They can also save loads of money and grief if the alternative is to try to shoehorn an application into a processor when it just won't fit.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

Look at that.

a ATmega16..

I think that is the correct one, it has 8 10 bit A/D converters on board.

if you browse around the on Atmel location, you most likely will find smaller family like the AVR's that could have only 4 inputs..

These processors work very nicely...

I have one that has 11 ADC's I just can't think of the part # at the moment..

Jamie..

"Jamie" wrote in message news:yWupo.20584$ snipped-for-privacy@newsfe24.iad...

ATmega* have 8 multiplexed inputs feeding exactly one (1) ADC. And it's quite slow, ~100kSa/s typical.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

that maybe the case but I would bet it could be suitable for the OP's app.

I do think it depends on the actual clock speed the chip is operating at. One could operate the chip at lets say 10 mhz and scale that down to

1 mhz if that is needed for something else...

If this is to be used as a signal sampler that has to be synchronize like a digital scope at the cutting edge of it's band width? I think the approach is off on the wrong foot, however, one could simply make up for the offset with some shift buffers in code to correct for the multiplex problem. Of course, the uC needs to operate fast enough to process all the other functions between ADC sample readings so that staggered sampling does not take place..

I suppose a uC with fewer ADC's onboard may behave more efficiently if the uC just reads all of them regardless if they're are all enabled in the hardware regs.

Have a good day..

How much dedicated hardware is there on video A/D and D/A cards ?

At standard resolution, one channel is sampled at 13.5 MHz (Y) and two at 6.75 MHz each (U/V).

Hi Vladimir,

I am looking at STM32F103R for a new design. It has 3 independent 12 bit

1MHz ADCs. Pretty nice looking chip, still not 4 channels though.

--

John Devereux

Interesting. I'm just starting to play with their STM32F100RB (that cheap $10 board they offer.) Haven't yet looked at the 103.

Jon

On a sunny day (Sat, 02 Oct 2010 10:16:33 +0300) it happened Paul Keinanen wrote in :

Video AD is usually one chip these days. On my Asus Nvidia card there is a Philips chip, it takes composite in, and bit-blits RGB into the screen, without any processor cycles, except for setup.

t's

.

Simultaneous A/D in necessary in this app because. . .

I am detecting a back scatter event with four separate photodiodes to capture four different scatter angles. Currently I am using a legacy design from our licensor which uses the 8051F061 by Silabs. The original design only acquired from two channels so "no big" as that is what the 8051F061 just loves to do. To get the latest four photodiode eval platform going I simply used two of the same legacy boards. I believe we have the triggering problem solved (Silabs tech support freakin' rocks BTW) but getting the results from the two 8051F061s to stay matched up is quite a challenge.

The A/D conversions are sent serially to two separate serial to USB converters and then pulled into a C# application. Sometimes they are off by one ringy dingy sometimes they are off by seven ringy dingies ;-) I added some Microchip SPI RAM to the circuits to act as buffers just in case this sort of weirdness happened but still wonder if the C# program will read these in with out mucking it up. I guess since the C# program can read the SPI RAMs one unique address at a time that should work. We were hoping we could skip using them but I guess we're gonna have to now. ;-) To bad I just used one of the pins of the SPI interface to trigger the "slave" 8051F061. Surface mount soldering without a microscope on .050 pitch is a word that rhymes with pitch.

Sorry to be such a "noob" and wring my hands here in public but I am more of technician than an electrical engineer and until we get the budget to hire one I kinda gotta go with my guts on this and apply bandaids as to the gushing wounds as they occur.

Thanks so much to everyone in this group who have let me post here and help me out.

Best regards, Ed V.