A/D's and LED decoder/drivers

[analogman]

List,

First let me say hello as I am new to this.

I'm trying to display a YIG tuned filter's tuned frequency(2-18Ghz) as a product of the control voltage (0-10V). The problem is I need 6-digits (7-segment led's) to get the resolution I want. I guess I will need at least a 24-bit A/D and an LED decoder/driver that will interface the LED's. I also want this to be stand-alone, not microprocessor controlled. All I can find so far are mux led decoder/drivers that will do the job. Also, all I can find are serial A/D's that are at least 24-bit. Is there any way I can use two 12-bit A/D's? I need parallel operation for this to be stand-alone don't I? I know the wiring will get complex but like my ID states, I'm basically an analog man. I don't have the capability at work to program chips. Any help will be greatly appreciated.

 

[itsmoked]

Gesh... This is just plain wrong... You can do this with a double hand full of parts or with a $7 micro and few parts.

If this is a one of-a-kind for you to use, just by an HP voltmeter $800 and be done with it.

If this is for production you really should use micro. You can get some very inexpensive development tools that will
get you good results.

Even if you were to do some sort of strange venier scheme to get two A/Ds to work together you'd still want a micro to supervise everything.

Keith Cress
Flamin Systems, Inc.- <

http://www.flaminsystems.com>

- kcress@<solve this puzzle>

 

[IRstuff]

Unlikely you can do it with 2 12-b A/D's. The integral nonlinearityis probably not good enough to make it work.

Can't even see how you are controlling the frequency to 1 ppm without digital controls. And if you have digital controls, you should drive your display directly from the digital

TTFN

 

[MacGyverS2000]

Also, 6 digits of resolution is 20 bits... bump it up another bit or two for comforts sake, if it makes you feel comfy.

Dan
Owner

http://www.Hi-TecDesigns.com

 

[analogman]

This is a one of a kind deal, not production. We usually buy cots for any critical test requirement. This is just to help get the YIG tuned close to freq and then for accuracy use a spec an. I guess I need to give the big picture. I'm taking the YIG tuned voltage which is tuned using a precision power supply (accuracy to 1mv), feeding it to a precision zero drift opamp, scaling the ouptut to 1.666667 to 1. This will give me the equivalent frequency the YIG is tuned to in dcv. This I want to feed to a 6 digit led display. I thought the best way to do it would be to a/d the op amp output. I know it's brute force and ignorance but it's the only option I have. Concerns are obviously thermal noise and resolution. Thanks for the replies.

 

[cbarn24050]

24 bits!!! you haven't got a prayer. 16 bits is nigh on impossible.

 

[Comcokid]

I once used a circuit that took a A/D output through two EEPROMs, and back out a D/A to do custom linearization. All it needed was a clock signal. I don't remember if it was 12 or 14 bits. I no longer have access to the information or remember what devices were used. Something like this would be easy to modify by taking the EEPROM outputs to LED drivers.

So, the answer is: Yes, it is possible. Probably somewhere out there in a app note or design idea such a circuit exists probably for pressure sensor linearization. The engineer who designed that circuit frequently sourced the circuits from published sources.

However, 24 bits is another issue, unless you are throwing away the last 8 bits.

 

[zeitghost]

Seems to me that itsmoked hit the nail on the head: buy a 6.5digit voltmeter. Less stress & more likely to work. Why reinvent the wheel?

 

[IRstuff]

Still seems weird. You are running this power supply from 1mV to 1000V and all the way through an precision op amp?

Why are you not using a frequency counter? If you're tuning a frequency, why not measure the frequency instead of some "equivalent?" There are just so many error sources to account for that your ultimate accuracy is probably no better than 0.1% Just your voltage scaling error alone...

As for A/D, there are >20-bit devices out there:

http://www.aldinc.com/pdf/mb1s.pdf,

a 5 1/2 digit DVM chip

http://www.chipcatalog.com/Cat/279.htm

http://www.cirrus.com/en/pubs/proDatasheet/CS42406_PP5.pdf

TTFN

 

[analogman]

Well, thanks to those(few)who had words of encouragement. I thought this was a forum for help, not one to get pummeled when your down. I understand those of you who want to do this a different way, but you don't understand the conditions I test in, budget restraints, and other factors that prevent me from doing what you've suggested. That's why I'm here in the first place. I know most of you are probably a hell of a lot smarter than me concerning this issue, but please consider why I'm asking what I'm asking before you post. And NOTHING is impossible, I'll just keep plugging away until I get it right. I've never failed before and don't plan on starting now. Thanks anyway,

 

[IRstuff]

A $4200 12-digit Agilent frequency counter (

http://www.hensleytechnologies.com/detail.aspx?ID=1007)

is equivalent to 84 hrs of $50/hr labor. I can't imagine that either you or your company believes that you can design, put together, and calibrate an equally precise and accurate approach for 2 weeks equivalent labor cost.

Part of engineering the appropriate engineering solution is analysis of alternatives. If you are unwilling to entertain alternatives, then you do your company a disservice, particularly if the resultant product may be more widely useable, more accurate, and has NIST-calibration traceability potential.


TTFN

 

[analogman]

Once again you fail to understand. I've got about 2 million or more dollars in Agilent, Tek, Lecroy, SRS, Miteq, Anaren, and many others in my inventory available for use. This is a sideline project I work on in my spare time, mostly at home. I wouldn't even consider wasting my time (or my employer's) if this were anything other than that. I design (mostly analog) circuits in my spare time to mainly interface test apparatus, free up really expensive equipment so they can be used elswhere when needed, and for various other reasons. Some of these I do to get back to speed on circuits I'm seriously rusty on (like digital).I use them as learning tools to expand my capability for potential future requirements. This is not a mission critical project, but it's a useful one when done correctly. I want to get this circuit as accurate as possible but it doesn't have to be perfect. That's why I came here, to get suggestions or advice on the circuit I want to build, not the one YOU want to build. I can see there are too many people here looking down on me from atop their high horses though... sorry I asked.

 

[IRstuff]

Sorry you asked too...

Had you been up front about the purpose of your project, you wouldn't have gotten half the responses that you did.

TTFN

 

[cbarn24050]

Thats just typical, some bozo posts an impossible problem, people try to help him with possible alternative ways to get somewhere near and all he does is complain.

 

[zeitghost]

I've got an AVO (VOM) with a bent pointer.

Would that help? )

 

[coconutalley]

Ok let's try to work with your idea:
If I understand correctly, you want to take 1V and A/D
convert to 24 bits. That means resolution of 1/16.7 million which is basically microvolts level. This is down in the noise. Most of your resolution would be wasted on noise.
You might want to amplify it up so that you can get out of the "noise floor" and use more of the resolution.
Say, you succeeded.
If you are not a digital guy [I understand] and don't want to use a micro, the next thing you can do is to use discrete logic to build a decoder that will take the ADC output and translate that to a 6 digit number to be displayed. You'll need the decoder to translate the number to those 7-segment LED drivers, which will display it.
Good luck--it can be done, the challenges will: get out of the noise floor, and then that decoder circuit.

 

[itsmoked]

With the money and abilities at hand, it cannot be done... PERIOD.

It's not that you are incompetent or lacking the will.. It is just horrendously hard. It is on the level with walking to the moon. It ain't gonna happen and isn't even going to come close to happening. I'm sorry. The best you can hope for is to buy or assign an expensive piece of equipment to it and then you will only get near not close to the indication you desire.

Keith Cress
Flamin Systems, Inc.- <

http://www.flaminsystems.com>

- kcress@<solve this puzzle>

 

[Skogsgurra]

analogman,

I need to ask if you have ever tried to make an error budget for what you are about to do?

First, there is noise (OK, you can filter that down to ridiculuously low levels - but that will mean measuring for very long time).

Second, you have thermal drift. Given the environment you are telling us that you have (home lab), I cannot imagine that you can keep temperature within a few tenths of a degree (F or C), which would be required not to have problems with the conducting and soldering materiels at hand for someone working at home.

Then, there are all kinds of external electric influences. Mostly radio and TV transmitters, but also CB, pagers, PLC (Power Line Communication), radar and lots of such stuff. I guess that the oscillator you work with also outpts its share of radiation.

So, if you divide your error budget fair between these you will have around 1/4 of your accuracy lost in each of them. That effectively rises resolution and accuracy demand another two bits or so.

As most poster say: It is the wrong way to go. It cannot be done.

We do not say that to put you down, but to help you (from hurting yourself, perhaps). If you study these fora, you will find that the tone is kind and polite. We have rules that allow us to keep it that way and I find your childish reaction ("I came here, to get suggestions or advice on the circuit I want to build, not the one YOU want to build. I can see there are too many people here looking down on me from atop their high horses though... sorry I asked.") very diverging, to say the least.

We are seasoned designers from all walks of life and technology that gather to share insights and help each other and others. You included. But we do not accept flames as soon as things don't go your way. Ever worked in a lab? Ever got advice from someone that knew the stuff better than you? Did you accept it or did you tell that guy to shut up? Just asking...




Gunnar Englund

http://www.gke.org

 

[coconutalley]

Analogman,

I tried to help you but...it only brought in more reasons not to do it your way.
The gents are correct in that you would not achieve good results. The alternatives suggested are much easier, quicker and cheaper.