DA Wandler

Well if you just leave the D/A converter alone the output value will remain constant for infinite periods of time.

Thats the definition of "optimal configuration"

Practically it depends on your requirements.

No.

The charging time is defined by the D/A converters (or some buffers) maximum output current (some mA)

Assuming you disconnect the D/A after the conversion the discharge current is given by the capacitor leakage, the opamp input leakage and the leakage of the switching element. These leakages can be in the pA so the time constant is much larger here.

Just tell us a little bit more what you want to do.

Georg

Hi Veronica,

Holding analog values for that long to any reasonable precision is going to be tough. Not that it can't be done but you are looking at film capacitors and very high impedance circuitry.

An easier solution would be to use a multiple DAC chip that has registers for each DAC. So, you just clock in each data word and let it sit there. Then select the next DAC and clock the next value into its register. That scheme will hold the outputs constant until the cows come home or until new data is clocked in. Such converters will have a clock and a 'selector'. For eight DACs there would be three binary selector lines, for sixteen it would be four and so on.

Look around at

for suitable devices. It isn't the only company but they usually offer a large variety.

Regards, Joerg

Am Thu, 11 Nov 2004 20:08:42 GMT schrieb Joerg :

I would also strongly suggest such a solution. Why store the analog values, if you have them already in digital form? There are many circuits who digitize a signal just because they have to store it somehow - storage is one of the things which are best done digital. Look also at

they have some nice Multi-DACs.

--
Martin

Thanks for your numerous answers so far. For a better understanding let me elaborate on my intention. What I want to do is to handle several outputs (with the analog representation of the digital value) with just one single D/A converter. That means: feed the digital values through a single D/A converter and switch the converter output to one analog hold circuit per channel. Therefore the goal is to hold the analog values! The analog values to be hold are DC, that's true. Again, would it be advisable to use a capacitor and op-amp? How should I dimension the capacitance and how can I affect the droop? Is it realistic to expect virtually no droop? Isn't it, that with a large time constant the charging time would be endless? Maybe the solution is nearer as I can see? Maybe there is another way to solve the problem. But this "one D/A converter for multiple output channels"-configuration should be seen as basic condition!!!

Many greetings, Veronica

[...]

It is possi- but not sensible. One D/A for each output channel is easier and cheaper and more realiable and smaller than an analog hold if we are talking hold times in the minutes range. You did not specify a precison, so a more detailed discussion is futile. Also you give only a partly specification of what you intend to do, so noone can really help you. Somehow the suspicion arises that you try to get an exam/homework question solved...

--
Dr. Juergen Hannappel          http://lisa2.physik.uni-bonn.de/~hannappemailto:hannappel@physik.uni-bonn.de  Phone: +49 228 73 2447 FAX ... 7869
Physikalisches Institut der Uni Bonn Nussallee 12, D-53115 Bonn, Germany     
CERN: Phone: +412276 76461 Fax: ..77930 Bat. 892-R-A13 CH-1211 Geneve 23

Veronica Matthews schrieb:

Hello,

that is definitly the wrong way! Use several D/A converters with an integrated digital hold register and you have no problem at all to hold the value even over a very long time. There are ICs with several such D/A converters on one chip.

An analog sample and hold that should load a value within 1 microsecond and hold it over 5 minutes is very, very difficult because of the extreme relation of sample and hold time, 1:3*10^8.

Bye

Am Fri, 12 Nov 2004 09:06:56 +0100 hat Juergen Hannappel geschrieben:

I think, she would like to do something cheaper than the others - but as we discussed already, she is doing that the wrong way :-) If it would be homework, perhaps we already gave the answer: Its not sensible :-)

--
Martin

Just to put my aim in perspective: I'm neither trying to fool you nor trying to get my homework solved (like a given individual presumed). Why I am talking about a basic condition with respect to the "one D/A converter for multiple output channels"-configuration is that this single D/A converter already exists in hardware. It is there, physical, for me to touch, already bought... And now I want to use this very D/A converter to feed several output channels. Of course I could buy a DAC for every channel but that's not my intention. The hardware setup does not allow to solder other devices on the board. So PLEASE just take it as it is! I want to solve the problem that way. So don't try to proselytize me like that jehovah's witnesses guys... ;-)

Hope you come up with more constructive suggestions!

Veronica

Hi,

Most of that "guys" here are thinking, that your problem ist not solveable by this way.

Example: Take a 10µF film capacitor. You have to charge it in 1µs to 5V (1MHz sample-frequency).

I = C * dU/dt | dU = 5V I = 50A You have to charge the capacitor with 50A. Do you think, this is possible? Not with a simple Op Amp.

Now we guess that the leakage current ist about 1µA

dU = I * dt/C |dt = 5minutes =300sec dU = 1µA * 300sec/10µF = 30V !!!!

In 5 minutes, the voltage decreases 30V. If you find components with a leakage currents of 1nA it still will be 30mV.

Do you think this is possible?

I think you won't get the answer you want to hear in this group.

Take a fast AD-converter, sample your signal from the existing DA-Converter, save your analog values digital in a fast microcontroller, FPGA and then use several DACs.

Michael

Veronica Matthews schrieb:

Veronica Matthews schrieb:

Hello,

we all tried to convince you that the use of sample and hold is the wrong way, this was very constructive.

If you really want to build a sample and hold which can sample in a microsecond and hold for 5 minutes, the best way would be to use an A/D converter, a hold register and a D/A converter.

Building such a sample and hold with only analog parts is nearly impossible because the requirements on isolation resistance, input current of an op amp and the resistance of an electronic switch in the closed and open mode are all far to heavy.

Having bought some parts already is no excuse for insisting on the wrong way to solve a problem, you would only waste a lot more time and more money.

Using several D/A converters is the easiest, cheapest and fastest way to solve your problem. It also gives a better performance and stability over time and temperature than any analog solution.

Good luck and bye

Circuits where one D/A was multiplexed to several S&Hs were published in the 80ies: Döpfer "256-Kanal-Analogausgabe für uP-Syteme" Franzis-"Elektronik" about 1986. Ich can scan that article, but there isn´t much to it. A RAM is continously read out to an 8 Bit D/A AD558 that updates simple S&Hs built of CD4051 and TL 084. There isn´t much said about accuracy, but 8 bit should be possible. The simple S&H will not "jump" from 0V to 5V, rather reach the final value more slowly, but that may be ok in many applications.

Then it was common to mux several analog inputs to one S&H followed by an "expensive" 12 Bit sucessive approximation type A/D. So the idea was pretty obvious. And anyone knew how to fix S&Hs.

MfG JRD

Hi Veronica,

The folks here aren't trying any proselytizing. They are right, you can't hold an analog value in a capacitor for five minutes. The droop will just be too much no matter how much you spend on that cap.

Ok, here is another suggestion. I know that's cheating but it will be nice and small. You only have to be able or find someone to program a micro controller: Take a micro controller with a built-in ADC. TI makes these in their MSP430 series but it is best to pick one that you or someone in your lab is familiar with. If you have never programmed a uC but know someone who can you'll be amazed how fast they can whip up working code for this.

Now you can measure and store whatever you want, for five minutes, an hour, or until next year. If the uC doesn't have a DAC (mostly they don't) the output can be through a PWM channel followed by an RC lowpass. If you need all analog values present in parallel that can also be done using a simple multiplexer such as a CMOS "8 in 1" chip. This needs three selector lines and the PWM routine would work 'round robin'.

Regards, Joerg

"Uwe Hercksen" schrieb:

extreme relation of sample and hold time, 1:3*10^8.

Hello,

1. if you have a 1MHz-processor, then the time of voltage-output is more than 1µs.

1. it is possible to store the voltage for some ms in a cheap sample and hold circuit. This is enough to feed the following circuit.

2. look here:

you can store the voltage for more than one hour (!) with an error of only 5mV, with old-fashioned analogue parts of 1985! Try newer ones! So your problem is solved. (I hope, somebody will translate the text for you.)

Martin Siegwarth

Am Sat, 13 Nov 2004 13:11:14 +0100 schrieb Martin Siegwarth :

OK, this is a really fine S/H circuit. I just saved the Artikel for possible future use. Although normally I just tell the programmer (Person, not device :-) ) to tell the Microcontroller to hold the value. And with your two stage concept it seems possible to solve the problem of veronica.

--
Martin

Ich habe mal eine Dunkelkammer-Belichtungs-Automatik mit einem LM 3900 gemacht. Ein 10µF (non-Elko) wurde in einer abgleichbaren Schaltung (Seite 16 im Daten- blatt) so hingetrimmt, dass er weder auf- noch ab- wanderte. Die anderen 2 (Norton-) OpAmps wurden als Komparator und Mess-OP's genutzt. Das als PDF hier liegende Datenblatt sagt zwar "Februar

1995" aber ich bin mir sicher, diese Schaltung vor 1975 gelötet zu haben.

Als LDR nutze ich einen speziellen grün-empfindlichen VALVO-Typ, den es nirgends zu kaufen gab. Aber ein Be- such direkt in HH ['Chile'-Haus, Burchardstrasse (?)] brachte mir den Kontakt zu einem der dort arbeitenden Ingenieure und damit ein paar 'Muster'...

Mit nostalgischen Gruessen, Holger