ok, so the memory fet idea sucks.

[Brian Marshall]

so what better ideas are there, for "memory" without getting in to really complex digital stuff and dedicated IC's?

anything?

[Peter Snowberg]

I don't think the memory FET idea sucks at all. It just "suffers" from analogness.

Without doing digital, you won't find a circuit that is able to keep a value in memory that is stable for any length of time.

Well.... there is one way without going digital, but it's even more complex than using an A/D, a latch, and a D/A. A couple years back, I was restoring some vintage 1954 U.S. Army gear that was used for gun directing and missile guidance. Their solution was to use the input signal to control a servo-driven pot. If the input went wacky or away, the servo would halt and the pot reading remained the same. It originally came out of Bell Labs in 1943 from a phone line testing chart recorder of all things.

Sorry.... but if you want consistency.... digital is it.

Take care,
-Peter

[Brian Marshall]

Somehow i knew it would be you that would answer this, and i knew you would tell me something i didnt want to hear.

any good books for digital stuff that is easy to follow.

I don't think the memory FET idea sucks at all. It just "suffers" from analogness.

Without doing digital, you won't find a circuit that is able to keep a value in memory that is stable for any length of time.

Well.... there is one way without going digital, but it's even more complex than using an A/D, a latch, and a D/A. A couple years back, I was restoring some vintage 1954 U.S. Army gear that was used for gun directing and missile guidance. Their solution was to use the input signal to control a servo-driven pot. If the input went wacky or away, the servo would halt and the pot reading remained the same. It originally came out of Bell Labs in 1943 from a phone line testing chart recorder of all things.

Sorry.... but if you want consistency.... digital is it.

Take care,
-Peter

[Peter Snowberg]

Sorry to be the bearer of bad news.... well... to a digital nerd like myself it isn't all bad, just different.

Normally this sort of thing is done with a microcontroller and some programming. The good news is that this can be done without needing a microcontroller.

These days the cheapest and best way to go is to use serial I/O on the converters. If you can send a "take snapshot" sort of digital signal, it's really easy to wire two converters to each other along with a simple oscillator (like a 7555) and maybe a counter chip to control the process depending on the converters used. The D/A will absorb the value from the A/D and then hold it.

If you can give me some more info on your application I may be able to find an easy solution for you with something as simple as two or three 8 pin chips. PM me if you don't want to go public. I just need to know some basic stuff like minimum resolution and input/output voltages. 12 bits will get you 4096 values which is enough for most apps but for fine control you might want to go with more bits like 16. It depends on what you're controlling.

Sorry.... I don't know of any books to recommend. When I'm trying to figure out something new, I go right for the datasheets. I've already wasted the time so you don't have to.

Take care,
-Peter

[gez]

Brain, it doesn't necessarily suck , but perhaps you could enlighten us a little and explain what you're trying to do? (you don't have to go into huge detail and give away anything).

Perhaps it is fine for what you're trying to do?  Things like sample and hold techniques have been used to good effect in many analogue circuits.  Even if it is inappropriate, perhaps one of the gurus here can come up with a solution that you DO want hear!

[Brian Marshall]

Sorry if i was unclear.  

I dont really have an aplication in mind, just a general idea.

Say i have a pedal with 6 knobs, and i want to make it have 2 knobs... one 6 position rotary switch, and 1 knob to adjust the selected parameter.

In general this is just something i want to know a little more about.  

Bri

[R.G.]

At the risk of being repetitive - why not use a PIC?

I realize that you said you don't want to use complex specialized digital chips. But PICs are complex **unspecialized** chips that could well be characterized as "whatever you want" chips.

An 18 pin PIC costs under $3.50 in ones, and if you can do it in six I/O pins or less, you can get a baby eight pin PIC for $1.90. You can program them in C, or in Basic, or in the assembler language, which is only 35 instructions. You do have to have a programmer, but that can be as cheap as a few bucks for the basic PICs.

What I think you were doing is to have the "cap" get the value of a voltage from a pot and then hold it. In a PIC, you would use the pot to set the current into the capacitor and run one I/O pin to the junction of the R-C. To read the pot, you set the I/O pin to "output", pulse a "0" long enough to run the cap to 0Vdc, then set the I/O pin to input and read the cap voltage as though it was a digital input. The PIC input is a schmitt trigger, so it will have a definite point where it flips from 0 to 1 as the voltage on the cap ramps up through the resistor. The PIC reads the time involved in ramping up, and can subdivide the time into one of 256 values.

Then to output the value, you set on the of the internal timers to put out a PWM voltage with the timed value, from 0 to 255. That PWM value goes into a simple R-C filter and you have a stable output voltage.

This process isn't fast enough to digitize audio, but it's plenty fast enough to digitize control signals.

It's not widely appreciated, but a PIC is often the *cheapest* way to do a lot of controls, if they're more complicated than a simple pot.

But I'm preaching again...

Sorry...

[Brian Marshall]

I'm scared of programming  :cry:

not really.  i used to be pretty proficient with basic long ago.

seems like a lot of work though.  im sure once you get started it aint that hard though.  Just gotta open my mind a bit and learn.

[Peter Snowberg]

Rev. R.G. is big on the PICs with PWM, but I tend to shy away from that approach because of the noise issues and the resources it requires in the already limited microcontroller (timer....interrupt...). I do however think that using a PIC is a GREAT way to go.

If you want to make adjustments to several parameters, these days you can get multi-channel D/A converters which sure cuts down on cost and board real-estate requirements.

R.G. is talking about using an "integrating" A/D to get the pot setting. This works well for low resolution signals, but if you want to affect something like the speed of an audio oscillator, you'll quickly find that 8 bits is not enough to give you analog feel. You can get better resolution with a slightly different technique using the schmidt trigger input called "dual slope conversion". I'll leave research on that to you if you want to persue it. (Here is a Circuit Cellar article on the subject that's worth a read)

I would tend to just use a real-deal A/D with good resolution, but that's because I'm lazy and I want results that I can solidly predict from a spec sheet. Dual slope converters are a VERY good thing to check out though. Using a PWM output and any integrating converter for A/D on the same chip is just asking for a headache.

With a single channel A/D, a small pic, and a multi-channel D/A, you get a complete solution in three chips and NO PWM noise to contend with.

Even something like a 8 pin PIC could be used here if you use a pseuo-analog approach for the rotary (resistors between each pole so that position 1 is like 10K, 2 is like 20K and so on) and then use a dual channel A/D. If all the proramming is just shifting bits from external converters the task gets MUCH easier. That also gets you down to the sub $2 PICs.

If you use a PIC with EEPROM you can also save the settings and recall them with the touch of a button.

Take care,
-Peter

[puretube]

looks like we`re back at the digi-pots discussion from a little while ago...