"Inverse" VDR

[benta]

Hi:
I'm looking for a component that probably doesn't exist, but I'd like your input all the same.
Application is a tachogenerator for a motor with a wide rpm range.
The TG is driving two optocouplers with antiparallel LEDs for isolated mesurement (forget about the secondary side).

The problem is, that at low rpms the output voltage from the TG is only 3 Vpp, meaning that I need a rather low resistor value for the LEDs.
However, at max. rpms, voltage from the TG is 60 Vpp and the chosen resistor value pulls a lot of current from the TG.

Question is, does something like an "inverse" VDR exist, meaning a resistor the increases in value as the voltage over it increases?

Thank You,

Benta.

 

[BrianG]

It always helps to provide full information of the intended application. You say "forget about the secondary side" but clearly if your tachogen has a d.c. output it produces a linear output voltage proportional to r.p.m., so how do you intend to compensate for your special non-linear resistor or the non-linear transfer function of the opto-coupler itself?

If your tachogen has an a.c. output then perhaps you are using the a.c. frequency from the tachgen to measure its r.p.m. via some sort of counter? If so, to limit the dissipation in the optocoupler and resistor, a good way to solve this is to use a constant current source. This will give a constant current over a wide range of supply voltages and may have the advantage of producing a square wave into your optocoupler input.

 

[IRstuff]

That's known as a constant current source

TTFN

 

[nbucska]

BENTA:
Please read my FAQ !!!!!!

if you divide the TG output by 12, you get 1/4 to 5 volt
which seems to be a reasonable range.

What is the RPM range and the required accuracy?
Accuracy vs RPM? ( do you need same at low speed? )

Do you need position or only speed control?

Did you consider using digital encoder instead of TG ?



Plesae read FAQ240-1032
WEB: <

http://geocities.com/nbucska/>

 

[benta]

Sorry guys, rereading my post I see I took something for granted:
I'm using the tacho-generator for frequency measurement. It's an 8-pole tacho on a DC motor (motor and tacho are given, it's an existing machine).
Frequency generated by the tacho is 50...1100 Hz, I'm not concerned about position, only speed. Through the optocouplers I have a frequency counter on the secondary side scaled to show correct rpm.

IRStuff: Yes, a constant current source was my first take also. However, I run into problems on low rpms, where there is not enough "headroom" (voltage) to drive both CCS and LEDs. I've looked at using JFETs for current source, but due to the fact that I'm measuring AC I run into problems. Two JFETs plus two diodes would be a solution, but again I don't have enough voltage. Also finding 60+ V JFETs is not easy.

This is the background for my request for a different solution/component.

Thanks,

Benta.

 

[IRstuff]

Then you need have boost for the low voltage and clamp out the high voltage condition:

http://para.maxim-ic.com/results.mv...(V)<!--#I-->&av_8=5.0&op_8=ge&tree=master&p=1

TTFN

 

[benta]

IRStuff, I'm not quite certain I understand. Should I build an extra isolated power supply to boost the output from the Tacho?
The tacho generates enough voltage and current to drive the LEDs, it's just that at the higher voltage the output current becomes too high.

Benta.

 

[BrianG]

OK, not a single component solution, but how about using a depletion mode FET? The idea is to use it in conjunction with a diode and capacitor to store the instantaneous r.p.m. voltage, plus a resistive divider to apply this voltage to the gate of the FET. As the tacho speeds up the FET turns off progressively (might need a zener to limit the range, together with some experimentation with the resistors, to get the gate voltage range correct)

Supertex DN2xxx and DN3xxx series have breakdown voltages of several hundred volts and on resistance at VGs = 0V of 20-40 ohms and can handle a few hundred mA Id.

 

[ijl]

So, basically you have an AC signal with a varying frequency and a varying amplitude. The amplitude is too low at low frequencies, and too high at high frequencies. Correct? If so, then this is a common problem in electronics.

A simple solution might be the following: Take two small-signal diodes (or two red LEDs!) and a resistor of about 10 - 50 kOhm. Connect the diodes anti-parallel. Connect one end of this anti-parallel diode-combination to the ground of the tacho, and the other end to the resistor. The other end of the resistor is connected to the signal output of the tacho. (If the output of the tacho is symmetrical, then it does not matter, in which way the diode-resistor combination is connected to the tacho.) The diodes will limit, or clip, the amplitude of the signal from the tacho, so that the amplitude over the diodes is about 0.6V (or about 1.6V for the LEDs). This clipped signal can then be amplified to the desired amplitude with a simple amplifier, using for example an operational amplifier.

The output of the amplifier may be connected to the optocoupler using a suitable resistor. Depending on the amplifier, the optocoupler, or the levels of the supply voltages it may also be possible that you have to add some other components (a transistor, for example) between the amplifier and the optocoupler to match the signal from the amplifier to the input of the optocoupler.

 

[kontiki99]

I think they used to use ballast resistors in old auto ignition circuits. These were designed to heat up and increase the resistance in a predictable way as the voltage went up.

Some of the early analog designers used light bulbs in circuits to do this same thing in oscillator circuit feed back loops I believe.

 

[benta]

ijl:
Yes, your problem description is correct, though I do have just enough signal at the low voltage level, it's the high level that bothers me.
Clipping and ampifying is of course an option, but I really do want to avoid having to power the tacho side if possible, as this makes the circuit much more complicated.

I've now found a couple of 100 V JFETs. Shorting gate and source should give me a couple of milliamps which is all I need for the optocouplers. Antiparelleling them together with a couple of Schottkys might just have low enough forward voltage for my low end. For the high end they should work as current cources.

Thanks,

Benta.

 

[waross]

Why don't you change the resistance to an inductance. It will increase in impedance as the frequency increases.
respectively.

 

[benta]

Waross, I thought about that one as well.
Problem is, I'd need a couple of Henrys to get any effect, which I haven't found practicable.

Thanks,

Benta.

 

[ijl]

I liked the clipping idea, so here is another try without amplification. Replace one of the diodes (the "forward one") with a blue LED (or maybe green LED, or a zener), and use a small resistor of about 1 kOhm. Connect the optocoupler to the diodes with a small resistor. As a side effect, you get a magical blue glow with an intensity that varies with the speed

 

[ijl]

Well, it seems that I am writing faster than I am thinking. The LED does not solve the problem of the high current drain from the tacho. It only moves the current out of the optocoupler.

 

[e2zn]

I'm not familiar w/TGs, but how much is too much current?

Using an RC LPF of 1kohm and .47uF, 50Hz is unattenuated, and 1100Hz is down almost 12dB, giving about 15Vp-p.

Current through R at 50Hz would be 3mA, and at 1100Hz 45mA. Would that be too much for the tach, or too little for the optos?

e2zn

 

[benta]

OK, IRStuff's original post ("That's known as a constant current source") sent me searching again.

This might work:

http://www.vishay.com/fets-small-signal/SSFsgcrdP/

I'll try to replace the LED series resistor with two current regulator diodes plus two Schottkys, all in "antiparallel". This should give me an AC current limiter.

So, I've saved one resistor, total balance three extra (small) components.

Hope it works, will let you know when I get the samples.

Thanks,

Benta.

 

[benta]

I thought I'd share the results with you on this problem.
I got the constant-current diode samples, did some calculations and at the end it turned out to work extremely well on the breadboard prototype.

The solution was two constant-current diodes, helped by two low-voltage drop Schottkys (the CC diodes don't like reverse voltage) driving a dual optocoupler with the LEDs in antiparallel.
Given that I only have around 0.5 V "headroom" to drive the LEDs at low voltages, I absolutely see no other way of doing this.

Schematic here (secondary side of optocoupler not shown):

http://i69.photobucket.com/albums/i56/Benta/opto_int.jpg

One thing that's extremely nice about this solution is that through the capacitor I'm able to control the duty-cycle of the output. This is important, as the turn-off time of the output transistors is lousy at very low currents (where I'm operating), and the off time of the LEDs is extremely short when running off full-wave rectified AC. The capacitor converts the whole thing into a trapeze- waveform with the crest time only controlled by cap. value.

Regards,

Benta.

 

[melone]

Why not use a Voltage-to-Frequency converter, and just modify the modulation that you provide to your LED?

 

[benta]

Why should I? The tachogenerator already generates the frequency that I need to measure.
This circuit works perfectly and it has the following advantages:
1: It's dead simple.
2: I do not need to power the tacho side.

So, this was not a question, just the solution to an earlier problem that I could share with you.

Regards,

Benta.