A valuable device that is not as well-known as perhaps it should be.
Recently, we saw the awesome and versatile water tank monitoring project by Ashley Woods. In it, the pressure sensors used to monitor water levels run as a current-variable device, rather than the variable voltage that so many Arduino and Raspberry Pi sensors use. It occurred to us that current-based sensors are rather under-represented when in fact they are highly useful, albeit a bit harder for less experienced makers to use.
In turn, that led to the constant current source, another circuit object that receives little attention. However, they’re more valuable than many people realise. Even within the DIYODE office, we realised how little attention we pay to these things at times, and decided to correct that by dedicating this month’s Classroom to the LM334Z, the most common and accessible constant current source for makers. This is because it is the device that is sold over the counter by the major electronics retailers, and many of the online maker-oriented suppliers as well.
Of course, if you’re an engineer or a trade-experienced maker, you know that there are many other devices available. Some of them are probably better than the LM334Z. However, that’s only helpful if you’re comfortable with the trade-oriented, extensive websites of Element 14, RS Components, and the other online players who operate Australian websites but have little if any stock and staff in this country. For most people, the LM334Z is the constant current source that they can get their hands on.
WHAT IS A CONSTANT CURRENT SOURCE?
The operative word here is ‘constant’. Technically, a resistor is a current source as well, but we know that the current from this varies with voltage, and so that’s not what we’re interested in here. While there is more to it in the long run, a constant current source can be summed up fairly easily. It is a device or circuit which has a fixed current passing through it, regardless of changes in load resistance and most importantly, supply voltage.
In reality, there is no such thing. Any circuit or device has some variation in its output, with temperature being the biggest factor in the case of dedicated constant current sources. Also, very high impedance loads will change the game. Impedance can be ignored for practical purposes in the majority of cases, unless you for some reason want to feed a constant current source into the input of an operational amplifier. Given that Op-amps are designed with the highest impedance that can be practically made, and are voltage-driven, we can’t see why you would want to do this.
However, we don’t know everything and there may well be a reason we haven’t thought of for you to want to feed a constant current source into a very high impedance load. There are situations where you can bias the inputs or offsets of some Op-amps with a current source, but these are not wired as a direct, point-to-point input that would have an issue with high impedance. Rather, they are usually dedicated bias inputs, or wired as a divider arrangement across a standard input.
MEET THE LM334Z
LM334Z ELECTRICAL SPECIFICATIONS
One challenge in discussing this device, and many others, is differences between manufacturers. The physical examples we bought were National Semiconductor branded devices. National Semiconductor has been owned by Texas Instruments since 2011 and no longer exists as a company, but the branding still exists.
This is because TI retains part numbers, order codes, and related material exactly as it used to be before acquiring a company, so customer orders are not impacted and don’t have to change. The result is that you can still buy items from the Texas Instruments brand, and end up with a National Semiconductor product, if that’s the code you order. This does not indicate new old stock, and is exactly what happened to us when the manufacturer was listed as ‘Texas Instruments’ on the order page.
If this happens, the products are considered identical. Other brands may not be exactly the same, but if they are different, they will have a different listing even if the brand is owned by another.
However, other manufacturers have produced them and many can still be bought as new old stock even if a given manufacturer has stopped making them. As is always the case, we recommend figuring out the brand of your device and downloading the datasheet from the correct manufacturer. Some details differ, like the maximum throughput voltage, which is 40V for National Semiconductor and Texas Instruments devices, but 30V for ST Microelectronics devices. ST still has the LM334 as a current product, no pun intended.
The LM334 is a dedicated constant current source that is sold at the retail level in a TO92 package. As is usually the case, other packages are available from commercial suppliers, and you may chance across these from time to time. However, as this is rare for most makers, all of the information here relates to the TO92-packaged LM334Z. Aside from some of the specifications, most information carries across the device range.
The LM334Z is a three-pin device looking to the casual observer to be a transistor. In fact, it isn’t, but it contains some. The three pins are for voltage in (V+), adjustment (ADJ), and voltage out (V-). At a minimum, the device can be used with only one external component: A resistor, between the ADJ and V- pins, to set the output current. We’ll describe this in detail shortly, including some of the maths required. First, however, it is worth unpacking some of the specifications from the table.
You might have noticed that we use the term ‘throughput’ voltage. This is because the device gets used in series, and passes, as near as makes no difference, the full voltage of the current through it. Only the current is limited. This term is different from an ‘operating’ or ‘supply’ voltage of a device, which implies consumption by the device. The maximum is between 30V and 40V, depending on brand, so make sure you have the right datasheet and read the right line for your version! Many datasheets cover the LM134 and LM234 as well, which may differ. The reverse voltage is 20V. The datasheets specifically state that the LM334 can be used as a rectifier and current limiter in one, should the need arise, but the lower reverse voltage means this cannot be so at the same voltage range as in a forward-biased DC situation.
Output current is between 1μA and 10mA. However, as we will unpack later, temperature considerations come into play. The 10mA maximum current has the device dissipating quite a bit of heat, and may not always be practical. The device has a maximum power dissipation of 400mW, but this is for the die junction inside the device. In the thermal characteristics section, we’ll describe why this is not a real-world figure.
Regulation varies quite a lot between models and situations, so if this is important to you, carefully read the data sheets. It varies with temperature, voltage, set current, and other factors. However, even in a worst-case scenario, the variation is so small that only an engineer is likely to care. Most makers will have trouble measuring the difference. We certainly don’t have equipment sensitive enough here in the DIYODE workshop.
SETTING THE CURRENT
Setting the output current of the LM334 is achieved with a single external resistor, connected between ADJ and V-. For optimum precision, this resistor should be mounted as close to the terminals as practical. Metal film resistors are recommended, but notes in the datasheets suggest wire wound resistors for the best precision possible. This is because of thermal characteristics, which we discuss further on. The formula for this is:
ISET = 67.7mV / RSET, where ISET is the set current in amps, and RSET is the value of the set resistor in Ohms.
This means the resistor will never be smaller than 6.8Ω, which will come very close to 10mA for ISET. However, even at 1% tolerance, there are situations where metal film resistors are still not accurate enough. You can measure one until you find one with the right value, or use some form of trimpot. However, in many commercial cases, the ‘resistor’ is formed with a carefully-designed PCB trace. That’s only valid for Makers who are designing their own PCBs, so most of us will be using metal film resistors.
If you’ve been reading along in the datasheet, you might have noticed that the device current is a combination of two figures: The current through the ‘Set’ resistor, and the bias current through the LM334 itself. There is a relationship to these currents, and each manufacturer’s datasheet provides a graph to show the non-linear relationship. This graph will give a number to substitute for the letter n in the following equation:
Remember, in algebra, any time that two values are next to each other with no symbol, they multiply. Also, don’t forget order of operations. We get several inquiries each month that we publish equations, for people unable to get the answers we do. In most cases, it is because many people, no matter their level of knowledge in other areas, forget details like order of operations when they do not do maths often enough. Some calculators don’t do it automatically, either, though this is usually a formatting issue in the input.
Note the non-linear x-axis scale, which appears to be logarithmic.
So, make sure you do the divisions within the brackets first, then multiply those numbers together. Note the figures - RSET is the value of the set resistor in Ohms; n is the value from the ‘Ratio of ISET to IBIAS’ graph; and VR is the voltage from the ADJ pin between it and ground, and is related to temperature. Both here and elsewhere in the datasheet, we see that VR has a linear value related to the temperature of the junction in Kelvin. See the section on the ADJ pin for more details.
THERMAL CHARACTERISTICS
There is a very marked and defined effect of temperature on the LM334. It can actually be used as a temperature sensor, but more on that in a relevant section. What is important about this is that the value of the current output will change depending on temperature. The formula above for setting the current is valid for a junction temperature, not ambient, of 25°C. That means even if the outside of the case is in a freezer, if there is enough current drawn to heat the junction deep within the case to above 25°C, then the current value will be different to what is calculated. The reverse is true, too. If the core is cooler, changes in temperature on the outside of the case need time to reach the junction.
The datasheets warn of significant effects on regulation for any current above 100μA. This is one of the only times voltage becomes important in using the LM334. For every volt increase at 1mA, the junction temperature increases by 0.4°C. That value is for still, unmoving air, but remember, it is a junction temperature value, too. The thermal resistance of the case means this won’t immediately dissipate to air.
The LM334 has a positive temperature coefficient, which means its internal resistance increases with heat. It is stated in most of the datasheets we checked as 0.33%/°C. There can be manufacturer variations. In other words, every increase of 1°C brings a 0.33% change in current. So, at 12V, and 1mA output, that’s 12V x 0.4 x 0.33 = 1.584%. Note that this is the amount the current is reduced by, even though it is a positive number.
The biggest effect of temperature on the LM334 is seen in power dissipation. Any current passing through the die or junction inside the device causes heating. The datasheets give thermal resistance figures, which means the ability of heat to pass from the junction to the outside world. The figures are quoted for ‘junction to case’ and ‘junction to ambient’. The heat escape pathways are vai the electrical connection leads, any tabs fitted, or the case. Some cases, however, carry heat, and some do not. Ceramic cases do, while most plastic cases do not.
There is a huge clue here in the fact that the listing for the TO92 package does not include a figure for ‘junction to case’, but rather lists ‘N/A’. The plastic case is an insulator and therefore the only way for heat to escape is through the leads of the device. The figure for this is given for packages with 0.125mm leads and 0.4mm leads. As we did not find any retail sales of packages with 0.125mm leads, it feels safe to only discuss the 0.4mm figure.
At 180°C/W, this figure means that every watt of power dissipated in the device will generate 180°C of heat, as the leads provide only a limited cooling path. Of critical importance, this assumes leads soldered to copper tracks with the expected amount of heat sinking capability for a typical such situation. Solderless breadboards are a whole different story. They only make contact with the sides of the legs, and only provide a limited ability to sink heat.
The LM334 has a maximum junction temperature of 100°C. It should be operated below that. At best then, assuming 90°C as an acceptable upper limit, 180°C/W turns into 0.5W of dissipation possible. At the 40V, 10mA maximum, that’s 40V x 0.01A = 0.4W. That’s exactly the same as the 400mW power dissipation listed in the datasheet, meaning there is probably some other factor we haven’t considered in the calculation based purely on the thermal resistance figure. This illustrates the value of reading the whole datasheet. Given that this figure assumes ideal circuit board heatsinking, and that the device is known to have less regulation as soon as junction temperature increases, it makes sense not to max out the device at all, not even close. The hotter the junction, the more the power rating must be derated, and the less the device will regulate.
TEMPERATURE COMPENSATION
With the temperature having such a marked effect on output, both in absolute terms and in terms of the accuracy of regulation, ambient temperature must be considered. The datasheets show a way of compensating for ambient temperature changes by adding a further resistor and a diode. This is for cancelling out drift in output caused by ambient temperature rise, and will do nothing to negate the effects of junction temperature rise due to power dissipation. This circuit relies on the fact that a forward-biased silicon diode has a negative temperature coefficient, which means it becomes more conductive as it heats.
In this case, there is more to setting the current than the equation above. The total output current is the sum of I1 and I2, which are almost half the total current each. The difference is that I1 includes the bias current, while I2 does not. The value of VR is generally increased by 5.9% for calculation purposes to include IBIAS in I1, but you can go back to the previous formula and calculate it specifically. In the datasheet, a full explanation of the maths involved is presented, but the recommended ratio for R1 to R2 is 1:10. Assuming a silicon diode forward voltage drop of 0.6V, the end result is a formula for ISET as follows:
If you measure your diode and it is different from this by a concerning value, head to the datasheets and insert your own numbers into the formula there. Once you know the set current and the value of R1, you can calculate R2 from the ratio of 1:10.
THE ADJ PIN
The ADJ pin is the adjustment voltage, and is used in conjunction with the Set resistor to set the current through the LM334. The voltage at this pin is labelled VR. VR has a linear value related to the temperature of the junction in Kelvin, an absolute measurement starting from 0°K (-273°C). This figure is 214μV per °K. The unit size of the Celcius and Kelvin scales is the same, unlike Fahrenheit. They just have a different starting point.
So, if 0°C is 273°K, then 25°C is 273°K + 25 = 298°K. To find VR at 25°C, the value of 214μV/°K is just multiplied by whatever temperature you need. Seeing as the datasheets are all figured for 25°C:
VR = 298°K x 214μV = 63 772μV, or 63.772mV.
In most scientific literature, the unit for the Kelvin scale is just written as K, not °K. It is still a degree, which is just a term for ‘equal sized unit of increment’, but the ° symbol is not usually used. However, it is not fully incorrect to use it. The reason we have done so here is the same reason the datasheets use it: Both ‘K’ and ‘k’ are commonly used for other units or as a multiplier in datasheets, and adding the ° symbol adds clarity and instantly flags the label as a temperature label.
ESD PROTECTION
The LM334 has a degree of resistance to damage by ElectroStatic Discharge (ESD), although only to a value of 2000V. A note in the datasheet indicates that this is based on something called the ‘human body model’, a 100pF capacitor discharged through a 1.5kΩ resistor. This is a standard based on a United States Military testing standard, and is designed to reflect the ability of a device to withstand the manufacturing process. A 100pF capacitor is charged to a specified test voltage, which is where the figure of 200V comes from in this case, and discharged through the 1.5kΩ resistor.
However, in the uncontrolled real world, voltages may be much higher than that. Human skin should have a much higher resistance than that, too, but this is not always guaranteed. Even at higher resistance values for human skin, ESD can still be in the order of tens of thousands of volts, albeit at miniscule currents. It depends on the humidity of the day, the material of your clothing, your shoes, the flooring, dozens of factors within your body, and other factors besides.
Many datasheets we checked, such as Texas Instruments’ NE555 and On Semiconductor’s BC547, do not include an ESD figure, and we generally consider these devices to be not ESD-sensitive. Therefore, we feel that the inclusion of the figure indicates care should be taken and at the very least, basic static protection procedures used.
OTHER POINTS TO KEEP IN MIND
The thermocouple effect of having dissimilar metals joined is flagged in the datasheet as a potential effector of the sense voltage at the ADJ pin. If the resistor, with different metals at its junction and within its make-up, is mounted too far away from the device, they will be affected by ambient temperature differently. Additionally, the effects of high-resistance joints is highlighted, with sockets recommended against. For more specialised users, there is information worth reading in the ‘Application Information’ section of the datasheet, but for most users, this summary should be sufficient.
Output impedance changes with frequency, indicated by a graph in the datasheet. As most uses are DC, this is only an issue for some people. Of more concern is the fact that regulation is not instant. There is another graph which indicates this, so applications where power is applied and removed from the device regularly need to be considered.
ALWAYS READ THE FOOTNOTES
Footnotes exist for good reason. They explain details and flag problems, and are not included for fun, or to reference sources like an academic article. Always read them and take note of what applies to you. Some may completely alter your understanding of the information presented in a table or graph.
One great example from the LM334’s datasheet is that all figures are for a junction temperature of 25°C. However, also in the same footnote, it is revealed that the test conditions involve pulsed power to ensure no temperature drift. That means that all figures presented are likely to be different for everyday users in situations where current is constantly applied.
USES
LED DRIVER
One of the most common uses of the LM334 online on forums and circuit-sharing websites is as an LED driver. In many cases, this is pushing the limits of the device as people try to squeeze the full 10mA out of it. However, there are valid points to this concept and some interesting facets of LEDs we often forget about.
Most of the time, we limit current to an LED using a resistor. We have to find (or assume) is forward voltage drop, and calculate precisely with this and the operating current to find the resistor value. This can lead to makers forgetting that in fact the figure of a voltage drop is just that: A voltage drop. It is not a supply voltage. The LED can have a much higher voltage passing through it if the current though it is limited.
This means that in situations where voltage may vary, having a constant current source is a good way of operating an LED, as at no point does the current through it, which is what it is actually sensitive to, change. Even if the voltage doubles, the current does not fluctuate. If this occurred with a resistor as a current-limiter, a doubling of the voltage would translate to a doubling of the current, because the resistor value stays the same.
The LM334 is also good for driving specialised high-efficiency LEDs, many of which operate at 2mA or less. At these currents, with a decent supply voltage, a regular resistor may not even be good enough due to tolerance: a small percentage either way could actually be more than the LED can handle. Because of this, have a look at the ‘Output Current Boosting’ section for a better way of driving LEDs.
ALTERNATIVE LED DRIVER
There is another way of using the LM334 as an LED driver, and it is related to some of the other circuits in this article. We are fairly sure it originated with Bob Pease, something of a legend in the analog electronics community. This one can comfortably supply 10mA at 9V to an LED, and would likely function even higher.
The limiting factor is the power dissipation in the transistor PN junctions, and the power dissipation for the chosen component. Because we haven’t built it, and a video explaining it by Bob Pease himself still floats around, we haven’t altered it in any way. We’re just relaying it, presenting it with its original component values and choices.
TEMPERATURE SENSOR
The consistency of the temperature drift of the LM334 makes it useful as a temperature sensor. Because it draws a constant current, it is immune from the issues of long wire runs that many voltage-based sensors suffer from. For example, a thermocouple outputting only millivolts at room temperature would easily have this signal degraded by a long twin-wire run.
Other sensors output a voltage as a percentage of the input voltage, but if this is reduced by a long wire run, then the reference voltage will not arrive at its destination with the correct value.
This circuit still uses three wires: Two for the power and one for signal. However, the constant current nature eliminates the issues of supply voltage drop. Care will be needed to ensure the signal wire is of large enough cross-section to avoid significant drop across that. Because the output is absolute, it can be scaled to Kelvin, and the component values shown will achieve an output of 10mV per degree Kelvin. So, for 25°C, the output will be 2.98V.
That circuit has an output impedance less than 100Ω. For even lower impedance, less than 2Ω, a transistor and extra resistor are needed. Again, with the values shown, the output will be 10mΩ/°K. VIN in both cases must be 5V or greater. Resistor values should be selected from 1% metal film devices.
Some of the values shown are not in the retail series, so the nearest value should be chosen and individual resistors tested to get as close as possible. It is tempting to use 10% devices to get closer to the stated values, but these will be carbon film and are still not common in retail. We have not tested the effect of this variation in value.
CURRENT REFERENCE
The most basic of circuits is as a current reference source. Because these are generally below 1mA, no output components are needed. It is just the basic circuit presented in the ‘Setting the Current’ section, and the maths in that section applies. The difference is that while the text implied we were setting the current to 10mA as the maximum in that section, here, values will be kept below 1mA in the main.
This circuit has uses as a reference, and as a bias current supply. In some amplifier designs, a bias current is needed to feed into a complementary or balanced transistor pair. As acknowledged earlier, some Op-amps and other ICs have bias inputs that are current-driven.
Further to that, the basic circuit is also useful for another task we rarely think about. Many makers are familiar with RC (resistor-capacitor) timing circuits. We covered these in a Classroom in Issue 44, when we discussed ways to test capacitors. In that issue, we covered the timing constant, and discovered that the charging curve on a capacitor is curved, because as the voltage across it increases, the current decreases.
However, because the LM334’s current output is independent of voltage, it can be used to charge a capacitor linearly. This is quite a useful point, so much so that we’re going to cover it as a hands-on build at the end, where we make a ramp generator circuit.
OUTPUT CURRENT BOOSTING
The output current of the LM334 can be boosted with the accompanying circuit. The 100Ω resistor and 100pF capacitor are guide values. The datasheet suggests no values, we got these from other research and experimentation. They exist for stability, and so can be adjusted by experimentation. The transistor suggested by Texas Instruments and National Semiconductor is a 2N2905, however many PNP transistors would work. Just be careful to monitor output and change values of components to keep stability and regulation. The output current will be limited by the power dissipation of the transistor, so check the datasheets of the device you wish to use.
DON’T!
There are several examples floating around the internet of a current-boosting circuit like this one. At first glance, it looks good, and some people will defend it until they are blue in the face. However, the regulation of current in this case relies on the gain of the transistor, it’s beta or Hfe figure. Herein lies the challenge.
One of our favourite transistors here at DIYODE is the highly versatile BC338, and its PNP counterpart that would suit this circuit, the BC328, has a gain range of between 100 and 630. That means, from any given batch of examples, one BC328 might have a gain of 100, while the one next to it may have a gain of 600, and the one next to that a gain of 200.
This means that the circuit as presented is not reliable. If it were used as, say, an LED driver, one LED could be 6.3 times brighter than another. While the difference in upper and lower gains is not as big for some transistors, it is still never good enough for a regulated circuit such as this. Don’t use In the datasheets, there is an example provided of a ramp signal generator, also known as a sawtooth generator. However, that circuit relies on an external reset trigger. Our immediate thought was to use an NE555 timer IC to generate the reset pulses, but we decided to both brainstorm and look around the internet for other ideas.
Hands On:
Ramp Generator Example
| Parts Required: | Jaycar | Altronics | Pakronics |
|---|---|---|---|
| 1 x Solderless Breadboard | PB8820 | P1002 | DF-FIT0096 |
| 1 x Packet of Breadboard Wire Links | PB8850 | P1014A | SS110990044 |
| 1 x LM334Z | ZL3334 | - | - |
| 1 x NE555 | ZL3555 | Z2755 | - |
| 1 x LM358 | ZL3358 | Z2540 | - |
| 1 x BC547 NPN Transistor * | ZT2152 | Z1040 | DF-FIT0322 |
| 1 x BC557 PNP Transistor * | ZT2164 | Z1055 | DF-FIT0322 |
| 1 x 150Ω Resistor * | RR0552 | R7538 | DF-FIT0119 |
| 2 x 1kΩ Resistor * | RR0572 | R7558 | DF-FIT0119 |
| 1 x 3kΩ Resistor * | RR0583 | R7569 | DF-FIT0119 |
| 1 x 6.8kΩ Resistor * | RR0592 | R7578 | DF-FIT0119 |
| 1 x 68kΩ Resistor * | RR0616 | R7602 | DF-FIT0119 |
| 1 x 100nF Capacitor * | RM7125 | R3025B | DF-FIT0118 |
| 1 x 10µF Capacitor * | RE6066 | R4767 | DF-FIT0117 |
| 1 x 100µF Capacitor * | RE6130 | R4825 | DF-FIT0117 |
| 1 x LED * | ZD0152 | Z0800 | ADA299 |
Note: Quantity required, may only be sold in packs. A breadboard and prototyping hardware is also required.
In the end, we stuck with the NE555 option. It has several advantages and disadvantages when compared to other designs. While other designs are self-triggering, and the NE555 version has to be calculated, we felt having two independent halves would allow greater observation of changes and effects. In addition, few makers are unfamiliar with the NE555 Astable circuit, while many of the other options, though simple, are still less familiar.
Assembly should be straightforward, based on the schematic and Fritzing. Unlike a PCB-based kit, work in sections from left to right, rather than fitting all the wire links, then all the resistors, and so on. This is because sometimes spacing will change, based on how you bend your component legs, but it also keeps mistakes to a minimum. Note carefully that in both the Fritzing and photos, the 6.8kΩ resistor at the LM334 is end-on, so its colour bands cannot be seen in the images and the resistor itself may be missed.
The circuit consists of four sections: The NE555 Astable section, generating the reset pulses; the inverter section, because we want a short high output pulse and a long low section, not practical with the standard Astable NE555 arrangement; a ramp generator based around the LM334; and a voltage follower based on an LM358 Op-amp, to boost the output level and turn the current-driven ramp circuit into a voltage driver for loads.
The NE555 Astable circuit is the standard version. We chose resistor values of 68kΩ and 3kΩ, with a capacitor of 100μF, to give us a high time of 4.9 seconds and a low of 0.2 seconds. Feel free to play with these values, just be mindful to keep the low period quite short, for reasons that will become clear shortly.
After that, the output of the NE555 meets an inverter circuit. This is formed from a PNP transistor controlling the base of an NPN transistor. We chose the BC557 and BC547 respectively, but any general purpose transistors will work. The idea here is that the NE555 cannot have a low period longer than the high, yet we want exactly that. Instead, we have the NE555 with a 4.9 second high period, during which the PNP transistor is held low by virtue of the base being fed, via a 1kΩ current-limiting resistor, with the output voltage from the NE555. When the output briefly goes low, the PNP transistor can now conduct from its emitter to its base and to ground, via the low Pin 3 of the NE555, which sinks current when it is low, rather than just being ‘off’. As soon as the PNP transistor conducts, it passes current via the 1kΩ resistor to the base of the NPN transistor, which itself now conducts. The circuit had to be arranged this way because the ramp circuit needs to be controlled by a low-side switch.
Next comes the ramp circuit itself. It’s pretty much straight from the datasheet, but with component values added. We chose a 6.8kΩ resistor and 10μF capacitor, but you can alter these. Using the formula from earlier in the article, and remembering to convert to base units, we have:
0.0677V ÷ 6800Ω = 0.000009955A.
Converted down, multiplied by one thousand twice, we end up with 9.955μA, or as near as makes no difference, 10μA.
That charge current is fed to the 10μF capacitor, the voltage across which rises linearly, as opposed to the exponential curve we’re used to from a normal RC circuit. This rising voltage goes to two places. Firstly, it goes to the collector of the NPN transistor which, most of the time, is off and not conducting. It is also sent to the non-inverting input of the LM358 Op-amp.
The operational amplifier is a dual device, but we're only using one half. The inverting input is tied to the output, which turns the device into a voltage follower. This means that whatever is at the input is transferred straight to the output. In effect, it is a buffer, but it is used in this case to drive a load, because any significant current flowing out of the ramp generator will affect the timing. There is only 10μA to charge the capacitor with in the first place! The high-impedance input of the amplifier ensures this circuit is not affected. The output can source enough current to run the LED connected to it as a visual indicator, or drive another circuit.
The idea is that the NE555 and ramp generator both start their timing periods when power is first applied. They will not, however, be perfectly synchronised, but fairly close. Because of the inverter circuit, every 4.9 seconds, there is a 0.2 second high time, during which the NPN transistor conducts. It dumps the current from the capacitor straight to ground, limited only by the capacitor’s own internal resistance. This terminates the rise in the voltage across the capacitor, and drops the signal level to zero again.
We fitted the output of this circuit with an LED for a visual indication, but because the LED has a minimum forward voltage and a non-linear response to brightness versus current, it’s far from perfect. Really, you’ll need an oscilloscope to see what’s going on in this circuit. We used one of our new favourite tools, the Sensepeak PCBite holder and probe kit from K&A Electronics, which made this job a lot easier.
There are significant imperfections in this circuit. It was chosen for its ease of fiddling and alteration, rather than as a really useful circuit. In a real situation, a breadboard is completely unsuitable, as discussed above. Even at 10μA current on the LM334, we’re starting to risk internal heating through lack of heat sinking through the legs and the tiny spring-clips of the breadboard. Breadboards are also notoriously noisy. The use of the independent reset pulse is unrealistic too.
If you do want to use a circuit like this, we suggest some of the designs to be found online where the other half of the Op-amp is used to trigger the transistor and bleed the capacitor, based on the rising voltage and not an independent timing.
Lastly, the output would drive a circuit, and not an LED. An example would be a voltage-controlled oscillator with an audio output, which would ramp up then start again. This sound is familiar to anyone who has had to evacuate a commercial building or heard the same in the news media. Having said that, using an independent timer like the NE555 circuit would mean that you could introduce ‘off’ periods between the ramps, for a different sound altogether. Just increase the 0.2 second low time, by changing the resistor values.