Full Stack Electronics · Stage Three · Experiment 8
Build the 555's second comparator from the other kind of transistor. Then measure how close to ground the first one keeps working.
The comparator finished in Experiment 7 is the 555's threshold comparator. It watches for the timing capacitor to rise past 2/3 VCC, 3.33 V on a 5 V supply. The 555 has a second comparator, the trigger comparator. It watches for the capacitor to fall below 1/3 VCC, 1.67 V, and starts the next cycle. Its input falls towards ground, so this comparator has to work closer to 0 V than the threshold comparator.
In Ken Shirriff's analysis of the NE555's die, the threshold comparator has NPN input transistors and the trigger comparator has PNPs. The probable reason is where each pair's emitters sit. An NPN pair holds its emitters a base-emitter drop below its inputs, and its tail needs room between them and ground. Experiment 7 measured the NPN pair's low level at 1.12 V with V− at 1.67 V, so its emitters already sit about 1 V above ground. As both inputs fall towards ground, that room runs out. A PNP pair holds its emitters a drop above its inputs, and its tail hangs from +5 V, where the room stays.
So this experiment builds the trigger comparator. It is the circuit of Experiment 7 with every NPN swapped for a PNP, every PNP for an NPN, and +5 V and ground exchanged. Its 1.67 V reference comes from power supply output 2. The first measurements check that it switches at 1.67 V with its output at 0 V and 5 V. Then the floor test lowers the threshold comparator's reference towards ground and measures where the NPN pair stops working.
The plan is to swap every transistor for the other kind and exchange the rails. To see what that does to each stage, start from the circuit as Experiment 7 finished it. Figure 8-1 is the whole threshold comparator. The NPN pair Q1 and Q2 has the NPN tail mirror Q3 and Q4 below it, and the PNP mirror Q5 and Q6 above it as the load. The PNP output stage Q7 reads node A from the +5 V line.
Figure 8-2 is the trigger comparator, drawn on the same skeleton. Each stage moves to the other rail, and +5 V stays at the top of the drawing. Q10 is drawn to the right of Q11, where it stands on the board. The seven transistors are numbered on from the threshold comparator's, Q8 to Q14. Each number is seven more than the transistor it replaces: Q8 does Q1's job, and Q14 does Q7's.
Put Figure 8-2 beside Figure 8-1. Each stage has its opposite number on the other rail:
| Stage | Threshold comparator (Figure 8-1) | Trigger comparator (Figure 8-2) |
|---|---|---|
| Input pair | Q1, Q2: NPN, emitters joined below them | Q8, Q9: PNP, emitters joined above them |
| Tail mirror | Q3, Q4: NPN, emitters on ground, 4.7 kΩ from +5 V | Q10, Q11: PNP, emitters on +5 V, 4.7 kΩ to ground |
| Load | Q5, Q6: PNP mirror on +5 V | Q12, Q13: NPN mirror on ground |
| Output node | node A, Q2's collector | node B, Q9's collector |
| Output stage | Q7: PNP, emitter on +5 V, 4.7 kΩ pull-down | Q14: NPN, emitter on ground, 4.7 kΩ pull-up |
Current still flows from +5 V to ground, but each stage now handles it from the other side. In the threshold comparator, the tail mirror stands below the pair and draws the tail current out of its emitters to ground. In the trigger comparator, the tail mirror hangs above the pair and pushes the tail current into its emitters from +5 V. The voltage limits also reverse, and the node B check measures that.
The last section put the trigger comparator's pair and tail next to +5 V, and the board does the same. In the threshold comparator, the pair and the tail mirror stand in the lower block, next to the ground rail. In the trigger comparator, they stand in the upper block, next to the +5 V rail. The NPN load and the output stage go in the lower block below them. The new circuit starts at column 16, one column clear of the threshold comparator, and takes columns 16 to 27 in both blocks.
Two things change from Experiment 7. The rail links of the drawings so far sit in columns 28 and 29, right against the end of the new circuit, so they move to 38 and 39. The NPNs of the lower block are turned round, with their flat faces towards the far edge of the board. So each NPN's collector comes first and stands one column right of the PNP collector it joins, a short diagonal jumper away. Figure 8-3 shows the upper block: the input pair and the tail mirror. Its PNPs stand the same way round as Q5 to Q7, with legs reading emitter, base, collector from left to right.
The steps follow Figure 8-3. Start with the power off.
The collectors of Q8 and Q9 are still unconnected. In the threshold comparator, the pair's collectors meet the PNP load. Here they meet the NPN load, in the lower block below the pair. Figure 8-4 adds the load, the output stage and the leads.
Each lead reaches its base through a 1 kΩ, for the reason Experiment 7 found: a base that conducts at the end of an instrument's lead can oscillate. The threshold comparator's inputs in 10j and 13j stay empty until the leads come back for the floor test.
The next section switches the power on. If OUT then stays near 0 V whatever the input does, Q14 is on all the time. Check that Q14 is a 2N3904 and that the pull-up goes to the outer bottom rail. If OUT stays at 5 V whatever the input does, check the jumpers from 18e to 19f and from 19i to 17i. Also check that Q8 to Q11 are 2N3906 with their flat faces towards the near edge. The 4.7 kΩ on Q10 must go to the outer top rail, ground: on the inner +5 V rail, the tail would carry no current.
The new comparator is complete. The first check is the one that Experiment 7 made on the threshold comparator at 1.67 V: a triangle through the reference, and OUT on the oscilloscope. OUT should switch between 0 V and 5 V each time V+ crosses 1.67 V.
Figure 8-5 is the capture for this check. OUT switches between 0.0 V and 5.00 V each time V+ crosses about 1.7 V, near V−. It is high while V+ is below V−, the same way round as the threshold comparator's OUT. For the trigger, that is the useful sense. Its output goes high while the input is below 1/3 VCC, which is when the 555 has to start a new cycle.
The last section expects OUT at both rails. This section explains why, and what the reflected circuit does to the limits that Experiment 7 found at node A. Figure 8-6 marks the steps on the circuit. The tail is drawn as one current source hanging from +5 V.
Step 4 makes two claims about node B. With Q14 connected, node B stops near 0.65 V. Without Q14, it would rise to about 2.3 V. You can check both by moving the probe to node B and removing one jumper.
With Q14 removed, nothing takes current from node B. So the whole tail current, up to 0.93 mA, leaves through Q9's base into V−. Through the 1 kΩ, that current would lift Q9's base by up to 0.93 V. For this check, power supply output 2 connects to Q9's base column directly, as in my first build, so V− stays at 1.67 V. Figure 8-7 shows the leads.
With V+ above V−, node B rises well above 0.65 V and stops at 2.40 V, far below +5 V. That is near V− + 0.65 V, about 2.3 V, but a little above it. My first build read 2.44 V here. The collector cannot rise past its emitter, so the emitter itself sits higher than the sum. With nothing taking current from node B, Q9's base carries the whole tail current. Its base-emitter drop is then more than the round 0.65 V: at least 2.40 V − 1.67 V = 0.73 V. Experiment 7's floor showed the same effect: 3.33 V − 2.64 V = 0.69 V.
If node B's high level climbs and falls with V+ instead of staying level, power supply output 2's lead is still on the free leg of the 1 kΩ. The tail current through the 1 kΩ then lifts Q9's base until Q8 starts to share the current. Q9's base follows V+, and node B follows it about 0.6 V higher. Move the lead to the resistor's lower leg, where it enters 17b.
Now reconnect Q14.
With Q14's base on it, node B's high level falls from 2.40 V to 0.72 V, Q14's base-emitter drop. My first build read about 0.9 V here. Q14 then turns that small swing into the full swing of OUT in Figure 8-5. Q7 did the same at node A in Experiment 7, where the low level sat 0.70 V below +5 V. The table lists the expected and measured levels, and the figures that capture them.
| Level | Expected | Measured | Capture | Set by |
|---|---|---|---|---|
| Node B high, Q14 disconnected | about 2.3 V | 2.40 V | Figure 8-8 | Q9's emitter, V− + 0.65 V |
| Node B high, Q14 connected | about 0.65 V | 0.72 V | Figure 8-9 | Q14's base-emitter junction |
| OUT high | 5 V | 5.00 V | Figure 8-5 | the pull-up to +5 V |
The node B check explains how the twin reaches both rails. But Experiment 7 already ran the NPN threshold comparator with V− at 1.67 V, and its OUT read 0.0 V and 5.00 V there too. So on this board, at 1.67 V, the NPN comparator does the trigger's job as well. With V− at 1.67 V, the NPN pair's emitters sit near 1 V whatever V+ does. When V+ falls to 0 V, Q1 switches off.
The two pairs differ when both inputs are low. Figure 8-10 puts both inputs of each pair at 0.3 V. That is below 0.42 V, where the NPN comparator stopped switching in the floor test.
In the NPN pair, the emitters sit one base-emitter drop below the inputs. The tail mirror below them needs some voltage across it to keep its current. With the inputs at 0.3 V, the full 0.65 V drop puts the emitters at −0.35 V on paper, below ground. The emitters cannot go there, so the tail runs out of room and the pair runs out of current. The PNP pair has the same limit at the other end, near +5 V. Near ground, its emitters sit above the inputs, and its tail has most of the supply.
Where does the NPN pair stop? A first estimate adds one base-emitter drop, 0.65 V, to a couple of tenths for the tail mirror: about 0.9 V. The next section measures it.
The last section estimated the floor at about 0.9 V. The measurement uses the threshold comparator: its reference falls in steps while V+ keeps crossing it. Power supply output 2 moves back to the threshold comparator's V−, and the generator and the probe move with it. Figure 8-11 shows the leads.
At these voltages, the value you set on the supply and the voltage at its terminals can differ by a few tens of millivolts. That is the size of the steps. Read the measured output on the supply's display, or a multimeter from 13i to ground, and use that reading.
Figures 8-12 to 8-14 are three of the steps. The comparator switches cleanly well below the 0.9 V estimate. Then, over less than a tenth of a volt, OUT's pulses lose their height, and OUT stops switching. The table gives this board's readings beside those of my first build.
| V−, measured | OUT on this board | OUT in my first build |
|---|---|---|
| 0.56 V | full-height pulses, 0.0 V to 5.04 V (Figure 8-12) | full-height pulses with flat tops |
| 0.50 V | not measured | pulses, their tops starting to round |
| 0.48 V | pulses only 1.08 V high (Figure 8-13) | short, rounded pulses |
| 0.45 V | not measured | small bumps where V+ is lowest |
| 0.42 V | a step of about 0.1 V, no switching (Figure 8-14) | flat, no switching |
So the NPN pair's floor is near 0.45 V, about half the first estimate. On this board at 5 V, the trigger's 1.67 V reference sits more than a volt above that floor. So an NPN comparator would do the trigger's job here.
The NE555 has less margin, and that is probably why its trigger comparator is PNP. The chip puts two transistors on each input, a Darlington pair, which adds a second base-emitter drop. That would lift this floor to about 0.45 + 0.65 = 1.1 V. At the NE555's minimum supply of 4.5 V, the trigger's reference is 1.50 V, and the control pin can pull it lower still. A PNP pair keeps its tail's room at all of these levels.
The 0.9 V estimate assumes that the tail keeps its full current, about 0.93 mA, however low the inputs go. As V− falls, the emitters fall with it, and Q4, the tail's output transistor, has less and less voltage across it. When Q4 saturates, it passes less current. A smaller current needs a smaller base-emitter drop, about 60 mV less for each factor of ten. So the emitters can follow V− further down than the estimate allowed.
OUT keeps its full height while the tail still gives Q7's base the few microamps it needs. The pulses shrink when the tail current falls below that. That point works out near 0.55 V. On this board it lies between the 0.56 V reading, with full-height pulses, and the 0.48 V reading, where the pulses reach only 1.08 V. This explanation fits the readings. I have not measured the tail current near the floor.
Q7 lifts OUT to +5 V through the 4.7 kΩ pull-down. The current that Q7 must pass is
If Q7's current gain is about 100, its base needs
While V+ is below V−, that base current is the tail current that Q2 takes out of node A. From the full tail current to 11 µA is a factor of
The number of factors of ten in 85 is its logarithm to base ten:
At about 60 mV for each factor of ten, the pair's base-emitter drop shrinks by
So the drop falls from about 0.65 V to
Q4 still needs a few hundredths of a volt across it, about 0.02 V. V− must sit at least this far above ground for the tail to pass 11 µA:
Below about 0.55 V, the tail passes less than Q7's base needs, and OUT's pulses lose height. In the table, the tops are flat at 0.56 V and start to round at 0.50 V.
In Ken Shirriff's analysis of the NE555's die, the threshold comparator uses NPN transistors and the trigger comparator uses PNPs. In his words, this "allows the threshold comparator to work near the supply voltage and the trigger comparator to work near ground." The chip also puts a Darlington pair on each input, so its inputs sit two base-emitter drops from the pair's emitters. For an NPN pair, that would raise the floor measured here by another drop, to about 1.1 V.
The floor test is complete. Leave the leads where they are: power supply output 2 and the generator on the threshold comparator's 1 kΩ resistors, and the probe on 11e, where Experiment 9 starts. Power supply output 2 has fed one comparator at a time. Experiment 9 replaces it with references made on the board, one for each comparator.
No flaw joins the list in this experiment. The trigger comparator is the threshold comparator turned over, and it keeps the fixes of Experiments 4 to 7. The NPN floor near 0.45 V is a question of margin in the NE555. On this board the trigger works with either pair at 1.67 V, so the floor stays off the list.
| # | Flaw | What it means | Found | Status |
|---|---|---|---|---|
| 1 | No reference input | The threshold is the transistor's own 0.6 V. Nothing outside the circuit sets it. | Experiment 3 | Fixed in Experiment 4: the differential pair |
| 2 | Wide transition | The output needs a band of input to travel from high to low: about half a volt with one transistor, about a quarter of a volt with the pair. | Experiment 3 | Fixed in Experiment 6: the active load |
| 3 | Thermal drift | The base-emitter voltage falls about 2 mV per degree, and the threshold falls with it. | Experiment 3 | Reduced in Experiment 4: two matched transistors drifting together |
| 4 | Tail current follows the common-mode level | A resistor tail passes more current as both inputs rise, and the output moves with it. | Experiment 4 | Fixed in Experiment 5: the current mirror tail |
| 5 | Output low trapped a diode drop below the reference | The output's low level sits a diode drop below the reference instead of at ground. | Experiment 6 | Fixed in Experiment 7: the PNP output stage |
Three questions on this experiment: node B's ceiling, what Q14 does to it, and why the trigger comparator uses PNPs.
1. With Q14 disconnected, node B cannot rise above about 2.3 V because…
Q9 saturates when its collector reaches its emitter. The pair holds its emitters one base-emitter drop above the lower input. With V− at 1.67 V, the ceiling is near 2.3 V: Experiment 7's floor at node A, reversed.
2. With Q14 connected and V+ above V−, node B stops near 0.65 V because…
When node B is about 0.65 V above ground, Q14's base takes Q9's current. Node B stays there, and Q14 switches on hard and pulls OUT to ground.
3. On this board the NPN threshold comparator switched cleanly with V− at 1.67 V. The NE555 still builds its trigger comparator from PNPs, probably because…
The floor test stopped this NPN comparator near 0.45 V. A Darlington input adds another drop, to about 1.1 V. At the NE555's 4.5 V minimum supply, the trigger's reference is 1.50 V, and the control pin can lower it further. A PNP pair holds its emitters above its inputs and hangs its tail from +5 V, so its limit is at the top of the supply.
The board now carries both of the 555's comparators. The threshold comparator has its NPN pair next to ground and takes 3.33 V. The trigger comparator, the same circuit turned over, has its PNP pair next to +5 V and takes 1.67 V. The threshold comparator's output switches between 0 V and 5 V, and the trigger comparator's output should do the same. Both outputs are high while V+ is below V−.
The floor test measured how close to ground the NPN pair keeps working: near 0.45 V, on this board and in my first build. The NE555's Darlington inputs and lower supplies leave less room than that, which is probably why its trigger gets the PNP pair.
Both references still come from power supply output 2, one comparator at a time. Experiment 9: The Voltage Divider makes them on the board from the 5 V rail and measures how they follow the supply. It also shows why that removes the supply from the 555's timing.