SCR 12V to 5V USB Converter

SCRs are the switching devices for this unusual DC to DC converter that is suited for the ubiquitous 5V USB power source as well as numerous other applications. This novel power circuit topology uses an intermediate resonant link that facilitates self-commutation for both SCRs – while SCRs are easy to turn on, they tend to be difficult to turn off in DC applications. This discussion covers primarily the relatively complex control circuitry -the power circuit and theory were previously covered in this article:
SCR DC to DC converter. All circuitry consists of discrete components and should be relatively easy to simulate.
Schematic of the 12V to 5V USB Converter Circuit

Not for the novice or fainthearted
Getting this to work took all my expertise. Fortunately, you can benefit from this published work, but be assured that it cannot work unless EVERYTHING is in order simultaneously –that is why I cannot accept the absurd theory of evolution (incredibly complex life forms developed and optimized by themselves…).
The occasion for this design
It seems that in the Philippine Islands, there is one tough electronics instructor who seems to have a penchant for thyristors. One assignment given to his students was to come up with a solar charge control and USB power supply regulator using SCR power devices. Eventually, this request filtered down to me via electroschematics.com. Perhaps some wondered how I came up with the unorthodox idea for the SCR solar charge controls previously posted. At first, I thought this was crazy, but eventually realized that he was simply attempting to get his students to stretch their minds, attempt to think outside the box and realize that there are numerous means of solving virtually any problem. For me it has been a challenge and lots of fun.
SCR drivers
Two programmable unijunction transistors (PUTs) are employed to drive the SCR gates. The PUT is well suited for driving the primary of a pulse transformer. A 0.01uf capacitor discharged into the primary of the 1:1 pulse transformer is sufficient to generate a 70mA, 20uS gate pulse. In observing the gate pulse oscillograph, it can be seen that the relatively slow gate pulse rise time leaves something to be desired —this is the result of transformer leakage inductance caused by poor primary to secondary magnetic coupling.
Control Algorithm
SCR1 fires when DC output < 5V, AND SCR1 is forward blocking, AND 70uS time delay has elapsed.
SCR2 fires when SCR1 is reverse blocking, AND 70uS time delay has elapsed.
Implementation of this algorithm results in operation much like a well-timed sewing machine. The delay function is required to provide sufficient time for the thyristors to commutate –according to the specifications on this particular device, turn-off time is 50uS. Many power devices do not have a turn off time (Tq) specification.
How it works
For firing SCR1, the minimum delay time is a function of the charging of C8 –this delay is extended by the effect of the voltage regulator. D3 is the TI TL431 adjustable shunt voltage reference that is strapped for minimum (2.5V). Q7 compares this reference voltage to the feedback voltage through the feedback voltage divider that consists of R16 & 17. Q6 both inverts the feedback signal and references it to common potential so that it may shunt C8. Q5 turns on when there is voltage across the capacitor network –it essentially tells the firing circuit that the capacitor network is empty and needs a new charge. Q5 & 6 also form an AND function so that both conditions must be satisfied before C8 is allowed to charge. When C8 charges to approx 8.6V,Q8 fires and triggers SCR1. SCR1 subsequently charges the capacitor network and does not stop at 12V because at that point the inductor is fully charged and must keep conducting until the capacitor network voltage reaches approx 24V.
When the voltage across the power supply output reaches 1.2V, Q6 turns on and shunts the gate of Q10 thus turning off the startup timing capacitor. Q6 also turns on under short circuit conditions thus limiting short circuit current.
For firing SCR2, Q4 is the firing circuit. When the voltage across the capacitor network exceeds 12V, Q1 turns on, Q2 turns on and C10 is allowed to charge –at this point, the 70uS timing period commences. Actually, SCR1 does not block until L1 is fully discharged, but it will happen after another quarter-cycle of the resonant tank elapses (16uS). This allows another 54uS to elapse before SCR2 can fire and that satisfies the turn-off time spec for SCR1 (Tq).
Specifications

• Output voltage: 5V (may be trimmed by varying R17)
• Voltage regulation: Approx 1% no load to full load
• Response to load voltage changes: Immediate
• Output current: 500mA
• Efficiency @ FL: 55% (higher at 14V input voltage)
• Short circuit current: 1.3A (may be continuous without damage)
• Current limit: Not measured, but exceeds 1.3A

Output capacitor
C7 must be a low ESR (Effective Series Resistance) type, or must consist of parallel devices in order to reduce output voltage ripple. The Panasonic device recommended on the schematic is a potential candidate. The capacitance may actually be much greater than 1000uf without issues.
Bugs encountered
All issues involved the SCR1 driver. Initially, it would not wake up with a load connected. This was traced to SCR discharging the resonant tank into essentially a short circuit because the voltage across C7 was very low. At this point, the free wheeling current flowing through D1 kept SCR2 alive until the subsequent firing of SCR1. This resulted in both SCRs conducting simultaneously —not a good situation because it causes the load voltage to approach 12V when it is supposed to be 5V. This was corrected by adding C9 to delay the turn-on of SCR1 by about 240uS —normally it is about 70uS. The low voltage conditioning circuit consisting of Q9 & Q10 senses when the load voltage is less than 1.2V —below this voltage, Q10 is connected in parallel with C8 thus delaying the SCR1 firing circuit. When the voltage exceeds 1.2V, C9 is disconnected and the timing becomes normal.
The second bug involved latch up of Q8 —this was apparent when the 12V power input was teased (very severe and aggravating condition). This caused the output voltage to die and would not restart. This was solved by removing the back diode that was connected across the primary of T2, and reducing the impedance of the Q8 gate bias network that effectively increase the valley current (reset threshold) of Q8.
With these issues corrected, operation was extremely reliable and robust.
Conclusions
While it works great, this busy circuit is definitely not a preferred solution. The efficiency is not that great (55%) –perhaps it can be further optimized. However, it is a great experimental circuit that will expand the experimenter’s knowledge in many areas. On the other hand, there may just be a niche application where this circuit is a good solution.
Acknowledgement: Thanks to RencoUSA Inc., Sanford FL, for samples of the mini-drum inductor (RL-5480-4-22). I have always appreciated their excellent support and competitive prices.

Photos

Observe how much the circuit changed from the original concept.

This took up the entire length of the protoboard –busy circuit!

Check out the LEM current transducer that I use for measuring current –the current is actually inductor current rather than capacitor current. The inductor current includes the free-wheeling current through D1.
For the future

• High voltage SCR DC – DC applications
• LED Chaser firmware
• What is the difference between an operational amplifier and a comparator