Clamping circuit discussion

[sh9]

I am referring to the clamping circuit as decribed in chapter 6 of the OSD335x tutorial series pdf document Rev. 9.

The description explains how the circuit clamps the voltage to 1.55 V, but I don’t think it can do that while significant current flows through it. I believe that Figure 33 shows an impossible situation. It shows current flowing through the TLVH431 while there’s no voltage across it.

In my understanding, the TLVH431 needs at least 1.25 V between cathode and anode to operate. The clamping voltage is higher than that by at least the base-emitter voltage of Q1, which will be around 600-700 mV. It follows that the minimum clamping voltage will have to be around 1.9 V, which is still within the maximum 2 V dictated by the AM335x, but the margin is of course much smaller.

As the TLVH431 supports a clamping current of up to 80 mA, it might be possible to use it without an additional transistor. In this case the clamping voltage could be lower, down to 1.25 V.

Please check if my understanding of the circuit is correct, or if I’m missing something.

[Neeraj Dantu]

Sh,

TLVH431 controls the state of transistor Q1 being ON/OFF. The reference voltage input to TLVH431 is established by the voltage divider comprised of R40 and R41. The voltage divider makes it such that if the voltage difference between VDDSHV_3P3V and SYS_RTC_1P8V is > 1.55V, the reference voltage threshold of TLVH431 of 1.24V is met at it’s VREF input(Node A). This changes voltage at node B(base of Q1) turning transistor ON/OFF based on TLVH431 turning ON/OFF. Note that the voltage divider(R40 and R41) controls the threshold voltage here.

Please let us know if you have further questions.

Best,

Neeraj

[sh9]

Neeraj,

Your description is a “digital” view of the circuit, which is only valid as long as the circuit conditions permit it. I posit that this is not the appropriate way to describe the operation of the circuit. For example, Q1 is not “ON” or “OFF” in this sense. Sure, it can be off, but when it is on, it operates within its “linear” region, not in the saturated region. It can’t be regarded as a switch.

The TLVH431 likewise is working in a linear fashion, and can’t be regarded as a switch. While “ON”, if you want to call it that, it conducts current from cathode to anode, but it is unable to pull the cathode voltage lower than the voltage at the reference pin.

So my point is, you are oversimplifying if you view the circuit in terns of switches being on or off. It is essentially a linear circuit, which can only be described correctly when operating limits and characteristics are being taken into account.

To put some detail to this. You write:

“The reference voltage input to TLVH431 is established by the voltage divider comprised of R40 and R41. The voltage divider makes it such that if the voltage difference between VDDSHV_3P3V and SYS_RTC_1P8V is > 1.55V, the reference voltage threshold of TLVH431 of 1.24V is met at it’s VREF input(Node A).”

This is entirely correct.

“This changes voltage at node B(base of Q1) turning transistor ON/OFF based on TLVH431 turning ON/OFF.”

This is the problematic part. Neither Q1 nor the TLVH431 simply go from OFF to ON at this point. Rather, it is an amplifier in a feedback loop. Its operation is like this: The TLVH431 starts to sink current from its cathode to its anode, and part of this current flows through the base of Q1 (R43 takes the other part, but that’s not important for now). Q1 being a current amplifier, an amplified current thus flows from emitter to collector, which is the clamping action that we want. As a result of this, relying on the input impedance of the source we want clamped, the circuit regulates the current through Q1 such that the voltage at the reference pin of the TLVH431 stays at 1.24 V. However, for this to work, both TLVH431 and Q1 need to operate within the linear range of their operating conditions. The TLVH431 needs at least 1.24 V between cathode and anode. A silicon transistor needs about 600-700 mV between emitter and base. Those two voltages are in series with respect to the source that we want to clamp, so the circuit isn’t working in its linear range when we set the clamping voltage to 1.55 V, because that’s not enough for providing the sum of 1.24 and 0.65 V.

The result is that the circuit may appear to work, but its clamping characteristic will be weak, because the available gain is very low. You will find that depending on the source inpedance, the resulting clamping voltage would be significantly higher than 1.55V, i.e. the circuit’s own impedance is quite high for something that is supposed to clamp.

Note that a simulation might not tell you the true story here, depending on the accuracy of the TLVH431 model in conditions outside the recommended operating conditions.

A correctly dimensioned circuit would keep the operating conditions of the TLVH431 within the recommended range, even during clamping action. This means that there would have to be at least 1.24 between cathode and anode in this situation. If a simulator shows a lower voltage, the result should be viewed with suspicion.

[sh9]

I have built the clamping circuit, with the small change of replacing the TLVH431 with the LMV431, because I only hat the latter at hand. They should behave very similarly. The other parts are as shown in the tutorial document.

I put the thing into my trusty old curve tracer, and plotted the I/V curve, similarly to what you would do with a diode. I have attached a snapshot made with my smartphone. You can see the display settings at the right, i.e. the horizontal scale goes up to 2 V, and the vertical scale up to 20 mA.

As you see, it doesn’t take much current to get beyond 2 V, which occurs at around 20 mA. The LMV431 can only sink 30 mA, but if the object was to keep the voltage below 2 V, this would have been enough, i.e. the transistor could have been omitted.

Note the kink at around 1.55 V, which shows that the clamping action does indeed start at the right voltage. At that point, there’s practically no current through the transistor, the little current you see is through the 500 Ohm resistor. An ideal clamp would produce a line that is very close to vertical. Instead you get a curve, which from there on is similar to a diode curve. That is because of what I described above. The result is a very poor clamping circuit.

Attachments:

1. 

   Clamping-circuit.jpg
   

[Neeraj Dantu]

Thanks for the detailed analysis of the clamping circuit.  We agree that this is probably not the most ideal clamping circuit.  However, it should be good enough to accomplish its design goals.  The AM335x requires that if the VDDS and the VDDSHVx input voltage rails are ramping down simultaneously then the voltage difference between these rails should be within 2V of each other or there could be some slight damage to the IO circuits which could over time result in reliability issues (see Section 6.1.2 of the AM335x datasheet).  

 

  During power down, the TPS65217C cannot guarantee this condition since it cannot control the loading on the power output rails.  The TPS65217C will shut off LDO4 (SYS_VDD3_3P3V on the OSD335x-SM, which is generally used to power the VDDSHVx inputs) at STROBE 4 which is is much earlier than LDO1 (SYS_RTC_1P8V on the OSD335x-SM, which is generally used to power VDDS) at STROBE 15.  However, it cannot guarantee the down power ramps of each rail and therefore, the clamping circuit is a precaution against violating the 2V requirement.  Given that each rail is being powered down, the currents involved to keep the difference between the voltage rails < 2V should be small.  Additionally, there are many situations where there is no need at all for the clamping circuit.  

 

  The clamping circuit itself was developed by Texas Instruments (see http://www.ti.com/lit/ug/slvu731b/slvu731b.pdf Section 5) and used in the reference designs for the OSD335x to showcase the most robust design possible for our customers.  In general, we recommend customers putting down the footprints for the clamping circuit during prototyping and test if it is necessary to populate the clamping circuit for production.  If you have a better recommendation for the clamping circuit, please let us know so that we can add that to our discussion of the clamping circuit.  Again, thank you for digging into this, please let us know if you have any additional questions and we look forward to working with you on your design.

[sh9]

Neeraj,

thanks for your detailed reply.

Regarding my recommendation for a better clamping circuit, I don’t have a tried and tested design, but I can offer two starting points for designs you’d need to verify if they suit your requirements.

The simplest solution would apply if the clamping current doesn’t exceed 80 mA. You wrote that the currents involved should be small, so the chance of the simple circuit being adequate should be good. The circuit would use the TLVH431 to sink the current, i.e. there’s no transistor Q1, and the 500 Ohm resistor R43 is replaced by a piece of wire. This is the simple adjustable shunt voltage reference circuit. It clamps hard, so for a better margin I would choose a clamping voltage somewhat higher than 1.55 V, say 1.8 V, by adjusting the values of R40 and R41.

If a circuit is needed that can sink more current than 80 mA, one may want to look into using a part like the LT6650, which is a reference plus an amplifier. It is similar to the TLVH431, but has separate pins for input and output voltage, so it can drive the output lower to accomodate an external transistor like the 2N2907A. I trust the reader to be able to come up with a modified schematic. I’m unaware of a similar part in TI’s portfolio, but there might be one.

Regarding the questionable clamping circuit originating with TI, you may want to relate my findings back to them, so they can have a look at their application note, and perhaps come up with a new revision.

Cheers

Stefan

 

[Neeraj Dantu]

Thanks Stefan! We will forward your feedback.

Best,

Neeraj

[Author]

Posts