Output protection for DIY references

[e61_phil]

Hello,

hopefully I haven't overlooked another thread with the same topic. I think this topic is already discussed anywhere here in the forum but hard to find.

The question is: How to build a good output protection for a DIY 7V/10V reference. The Fluke 732C can withstand nearly everything what a calibrator can source (from short circuit to 1000V). This would be nice to have for a travel reference.
The reference should withstand a short or a voltage from outside without damage nor any change in the output voltage (after releasing the short )

Does anyone here have some good circuit idea for a protection like that?

Best regards
Philipp

[Kleinstein]

Protecting to a 1000 V level can be tricky.

Up to about 600 V one could use a pair of depletion mode MOSFETs with a resistor in between. This could be viable for output that should deliver up to about 1 mA.  It might still need extra ESD protection.

[Andreas]

Hello,

I think a calibrator can deliver max around 15 mA on the voltage output (more is not necessary to drive a 100K KVD)
The TransZorb diode (11V standoff voltage) at the output will have to withstand only around 200 mW as maximum. Right?

The cirquit diagram of a 732B is available: usually a short cirquit protected transistor pair after a chopper amp together with the SA11A.

with best regards

Andreas

[TiN]

57xx can do 50mA <220V ranges, 20mA on 1.1kV.

[chuckb]

Up to about 600 V one could use a pair of depletion mode MOSFETs with a resistor in between. This could be viable for output that should deliver up to about 1 mA.  It might still need extra ESD protection.

The dual depletion mode MOSFETs can also protect the path between the Zener and the output driver. Many opamps have diodes connecting the inverting and noninverting inputs. When the output is shorted to ground, the Zener is grounded also. Two Infineon BSP135s with a 1000 ohm current sense resistor will limit the current between the zener and the output stage to 1ma. Or you can use a 10k ohm resistor to do the same current limiting.

The BSP135 has the lowest turn on voltage that I have found so far.

Another option is to use the OPA189 Zero drift opamp for the output stage. It does not have diodes between the input pins.

[T3sl4co1l]

I'm quite fond of using a buffer op-amp with a series resistor between amp output and connector pin; feedback wraps around the outside (use a remote sense connection if you like) so the DC value remains correct, but EMC components can be added so that, not only is the amp's output current limited by saturation and the series resistor, but ESD is clamped without putting current (beyond ratings) into the amp.

For higher voltages, the depletion FETs others have mentioned, would do a fine job in place of the series resistor (or augmenting it).  You'd need an MOV or GDT to handle ESD/surge, and you'd want to add some zeners and resistors and filtering to the transistors to ensure there's no nastiness with blowing them up under transient conditions due to their capacitance.

Not so great for AC performance, but should be good for a DC ref.

Tim

[Andreas]

57xx can do 50mA <220V ranges, 20mA on 1.1kV.

Hello,

Ok then it will be clamped to ~15V at the SA11A within 732B which gives 750 mW
for the poor 3W diode before the 5730 switches off due to overload.

I see still no problem as long as the buffer can withstand these 15V.

with best regards

Andreas

[IanJ]

I'm quite fond of using a buffer op-amp with a series resistor between amp output and connector pin; feedback wraps around the outside (use a remote sense connection if you like) so the DC value remains correct,

Which is what I do with the PDVS2 and use an LTC1250 op-amp which has indefinite short circuit protected output (~20mA).
Can still be abused though to point of failure.

Ian.

[e61_phil]

Thanks for all your replies!

I attached a first version for further discussion. One have to add some decoupling caps of course.

U1 could be a OPA189. One could think about a higher voltage than ground for V- to reduce the common mode voltage.

R1 und R2 are the resistors from the precision divider to lift the voltage to 10V (VHD200 for example).

U2 is a jelly bean R2R output opamp which  delivers the output current and dissipates the heat to keep U1 cool. I added some amplification to keep U1 away from the rails. U2 could be directly sourced from the batteries (12V lead in my case). While U1 is driven by the LDO (10.5V).

The output protection is a simple TVS diode and R3.

The difficulty is to operate with 10.8V supply (12V lead acid battery discharged) only, I think. Therefore, I didn't put a transistor pair in the output. The current limit circuit will eat up to much voltage.

Edit: Added common for 7V reference

[Andreas]

Hello,

I ask myself: if you use a OPA189 with 5nV/deg C offset change why do you need another buffer to keep load changes away?

If you are using a (low ohmic) lead acid battery which can absorb some load then you could also use 2 normal diodes for output protection (provided the battery is always connected):
a SA11A with 11V standoff voltage has around 12.5V+/-5% breakdown voltage (1mA) and perhaps 15V at 50 mA so the 100 Ohms will conduct about 35 mA into the output protection diodes of U2 at minimum battery voltage.

In any case I would use a around 100nF low inductive capacitor across the output pins to protect against EMI. But in this case R4 should perhaps go to the other side of R3 to avoid oscillations.
Be aware that standard TVS diodes also have relative high capacitances in the order of some 100 pF. So some OP-Amps which have only 1nF maximum load capacitance stability may have problems if longer screened cables are used together with the zener.
The 100nF also builds a capacitive divider against ESD. (the ESD generator has around 150 pF) This is one reason why I did not use a TVS up to now in my output cirquits. But of course the 100nF do not help against a calibrator output.

Further EMI measures: Either some Ferrite beads (like in my AD587LW reference). But they have rather high DC resistance compared to a 732B. Or the solution from Cellularmitosis (some Ferrite cores within wiring which are low ohmic).

with best regards

Andreas

[chuckb]

I learned yesterday that the choppers need a low impedance at their input pins. The OPA189 with 100k to the non inverting input had a 650uV offset. With a 20pf or 2000pF COG cap to ground the offset was less than 0.2uV. Other op amps like the LTC2058 and he ADA4522 need the full 2000pF to get below 1uV. I don't know yet if higher capacitance helps them more.

[e61_phil]

@Andreas: You're right. There is only 4K temperature rise and therefore almost no change in offset..
How do you calculate the 35mA? If the output is driven to the clamped 12V, the OP will try to sink as most current as it can. Therefore I would assume 12V/100R = 150mA? But the current should be limited by the OP.


I played around with the 100nF output cap, but I don't like the results . If you want to get that stable you have to compensate quite early in frequency (see attachment).

@chuckb: My feedback divider has around 7k output impedance. Do you think that is already a problem?

[Andreas]

How do you calculate the 35mA? If the output is driven to the clamped 12V, the OP will try to sink at most current as it can. Therefore I would assume 12V/100R = 150mA? But the current should be limited by the OP.


I played around with the 100nF output cap, but I don't like the results .

Hello,

my idea was:
the zener clamps to around 15V@50 mA
the 15V go from the zener via the 100R over the OP-Amp internal output ESD protection diode to the +VBAT (10.8V when empty) rail
so we have 15V - (10.8 + 0.7V) = 3.5V difference.
Ok of course I did not regard that the output also will draw some current.

If you have a better idea against EMV it is welcome.
Earlier in this thread I had about 30uV difference between unbuffered (filtered) and buffered (not filtered) reference output before I added the 100 nF.

By the way: it is time to develop a home lab compatible test method for checking the effectiveness of EMI measures.
I have in mind coupling some RF between the reference and the DMM onto the cable via a 100pF - 1nF capacitor.
Perhaps we would also need a decoupling network on the DMM side.
Unfortunately I have only a signal generator up to 20 MHz. I think that some 10kHz to around 300 MHz would be fine.

with best regards

Andreas

[kj7e]

Hello,

hopefully I haven't overlooked another thread with the same topic. I think this topic is already discussed anywhere here in the forum but hard to find.

The question is: How to build a good output protection for a DIY 7V/10V reference. The Fluke 732C can withstand nearly everything what a calibrator can source (from short circuit to 1000V). This would be nice to have for a travel reference.
The reference should withstand a short or a voltage from outside without damage nor any change in the output voltage (after releasing the short )

Does anyone here have some good circuit idea for a protection like that?

Best regards
Philipp

Here was my solution;
https://www.eevblog.com/forum/metrology/ltz1000-10v-buffer/

Video in post 21 shows the output protection at work.

[e61_phil]

That seems to be the thread I overlooked..

Did you saw any EMI problems with that circuit? And it cannot sink current.

[kj7e]

Not seen any EMI issues from the buffer.  While it may not sink current, its fairly immune to back feeding a voltage, if you connected the output to a 12v power supply for example.

[MiDi]

I learned yesterday that the choppers need a low impedance at their input pins. The OPA189 with 100k to the non inverting input had a 650uV offset. With a 20pf or 2000pF COG cap to ground the offset was less than 0.2uV. Other op amps like the LTC2058 and he ADA4522 need the full 2000pF to get below 1uV. I don't know yet if higher capacitance helps them more.

First time I read about real impact of charge injection at chopper inputs 
It is weird that this is not even mentioned in datasheets.

[MiDi]

Here was my solution;
https://www.eevblog.com/forum/metrology/ltz1000-10v-buffer/

Did I get it right that the ocp and protection against back feeding voltage relies on an opamp with "indefinite" differential voltage capability between inputs?

For protection against back feeding voltage the base-emitter junction of bjt would be reverse biased into breakdown, how do you know that it is in soa?
(asked me the same for Dave's bodge board for 121GW)

You mentioned that the LT1010 was rejected, I would appreciate if you could tell what the drawbacks are.

I like the simplicity of your design 

[kj7e]

Indefinite short, yes.  Momentary feedback of a modest level, such as 12 to 15 volt or so.  I make no claims to how it could handle an accidental 30v or more.

I could not get the noise level down on the LT1010 and the LT1250 + 2N5088 seemed even cleaner/simper to implement.

[Echo88]

Regarding LT1010-rejection: https://www.eevblog.com/forum/metrology/show-me-the-stability-of-your-voltage-references/msg1349247/#msg1349247