INA105: INA105 to make a low-noise Howland current source

Dear Tamara,

thank you so much for the reference to the newest app note on Howland current source.

However, my issue with current noise still applies: the Howland current noise is dominated by the Rs value. The only way I see to lower the noise is increasing Rs and at the same time using a very high power supply value, which is quite an unfeasible approach.

Could you please suggest me any other solution? I am aiming at sourcing/sinking up to 100uA into a .

50 ohm load to ground with a current noise comparable to the current noise of 1M-10M resistor.

Thank you so much for your time.

Alberto

Hi Alberto,

hmm, compared to the voltage noise of a plain 50R resistor the additional noise contribution of your Howland source pump seems to be ultra low:

Kai

Dear Kai,

thank you for your (very correct, Indeed!) analysis.

I realized that I gave a misleading description of my setup.

I am going to use the Howland current pump to inject current into the TIA shown in the attached TINA file:

embedded_gain_TIA.TSC

The Current Source will be used to bias active and/or passive components for our 1/f noise analysis.

According to our lab tests, the TIA has a voltage PSD of 1,5 uV/sqrt(Hz) flat in the band 1Hz -> 100 kHz.

The noise of the Howland pump is dominated by the 100k resistor, which is very high for our application.

That's way I need a way to inject current with very low noise in our TIA. With the present setup, I am going to study

the Howland noise, more than my sample noise ...

Increasing the power supply of the Howland pump would help, however I am wondering if you can suggest me any

solution capable to source/sink current at very low noise. Any suggestion is extremely welcome !

With my best regards,

Alberto

Ciao Alberto,

so this is your concern?

alberto_opa657_3.TSC

Kai

As already discussed, increasing R9 can considerably decrease the noise:

Kai

But this does not fully solve your problem, right? :-)

Kai

Dear Kai,

thank you for integrating all the pieces of the puzzle!

The entire circuit works great with input currents in the range up to 1uA.

In some applications, I need to bias the device with higher currents (up to 100uA). As a consequence, the Howland
resistor R9 should drop to 100k. Again, as a consequence, the 'current subtractor' should use a 100k resistor, in order to be able to cancel out the static bias current and let the TIA to amplify just the low frequency noise. By doing that, I will see in input the thermal noise of the 100k resistors.

I guess, that's the best we can do. Or can we do any better ?

Thank you so much for time and extremely appreciated support.

Dear Kai,

you get my point, indeed...

I need higher current with lower noise ...

Am I playing against Boltzmann Law ?

Thank you once again!

Alberto

Hi Alberto,

for lowest noise you coud do it this way:

alberto_opa657_5.TSC

But you have to deal with dangerous high voltage (1kV)!

Kai

Dear Kai,

thank youn again for your interest in my case (against Boltzmann !).

Following to your suggestion to roam in the HV domain, I have found these two references:

According to such a modified and 'bootstrapped' Howland circuit, I should be able to do something in the 100V range. Not quite what I need,

yet better than a 1-kV electric chair for all my (electronic) sins !

Do you think it could be a viable solution? Could I dream of high current and low noise by using 100V and 1M ? It would give me a quite  honest 100uA.

How does noise from power supply transfer to the output ?

Any suggestion from you is (as always) very welcome!

Best,

Alberto

Hi Alberto,

I have two basic questions:

A current source is commenly used to drive a current into a load without being affected by the voltage drop across this load. This would be important, if the voltage drop across the load is relevantly high. But in your case the current is driven into a TIA which presents a very low impedant input. 100µA times 50R is only 5mV. That's your whole voltage drop across the "load" and is entirely negligible compared to the 100V of 1000V voltage source of my last scheme. So a voltage source plus a 1...10MOhm current limiting resistor would nearly exactly act like a current source, with the advantage, that's it's way simpler to construct a HV voltage source than a HV Howland current pump :-)

Another issue is the current you want to generate. I guess you just want to bias your sample with a pre-current, so that the DC servo is still able to "close" the loop? This would mean that you don't want to generate complexly changing currents? You would only alter a DC current up to the moment the DC servo can close the loop?

Am I correct?

Kai

Dear Kai,

you are definitely correct:

->I guess you just want to bias your sample with a pre-current, so that the DC servo is still able to "close" the loop? This would mean that you don't want to generate complexly changing currents? You would only alter a DC current up to the moment the DC servo can close the loop?

Yes indeed! I do need to bias my sample with a fixed dc current. For example, I need to force a given Ids into a FET channel in order to measure

the 1/f as a function of the bias point. The DC servo should be able to close the loop and let the TIA to amplify just the very-low-frequency noise, not

the bias current to set the FET operating point. I mounted your servo. It works GREAT !!!!

I do not need to generate complex current trends/waveform. Just the static current to bias my sample.

For many new materials, 1/f noise does change with the current flowing into the sample. I need to force 0.1uA, 1uA, 10uA or 100uA into the

sample and, for each current, to measure the 1/f noise. Each measure takes many hours and the current generator should be stable during the run.

-> But in your case the current is driven into a TIA which presents a very low impedant input. 100µA times 50R is only 5mV. That's your whole voltage drop across the "load" and is entirely negligible compared to the 100V of 1000V voltage source of my last scheme. So a voltage source plus a 1...10MOhm current limiting resistor would nearly exactly act like a current source, with the advantage, that's it's way simpler to construct a HV voltage source than a HV Howland current pump

The current is injected into my sample, which is then connected to the TIA input. For example, I need to inject 5uA into a drain-source channel of a FET,

with a given Vgs. The source is connected to the TIA input (virtual ground). In such a setup, the Vds is set by the FET according to its (Id, Vds vs Vgs) curve. In this case, the compliance of the current generator could be an issue.  Some new materials also have very high resistivity, so they develop quite a high voltage even a bias of  few uA. However I do not plan to exceed 10-20V across my sample.

=============================================================================================

If you feel a HV generator is the best for my application, could you please send me any reference about low noise HV power supply ?

When I was a teenager, I used to build valve radios. Many years have passed (too many...) and now I have no clue on how to build a realible

1 kV power supply for my setup ...

I am very, very grateful for your help.

With my best regards,

Alberto

Hi Alberto,

I would do it this way:

alberto_hv_1.TSC

alberto_hv_noise.TSC

Kai

Hi Alberto,

some words on the circuit:

The two back-to-back mains transfomers provide galvanic isolation from mains, and if you take safety class II transformers, you even don't need to connect the 0V of circuit to protection earth. With safety class I mains transformers, on the other hand, you would have to do this. An advantage that should not be underestimated in your application :-)

When it comes to fusing, standards have changed some years ago. Today most mains transformers have to have the fuse sitting in the secondary winding. This has to do with the soft characteristic of these transformers. The effect is that a short at the output side cannot be properly detected by a fuse in the primary winding. With the TALEMA's toroidal mains transformer for PCB mounting, I usually take, this is no issue and the fuse is allowed to be placed in the primary winding. I'm talking about this one:

Another advantage of this toroid transformer is its extra low magnetic stray field :-)

R1 simulates a bit the internal winding resistance of mains transformer. Because of this it should not be extra put in your circuit, (dnp = do not populate).

The circuit is optimized for your long-lasting measurements. So to deal with mains voltage fluctuations I have taken two cascaded zener diode chains. This decreases the output voltage changes resulting from mains voltage changes by a factor of two compared to the single zener diode network.

The IRF840 is easily available and withstands 500V. To prevent higher frequency noise and ripple to be transmitted by its rather high drain to source capacitance, R2 + C2, C4 and R6/R11 + C5 form hefty low pass filters.

Later more, Nikolaus is knocking on the door...

Kai

Dear Kai,

thank you again for your support.

I was wondering about the meaning of R1. Now, it is clear.

In Italy we have the power grid at 220V, so I am forced to put in series the two primary windings. Does it change

anything in your scheme?

The low pass filter is very welcome, indeed. Beside variations in the supply, my worst enemy in this kind of measure is the 50Hz noise.

Any suggestion to kill it is extremely appreciated.

Many many thanks !

Best,

Alberto

Hi Alberto,

we here in Germany have 230VAC too, like you in Italy. So I also have to put the two 150VAC windings in series.

Beside providing some pre-load, R5 and R8 are used to discharge all the caps when powering down the circuit (by removing the mains voltage at the input). During powering down Z4 protects the IRF840.

R7 is the obligatory 1k gate filter resistor to slow down the IRF840 a bit and to improve stability.

The circuit does not contain an extra short circuit protection, because the short circuit current is limited by R6/R11 anyway. For safety reasons I would mount R12 directly behind R6/R11 and put it into the same enclosure as the rest of the circuit. Route the right side of R12 to the output connector. This limits the short circuit current and touching current to under 240µA which should no longer present a problem. Put the whole circuit into an enclosure so that not anybody can touch the circuit inside!!

To avoid thermal runaway of the zener diodes' threshold voltages and the gate source voltage of IRF840 provide good cooling. Even the use of a cooling fan could make sense, if lowest thermal runaway is desired.

Zener diodes show considerable manufacturing tolerances. So try to get the much more precise BZX85B (+/-2%) types instead of BZX85C (+/-6%) types and/or hand select the zener diodes for lowest tolerance.

To be able to handle +/-20% mains voltage fluctuations, the circuit generates "only" +240VDC. So a 2M4 resistance is needed to create a current of 100µA into a zero Ohm load. As 10k + 10k + 1M are already sitting in the circuit, another 1.38M (1M2 + 180k, e.g.) have to be connected outside. Mount these resistors as close as possible to the zero Ohm input.

When you need a current of 10µA, on the other hand, choose a total resistance of 24M. This will allow lowest noise operation :-)

Kai

Dear Kai,

thank you for all the explanations !

I am going to make a lot of noise (measurements...) with my new setup!

Thank you so much! I am really grateful !

Best,

Alberto

You are welcome :-)

Kai