High Impedance Voltage Amplifier Designs

Prof. Shyam:

Two questions:

1) Could your signal source be characterized and used as a charge source? (It sounds as though it could.)

2) If so, have you considered using a charge amplifier configuration?

Last time I raised that subject, some controversy arose, but that LOW IMPEDANCE configuration could reduce some of your error and noise problems. I'd be glad to discuss it further.

DickL

Dear DickL

Your point is valid and actually applied in some of the designs such as PM Tube, Channel Electron Multiplier, MCP and Photodiode charge amplification. Charge is injected into very low impedance generally 50 Ohms due to cable impedance matching. Unfortunately in such fast devices, charge often lasts only for few ns and pulse shape is not important.

In current design that is being proposed here, the pulse shape is very important. Take the case of shockwave sensing through piezoelectric pressure transducer and you wish to retain the shape of the pulse in frequency domain 100kHz to 10MHz, like the way you do in broadband amplifiers. In fact, this design is of high impedance broad band amplifier.

Charge source right now is typically 1us deliberately constructed using a 1us chopped constant current ion beam. Both time and charge is precise. I can change the parameter if I wish. This simulation device is also an experimental device for sensing unknown current / charge pulses. Charge integrating device often used in nuclear detector is shown below is not a good circuit for current application.

Dear Prof. Shyam:

Remember that the low impedance of the charge amplifier is capacitive, not resistive, and time constants can be long. Cable impedance matching is not a consideration because the cable is essentially "shorted out" by the large input capacitance of the amplifier (Cfeedback X Open loop gain). You might find the characteristics of this amplifierinteresting:

http://www.amptek.com/a250.html

Best regards.

DickL

I agree. I am not using a capacitor to hold charge but 1G Ohms sure will have 0.25pF capacitance. This capacitance will work on early edge of the signal and also in the end discharge of capacitor.

I have not used feedback capacitor and hence, capacitor charging does not stabilizes the amplifier back to zero. Resistive element works in the flat zone of the pulse and causes 1pA/mV leakage in the sensor capacitance charge. Assuming that source capacitance is 100pF and leakage resistance 1G Ohms, the time constant is 0.1s which is very large for 1us charge pulse and I do not expect any visible pulse shape change due to leakage current. 0.25pF will not add too much capacitance to source capacitance of 100pF. JFET Gate source capacitance is 4pF which is used for charge sensing is 4% of the detector capacitance. This means only 4% charge is used for sensing by the JFET. Transfer of charge will take place in <5ns which is rise time of the input pulse.

This is all working as per plan. However, noise is the problem so I am considering expert advice of Rhabe to change the shield to u-metal shield.

Dear Shyam,

be careful with µmetal, it is very sensitive to any stress and strain (both!).

And: be sure to have electrically conducting layers of shield and magnetically conducting layers of shield separated by electrical insulation.

If no magnetic shield is suitable then two (or more) layers (isolated) of conductive material is much better than 1 only. And magnetic shielding may be ineffective at high frequencies as magnetic data µ, hysteresis, Hc, Bsat, Br are functions of frequency.

I had much better performance from using very low carbon steel in replacing not efficient µmetal, but this was 0 to 1 MHz. Cheapest and easy to get material was sheet for deep drawing automobile parts. Lowest yield stress will give best magnetic performance.

At your frequencies electrical shielding should be sufficient but only 1 layer is often not enough.

RHABE

Dear Rhabe

Can you explain, why multiple layers will work better for electrical shielding? Is it due to increased surfaces?

My noise frequencies are 10MHz - 900MHz range and perhaps some Blue tooth 2.4GHz also may come in. 900MHz being cellular frequencies here and may have high radiation levels, they will be dominating. 10MHz looks is coming from broadband copper cables.

This is industrial use amplifier so how do I place these multiple layer shielding on it. It there any commercial manufacturer of small casing for such RF shielding.

I want to go to the bottom of this noise level.

Noise sources have clear signatures and are repetitive. Noise and signal are not synchronous and noise can appear randomly as per its own repeat frequency as seen in third picture. Noise riding on peak was selectively captured.

Noise alone

Noise signature existed even when power source was battery and not regulated power supply. However, using battery its strength came down so I can say that power supply acted as an antenna for the noise source.

Dear Shyam,

with one shield only the electromagnetic wave is exciting eddy currents in the shield and the magnetic field of these eddy currents is inducing the noise currents and voltages in the circuitry.

With two shields (isolated) the first layer eddy currents excite eddy currents in the second layer (etc) and as the magnetic/electric field of the first layer eddy currents is much lower than the primary incoming (noise) wave the shielding will be much better with some shields stacked upon each other.

I am not sure about best choice of material as usually copper or aluminum is used but the eddy currents are not well damped in these so I suspect that a higher resistivity may be better. (Copper oxide on the inner surface of waveguides is very bad because very bad conductivity. Silver sulphide is not so bad as having good conductivity.)

Similar problem: thermal radiation shielding.

And: be very cautious with incoming and outgoing cabling also if very short. We blasted early laser diodes (as cost was in the many 100$ each) with only 2cm of cabling to the battery. Shields on cables are not very good, braided shields are leaky and generate noise by rubbing at any of the wires contacting points (many millions). So if critical either rigid shields or conductive-plastic shielding should be tested.

RHABE

Dear Rhabe

I think you have a point. Looks like you are a Professors. Good. Your students will be much better engineers. I agree.

I will try these ideas as they are valuable.

Hi Shyam,

if you are shure that the amplifier and its connecting network is not significantly contributing to the measured noise (?) then I see many more maybe sources:

How is cabling to and from the amplifier, length, type, cable noise by seismic motion?

Most of the noise on the yellow trace seems to be quantisation noise, so a better resolving input amplifier should be used.

If cabling (rigid microwave cabling?) and amplifier is good then rework your box: it should have two walled structure or better three-walled.

The internal surfaces all be highly polished (as in hollow waveguides) not to generate electromagnetic stray fields. Did you band-limit the structure of your shielding pot to the frequencies you want to measure? It will work as a pot-resonator if big enough and it will transmit some of the currents and voltages that flow only in a very thin surface layer.

So try the next box: copper-iron-copper, or (I never tried but I suppose to be good) titanium - iron - titanium. Any layer isolated from the next and tied together at the point of zero potential - if existing else a floating zero to establish.

Aluminum is not too good: too good a conductor so low damping.

What about the type of capacitor you used?

In principle I agree with DickL that a charge amplifier should do best - where is the distortion that you mention coming from?

RHABE

Dear Rhabe

I use high quality coaxial shielded cables for signal and also for power supply.

Casing are still Aluminum 3mm thick.

There may be some minor leaks at sliding joints even though casing is almost light tight.

u-metal shield perhaps is a good idea which was used in 11 track Tape Recorder in olden days. I have seen some of those high speed recorders at similar high frequencies using FM. I think 500kHz was the frequency used. I have used u-metal shield for PM Tubes. Basically it is magnetic and electrical shield in one.

I will have a look at broadband communication signal of my Telephone line now being used for Internet. This line is only 50 cm away from my experimental set up. If this is what going in my amplifier then amplifier shielding sure is of bad quality. I need to improve it.

My friends used ordinary MS boxes and their amplifiers work better may be one reason to account for this. I think your point is good one and I consider it an expert advice.

Dear Shyam:

You may want to add some ferrite cores to your power and input lines in addition to better shielding. They are available with defined bandwidths and are low cost.

Additionally, your 100pF capacitor will inject 6.4μV of noise at 300'K according to the equation: where KB is Boltzmann's constant, T is temperature in Kelvin and C is capacitance. This will add to your noise floor.

Reducing the capacitor electron noise comes down but voltage noise increases. For low voltage amplifier larger capacitors are better and for low charge measurement by charge injection method, low capacitance helps much better. Using large resistance to form voltage from low charge has a problem of resistance noise but it increases by square root and voltage increases by XR and hence large resistances from 1M to 100M yield good results.