PI noise

[green]

Lately I've been using an integrator and two post amp stages with a .01usec feed back TC each stage followed by a negative absolute value and a 2.3Hz 5pole butter worth filter. Recorded some scope pictures with the integrator and post amp feedback TC changed to .033usec without the abs. value and 5 pole filter. R7 and C4 are the feedback RC for the integrator, 10k input resistors in series with input switches(1C integrator). A photo cell was mounted 8 inches above the coil, a 200mm diameter disk was glued to the bottom of the water bottle swinging from a pendulum, US nickel stuck to the disk. 200mm/.2second=1meter/second. The nickel swings across one end or from end to end across both. Bottle was pulled back about 18 inches and let go then caught on the rebound, crosses coil twice, I think I like the circuit without the absolute value and 5 pole better.

[Davor]

1C integrtors require a fast op amp, or otherwise they can't cope with charge injection pulses. I'd expect this to be noisy. A minipulse differential integrator is an interesting solution, but it requires a fast op amp, as it combines 1C (inverting) and 2C (non-inverting) paths. Charge injection may be flattened a bit by adding gate resistors (gate stoppers) so that garden variety op amps may work well here. Otherwise, aim for op amps with well over 1 MHz gain bandwidth product.
Eric, thanks for the C12 tip. I wouldn't use tantals though - too unpredictable, and also slavery promoting. Your anagram is so cool, but mine is better...

Vodka Rivers

[Ferric Toes]

1C integrtors require a fast op amp, or otherwise they can't cope with charge injection pulses. I'd expect this to be noisy. A minipulse differential integrator is an interesting solution, but it requires a fast op amp, as it combines 1C (inverting) and 2C (non-inverting) paths. Charge injection may be flattened a bit by adding gate resistors (gate stoppers) so that garden variety op amps may work well here. Otherwise, aim for op amps with well over 1 MHz gain bandwidth product.
Eric, thanks for the C12 tip. I wouldn't use tantals though - too unpredictable, and also slavery promoting. Your anagram is so cool, but mine is better...

Vodka Rivers

You are a savvier Dork than I thought. Yes, rivers of Vodka are much better than rusty toes.

I haven't found charge injection from the gate drive to be a problem. In the Minipulse days things were not too critical but even in later designs with delays and sample pulses of just a few uS I didn't see a big problem. The fast spikes at the beginning and end of the sample could be seen on a scope but disappeared if you followed the integrator with a simple RC LP filter. Provided the gate drive is low impedance both ways then the charge injection is followed by charge suction, therefore the effect cancels. This is the situation with CMOS gates using a + and - supply. At least that is my story.

Erotic Serf.

[green]

Two questions.(1)Is there a better place to connect the switches with a 1C integrator, between the preamp and integrator input resistors or between integrator input resistors and integrator - input? (2)Can you do GEB with a 2C integrator, with hold or normal?

[Ferric Toes]

Two questions.(1)Is there a better place to connect the switches with a 1C integrator, between the preamp and integrator input resistors or between integrator input resistors and integrator - input? (2)Can you do GEB with a 2C integrator, with hold or normal?

If you are using a CMOS 4066, or similar, it is what is known as a bilateral switch in that it can conduct either way and input and output can be swapped. The switching function is obtained by +Vdd or -Vss on the control gate. In my case, I normally use +5V and -5V for the RX rails and signal ground at 0V. The same supplies are for the CMOS, which means that you can put the gate where you like, either before the integrator input resistor or after. If you want the hold arrangement then it has to be after, so as to isolate the capacitor from the feedback resistor. I prefer the hold arrangement as there is less leak down of the capacitors as the parallel resistor is out of circuit. Good capacitors (matched in the case of 2C) are a must and also an opamp with very low input bias current.

Either the 1C or the 2C can be used for GEB but you have to do a polarity inversion somewhere to subtract out the ground signal. You can do it in the preamp by the addition of an inverter using a 1C. Or do it in the integrator by using the 2C. Three or four sample pulses are required to sample the decay signal at different points in time and on the GEB channel you normally require additional gain, hence the noise issue.

Eric.

[Orbit]

Hello Eric !
Do you's 4066 Better from J113 as in the case gs 4 ?

[Ferric Toes]

Hello Eric !
Do you's 4066 Better from J113 as in the case gs 4 ?

Hi Orbit,
Yes, I have not used J113/J112 as analogue switches for a long time. They had to be matched for ON resistance whereas 4066 has 4 switches in one package so the match is already good. The rest of the pulse generator driving the 4066 needs to be between +5 and -5 rails.

Eric.

[green]

If you are using a CMOS 4066, or similar, it is what is known as a bilateral switch in that it can conduct either way and input and output can be swapped. The switching function is obtained by +Vdd or -Vss on the control gate. In my case, I normally use +5V and -5V for the RX rails and signal ground at 0V. The same supplies are for the CMOS, which means that you can put the gate where you like, either before the integrator input resistor or after. If you want the hold arrangement then it has to be after, so as to isolate the capacitor from the feedback resistor. I prefer the hold arrangement as there is less leak down of the capacitors as the parallel resistor is out of circuit. Good capacitors (matched in the case of 2C) are a must and also an opamp with very low input bias current.

Either the 1C or the 2C can be used for GEB but you have to do a polarity inversion somewhere to subtract out the ground signal. You can do it in the preamp by the addition of an inverter using a 1C. Or do it in the integrator by using the 2C. Three or four sample pulses are required to sample the decay signal at different points in time and on the GEB channel you normally require additional gain, hence the noise issue.

Eric.

Thanks for the reply.

Tried the switch both places with the 1C integrator and didn't see a difference in the noise but wondered if I was missing something. I think I understand the 1C integrator, gain=R fdbk/R in*sample rate*sample time, frequency cutoff=1/2pi/R fdbk/C fdbk. Tried to come up with the gain and frequency cutoff formulas for the 2C with hold. Sample time wasn't in the gain formula but was in the frequency cutoff formula so I didn't see how to do GEB. What are the gain and frequency cutoff formulas for the 2C with hold?

green

[Ferric Toes]

Thanks for the reply.

Tried the switch both places with the 1C integrator and didn't see a difference in the noise but wondered if I was missing something. I think I understand the 1C integrator, gain=R fdbk/R in*sample rate*sample time, frequency cutoff=1/2pi/R fdbk/C fdbk. Tried to come up with the gain and frequency cutoff formulas for the 2C with hold. Sample time wasn't in the gain formula but was in the frequency cutoff formula so I didn't see how to do GEB. What are the gain and frequency cutoff formulas for the 2C with hold?

green

I would have to try and find my notes from many years ago for the 2C with hold. Not easy. Frequency cutoff is varied by the internal
loop gain, the more gain the higher the cutoff frequency. I seem to remember that the integrator gain remained the same.
How have you done GEB so far?

Eric.

[Orbit]

Thank you Eric !

[green]

I would have to try and find my notes from many years ago for the 2C with hold. Not easy. Frequency cutoff is varied by the internal
loop gain, the more gain the higher the cutoff frequency. I seem to remember that the integrator gain remained the same.
How have you done GEB so far?

Eric.

Schematic I posted awhile back. There have been some changes but operation is the same. 1kHz oscillator starts sequence, take (A)target and (B)ground EF samples then turn coil on, then take (A)target and (B)ground samples. Switch to disable GEB samples if no ground problem. Target delay and target sample time adjustable, GEB delay and sample time fixed, adjust target sample time to cancel ground signal. GEB delay maybe 1 to 5usec. GEB sample 100usec.

[green]

[Tried the switch both places with the 1C integrator and didn't see a difference in the noise but wondered if I was missing something.]

I didn't notice a difference before. Yesterday I disconnected the leads from preamp out to the switch inputs and connected them to common to test integrator noise. Noticed if I touched common or either of the power supply pins I had a small change in output. Rewired the circuit with the switches to integrator -input, resistors to preamp. That problem fixed, don't think it changed the operating noise level. Ordered a different amplifier and switches to try to reduce noise level.

[Ferric Toes]

Can you draw a timing diagram that starts with the TX pulse (coil on) and then all sample pulses, delays and widths from that point. Just a typical setup will do. I want to see the relationship of each part of the cycle along to the next TX pulse, particularly how far out your GEB sample is from the 'off' edge of the TX.

Eric.

[green]

Can you draw a timing diagram that starts with the TX pulse (coil on) and then all sample pulses, delays and widths from that point. Just a typical setup will do. I want to see the relationship of each part of the cycle along to the next TX pulse, particularly how far out your GEB sample is from the 'off' edge of the TX.

Eric.

Some scope pictures. The 1kHz oscillator starts the control sequence. Takes EF samples before turning coil on. Target sample time was adjusted to cancel some hot ground. Times are what I've been using, not saying they are optimum. 1C integrator

[green]

Can you draw a timing diagram that starts with the TX pulse (coil on) and then all sample pulses, delays and widths from that point. Just a typical setup will do. I want to see the relationship of each part of the cycle along to the next TX pulse, particularly how far out your GEB sample is from the 'off' edge of the TX.

Eric.

Got a PM from a member reminding me I didn't start the sequence at Tx on. All scope pictures are external triggered when Tx command turns off, T at top center of screen. Tx on starts when second EF(100usec, normal out) sample turns off. Using the upper picture, time between Tx off and first EF sample is around 750usec. Lower picture, GEB sample starts 20.7usec after turn off.