The Gain Change Game

[Tinkerer]

Tweaking the HH

Hey Tinkerer and J.L. King.

Let's see if I'm starting to get it. First to Tinkerer's thoughts:

"OK, I downloaded the HH Rev D1.
I see no bucking coil. With a well balanced IB coil, things look quite different than a mono coil. Please post a scope screen shot of the TP 4 output."

So here's what the coil looks like:

[ATTACH]6916[/ATTACH]

Now to mix Tinkerer's thoughts with those of J.L.King:

I have kludged together the following with an LM4562 and a dip header and I think both gains are now at 33, correct?:

[ATTACH]6919[/ATTACH]

From the following 'schematic' (R13;HHvD1 disabled):

[ATTACH]6917[/ATTACH]

(Back to Tinkerer's question) the result is the following wave form at TP4 - volts/div=.2, time/div=5uSec (HHvD1):

[ATTACH]6918[/ATTACH]

So now my question for both of you...how can I improve this? Am I on the right track?
The wave form is cleaner (noise wise) and seems to be the same as before (except for that slight dip just before the climb back to zero, though MUCH quieter). Anyone else have any thoughts on this?

Thanks folks, GTB

Your coil looks good and you still have the possibility to improve the balance.

Looking at the scope picture:
I assume that you are using a x10 probe. If this is the TX and RX signal, you are using a 18us Tx pulse.

To be able to Discriminate FE, you will have to improve the balance.
Try to move the loose part of the lead of the bucking coil, while observing the scope. If you see any movement, adjust so that the TX pulse goes to 0V.
If this does not work, try adding one more turn to the bucking coil and move this around.
If this does not work, you have too much feedback, meaning too many turns on the bucking coil.
If you do not want to undo your coil you have the option to reduce the current in the bucking coil.

Tinkerer

[KingJL]

Hey there, J.L. King,
Is this what you meant? It's what I'm trying now and the gain is definitely better.

[ATTACH]6921[/ATTACH]

Would a cap across the 2nd stage gain resistors R2 or R3 help?

Definately not across R3. The need for a cap across thefeedback depends on the charactaristics of the op-amp. I have not studied the one that you are using, so I can't say. If the op-amp stages are stable without them, then I wouldn't use them. For example, an AD797 working at lower gain need the or else they will tend to oscillate due the the very fast slew rate. At a gain of 1 it needs ~35pf to tame the beast. At a gain of 10 it needs ~15pf. At a gain of 40 it really doesn't need any. So look at the datasheet for your op-amps. Usually they will state the required shunt capacitance needed in the feedback. Looking at your signals that you posted, I suspect that you do not need the caps across the feedback.

Regards,
J. L. King

[KingJL]

Hey there, J.L. King,
Is this what you meant? It's what I'm trying now and the gain is definitely better.

[ATTACH]6921[/ATTACH]

Yes. Sorry about the 'pin 2' business. Glad you sorted that out!

Regards,
J. L. King

[GT Blocker]

J.L.King

Thanks for the help on the gain question for the pre-amp. I still have one question about the pre-amp gain now that I understand the cap/gain part.

In the schematic above, how do R2 and R3 relate to each other? In other words, if I were to change R2/33k to 47k, would R3/1k change as well? Are the two values 'co-dependent' so to speak. If so, is there a formula to determine their relationship?

My next question has to do with the integration stage. I was under the impression that the lower the integration stage gain, the better (lower noise/error amplification during integration). Is this correct?

Would there be any benefit to lowering the gain in the integrator stage and making up for it in the final amp stage(s)?

Since the integrator is next in line in the gain change game, thoughts?

Thanks, GTB

[KingJL]

Thanks for the help on the gain question for the pre-amp. I still have one question about the pre-amp gain now that I understand the cap/gain part.

In the schematic above, how do R2 and R3 relate to each other? In other words, if I were to change R2/33k to 47k, would R3/1k change as well?

Yes if you wish the gain to be the same.

Are the two values 'co-dependent' so to speak. If so, is there a formula to determine their relationship?

Yes. In your schematic Rf is R2 and Rg is R3. The formula for gain of a non-inverting configuration is G = 1 + (Rf/Rg).

My next question has to do with the integration stage. I was under the impression that the lower the integration stage gain, the better (lower noise/error amplification during integration). Is this correct?

I'll pass on that one. I do not believe there is anything wrong with the gain of the HH integrators. You might want to give some thought to changing the filtering parameters though. The integration time of the HH design is a little fast for my taste. But that is stictly an opinion.

Would there be any benefit to lowering the gain in the integrator stage and making up for it in the final amp stage(s)?

I don't think so. But you would see benefit by feeding the output of the integrator through a LPF (w/co at ~10-15Hz). Even a single pole passive RC 160Hz LPF makes a significant difference (I fed the output of the integrator through a 10K to the .47 uF coupling capacitor of the autotune stage, and connected a 100nF capacitor between the junction of the 10K & .47uF and ground giving me a c/o of 159Hz). I also did some breadboard test lab setups feeding through a 3 order active LPF (10-15Hz) with excellent results.

Regards,
J. L. King

[GT Blocker]

so far, so good

I've implemented J.L.King's LPF as per above and there is a definite improvement. The dual amps on the pre-amp with the lower gains and the LPF are worthy mods. The 3 order active LPF sounds worthy as well, but I have no idea as to how to implement that (sigh...more research).

A few questions for Tinkerer,
In post #9 of this thread you mention, "With a gain of 100 in the next stage, the difference between sample 1 and sample 2 will be amplified by 100. Therefore if you sample too early, when the decay curve is not close enough to 0V, the detector becomes unstable when the second stage saturates."

Could you expand on that?

Also in your last post you mention,"Your coil looks good and you still have the possibility to improve the balance.

Looking at the scope picture:
I assume that you are using a x10 probe. If this is the TX and RX signal, you are using a 18us Tx pulse.

Correct. The output at TP4, dual coils (as in an above post).

To be able to Discriminate FE, you will have to improve the balance.
Try to move the loose part of the lead of the bucking coil, while observing the scope. If you see any movement, adjust so that the TX pulse goes to 0V.
If this does not work, try adding one more turn to the bucking coil and move this around.
If this does not work, you have too much feedback, meaning too many turns on the bucking coil.
If you do not want to undo your coil you have the option to reduce the current in the bucking coil. "

Could you post a scope shot of what you mean?

I have been very careful about observing the TP4 wave form while tuning the bucking coil. When you say,"adjust so that the TX pulse goes to 0V", do you mean TP2?
I found that I had to tune the coil for both long and short pulse widths. Tuning at the minimum pulse width without checking the coil's wave form at the longer widths showed the balance to be off considerably at the longer timings. Re-balancing the bucking coil for the longest pulse widths seems to be the answer. Am I wrong?

Thanks, GTB

[KingJL]

I've implemented J.L.King's LPF as per above and there is a definite improvement. The dual amps on the pre-amp with the lower gains and the LPF are worthy mods. The 3 order active LPF sounds worthy as well, but I have no idea as to how to implement that (sigh...more research).

Attached is a solution for a 3rd order LPF with c/o 15 Hz, gain ~1. For the opamp you can use TL07x(x = 2,4), TLC07x(x = 2,4) TLC27Mx (x=2,4,9), etc. I would start to watch for current drain of the components. The TLC27Mx series has an Iq of only 0.28ma per device wheras the TLC07x have an Iq of 2.5ma per device. The 27M series has a low GBW, but for audio frequencies are more than suitable. For the aforementioned LPF, they are superb.

Regards,
J. L. King

[Tinkerer]

IC8 gain

A few questions for Tinkerer,
In post #9 of this thread you mention, "With a gain of 100 in the next stage, the difference between sample 1 and sample 2 will be amplified by 100. Therefore if you sample too early, when the decay curve is not close enough to 0V, the detector becomes unstable when the second stage saturates.

Could you expand on that?

A lot of good questions. I will answer one at the time.
Consider your sample #1 to be at a level of 50mV.
The second sample at 0V. This can be adjusted with the level adjuster (R16,17,18,19,C27)
With a gain of 100, you would have 5V at TP5, meaning the Opamp is saturated.
Another example: you have adjusted the level to give 0V at TP5. Now you locate a large target that gives you 50mV at TP5. Again the Opamp is saturated.
So this is the limitation of the gain of IC8.

Tinkerer

[Tinkerer]

Balancing the IB coil

To be able to Discriminate FE, you will have to improve the balance.
Try to move the loose part of the lead of the bucking coil, while observing the scope. If you see any movement, adjust so that the TX pulse goes to 0V.
If this does not work, try adding one more turn to the bucking coil and move this around.
If this does not work, you have too much feedback, meaning too many turns on the bucking coil.
If you do not want to undo your coil you have the option to reduce the current in the bucking coil. "

Could you post a scope shot of what you mean?

I have been very careful about observing the TP4 wave form while tuning the bucking coil. When you say,"adjust so that the TX pulse goes to 0V", do you mean TP2?
I found that I had to tune the coil for both long and short pulse widths. Tuning at the minimum pulse width without checking the coil's wave form at the longer widths showed the balance to be off considerably at the longer timings. Re-balancing the bucking coil for the longest pulse widths seems to be the answer. Am I wrong?

Here is a drawing of the balanced coil output at TP4. I don't have a HH to do the actual picture.
The wave shape depends somewhat on the amount of power in the coil and the coil specs.
The coil can only be balanced for one specific coil current and\or pulse width.
Anytime you change the coil current you have to adjust the damping and the balance of the coil.
On the drawing I show the sampling point where the signal crosses the 0V line.
At this point, the signal goes positive for FE or negative for all non magnetic metals, or the opposite, depending if the coil or preamp is inverting.
I also show the effect of a FET transistor blocking the Flyback. In this case it is a shunt to ground. It can also be in series to block the signal totally.

Tinkerer

[GT Blocker]

Filter Q for J.L.King

Attached is a solution for a 3rd order LPF with c/o 15 Hz, gain ~1. For the opamp you can use TL07x(x = 2,4), TLC07x(x = 2,4) TLC27Mx (x=2,4,9), etc. I would start to watch for current drain of the components. The TLC27Mx series has an Iq of only 0.28ma per device wheras the TLC07x have an Iq of 2.5ma per device. The 27M series has a low GBW, but for audio frequencies are more than suitable. For the aforementioned LPF, they are superb.

Regards,
J. L. King

Hey there again, J.L.,
As I was assembling the components for the above filter, I ran into a passives shortage. I was gettin a bit discouraged (considerings all of the parrallel/series combos), when I noticed in your attachment that the filter type was a Bessell.
I remembered that I had a few MAX292 samples around from my failed sine wave output attempt. These are 8th-order Bessell filters that need only a cap to set the cutoff freq. The data sheet is here (5 pgs): http://datasheets.maxim-ic.com/en/ds/MAX291-MAX296.pdf Sooo....I tried this:



The cutoff freq should be ~33hZ with a 100pF, right? (I haven't been able to find a 220pF in my 'junk' to try ~15hZ C/O.)
Would there be that much difference? This filter seems to be working great so far....

And finally, here's the basis of this test:



Would the actually be any advantage to the MAX294 with anti-aliasing in this app (HHvD1)?

Thoughts?

Thanks, GTB

[KingJL]

Hey there again, J.L.,
As I was assembling the components for the above filter, I ran into a passives shortage. I was gettin a bit discouraged (considerings all of the parrallel/series combos), when I noticed in your attachment that the filter type was a Bessell.
I remembered that I had a few MAX292 samples around from my failed sine wave output attempt. These are 8th-order Bessell filters that need only a cap to set the cutoff freq. The data sheet is here (5 pgs): http://datasheets.maxim-ic.com/en/ds/MAX291-MAX296.pdf Sooo....I tried this:

[ATTACH]6955[/ATTACH]

The cutoff freq should be ~33hZ with a 100pF, right? (I haven't been able to find a 220pF in my 'junk' to try ~15hZ C/O.)
Would there be that much difference? This filter seems to be working great so far....

And finally, here's the basis of this test:

[ATTACH]6956[/ATTACH]

Would the actually be any advantage to the MAX294 with anti-aliasing in this app (HHvD1)?

Thoughts?

Thanks, GTB

Can't see any advantage. What do your tests show?

Regards,
J. L. King

[GT Blocker]

follow up

I guess that I worded that clumsily. I was wondering if there is much to be gained from a 33Hz to a 15Hz cutoff in the first question. Guess when I get a 220pF I can give it a whirl.

As far as test results, most have been subjective tests.

I'm not sure of the signal TPs and wave forms to answer any objective questions, atm. As I've said, I'm new to this analog game.
I really have no idea just what 'anti-aliasing' is, for that matter. As for an elliptical filter, yah right.
I just wondered if there was an obvious advantage, given the relative simplicity of the mod.

I did notice an immediate change (decrease) in background noise at TP4 after the dual op amp mod in the front end, so I'll start checking upstream and see what I can learn.

Thanks for the help, GTB

[GT Blocker]

Hey again J.L. King and Tinkerer

J.L. King,
The Max 292 Bessel works great at the 15Hz cutoff. For some reason, the Max 294 elliptical version doesn't work at all, even though the pin-outs and passives remain the same. Weird. Perhaps that is where the anti-aliasing comes in...more tests are in order.

Thanks for the help, GT Blocker

Hey Tinkerer, I forgot that I moved the coil part over to coils, duh...oops Part two is there.............

[KingJL]

The Max 292 Bessel works great at the 15Hz cutoff.

Great to hear that. I have some MAX260's that I plan on using in a project that I have going.

For some reason, the Max 294 elliptical version doesn't work at all, even though the pin-outs and passives remain the same. Weird. Perhaps that is where the anti-aliasing comes in...

I don't think that the reason it doesn't work at all has anything to do with anti-aliasing.

Regards,
J. L. King