TGSL Experiments

[golfnut]

noise in preamp

just ran some models for IGSL pre amp



LM388 4.5nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 100uV

mod trial- 220 10k output noise at 1Hz = 14uV





LT1115 0.9nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 6.6uV
mod trial- 220 10k output noise at 1Hz = 0.3uV


The 1115 is 8pin dip - may get one and try it. Hassle to wire in tho - shame not same pinout.

http://www.linear.com/product/LT1115

S

[der_fisherman]

Commenting further on my last graph of the null phases:

1. I shouldn't have said switching the RX leads didn't change much. The graph looks pretty much the same, but now the phases corresponding to more overlap of the coils are the more positive phases, the opposite of the previous graph with the RX leads reversed. I guess this isn't any surprise.

2. Note that the previous two graphs were for a configuration using a standard USB cable. My next graph will compare data using a Belden cable to the USB cable for the non-grounded RX case, for the heck of it. I don't expect too much difference, possibly not even in the margin of error of my data points, but let's see.

3. By the way, I know my theoretical model of two sine waves 90 degrees apart is not exactly correct because my null signal becomes quite distorted near the minimum null voltage, doesn't look like a nice sine wave.

In fact, to measure phase, I measure the positive-going zero-crossing difference for one estimate and the negative-going zero-crossing difference for another, and average the two, because the waveform is so non-symmetric. Maybe dumb approach, but that's what I do.

To normalize the zero crossings, I center both waveforms (TX and RX) vertically on my oscilloscope center x-axis gridline, making the top and bottom peaks equidistant from the x-axis. This may not work so well for the RX null signal at minimum when the top half of the waveform may look sine-like and the bottom flattened, or vice versa, etc.

But hopefully averaging the two null-crossing phase differences help reduce the errors.

I noticed that Dennis the Mennis, in his tests, seems to have a nicer looking null signal as it moves across the minimum. Perhaps it is just noise in my workshop causing distortion.

Dennis the Mennis and I seem to be studying this null phase characteristic simultaneously, so it will be interesting to compare his results to mine, and the combined data may be helpful.

By the way, I'm not doing this to make a big deal out of null phase. In fact, I generally thought the null phase was not critical and depended on peoples shields, cables, etc. But it seems to be a topic of interest and so deserves some study in hopes of debunking any myths, and learning useful practices, etc.

-SB

I am watching, listening and reading, I also hope I am learning.
thanks to you all.
Andy

[simonbaker]

just ran some models for IGSL pre amp



LM388 4.5nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 100uV

mod trial- 220 10k output noise at 1Hz = 14uV





LT1115 0.9nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 6.6uV
mod trial- 220 10k output noise at 1Hz = 0.3uV


The 1115 is 8pin dip - may get one and try it. Hassle to wire in tho - shame not same pinout.

http://www.linear.com/product/LT1115

S

Looks "sweet".

It would be nice to have a central place on the forum we could catalog candidate parts like this with comments why chosen, so we can rummage there for ways to spruce up our MDs.

-SB

[Jerry]

just ran some models for IGSL pre amp



LM388 4.5nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 100uV

mod trial- 220 10k output noise at 1Hz = 14uV





LT1115 0.9nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 6.6uV
mod trial- 220 10k output noise at 1Hz = 0.3uV


The 1115 is 8pin dip - may get one and try it. Hassle to wire in tho - shame not same pinout.

http://www.linear.com/product/LT1115

S

Those are some interesting numbers that get even more interesting as I study them.

Just dropping the impedance with the LM388 gave a 17 db improvement.

Dropping the impedance on the LT1115 however gave a 27 db improvement.

Comparing the high impedance (4.7/220k) for each device gives a 27 db improvement between the LM388 and LT1115.

The same comparison with low impedance (220/10K) gives a 33 db improvement.

Here is a handy calculation for decibels.

http://www.sengpielaudio.com/calcula...lification.htm


Good food for thought.

Thanks for posting.

Jerry

[der_fisherman]

Changing of components

just ran some models for IGSL pre amp

LM388 4.5nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 100uV

mod trial- 220 10k output noise at 1Hz = 14uV

LT1115 0.9nV/root Hz

as drwg- 4k7 220k output noise at 1Hz = 6.6uV
mod trial- 220 10k output noise at 1Hz = 0.3uV


The 1115 is 8pin dip - may get one and try it. Hassle to wire in tho - shame not same pinout.

http://www.linear.com/product/LT1115

S

What does that mean to me as an owner of an unfinished IGSL? Should I make/plan such changes or is it too early for that.....?
Also, I don't remember seeing the LT1115 anywhere, or is it seen as a possible alternative to another chip? The LM388 for example?
I would be most grateful for some simply put infos so that I may also (try to) grasp the importance.
Thanks in advance.
Andy

[golfnut]

What does that mean to me as an owner of an unfinished IGSL? Should I make/plan such changes or is it too early for that.....?
Also, I don't remember seeing the LT1115 anywhere, or is it seen as a possible alternative to another chip? The LM388 for example?

Currently I havnt seen a system design/budget lineup for the IGSL - I dont know if there are any noise issues or gain issues, or dynamic range, sensitivity, etc

For a general purpose IB KIT IGSL its good fun and does ok. Id get yours going as is, experimental modding may slow you up and you may loose valuable detecting time asking more engineering questions


If preamp noise was proven to be a limiting factor for small target detection, this type of mod may well help.

Currently Id say it was an exercise in learning about noise from resistors and a look at quieter devices - and sharing useful stuff to readers.

S

[golfnut]

often really high spec - opamps are only one amp in an 8pin package


ease amp mods with this -- adapter $9

from here.. http://www.audiophileproducts.com/opamps

[simonbaker]

more null phase; USB vs Belden 8723 cable

Okay, some more null phase graphs! (roar of crowd)

Basically this time I compared nulling with my coils hooked through a USB cable vs. a Belden 8723 cable.

I have some pretty ugly connections (twisted wire connections at the coil end; vivisected Belden to USB connector for the PCB end...), but some useful differences came through.

The Belden cable has some very enthusiastic shielding, meaning each RX and TX pair is wrapped in a shield, then both pairs wrapped in an outer shield.

The USB cable has a good double shield (foil and braid), but all four wires are interior to the shield.

Comparing the USB to the Belden, it seems the Belden gets a much deeper null and the plots look closer to a theoretical ideal transformer, which would be two vertical lines separated by 180 degrees.

I tested with grounded and ungrounded RX lead as before. There was the characteristic deeper null in the ungrounded configuration, and a phase shift of about 9 deg between the two cases.

Conclusion:

Again, it seems the null phase shifts very rapidly at the null point and I feel it is due to the presense of another coupling between the RX and TX leads that is about 90 deg shifted from the magnetic coupling of the coils.

Comparing the USB against the Belden, it would seem that for unshielded RX, TX coils, the cable is causing the additional coupling to varying degrees, since the two cables caused different null voltages.

I will be interested to see how adding shields to the coils affects the null also -- is the null voltage dominated by the shields, or the cable, or equally by both?

I still do not have any feeling that the exact null phase is important; which side of the minimum you choose may have some importance which I don't know at this time.

Well, something to look at anyway. Excel graphs can be pretty...

The Belden cable is slightly longer than the USB cable, around 6 to 9 inches I would guess.

-SB

[Dennis the Mennis]

Commenting further on my last graph of the null phases:

1. I shouldn't have said switching the RX leads didn't change much. The graph looks pretty much the same, but now the phases corresponding to more overlap of the coils are the more positive phases, the opposite of the previous graph with the RX leads reversed. I guess this isn't any surprise.

2. Note that the previous two graphs were for a configuration using a standard USB cable. My next graph will compare data using a Belden cable to the USB cable for the non-grounded RX case, for the heck of it. I don't expect too much difference, possibly not even in the margin of error of my data points, but let's see.

3. By the way, I know my theoretical model of two sine waves 90 degrees apart is not exactly correct because my null signal becomes quite distorted near the minimum null voltage, doesn't look like a nice sine wave.

In fact, to measure phase, I measure the positive-going zero-crossing difference for one estimate and the negative-going zero-crossing difference for another, and average the two, because the waveform is so non-symmetric. Maybe dumb approach, but that's what I do.

To normalize the zero crossings, I center both waveforms (TX and RX) vertically on my oscilloscope center x-axis gridline, making the top and bottom peaks equidistant from the x-axis. This may not work so well for the RX null signal at minimum when the top half of the waveform may look sine-like and the bottom flattened, or vice versa, etc.

But hopefully averaging the two null-crossing phase differences help reduce the errors.

I noticed that Dennis the Mennis, in his tests, seems to have a nicer looking null signal as it moves across the minimum. Perhaps it is just noise in my workshop causing distortion.

Dennis the Mennis and I seem to be studying this null phase characteristic simultaneously, so it will be interesting to compare his results to mine, and the combined data may be helpful.

By the way, I'm not doing this to make a big deal out of null phase. In fact, I generally thought the null phase was not critical and depended on peoples shields, cables, etc. But it seems to be a topic of interest and so deserves some study in hopes of debunking any myths, and learning useful practices, etc.

-SB

Hi Simon ,

I made a videoclip in which I've wanted to show what happens during 'nulling' or maybe

better 'adjusting' the in this project unshielded coils .

I'm working on clips with other types of shielding : Alu / Mylar from VGA cable / Mylar from rescue blanket .

This all happens while shifting the unshielded coils hardly 1 millimeter !

I've posted it on Youtube : http://youtu.be/fzCbYp1kRTo?hd=1.

Please not that this clip is still under construction !

The testpoints before the different clips start speak for themself I think .

All are made in All Metal mode , I've got the TGS pcb but won't be much different compared to the TGSL or IGSL and others simular .

* It starts with Tx vs Rx output LF353 : what I'm trying to point out here is the initial

shift differences between different shielding materials : I had many ( as others )problems

trying to get the wanted '20 degrees' . At this moment we know that this shouldn't be a

an issue if you reach it or not.

* the second and third part of the clip is the Rx output LF353 vs Uc15 ( and

Uc12) FET : these voltage can vary a lot again within that millimeter .

* The fourth part : Uc15 vs Uc12

* The fifth and sixth part and perhaps most important :Uc12 &Uc15 vs LM308 outputs

the (huge) influence of Uc12 and Uc15 on the noise ....


Allthough only unshielded coils are displayed in this clip yet , I think possibly explains a lot

of problems seen and heard on this site ( poor detection and fals alarms ).

It will depend also on your envirometal noise of course but the basic seems to start here.


Now I know why Ivconic was spending lots of time to get the ideal point of overlap , I remember him writing that you should have patience while doing that ....


to be continued ...


Ps : I'm working also on a Belden and an USB cable which I can easily swap for

testpurposes like Simon did .

Whenever I'm there I'll let you know the results.


kind regards ,


Danny


Dennis the Mennis

[simonbaker]

Hi Simon ,

I made a videoclip in which I've wanted to show what happens during 'nulling' or maybe

better 'adjusting' the in this project unshielded coils .

I'm working on clips with other types of shielding : Alu / Mylar from VGA cable / Mylar from rescue blanket .

This all happens while shifting the unshielded coils hardly 1 millimeter !

I've posted it on Youtube : http://youtu.be/fzCbYp1kRTo?hd=1.

Please not that this clip is still under construction !

The testpoints before the different clips start speak for themself I think .

All are made in All Metal mode , I've got the TGS pcb but won't be much different compared to the TGSL or IGSL and others simular .

* It starts with Tx vs Rx output LF353 : what I'm trying to point out here is the initial

shift differences between different shielding materials : I had many ( as others )problems

trying to get the wanted '20 degrees' . At this moment we know that this shouldn't be a

an issue if you reach it or not.

* the second and third part of the clip is the Rx output LF353 vs Uc15 ( and

Uc12) FET : these voltage can vary a lot again within that millimeter .

* The fourth part : Uc15 vs Uc12

* The fifth and sixth part and perhaps most important :Uc12 &Uc15 vs LM308 outputs

the (huge) influence of Uc12 and Uc15 on the noise ....


Allthough only unshielded coils are displayed in this clip yet , I think possibly explains a lot

of problems seen and heard on this site ( poor detection and fals alarms ).

It will depend also on your envirometal noise of course but the basic seems to start here.


Now I know why Ivconic was spending lots of time to get the ideal point of overlap , I remember him writing that you should have patience while doing that ....


to be continued ...


Ps : I'm working also on a Belden and an USB cable which I can easily swap for

testpurposes like Simon did .

Whenever I'm there I'll let you know the results.


kind regards ,


Danny


Dennis the Mennis

It is a good start, and will be easier to follow when you add labels indicating exactly what case you are running at each time -- plus some separator, like a title page, between cases.

1. Did I notice something funny with the voltage at C15 as the signal continued to increase -- did the C15 voltage reverse near the maximum null voltage?

2. I also am very interested in your large increase of noise at the LM308 output when the RX signal moves away from the null point.

I don't understand it. My current thought is that it is not really EMI noise but possibly mechanical vibration that is amplified because the vibration modulates a larger residual null signal. Probably wrong. The other possibility is that the noise is some kind of jitter or modulation of the TX signal that is coupled more strongly when off-null.

What else could it be? I don't see why EMI noise would increase.

Can you try separating the coils completely and tilting them so they are minimum coupled, and check the LM308 output noise?

Regards,

-SB

P.S. I just tried the same null vs noise test on my setup (using a dfbowers PCB and unshielded coils, USB cable). My coils are mounted to plexiglass plates in my "nulling machine" as shown in a previous photo.

I did not get an increase in noise at LM308 output when I changed the coil overlap (unlike your test). It remained about the same for all overlaps and RX signal null voltages.

I think I am testing correctly.

Maybe you should try fixing your coils firmly when you move off the null and see if the noise voltage drops back down.

-SB

[golfnut]

Really clear Dennis,

I liked the green trace showing the mixer output DC offset shift with large signals. This is well documanted behavior of a direct conversion receiver. The Opamp input DC will shift here too, not great.


(As the Tx and Rx are on the same Freq the image lands on DC - it manifests itself as a DC component proportional to Rx signal.)

You can eliminate this behaior by AC coupling the fet mixer outpt to the opamp.

Or use an image reject mixer - like modern compact commercial receivers.



S

[simonbaker]

More coil nulling tests

Ok, nothing new here, just a repeat experiment with some unshield coils made by dfbowers. The previous tests used some early coils I made but not really to TGSL spec or quality.

So this graph shows RX voltage (at LF353 output pin 7) vs. phase as the coil overlap is shifted across the "null" point.

I compared three cables: a USB cable dfbowers provided (USB dfb), a USB cable I acquired (USB sb), and my Belden cable with extra paired wire shields.

My USB cable appears to have slightly thicker wires and more robust shield; the whole cable is slightly thicker than the USB dfb cable.

Conclusion:

The two USB cables were very similar with what looks like a slight phase shift offset between them -- not sure if that is significant or just measurement error. The USB sb cable may have slightly deeper/squarer nulling profile.

The Belden cable once again shows a very deep and square nulling profile compared to the USB cables. The minimum voltage is less than for the USB cables, and the phase stays constant longer over the voltage range, looking more "ideal" in my opinion.

The Belden cable looks very attractive for achieving a deep null and for seemingly introducing less "capacitive" (just a term, not sure what it is) cross-over signal.

However, this does not mean the Belden cable is better. The "capacitive" signal added to the magnetically coupled signal may not hurt detection at all. However, it seems to make a larger null signal, and would limit preamp (LF353) gain for some MDs. Our TGSL preamp gain however can handle a pretty hefty null signal it as designed.

Next Tests:

These graphs give a baseline for testing how various coil shields affect the same measurement (RX voltage vs phase).

I hope to test: kitchen foil shield; mylar-aluminum shield (space blanket); aluminum duct tape shield.

(Of course what we really want to know is how these cases compare in the field (depth, discrimination). I'm sorry I can't do that at this point; it would require building many coils with quite identical parameters and field testing in a noise-free locale.)

Random Thoughts:

One thought: could "capacitive" coupling in the cable be a source of "microphonic" behavior, where touching the cable or vibrating it causes chatter in our detector? And could the phase of the null signal have an influence on how well we can block such mechanical noise?

Construction of the cable may be the most important factor in avoiding "microphonics"; however, a smaller overall null signal may also help if vibrations modulate the null signal proportional to its amplitude.

And the effect of null phase on microphonic noise might be interesting to investigate. If your "capacitive" signal phase also centers near the ferrite ground-balance point, then perhaps microphonics are blocked similar to any ground modulation.

It would be interesting to experiment with an MD that has "microphonic"/vibration problems and see how the cable and null phase might affect it.

-SB

[simonbaker]

Mylar advice

I'm about to try shielding some coils with aluminized mylar from an emergency blanket.

I'm improvising since I haven't tried it before.

I cut a long strip off the side about 1.5 inches wide with the idea of spiral-wrapping it around the coil.

I confirmed that only one side is conductive, and my plan was to first spiral-wrap a bare wire around the coil as a lead, and then wrap the mylar over that with the conductive side "down" (pressing on the lead wire), then tape over it.

I could do it the other way too, first the mylar, conductive-side up, then spiral-wrap the lead wire over the mylar, then the tape. Maybe get a better contact that way.

I discovered something troubling. I found the my mylar strip had conductive "breaks" along the length, dividing it into separate conductive sections. The breaks seem to be where the aluminum wears off at wrinkles due to the folding.

This probably isn't a problem if my lead wire goes end-to-end. I sure wouldn't want to attach the lead at one end.

Because of this, I think I'll opt to first wrap the mylar, then the lead wire tight over it, then tape to insulate the two coil shields from each other. Of course leave the gap in each shield too!

I'm interested in others' experiences with aluminized mylar.

-SB

[Comp]

Sorry, cant help, but thanks for the info on the emergency blanket!!!! I am awaiting delivery of a couple, with which I am to shield my IGSL coils,... thanks for the warning!!! Will dounle check mine, before I tape up/pot up!!!


Thanks, Good Hunting, Fred

[simonbaker]

mylar shielded coil

This mylar stuff is a bit of a witch to work with...

Tried wrapping a strip. I estimate it may overlap such as to give 3 layer thickness, not sure.

-SB