Auto-tune modifications?

[KingJL]

Has anyone successfully modified the auto-tune section for better performance. There appears to be significant potential for improvement in this area. TP 5 indicates plenty of dynamic range remains (that follows target strength) between 0 and +5 left even after the signal at tp9 becomes erratic. It seems that the dynamic range of the circuit of ic8b is very limited. I would be interested in exploring options in this area. Maybe something to give a more lograrithmic response (high sensitvity at lower signal levels, with senstivity decreasing as the voltage at tp5 increases ). Maybe more filtering is required in IC8a, ic8b to take advantage of the target signal dynamic range of in ic8a.

Thoughts?

[KingJL]

Autotune Instability

Upon further investigation, I find that even though there is reserve dynamic range of IC8a, the jitter/noise/interference peaks that are bouncing at the output of IC8a are enough to overwhelm IC8b with it's high gain.

I followed the advice that I have seen from REG on this forum and reduced the pre-amp gain. It may seem counter-intuitive, but it works. In my case I reduced my second stage pre-amp gain from 150x to 100x (not the easiest chore on a SMT board). Much cleaner signals. 100% improvement!!!

I am thinking that a high-pass filter after the pre-amp might also provide improvement, especially in noisy environments. I will have to determine what the c/o freq should be. But for now, the system is stable enough and seems to be performing well enough to put this dual board HH through some real world tests (out in the field).

[KingJL]

I am thinking that a high-pass filter after the pre-amp might also provide improvement, especially in noisy environments. I will have to determine what the c/o freq should be.

My initial determination is that a corner frequency of 284Hz using a C of .1uf and an R of 5600 ohms, should provide significant attenuation of the predominant environmental interference of 60Hz.

[Tinkerer]

High pass filter

My initial determination is that a corner frequency of 284Hz using a C of .1uf and an R of 5600 ohms, should provide significant attenuation of the predominant environmental interference of 60Hz.

Hi KingJL,
The criteria for the high pass filter at the output of the preamp, is not the corner frequency but the point where the decay of the signal is slowed down so that you need to sample later.
An earlier sample has more sensitivity to small targets.
Is your coil shielded? The shield will reduce the noise very much.
Tinkerer

[KingJL]

Hi KingJL,
The criteria for the high pass filter at the output of the preamp, is not the corner frequency but the point where the decay of the signal is slowed down so that you need to sample later.
An earlier sample has more sensitivity to small targets.
Is your coil shielded? The shield will reduce the noise very much.
Tinkerer

Tinkerer, I can sample pretty early with things the way they are. I can sample at 7 us after trailing edge of TP1 (and if you look at the trailing edge of the current pulse on Q3 source it 5 us) and still detect targets while remaining stable. I attribute this to both the fast coil design by bbsailor and the fact that I took the time to build a well balanced DD coil. And yes, it is shielded. To get any better results in the coil area will probably require me to build one using litz wire (which I will probaly explore shortly). I still feel as though some high pass filtering may be in order. The filtering is to decrease 'environmental' noise and make operations even more stable with fewer yet false targets/signals. The high pass filter with a corner frequency of about 284 Hz has less than a 3% effect on the lowest frequency that we are concerned with. That being the lowest operating frequency (in my case about 600 Hz). Yet it provides 6+ db attenuation for 60 Hz. If you are using this thing anywhere but a remote beach (especialy around any residential setting), you have probably got some environmental noise.

[Tinkerer]

High pass filter

KingJL,

7 uS is excellent. BBsailor makes the best coils.
A low pass filter is good to have.
The repetition rate of 600 (PPS) is not the frequency of the signal.
If your decay curve has a duration of 7uS, the slew rate of the transient relates more to a frequency of 70khz.
On the SM-PCB you do not want to change the components many times.
I suggest you try 1K Ohms and 150pf.
Tinkerer

[KingJL]

KingJL,

7 uS is excellent. BBsailor makes the best coils.
The high pass filter is good to have.
The repetition rate of 600 (PPS) is not the frequency of the signal.
If your decay curve has a duration of 7uS, the slew rate of the transient relates more to a frequency of 70khz.
On the SM-PCB you do not want to change the components many times.
I suggest you try 1K Ohms and 150pf.
Tinkerer

Tikerer,
I agree that 600 Hz is not the frequency of the signal. I was just saying as a worst case you do not need anything below the repetition rate. On the high side, the figure is the base rate of the fastest target (70kHz in your example) + all odd harmonics up to and including the 9th order odd harmonic for maximum signal extraction. In fact the bandpass of signals spans from about 2500 Hz for the largest iron to less than 1Mhz for the smallest gold (that allows for the decay time + up to and including the 9th odd harmonic to allow maximum intellegence extraction). I have not seen this documented anywhere, but I am allowing for 10% above and below my estimations. On SMT-PCB, I still have one extra if needed!! Seriously though, I am in agreement there. But in my case, I can implement a high pass filter without modifying my board. Remember I go from the second stage pre-amp to a VR for gain reduction if needed.

Good discussion!!

[Carl-NC]

JL,

First, I agree that the gains of the stages have never been optimized, so there is probably lots & lots of room for improvement.

Second, if you want to add a HPF after the preamp then Tinkerer is right, you just need to set it for minimal impact on the decay. As an example, if you have a 500uH coil and a 500-ohm damping R, then the decay tau is 1us (this ignores any effects of the clamping R). You want the HPF to have a tau much faster so that it does not slow down the decay. Let's say 10x faster, or 0.1us. That's 1k & 100pF.

- Carl

[KingJL]

JL,

First, I agree that the gains of the stages have never been optimized, so there is probably lots & lots of room for improvement.

Second, if you want to add a HPF after the preamp then Tinkerer is right, you just need to set it for minimal impact on the decay. As an example, if you have a 500uH coil and a 500-ohm damping R, then the decay tau is 1us (this ignores any effects of the clamping R). You want the HPF to have a tau much faster so that it does not slow down the decay. Let's say 10x faster, or 0.1us. That's 1k & 100pF.

- Carl

Carl,

In my case, with the split preamp, I wouldn't think anything I did with a high pass filter at the output of the second pre-amp could impact the decay time. Now I believe that an overly agressive (too low c/o frequency) low pass filter could effectively destroy it.

I believe with the values 1K and 100pf yields a corner frequency of about 1.5 MHz. I think a little high for a high pass filter in our application, but probably exactly right for a low pass filter in our application.

Where is my thinking skewed here?

[KingJL]

Clarification

JL,

First, I agree that the gains of the stages have never been optimized, so there is probably lots & lots of room for improvement.

- Carl

One thing I would like to be clear on in this discussion. None of this is a criticism of the HH design. Quite to the contrary, as designed, the HH is a good and very capable detector. Any and all problems occur when the design is altered. In my case, when I split the pre-amp, I also increased the total pre-amp gain by 50% (1500x vs 1000x as designed). Everything worked, but very noisy with a lot of false targets. What I found was that environmental noise was overwheming the autotune circuit of IC8b (I actually believe that the environmental signals were providing peak signal levels in the 2nd stage that were at the output limits). I decreased the gain of the 2nd stage to bring the total pre-amp gain back to 'as designed' (1000x), and now I have a stable and very sensitve PI detector. But as the HH was presented as an educational/experimental platform, I saw an area to explore and discuss different options. If I was building my platform again, I would put a high-pass filter with a tau of 560 usec between the two pre-amp stages. This would provide 2 benefits: (1) block/remove the DC Vos of stage 1, from stage 2; (2) Provide -13.7 db attenaution of 60 Hz. As it is, I am going to experiment with that size filter between the output of the 2nd stage and IC8. I believe that with a high-pass filter you want to pass the tau of you slowest desireble decay and anything/everything faster. A high pass filter with a tau of 560 usec will pass and not distort a signal with a decay tau of 1 (in fact by definition, will pass any tau 560 usec and faster). For a low-pass filter you do just the oposite design to pass the tau of your fastest decay and everthing slower. So for a low-pass filter you want a tau on the order of .1 usec. For a high-pass filter you want a much, much slower tau. For a band pass filter you want both your high and low pass filters. If you used the same tau for both, your bandpass would be extremely sharp (in effect you have no 'band', just pass the corner frequency).

But for the record, the HH, as designed, is a great little machine, works well and is a great platform for experimentation!

Regards,
J. L. King

[Carl-NC]

In my case, with the split preamp, I wouldn't think anything I did with a high pass filter at the output of the second pre-amp could impact the decay time. Now I believe that an overly agressive (too low c/o frequency) low pass filter could effectively destroy it.

I believe with the values 1K and 100pf yields a corner frequency of about 1.5 MHz. I think a little high for a high pass filter in our application, but probably exactly right for a low pass filter in our application.

Where is my thinking skewed here?

You are right, and I am embarrassed! I was thinking of LPF, not HPF.

For the HPF, you want the tau to be slow enough so that you don't start attenuating the decay. Again, make it at least 10X slower, say, 10us. That's 1k & 10nF.

Can you make it even lower? Yes, but you (possibly) need to be careful that the HPF doesn't have such a long tail that you build up a DC offset over time. This may not even be a problem, I don't know. But to avoid this, make the tau 10X faster than the pulse period. If you are running 1kHz, this limits the tau to 1ms/10 = 100us. That's 1k & 100nF.

One thing I would like to be clear on in this discussion. None of this is a criticism of the HH design.

Criticize away! HH certainly has its issues. But, as I've always said, it was never intended as a build-and-use detector, but rather as a learning platform for PI. I'm glad to see you're doing just that, and taking it in a new direction.

- Carl

[KingJL]

You are right, and I am embarrassed! I was thinking of LPF, not HPF.

For the HPF, you want the tau to be slow enough so that you don't start attenuating the decay. Again, make it at least 10X slower, say, 10us. That's 1k & 10nF.

Can you make it even lower? Yes, but you (possibly) need to be careful that the HPF doesn't have such a long tail that you build up a DC offset over time. This may not even be a problem, I don't know. But to avoid this, make the tau 10X faster than the pulse period. If you are running 1kHz, this limits the tau to 1ms/10 = 100us. That's 1k & 100nF.



Criticize away! HH certainly has its issues. But, as I've always said, it was never intended as a build-and-use detector, but rather as a learning platform for PI. I'm glad to see you're doing just that, and taking it in a new direction.

- Carl

Thanks for the clarification and the reply. Jeez, you had me scared. 30 years ago they took away my slide rule (I still miss it), and stuck a calculator in front of me that couldn't do log. I was lost! Then they told me that 3 significant digits of accuracy wasn't good enough. I was afraid that now they had changed the fundamentals that I learned long ago, and left me with nothing. Whew, what a relief!

In my experimentation, I will try a HPF with a tau of 100usec. (maybe about 150 usec since I run as low as 650 pps). I was initailly afraid to take it that high as we are well within the range of the tau of large iron, but with your recommendation those fears appear to be groundless.

Thanks again.

Regards,
J. L. King

[KingJL]

HPF Mod Success

I modified my dual board HH today to have a HPF between the output of the 2nd stage pre-amp and R23. C 100nF, R 1k (as per suggestion by Carl). Improvement very significant. Extremely stable; fewer false signals. I even decreased the setting of R35 to it's minimum resistance; still stable, even inside the house. Threshold setting works much better; has more control range. There appears to be less jitter/jumping at tp-9 caused by environmental noise.

For anyone that builds a split pre-amp, I would recommend that you put this HPF between the two stages. If I was building mine again, an HPF between the stages would be included as well as an HPF at the output of the 2nd stage.

[rajesh]

hammerhead ic7

dear friend king jl my hammerhead work fine means stable but while working whats will be better setting across volt at the ic>7> pin no - 3 and
4 ..........?

[KingJL]

I think I have discovered the one modification that greatly improves the operation of the HH. An active filter, 2 pole Bessel with 3.5hZ cutoff, followed by an inverter (to get the signal back to the correct polarity for the autotune circuit), placed after the output of the the differential integrator (IC8a). The output of the inverter feeds the existing autotune circuit. Solves all my instability problems and give smooth operation. Sweeeet!