AMX TX

[ivconic]

In meantime , LM317T (1.5A) that regulates voltage for TX supply; can be controlled by the mcu too.

[ivconic]

Since the CD4093 has 4 nand circuits; it's stupid not to use them all.
So one will serve for Off/On voltage of 2v for TX power supply, if necessary.
And the other 3 nands can be cleverly used to square the signal coming from the Atmega; further straighten and shape into an ideal square signal.
Because the Atmega does not give an ideal square signal, it is a bit irregular.
Crossing through nand it becomes perfect square.

[ivconic]

Variety of choices relating to power supply
The "SW" is only for testing purposes.

[ivconic]

WRONG!
Sometimes simulator may confuse unexperienced user...

[moodz]

WRONG!
Sometimes simulator may confuse unexperienced user...

Now your assimulation is complete ... resistors are futile.

[ivconic]

Works like a charm... in simulator!
In reality...?
...
Of course, this was all a little exercise. I have no intention of doing that. Because it's a waste of time.
The rest of the detector ... that is, the rest of the requirements from the criteria list are such that they will require a powerful ADC, a powerful MCU and a lot of code.
And in the end what do you get?
A PI detector that will detect gold nuggets. No discrimination.
If there is discrimination; it will hardly be reliable.
Instead; I can make a few commercial devices, sell them and buy Deus II with that money.
Because I'm not looking for gold nuggets anyway.
And in the areas where there are gold nuggets, in the surroundings; the soil is so polluted with ferro oxides, pieces of small rusty iron and similar garbage...
so I just don't have the will to deal with it.
If you (the team) manage to build that detector and it works as intended; and if you offer a kit at an affordable price: I might be your customer.
Just out of curiosity.
Good luck in your future work!

[ivconic]

Given that I do not have more detailed previous knowledge on these topics; I've been reading a lot about it in the last few days.
And based on what I read and what I understood; I came to the following conclusions.
The "most secure" dead time ("safe zone" I would call it) depends on the specific H-bridge circuit design,
including the characteristics of the mosfets and the load being driven.
In general, a dead time of 1-5% of the total switching cycle time is commonly used in H-bridge circuits.
To calculate the total switching cycle time for a 25 kHz switching frequency, we can use the formula:
T = 1 / f
where T is the period or total switching cycle time, and f is the switching frequency.
If f = 25 kHz, we get:
T = 1 / 25000
T = 0.00004 seconds = 40 microseconds or shorter 40uS.
A dead time of 1-5% of this time would correspond to a range of 400 to 2000 nanoseconds.
I must admit that I am somewhat surprised with these values, all the time I thought it will be in range of few tens of nS.
Or this "rule" does not applies to specific design we have here?
...
It's worth noting that the switching cycle time is different from the duty cycle, which is the percentage of time that the switch is in the "on" state during one cycle.
In our vocabulary this can be said as "pulse width".
The duty cycle can be calculated as:
D = Ton / T
where Ton is the time that the switch is in the "on" state during one cycle, and T is the total switching cycle time.
The duty cycle is a key parameter that determines the average power delivered to the load.
In our vocabulary, again; it is pulse width, commonly expressed in uS.
Playing with FelezJoo PI (my first PI where I could see on the LCD preciselly timings expressed in accurate values); I noticed significantly better and more stable
detection; as I increase pulse width up to some levels.
At particular case, the FelezJoo PI case; the default is set to 150uS for pulse width.
At 250-300pps when I increase pw to 180us and up to 200uS; detection on smaller targets improves significantly.
Power drain, logically, increases too.
Does the pulse width plays any role in this particular design with H bridge?
...
Current is a critical parameter in the design and operation of H-bridge circuits.
The current determines the power delivered to the load and the stress on the mosfets and other components in the circuit.
A higher current means higher power delivery and higher stress, while a lower current means lower power delivery and lower stress.
In H-bridge circuits, the current flow through the mosfet depends on the state of the switches and the load impedance.
When the H-bridge switches are closed in one direction, the current flows through the load and the corresponding mosfet.
When the switches are closed in the opposite direction, the current flows through the load and the other mosfet.
The mosfet must be able to handle the maximum expected current flow without overheating or failing.
The current also affects the efficiency and switching speed of the H-bridge.
A higher current requires higher voltage levels to switch the mosfets on and off, which can increase the switching losses and reduce the efficiency.
In addition, a high current can cause voltage drops and noise in the circuit, which can affect the performance and reliability.
To ensure proper operation and safety of the H-bridge, it's important to select mosfets and other components that can handle the expected current levels,
to properly design the circuit to minimize switching losses and noise, and to monitor and control the current during operation.
This can be done through various techniques such as current sensing and feedback control, overcurrent protection, and thermal management.
...
Current, constant current

In some applications, such as motor control and LED lighting, constant current sources are commonly used to provide a stable and predictable current to the load.
A constant current source can ensure that the load receives a consistent amount of current regardless of variations in the load impedance or other factors.
This can improve the stability and accuracy of the circuit and prevent damage to the load or power devices.
However, using a constant current source in an H-bridge circuit may not always be necessary or practical.
In some applications, a constant voltage source may be more appropriate, such as in power supply circuits or some types of motor control.
In addition, implementing a constant current source can add complexity.
Ultimately, the decision to use a constant current source in an H-bridge circuit should be based on the specific requirements of the application and the
trade-offs between performance, complexity, and cost.
So in simple words; it is "by case to case" situation.
What is the situation at this particular project?
Do we benefit from the constant current source or not?
If it is been said that higher current can cause voltage drops and noise in the circuit; I will simply assume that is mandatory to supply such TX from the constant current source as MUST.
So, instead monitoring current by the mcu, with means of some current sensing method; isn't better first to design constant current power supply for TX?
Or , as I did in some of my previous posts; add mosfet and BC337 and fix the current to constant value, not dependable on what's going on in H bridge else.
From all this I can pull conclusion that only thing about current here is important to keep it constant to avoid power loses and introduction of noise, am I right?
...

You know... one of you probably wrote this somewhere, answered all those questions, but watered it down with a lot of unnecessary rhetoric.
That's what I was also afraid of at the beginning of this thread.
That the individual ego will prevail and that the desire to show one's superior knowledge will cover the very essence of this topic.
That's why we have a bunch of threads, topics and many more posts, which are very difficult to read.
That is; it is difficult to extract from them as short, clear and concrete answers to direct questions as possible.
As well as clear explanations, where it is necessary to explain it further.

[lucifer]

Hi ivconic,
I'll try to provide some answers to your questions as per my knowledge and understanding of this TX circuit.

Given that I do not have more detailed previous knowledge on these topics; I've been reading a lot about it in the last few days.
A dead time of 1-5% of this time would correspond to a range of 400 to 2000 nanoseconds.
I must admit that I am somewhat surprised with these values, all the time I thought it will be in range of few tens of nS.
Or this "rule" does not applies to specific design we have here?

The dead time should be as short as possible to minimize losses but not shorter to cause shoot-through. During the dead time all FET switches are off and any energy stored in the coil gets dissipated. So you need this period short to save time and power required to re-charge the coil on the next cycle. How short exactly I assume will depend on your FET driver capabilities and the speed of the FET switches.

Does the pulse width plays any role in this particular design with H bridge?

I think it was already mentioned that the duty cycle is always 50%, so your pulse width is always 1/2 of the period. You can only change the period (frequency). Does the frequency matter? Yes, I believe it does. A higher frequency allows more samples to be integrated and increase the signal-to-noise (SNR) ratio. A higher frequency might hinder sensitivity of slowly decaying targets because current polarity is switched while the slow target has not fully decayed yet. Fast targets remain unaffected. Carl has posted a very good and detailed explanation of that in a thread called Optimizing target responses. A higher frequency will cause a higher power consumption as well. It's also possible to combine two different frequencies and process the RX signal separately for each frequency.

What is the situation at this particular project?
Do we benefit from the constant current source or not?

Yes, there's benefit. Using a constant current TX circuit the current in the coil rises and falls rapidly, like a square wave. The sharp current edges create higher eddy currents in the targets, hence they should provide a stronger signal to detect by the RX circuitry. Also the energy in the coil is not wasted (compared to the traditional PI), only its direction is changed between the cycles. Of course, there are some small losses and these produce the "tilt" that is to be compensated.

[Carl-NC]

I don't know the answer to the dead time question. So far, Paul has seen no need for any dead time at all, but I don't know how far he's explored this circuit. I plan to go up to 1200-1500V on the flyback, which is exceptionally fast and useful for small gold. But I also want to look at very slow transitions, such as might be used for a 100 amp/100ms tow-sled version for Atocha bars, the same as Eric Foster did for Mel Fisher. I have no idea what dead time might (or might not) be needed, but that's why I'm using 2 timer pins with dead time adjustment. I want the option.

Pulse width plays the similar role here as with a normal PI. A short pulse width is sufficient for low conductors while a wide pulse width is needed for high conductors. The difference here is that each pulse settles to a flat-top current almost instantaneously. In a traditional PI, the current exponentially rises. That's probably why you see an increase in sensitivity when you increase the pulse width: the coil current increases. In CCPI, a 20us-wide pulse is sufficient for a 5us target (sub-gram nuggets) and increasing the pulse width will not increase current or depth, and will in fact reduce depth by reducing the integration rate.

Constant current is the key to this whole concept. But there are 2 ways to achieve CC. One is to create a high-impedance CC source, apply it to a circuit, and get whatever voltage is developed. The other way is to use a voltage source with current monitoring and use feedback to servo the voltage to achieve the current you want. I'm using the latter method as it's simple and more power efficient. The use of CC is fundamental to the basic operation of being able to see a target response while the coil is still energized, not because of noise or power loss. However, better power efficiency (than a traditional PI) is a fortuitous fallout of the operation.

A white paper describing the theory of operation would help a lot, and if I ever get some spare time (!) I will try to write something up.

[Carl-NC]

A higher frequency will cause a higher power consumption as well. It's also possible to combine two different frequencies and process the RX signal separately for each frequency.

A nice benefit of CCPI is that power consumption is practically constant for any frequency. Transition losses (the cause of tilt) increases with frequency but they are a very small percentage. Somewhere I posted the 2-frequency TX waveform I'm considering:

[ivconic]

Thank you both; Lucifer and Carl, on additional clarifications.
I will focus only on the essential issues.
So the dead time is still something to be determined.
What we know for sure so far is that it will be in the range of few tens of nS to possibly a few hundred nS.
Using two pins to separately switch the positive and negative sides has therefore a visible advantage over my attempt to do everything with one pin.
Because in the case of two pins, each pin can be individually timed in the code and this gives a wider range of settings.
While my attempt lacks in that, because then the dead time is defined by the delay in the circuit that inverts the signal from the pin.
Whether it is a dedicated inverter circuit (4069 and similar) or a NAND circuit (4093).
Both involve a delay of a few tens of nS. I haven't checked the datasheet but I'm pretty sure it's pretty close to accurate.
Maybe in practice it will be quite enough and maybe that approach is good enough. But it may not be good with every choice of mosfet.
So the two pin selection is still a big advantage.
My idea was to practice a "poor man's solution" approach. And I'm not dissatisfied with what I saw in the simulator.
Duty cycle... pulse width... Generally 50:50% is the right solution.
But in this case, reading Carl's arguments; I think for the purposes of this project, Carl is right.
So let's accept that 20uS is the pulse width.
Should it remain constant or leave the possibility of additional settings in the code?
I think the choice of mosfet will also play a role.
We need to see how the "regular" IRF740/840 type behave.
Paul has already done a significant amount of work on that, so it's no mistake to assume that the IRF740/840 will be doing a good job there.
Constant current. I assumed it was of great importance. But I had to ask and get confirmation.
On one of the schematics I posted, you can see a logic level mosfet IRLZ44N with a BC337, in which the mosfet plays the role of a variable active resistance,
and the source resistor limits the maximum current.
By simply calculating that resistor, you can calculate the constant current that will flow between the drain and the source.
If the gate is pulled up high, then there is no regulation. Maximal constant current, defined by source resistor will flow between drain and source.
But if PWM is applied to the gate; then the current can be additionally regulated within the range defined by the source resistor.
It is an interesting solution because it is possible to fine-tune the constant current level.
If the maximum current allowed for the assembly is, say, 500mA... and fluctuations occur during operation;
the fast PWM control can immediately react and compensate for the current drop due to these fluctuations.
Now that I'm writing this, I realize that in that case, there should be room for compensation, so for a consumption of 500mA,
maybe a circuit should be dimensioned that will be able to deliver something more than that.
With the PWM signal then "lock" it to exactly 500mA and keep it that way until the need for correction appears.
Although that circuit is very simple; in practice it has proven to be very effective.
I used it to power and control super bright WRGB LEDs for various lighting effects.
In the same 12W "diode" there are 4 (WRGB) separate LEDs, each 3W.
But each with a special maximum current, different then others.
And I used such a circuit separately for each one led, each set to a different current.
So I checked that circuit in practice and it turned out great.
IRLZ44N don't need heathsink, it remains totally cool after hours of working with such leds.
Last I did with same constant current regulator was sport semaphore with led strips.
4 (common cathode) + 7(segments) = 11x IRLZ44N, all without heathsinks and all totally cool!

...
A white paper is good idea.

[ivconic]

A nice benefit of CCPI is that power consumption is practically constant for any frequency. Transition losses (the cause of tilt) increases with frequency but they are a very small percentage. Somewhere I posted the 2-frequency TX waveform I'm considering:

Me: "...So let's accept that 20uS is the pulse width..."
In the case of 25kHz yes.
In the case of 5kHz? 100uS?
...
Obviously the tilts will occur less at 5Khz and more at 25kHz, in overal timings scheme.
So the "dynamic" compensation is needed.
But that should not be the problem, right?
Actually; tilts will be more "visible" at lower freq.
And almost insignificant at higher freq.

[ivconic]

Now about the "tilts"...
So, Carl, how did you envision solving the tilt compensation problem? I guess snubber network is not a good solution?
"Tilts" or spikes are common in the waveform during transitions and are caused by parasitic capacitances and inductances.
Can be compensated by adding a snubber network.
Now... I wonder (I ask you) if adding a snubber will affect the RX signal?
It shouldn't (directly) if the searchhead has two separate coils, one for TX and one for RX, right?
But; the snubber network can affect the current in the coil (and RX indirectly).
The snubber network is connected in parallel with the mosfets, and it can create a path for the current to flow through
during the switching transitions. This can affect the current in the coil and the voltage across the mosfets.
Therefore, it's important to carefully choose the values of the snubber components to balance the tradeoff between reducing
voltage spikes and ringing and minimizing the additional losses in the circuit.
I honestly didn't even know about this solution until a couple of years ago, when Davor pointed me to it and gave me some solutions.
But it wasn't a project anything like this one. It was a "slow" H-Bridge solution with very high currents, 14-30A.
I don't see how you're going to solve that problem...unless you use a very special "switch" circuit to power the TX stage.
I haven't looked at the datasheet so I don't know if the solution from your schematic includes such a possibility?
In a few simple sentences; can you explain your solution?
....
White paper, indeed good idea!
You should have done it earlier... you see!

[Carl-NC]

The transition loss is identical no matter what the width of the previous or following pulse. Tilt is just the current trying to get back to the proper level, so apparent tilt will seem worse at a high frequency because there is less time to make up the loss. But it's really the same correction making up the same loss. Whatever tilt correction works at 20us should work identically at 100us.

[ivconic]

I forgot to complain publicly... in fact I lost all hope that complaining publicly would get any better...
But the forum works very slowly for me.
Everything else on the net, the vast majority of sites open for me very quickly, almost instantly, but
the Geotech forum pages do not, on the contrary, terribly slow!?
And when I click on the topic, it necessarily takes me to the first page, which loads slowly, in two steps,
first it loads one part, then it stops, and then it barely finishes loading.
And then I have to scroll all the way down, where I have a choice of which page to switch to.
When I click on the last page; everything goes slowly again, "spinning, spinning" for up to half a minute, sometimes a minute.
It's very tiring to follow the forum since you switched to the new version.
I know I'm always the one complaining about something... but this time it's really, really bad.
That's why it often happens that I don't get to see new posts while I'm writing my post.
Everything else on the internet works very well for me, any site from any country.
Therefore, it is not a "regional" or "locational" problem.
I have a constant bandwidth of 25 Mb/s for download and 2-5 Mb/s for upload...
Believe me, I have a very strong will and motivation to participate in this forum, since I suffer such torture...
and despite everything, I don't give up.
If you do not believe me; I can make a video of how it behaves? Big problem!
Sorry for going offtopic for a while...