FKK coils, free beach movement :)

[Davor]

I revisited a balanced preamp idea, and came up with a possible poor man's solution to a fine low impedance and low noise balanced preamp solution.

For some time now the RF front end development goes into one interesting direction, a LNA based on a capacitor cross-coupled common gate (or base) configuration. Because of cross coupling the sources with opposite phase gates, this configuration reduces noise, impedance, distortion, common mode, and power consumption ... simultaneously. With impedance at 30-50ohm it is a perfect match for RF systems ... but a bit lowish for a serious MD front end.

I rearranged this configuration a bit to use the best features, and came up to a configuration that has it all, and also has input impedance in kohm range. The best part is that no special transistor matching and no bipolar capacitors are needed, and noise performance better than 1.5nV/sqrt(Hz) is achievable with garden variety transistors at modest currents.

I did not test this configuration in real life yet. Due to easy to obtain components it is high on my to-do list. If it works in real life as well as it does in simulation, it will go into my IGSL front end.

Presented is a configuration with series resonance balanced coil at 16mH and a centre tap.

[Davor]

I rounded up values a bit to obtain square numbers, and made a comparative simulation of a normal IGSL front end and a front end with a 1:7 transformer. Both use 16mH search coils designed as centre tapped, hence two 4mH tightly coupled coils. The transformer front end is simulated with a voltage dependent voltage source that acts as an ideal amplifier. It gives me more freedom in playing with input parameters, e.g. input noise source.
It is designed to simulate both differential mode (target) response using V1, and common mode response (wet grass) using V3 and 10p capacitors. Simulation also includes a worst case calculation option that sweeps the resistors' values in IGSL front end. It is disabled by commenting a ".step param run 1 30 1" line. If uncommented it does 30 worst case iterations. It may be useful in observing common mode rejection of an op amp circuit.

Here goes the noise performance (optimised for 8kHz) of a transformer front-end. Spice file is included. Other configurations are possible by disconnecting ground here or there

[Davor]

Maybe one could be used "back to front" to step up the coil signal.

That's precisely the idea with mains transformers. In case of a 220:15+15 I have in fact 220:30 with ground tap, and it is an impedance ratio of 7.333^2=54 and it is precisely in a butter zone for a coil with R~20ohm and a preamp at 4nV/sqrt(Hz) ... it just places it there.

The best of all is that it also combats the common mode, so it is perfect for a FKK coil, with or without a centre tap.

[Davor]

Here is schematic of a Garrett that used a transformer in the receiver.
Not sure if there were any additional components in the coil?

Mouser carries a full line miniture of audio matching transformers for under $2.

Yes, that's a fine example. My local problem here is that I don't have any source of dedicated audio transformers at hand, but as mains transformers are wildly growing just about everywhere, I'd check those first. In case a plain old mains transformer can do the trick - hurray!

The main reason why transformers left the scene was a poor low frequency response, and that is not a problem at all with MDs.

BTW, a garrett example is actually a wee bit missing the point on this topic, as one of the Rx coil's sides is grounded, hence the common mode is converted to differential mode, and the coil's resilience depends solely upon the shield. In this example only impedance transforming feature is used, while common mode suppression is not. A FKK coil would surely use the common mode suppression capability.

[WM6]

Maybe one could be used "back to front" to step up the coil signal.

As Fisher Impulse.

[Altra]

Here is schematic of a Garrett that used a transformer in the receiver.
Not sure if there were any additional components in the coil?

Mouser carries a full line miniture of audio matching transformers for under $2.

[Gwil]

There was a time when small transistor radios used a transformer to couple the output transistors to a low impedance (often 3ohm) loudspeaker. This arrangement seems to be less popular now but the transformers are still available, and they are small and not very expensive. Maybe one could be used "back to front" to step up the coil signal.

[Davor]

I'll not draw anything as it is most probably very obvious for you. I'm in a middle of something at the moment, so I'm not in a mood of drawing.

A transformer is simply placed between a Rx front-end and a coil. The configuration is as per the coil configuration, preferably a split differential with a centre tap, but the beauty is in accepting just about any coil configuration there is. Secondary is a Rx single ended preamp.

The adding number of windings is a direct approach to what I'm saying, but there are some practical limits to that.

Example...

Say we have a 4nV/sqrt(Hz) input noise preamplifier. It's equivalent noisy resistor is of 1kohm.
A typical Rx coil of a nowadays IB CW VLF detector would have ~16uH and maybe some 30ohm, or less.

There are 2 possible connections of such a coil, a quasi-resonant with a parallel capacitor, and a series connection resulting with very low Q.

Now, we know the shortcomings of a quasi-resonant configuration, but it effectively provides some level of up-transformation, and consequently somewhat better S/N. It sucks at low frequency suppression, as it is effectively a low pass filter.

The series connection has inductance compensated by a series capacitor. Because it is terminated with high impedance, this tank is hardly resonant at all (Q<0.5) but it effectively fights the low frequency mains interference. It does not provide up-transformation and because the difference in 30ohm source with 1k internal noise source of a preamp, there is ample room for improvement. Ergo transformer.

The same effect would be possible with a coil winding at some incredible inductance and coil resistance at about 1kohm. But the common mode signal would be adequately rejected only if the preamp is a real differential one, such as instrumentation amplifier. A transformer rejects common mode signal by it's own right, and that's the reason I'm about to test it.

There is a 3 dB noise advantage of single ended preamps against the differential ones, so a balun helps as well ... again a transformer trait.

The only trouble is that real microphone transformers are extinct (or incredibly expensive), and I have no clue if the mains transformers can be convinced to behave.

[simonbaker]

I am considering a small mains transformer as a kind of Rx step up balun transformer. I'm not thinking about some super sexy high end microphone or MC transformer, but a cheap and cheerful small mains transformer, e.g. 220 to 2x15V small 2VA transformer. It kinda makes sense. Unlike a true microphone transformer, there is no real interleaving here to improve leakage inductance, but otherwise it should work just fine at VLF. With proper insulation and shielding it could do just fine.

The logic behind this is twofold.
Noise matching of the coil is quite demanding even nowadays, even with money to throw away, let alone using any garden variety opamp. Using step up transformation eases the noise requirements of a detector input.
Second, it may also be a good approach for making an existing detector aperiodic, and happy with balanced mode connection of a coil. Thus every "normal" detector could become happy with a FKK coil AND get better noise performance in a process as well.

This step up reasoning comes from the fact that many popular builds have a parallel capacitor at the input, which acts as a step up transformer, yet a frequency dependent one. Due to the phase transition problems it is slightly detuned and thus critical. In case one wants to make a front end less resonant or aperiodic, the transformation effect is lost so the front end quality relies on input preamplifier low noise capabilities. These tend to be modest in <100$ per chip range. Marrying the two approaches is possible using transformers.

I got some hints from the tape mic builders that it can work, so I'll give it a try. Their main concern is loss of low frequency signal, but that's completely irrelevant here.

There are losses in using a sub-optimally designed transformer, no question about that, but most of them are either curable or negligible compared to noise mismatch of, say NE5532 and a 6mH coil - it can only get better.

Compared to the effort of shielding a coil, and a cost of a small transformer is very tolerable. Paired with the noise benefits, I think I have a winner.

Hi Davor: can you give us a quick sketch to show where your transformer is in the MD circuit, just to make sure we understand correctly?

It seems like it might be a question I have wondered about -- I wondered if it was better overall to keep adding windings to the RX coil to increase impedance that way. But transformer maybe is different and is advantageous for other reasons.

Regards,

-SB

[Davor]

I gave this transformer approach another look and it seem to me as a missed opportunity among the metal detectors. True, I did not test it yet, my shack was under the snow till only a few days ago. Some say it's global warming (yeah, right).

So what's so wrong with transformers that they did not catch up is a bit of a mystery.

The role of a typical mic transformer is in impedance translation from low impedance of a microphone to a noisy preamp that behaves fine at high input impedances. However a transformer is in such case appears as an amplifier, it is completely passive and all it does is broadband impedance matching. In mic applications it is notorious for the lack of low frequencies, and some distortions at higher loading, but otherwise it is just fine.

Looking at the leaky transformer model at wikipedia, https://en.wikipedia.org/wiki/Leakage_inductance you are faced with a scary picture:



It shows leakage inductances as two separate entities and hints a design nightmare. In reality it is not so. A much more likeable model was given by December 2002 issue of Practical Wireless:



It truncates all the nasty bits to one side, and it suddenly looks much better. Adding a search coil to the primary will require something to be done about compensating the leakage inductance L1, but the rest of it should be just fine. A series capacitor should do.

Lp is the inductance of the primary and it is very large by definition. In pulse transformers it is usually set to at least 3 times the impedance, but here it is much more.
Rp represents the losses in the core, and it is of lesser importance for Rx purposes.
Secondary leakage is of no consequence, and losses are truncated to primary, so the model is just fine.

The most critical in power transformers design is the lower usable frequency. At any significant powers the core is near saturation. Surprisingly the higher frequencies are not losing in efficiency, but the power transformers are losing some power transfer capacity - the core is not running near saturation. It is of no consequence with Rx or mic impedance matching transformers, as the power transfer is negligible.

At much higher frequencies the losses and leakage inductance become problematic. How much so I'll have to find out. I'll let you know.

Otherwise, the transformer is an ideal balun, and impedance matching provides for low noise operation at low input impedances.

[Davor]

Mounting a transformer on a PCB is perhaps not such a good idea because to make it properly, you need to shield the transformer. Otherwise the leakage inductance is picking up EMI. Leakage inductance of small transformers' secondary may fall anywhere between 200uH and 2mH, maybe even more. In a relatively narrowband operation such as MD it is not a problem, until it starts picking up EMI. So a proper shield is a necessity. I'll wrap it in a few layers of aluminium foil.

Otherwise this transformer thing could be used to adapt many different types of coils to a simple single ended MD input. It can accept unbalanced, balanced, and split balanced with a ground tap (as I intend to continue using).

The olden microphone transformers were used to step up a 20 to 30 ohm microphone to ~47kohm amplifier input, with a microphone in a balanced mode to avoid hum from the mains. MD coils are not at all different from mic coils regarding matching to a preamplifier, but 47k is a bit high-ish for the nowadays op amp inputs. A good low noise input at 4nV/sqrt(Hz) is equivalent to 1k, but such preamps suck at current noise, hence 47k input impedance is too much. To conclude, olden mic transformers are not the best choice, in spite the fact that they are far superior regarding losses etc.

I have high hopes that a simple mains transformer will do well as an impedance matching device. Series connection of a MD coil and a capacitor yields a fairly broadband response. In fact a capacitor is simply removing the inductance, but such contraption is well below showing any resonant behaviour because of very small Q, far below 0.5. Extending this approach to encompass a leakage inductance as well could do the trick.

[satdaveuk]

You know Davor you may well have hit on a good idea there, it all makes sense and certainly worth moving forward with experiments,I look forward to hearing your progress on the matter with a very keen positive attitude
Do you intend to locate the transformer in the front end or on the main PCB?
Regards

[Davor]

I am considering a small mains transformer as a kind of Rx step up balun transformer. I'm not thinking about some super sexy high end microphone or MC transformer, but a cheap and cheerful small mains transformer, e.g. 220 to 2x15V small 2VA transformer. It kinda makes sense. Unlike a true microphone transformer, there is no real interleaving here to improve leakage inductance, but otherwise it should work just fine at VLF. With proper insulation and shielding it could do just fine.

The logic behind this is twofold.
Noise matching of the coil is quite demanding even nowadays, even with money to throw away, let alone using any garden variety opamp. Using step up transformation eases the noise requirements of a detector input.
Second, it may also be a good approach for making an existing detector aperiodic, and happy with balanced mode connection of a coil. Thus every "normal" detector could become happy with a FKK coil AND get better noise performance in a process as well.

This step up reasoning comes from the fact that many popular builds have a parallel capacitor at the input, which acts as a step up transformer, yet a frequency dependent one. Due to the phase transition problems it is slightly detuned and thus critical. In case one wants to make a front end less resonant or aperiodic, the transformation effect is lost so the front end quality relies on input preamplifier low noise capabilities. These tend to be modest in <100$ per chip range. Marrying the two approaches is possible using transformers.

I got some hints from the tape mic builders that it can work, so I'll give it a try. Their main concern is loss of low frequency signal, but that's completely irrelevant here.

There are losses in using a sub-optimally designed transformer, no question about that, but most of them are either curable or negligible compared to noise mismatch of, say NE5532 and a 6mH coil - it can only get better.

Compared to the effort of shielding a coil, and a cost of a small transformer is very tolerable. Paired with the noise benefits, I think I have a winner.

[Aziz]

Good job Davor. I like it.

I think, it's becoming a very nice Nikolaus gift.

Cheers,
Aziz

PS: Temperature drift could be an issue however.

[Davor]

I have yet another candidate for a balanced preamp, a frugal version. The worst of the parameters is CMMR, and it is ~40dB, but it does not depend upon the resistors accuracy, so I guess it is just fine. Decoupling is not shown. It (ab)uses the fact that the impedance at the inverting input of the op amp is ~0. The gain is determined with R3 and R8 and somewhat with the transistors current. I placed a capacitor in series with R3, mainly because I don't want to pick selected matched pairs of transistors - unmatched transistors same type will be just fine.

I think I'll go with this one. With real world BC337 (that are modelled with too high Rbb in LTspice) i expect input referenced noise to be ~1.5nV/sqrt(Hz). Not bad for 2 euro of material. I think this will go into my IGSL. I like frugal builds.