Spice simulation of NE5534 and real life slow recovery time

[nick_f]

Hi All,

I start doing some simulations of the Hammerhead Rx circuit and I noticed I am unable to obtain the slow recovery time from saturation of the NE5534, running with a gain of 1000. From my experiments, I get a delay of about 4us due to the recovery time, which I can't model in LTSpice.

Maybe I should ask first if other people have noticed this effect in real life.
By reducing the gain of NE5534 to 100, the output is faster with about 4us.

I also added my circuit from LTSpice. There are some models required and I found them on the great forum http://tech.groups.yahoo.com/group/LTspice/

I don't have much experience with Spice, if people notice I am doing something wrong, please let me know.

Regards,
Nicolae

[Carl-NC]

Probably the 5534 model is at fault. Most Spice opamp models cover 1st-order effects and some 2nd-order, and are focused on decent modeling in normal linear operation. I would never trust an opamp model to give me good results on overdrive recovery. You have to build a test circuit for this.

- Carl

[nick_f]

Probably the 5534 model is at fault. Most Spice opamp models cover 1st-order effects and some 2nd-order, and are focused on decent modeling in normal linear operation. I would never trust an opamp model to give me good results on overdrive recovery. You have to build a test circuit for this.
- Carl

Thanks Carl, that would make sense.

Regards,
Nicolae

[Aziz]

Hi Nicolae,

you can use the op-amp model OP27. It has implemented noise and more poles. But it takes more time to simulate and is prone to convergencing.
You can find the OP27 models almost everywhere in the internet.

Aziz

[hobbes_lives]

Note too that there is a note on the web about adding a missing component to the spice model, C2. The following image shows the difference with and without C2.

[bbsailor]

Look at the hump!

Nicolae,

Nice photo of the hump on the back of the coil discharge. The amplitude of the kink point is about 0.55V or in the area of the bias voltage on a diode.

The top section of the discharge has less slope because the coil discharge TC is governed by L/R which is the parallel value if Rd and Rin. In the Hammerhead this would be: Rd=680 and Rin=1000 with an effective parallel value of 405 ohms.

When the pulse gets below the diode bias of approximately 0.6V it releases the Rin from being in parallel with only an Rd of 680 ohms and the slope of the discharge gets steeper at this point.

What the hump shows you is the effect or two possible coil discharge slopes, one causes by a 405 ohm discharge path and another one with a 680 ohm discharge path.

The conclusion to all of this....
Use the adjusted value of Rd as an objective indicator of any potential speed improvement you make. When you minimize capacitance in the coil itself, the coax cable, the MOSFET and other circuit sources, you change the slope of the coil discharge path. This now allows you change your coil/circuit design and see change in coil discharge hump slopes.

A few things allow you to see the effects of faster sampling.

1. less noise

2. better signal

3. time over the target

4. speed of response when going over a target

5. number of samples integrated

6. The wait time for the amplitude of the back EMF pulse spike to become damped for the earliest sampling is also critical. High pulse amplitudes, due to higher coil current, take longer to damp and allow sampling.

7. Coil-size to target-size ratio. Faster sampling may only be seen when looking for some very small targets with low target TCs that respond better under faster sampling conditions. Then, the ratio of target size to coil size becomes another factor.

When you add in the gymastics requires to eliminate the effect of moving a coil in Earth's magnetic field, together with the changes (speed and amount) in ground responses, it is easy to see how much of a balancing act it is to pick the right detector design for the a paricular class of targets located in a well understood environment.

Nice technical photos.

bbsailor

[nick_f]

Note too that there is a note on the web about adding a missing component to the spice model, C2. The following image shows the difference with and without C2.

Thank you Hobbes_lives and Aziz. I will add the capacitor and retest it.
I will also look for the OP27 model tonight and run some simulations with it.

Regards,
Nicolae

[bklein]

As slew rate improves with higher supply voltage - were you perhaps running a lower supply voltage and could try +/-15 or so? Curious if slew rate is part of the problem in addition to the saturation. Also there are input protection diodes in the 5534 that could perhaps be bypassed with external circuitry - based upon some things read in the DIY Audio forum.

Barry

[nick_f]

As slew rate improves with higher supply voltage - were you perhaps running a lower supply voltage and could try +/-15 or so? Curious if slew rate is part of the problem in addition to the saturation. Also there are input protection diodes in the 5534 that could perhaps be bypassed with external circuitry - based upon some things read in the DIY Audio forum.

Barry

Hi Barry,
I think the saturation is far worse than the slew rate.
It could be possible that by using a higher power supply voltage to obtain a better slew rate (I haven't run such tests yet). My conclusion at the time was that it is better to use a two stage amplifier (i.e. 30 x and 30x instead of a single stage of 900x), and limiting the level at the input of each stage (using antiparallel diodes). The diodes placed in the feedback circuit also worsen the response of the amplifier and Tinkerer has found that by using specific LEDs instead of standard diodes to obtain optimised function.
You can see his preamplifier circuits here: http://www.geotech1.com/forums/showthread.php?t=15470
And note Carl's recommendation to split the amplifier stage (for the second circuit).

Regards,
Nicolae

[bklein]

Hi Nicolae,
I think you are right about the two stage gain approach. Been a long time now but I used this in guitar amp design in the 70's. I also worked on analog video switchers after that and - like discussed in the link you forwarded - very useful design concepts. Wish now I didn't toss those schematics...
Also used in the guitar amps are the back to back diodes/leds/transistors in the feedback. The SD2000 schematics shown on this site use some cool concepts - like gating the coil signal, the feedback diodes, dual gain stages... neat stuff.

Barry

[Tinkerer]

Hi Barry,
I think the saturation is far worse than the slew rate.
It could be possible that by using a higher power supply voltage to obtain a better slew rate (I haven't run such tests yet). My conclusion at the time was that it is better to use a two stage amplifier (i.e. 30 x and 30x instead of a single stage of 900x), and limiting the level at the input of each stage (using antiparallel diodes). The diodes placed in the feedback circuit also worsen the response of the amplifier and Tinkerer has found that by using specific LEDs instead of standard diodes to obtain optimised function.
You can see his preamplifier circuits here: http://www.geotech1.com/forums/showthread.php?t=15470
And note Carl's recommendation to split the amplifier stage (for the second circuit).

Regards,
Nicolae

The attached circuit has given the best results so far. There is no saturation of the opamp OP37. (DD coil)
The FE discrimination works fine, even for difficult targets like Crown Cork in any relative position, tin cans etc. It discriminates perfectly between an Alu soda can and a steel can of beans of similar size.
nails and such are easy.

Tinkerer

[bklein]

The attached circuit has given the best results so far. There is no saturation of the opamp OP37. (DD coil)
The FE discrimination works fine, even for difficult targets like Crown Cork in any relative position, tin cans etc. It discriminates perfectly between an Alu soda can and a steel can of beans of similar size.
nails and such are easy.

Tinkerer

Umm... I don't really want any of those things

Do you have this in either larger format or different format?
Being a JPG it doesn't view or print well enough to read circuit details.

Barry

[Tinkerer]

Discriminating PI

Umm... I don't really want any of those things

Do you have this in either larger format or different format?
Being a JPG it doesn't view or print well enough to read circuit details.

Barry

Hi Barry,

knowing what is buried down there let's you decide if it is worth the trouble to dig it up.
Attached the EAGLE FILE of the circuit.

The TINKERERS_V1 IS AN "OPEN SOURCE" design. Anyone can copy it, build it and improve on it.
Any help in improving on it is most welcome.

Tinkerer