PI Project update

[Anonymous]

Since I posted my "platform design" over a month ago, I've been able to work on it about 2 whole hours. I did incorporate some of the feedback, but not all.
Eric suggested an RC tank for the coil... I added a series R to the MOSFET source but there is no room on the board for a big honkin' cap, so I added pads for an off-board cap. Yeah, I know, it needs to be close to the board.
Eric also suggested slaving the B sides of the pulse generators off the same RC. I did breadboard this, and it works very well! So the layout has the option of doing this or not.
Added a cap to pin 5 of the 555 per JC's suggestion. Added an option to connect R21 CT to pin 3 of IC6 for offset adjust, also per JC.
Two concerns that were not addressed: potential noise injection from the 7660 charge pump and the VCO.
Also, I still have not entered proper component values into the schematic. Please don't tell me the R and C values are wrong, I know that!
I would appreciate feedback on the layout, and any other feedback on the design itself. Remember, I have a lot of quirky hooks in this thing so the experimenter can try out all sorts of configurations. Parts of the layout may look strange in this respect.
Quick tour of the layout:
Top left is the 555 main clock.
Top middle is the 74221 pulse generators.
Top right is the supply circuits.
Bottom left is the coil transmit and receiver amp.
Bottom middle is the sampler, integrator, and 2nd amp.
Bottom right is the audio.
Total board size is 4.5x3.3 inches, a little bigger than I hoped. My plan is to make 4 initial boards using one of those awesome iron-on PCB transfer sheets - if it works(!), I will consider sending out for some nice boards if there is sufficient interest.
I would also like to reduce the design by eliminating the massive flexibility and spin a surface mount version. This would be really tiny!
Thanks,
Carl
<CENTER>
<IMG SRC="http://www.thunting.com/geotech/misc/images/pi1a.jpg">
<IMG SRC="http://www.thunting.com/geotech/misc/images/pi1pcba.jpg">
</CENTER>

 

[Anonymous]

Hi Carl,
I was having problems with the charge pump putting heavy ripple on the supply lines (input and output) and I was using the MAX660 which is a higher current version of the 7660. Per Eric's suggestion, I put a 100uH inductor in to feed the charge pump. This reduced noise feeding back. I also put low ESR caps in the Charge pump circuit which makes a huge difference. Finally, I put an RC filter on the output again using low ESR caps which reduced the ripple coming out of the charge pump. Maxim's web site has a good tutorial on proper layout and filtering for charge pumps. I am now using the National LM2662 which is the highest current charge pump I could find at 200ma. I'm doing some things that require more current than the 7660 could supply.
Thanks,
Charles

 

[Anonymous]

About noise: I got used to put LC decoupling circs (100n+ferrite bead) almost everywhere (but my filed of activity is quite different) - perhaps it could help here as well.
How about the OP37 instead of NE5534?
Regards,
pp

 

[Anonymous]

Hi pawel,
The OP37 works great and is what I am using for my playing around right now, running a gain of 2000 with 500 ohms input, 1 Megaohm feedback with 2-3pf capacitor in parallel with the 1 Megaohm. Nice and stable and low noise, with good frequency response. Only cost a few dollars.
JC

 

[Anonymous]

One of the things I really strived for was to make the preamp setup extremely flexible. There are plenty of tweak components and several ways to hook them in so you can use just about any opamp.
- Carl

 

[Anonymous]

Hi JC,
I mentioned the op37 because of its popularity - I started my precision measurements with it. In my (non-pi) magnetic experiments the LT1128 was the best one - and it costs only two times more then 37 - but I don't know if the noise below 1nVHz-1/2 is really necessary in PIs.
Regards,
pp

 

[Anonymous]

Sorry, it was LT1028 !!! The mixture of LT1115 and LT1028 names reminds me how old I am - and I don't like it at all.
pp

 

[Anonymous]

1nV/rtHz is about the level where the opamp will begin to add thermal noise. Most designs I've seen use a feedforward R on the order of a few 100 ohms up to 1k, so let's say 2-4nV/rtHz. So you would want the opamp to be less than this.
Of course, this is thermal noise, and electrical noise may be much higher.
- Carl

 

[Anonymous]

Charles,
I did a little experimenting with the 7660 today. I prefer to stick with it because (a) it's sufficient and (b) it's easy to find.
Anyway, I set it up just like I'm using it in the PI: 12v in, 24v out, and a 7805 referenced to the +12v line, to get a +5v regulated line WRT the 12v line.
I used a 100-ohm resistor to simulate a 50mA load on the regulator. Looking at it with just a single o-scope channel, noise looked pretty bad. When I set up using A-B difference mode it cleaned up quite nicely.
What I found is that an LC filter going into the regulator made an almost imperceptible improvement. I found that an equally good solution was to just use decent size electrolytics. Taking them down in value made the ripple show up pretty quickly.
I already have RC bypassing on both supplies feeding the preamp, and these could be LC if that helps any. I agree, low-inductance caps should be used here. Beyond that, I could try to squeeze in an LC at the regulator but room is really tight. I may wait to look at this after I make the first proto boards - I'm sure there will be other layout tweaks!
- Carl

 

[Anonymous]

In PI's, there is usually quite a bit of resistance noise in the receiver coil circuit. That will usually determine the noise figure as it relates to preamp voltage noise.
In addition, it is necessary to pay attention to the op amp bias current noise which will be dropped across the receiver coil circuit impedance. All JFET and CMOS input op amps, and most bipolar input opamps are acceptable in this regard, but a few bipolars have input bias current noise levels which are unacceptable in this application.
Also pay attention to the current noise which will be dropped across the grounded leg of the feedback network.
When using uncompensated op amps, remember that you usually can't use capacitors in the feedback network or it will be unstable or will oscillate.
Beware current feedback amps. The manufacturers usually don't specify the noise current at the low impedance input, and guess what? If you knew, you probably wouldn't use it in a low noise application unless the grounded end in the feedback network was less than ten ohms. NOW THE GRIPE! I can't be the only engineer who's had to find this out the hard way. It really irritates me that a thing like this would happen in an industry which is generally characterized by trustworthy information. Do they think engineers won't notice that the amp doesn't work in their application? Do they think the engineers who get suckered will trust other products from that company?
While we're at it: other dirty secrets of op amps. There's a lot of cleverness in op amp output structures. Most of those those little bits of cleverness buy you something at the expense of something else. The op amp manufacturers are eager to tell you the advantages of their clever tricks, but leave it to you to find out that it won't work in your application.
If they would at least publish a decent internal schematic of the op amp, us old timers who actually understand op amp guts could figure out whether the disadvantages were likely to cause us problems; and, if so, what we ought to do to solve them.
I admit that once you've been had by CFA inverting input bias current noise, you say to yourself "well stupid me, of course!", but still, the manufacturer knows it's a fatal flaw in MANY apps, and let's be frank-- leaving this parameter off the data sheet is (literally) about a picoamp/rt.Hz away from fraud.
--Dave J.

 

[Anonymous]

Carl,
Knowing your associations I admire your tact.
Graeme

 

[Anonymous]

Hi Pawel,
Do we subtract 1028 from 1115 to get your age <IMG SRC="/forums/images/smile.gif" BORDER=0 ALT="">
The OP37 is not quite as fast as the NE5534. In my Deepstar, it would be a struggle to sample at 15uS with the OP37, whilst there is plenty in hand with the 5534.
The noise performance of the front end amplifier has only been of secondary importance to its bandwidth and recovery time. Unless you use a noise cancelling coil arrangement (which is very rare on treasure hunting PI's), the pickup noise on the RX coil far exceeds the contribution of the amplifier.
Eric.

 

[Anonymous]

Other way to avoid noise is using Oxigen Free Copper. Not only OA is guilty for noise, coil material too. Let's look at Hi-End audio technology.

 

[Anonymous]

Hi Dave,
Welcome back from the land of OZ.
You made this statement, before addressing current feedback amplifiers.
"When using uncompensated op amps, remember that you usually can't use capacitors in the feedback
network or it will be unstable or will oscillate. "
Surely you meant this statement to apply to current feedback amplifiers only (which you talk about in the next paragraph) and not voltage feedback amplifiers (5534/2,op37,lm709,lm31 which are the commonly used devices.
I have used alot of the Comlinear Corp. current feedback amplifiers for other work, where the high speed was only met by using them, and they do have their problems and stability is one of them (no feedback caps). Have to pay more attention to layout and proper circuit design, but past that I had good luck with them.
So what is your favorite flavor of op amp?
JC

 

[Anonymous]

The OP37 is a well known example of a decompensated voltage op amp. Such op amps don't like capacitance in their feedback networks, for reasons that are usually explained in the data sheets.
--DJ

 

[Anonymous]

Hi pawel and Eric,
You are right again Eric, the 5534 is a bit faster than the op37 (at gain= 1000) and this translates into about 5us of delay difference roughly. With a 700uh coil, and leaving no margin on delay, I am getting 10us with 5534 and 15us with the op37.
The offset voltage and bias current drift with temperature is better with the op37, if someone is using large unswingable coils or working in areas where swinging coils is akward, and is therefore DC coupled (or this is a switched option) this can become important, at gains of 1000 or more. If larger objects are the game, then the added delay is not so important.
The noise of the amplifier itself is also brought out by the large gain and wide bandwidth (from microvolts rms to millivolts rms). But if it is pretty much truely "noise" it will get integrated and filtered (averaged) by later stages to the point it is not so important in terms of voltage.
Popcorn noise, shot noise, impulse noise or whatever you like to call it, is more distructive in this application, but it is not something normally specified on datasheets. It is this random pulse noise that can be very bad (large spikes) on some amplifiers that may cause problems, and probably will not get integrated out on a second to second period. The orginal PMI series of OP07s,etc. and such were much quieter than amplifiers before them.
But then like Eric pointed out, unless you find a quiet place on the planet to test it, outside "noise" will dominate the whole equation.
JC

 

[Anonymous]

It isn't just one company which leaves the embarrassing CFA inverting input noise current spec off the data sheet. Guess I'm not the only one that got burnt and was unhappy about it: it's starting to show up now on data sheets.
The problem with poorly documented output structures is a widespread problem in the industry going clear back to the venerable LM324, which has a deadband wide enough to "walk a mile for a Camel" through. The deadband would have been OK if the manufacturers (it is widely sourced) had properly documented it.
Ever notice that the input voltage noise of comparators is almost never specified? This isn't because the noise level itself is an embarrassment, it's because the manufacturers don't know that comparator noise matters in some applications-- like metal detector phase references.
Carl is an engineer who, I suspect, feels about the same way I do about these things, probably having been bitten a few times himself. Although he may not be in a position to say anything about it publicly, it wouldn't suprise me if he's doing a little missionary work behind the scenes.
The electronic components industry as a whole, is dominated by good quality products and good data sheets. Little discrepancies like that CFA bias current noise thing stand out so prominently, simply because they're so uncharacteristic of industry custom.
--Dave J.

 

[Anonymous]

Hi JC,
For some faster applications where I wanted a sample delay of less than 10uS. I tried the AD8055. This amplifier worked very well except for the "flicker" noise. You could actually see the dc baseline jumping about on the 'scope. Only a few mV, but enough to give a rougher threshold on the final dc amplifier, compared to a 5534. As a confirmation of this, I ran several samples of the AD8055 with the dc amplifier connected to a chart recorder.
Eric.

 

[Anonymous]

Hi Dave,
The op37 data sheets show applications of this part with capacitors in the feedback loop. Besides I thought that putting a capacitor in parallel with the feedback resistor was a common way of compensating all voltage feedback amplifiers, and/or limiting the bandwidth.
Do you know of a decompenated part datasheet which does explain this? the op37 datasheet doesn't.
JC

 

[Anonymous]

Hi Eric and all,
Good information as usual, won't spend my money on that amplifier. Looks like someone at signetics years ago did a good job on the 5534 amplifier, since not much "modern" is knocking its socks off. The uA709 is even older and still holding on as well, cept the stuff in the metal cans is getting more expensive with time, and parts are going away.
Never thought pawels post on the op37 would generate so much interest. Glad to see things pick up a bit.
JC