Front end opamps, which specs matter?

[GT Blocker]

Hey again, all. I was looking at opamps for the front end of the HH (IC6 - NE5534) and began to wonder just which specifications are the most important. The top three seem to be bandwidth, input noise, and slew rate. Are there others that need to be considered? Which parameter is the most critical, the next, and so on? I have a few LME49710NA's to try 'cause it seems to me ->Ne5534: BW=11mHz, Noise=4nV/sq.Hz , Slew=13v/uSec Whereas -> LME49710: BW=55mHz, Noise=2.5nV/sq.Hz, Slew= 20v/uSec.
I wondered if I'm looking at the right things when considering other opamps. Is the slew rate that important, or the noise levels more so? Seems like anything over 1mHz on BW ought to be ok...though I have no idea what the lower BW limit might be. Any input on this one? Thanks...GTB

[Tinkerer]

Front end opamp

Hey again, all. I was looking at opamps for the front end of the HH (IC6 - NE5534) and began to wonder just which specifications are the most important. The top three seem to be bandwidth, input noise, and slew rate. Are there others that need to be considered? Which parameter is the most critical, the next, and so on? I have a few LME49710NA's to try 'cause it seems to me ->Ne5534: BW=11mHz, Noise=4nV/sq.Hz , Slew=13v/uSec Whereas -> LME49710: BW=55mHz, Noise=2.5nV/sq.Hz, Slew= 20v/uSec.
I wondered if I'm looking at the right things when considering other opamps. Is the slew rate that important, or the noise levels more so? Seems like anything over 1mHz on BW ought to be ok...though I have no idea what the lower BW limit might be. Any input on this one? Thanks...GTB

The NE5534 is a bit slow for sampling at below 10us. You need more BW and Slew rate. The specs of the LME49710 look OK, will probably work better, but only trying will really tell.

Tinkerer

[GT Blocker]

I'm still curious as to how they relate...

Thanks, for the reply, Tinker.
I had thought even the 5534 BW would be adaquate, which shows how little analog experience I have. I guess my real question is how do BW, slew, and noise lvls each come into play in the choice. I.E. - why choose the 5534 for the front end and not just use two more for the integrator,etc? Why the TL072/82?
As we've discussed, I'm looking for 7uSec- or so for small gold and I realize that this is hard to do. The better I understand each part of the circuit,tho, I think the better my chances of taking advantage of the newer IC's on the market. This slew/BW thing is something I'm just not getting, I guess. Not having as much noise passed up line is a no brainer. Perhaps the rest of my question is as well.
While I'm asking dumb questions, Is there any advantage to clocking the 7660 (IC1) from a seperate clock? The Maxim specs state that best efficiency is @ 10kHz. Does the adjustment of the Freq and PW (R2/R3-IC5) have any effect on the PWR supply stability? I am considering a daughter board 556 for the IC5 socket, with seperate clocks for IC1 and the rest. Waste of time? Anyone?

I'm sure I'll think of a bunch more stupid questions in the near future as I try to come up to speed......STAY TUNED (bad pun?)

[J_Player]

I'm sure I'll think of a bunch more stupid questions in the near future as I try to come up to speed......STAY TUNED (bad pun?)

Hi GT Blocker,
You are asking the right questions about the amplifiers. I can't help much with that, but after you have made your choices of circuitry, there is another bottleneck you should look at. The coil must be fast enough to keep up with the circuit. BB Sailor wrote an excellent article showing how to build a fast coil here: http://geotech.thunting.com/cgi-bin/...oils/index.dat

Good luck with your hammerhead,
J_P

[Carl-NC]

Hi GT,

I would add one more spec to the three you mentioned: overvoltage recovery. Actually, this often is not even spec'd. It is the time the output takes to come out of hard-railed saturation. Some opamps that otherwise look great have poor overvoltage recovery.

After that, the BW spec determines how much gain you can run. This spec is actually a gain-bandwidth product, so an NE5534 (GBW=11MHz) with a gain=1000 (HH) will have a BW=11kHz. This results in a response time constant (tau) of 14.5us. If you want to sample at 15us, then you really need a faster response, so either a higher GBW or a lower gain (for some reason I thought the NE5534 had a higher GBW ). Ferinstance, let's say that at 15us we want to be at 3*tau settling on the opamp, so that means tau = 5us, and BW = 1/(2*PI*tau) = 31.8kHz. For GBW = 11MHz, this limits the gain to 345. You can run a higher gain (as with HH) but the opamp settling won't be out of the way and can degrade sensitivity. So HH really could use a reduction in preamp gain, which Reg Sniff has pointed out previously. Or a better opamp.

After BW, look at slew rate. Usually it's not much of a factor. With a BW = 31.8kHz, the exponential response of the opamp output is limited to about 1V/us, which is well below the slew limit of 13V/us.

The integrator is actually a sampling integrator, so you want a very high input Z opamp, such as a JFET-input opamp. Also, you don't need the speed or power of the preamp.

You can do better both in performance and power consumption with newer opamps. I didn't use them in HH specifically because for many of our foreign friends, they're hard to get. So I used vanilla parts.

Also, JPlayer is right, to sample below 15us you need to pay a whole lotta attention to the coil; also the cable, and the MOSFET switch.

On the 7660, I clocked it off the master clock to make the noise glitch synchronous with the sampling. Yes, it is less efficient,but should give lower interference. Honestly, I don't think I've seen any difference.

- Carl

[GT Blocker]

Thanks for the replies...Hope I got this right.

"I would add one more spec to the three you mentioned: overvoltage recovery. Actually, this often is not even spec'd. It is the time the output takes to come out of hard-railed saturation. Some opamps that otherwise look great have poor overvoltage recovery."

Is there anyway to infer the overvoltage recovery from the specs that they do publish? Just wondered...

"After that, the BW spec determines how much gain you can run. This spec is actually a gain-bandwidth product, so an NE5534 (GBW=11MHz) with a gain=1000 (HH) will have a BW=11kHz. This results in a response time constant (tau) of 14.5us. If you want to sample at 15us, then you really need a faster response, so either a higher GBW or a lower gain (for some reason I thought the NE5534 had a higher GBW ). Ferinstance, let's say that at 15us we want to be at 3*tau settling on the opamp, so that means tau = 5us, and BW = 1/(2*PI*tau) = 31.8kHz. For GBW = 11MHz, this limits the gain to 345. You can run a higher gain (as with HH) but the opamp settling won't be out of the way and can degrade sensitivity. So HH really could use a reduction in preamp gain, which Reg Sniff has pointed out previously. Or a better opamp."

So if I understand this.......If I want to reach 7uSec, then I have a tau of 2.33uSec, correct? If that is correct, and my GBW is 55mHz on chip x, then the max BW for the chip should be 68.3kHz, correct? So...If I'm not wrong already....A gain of 800 would result in a BW of 68.7kHz, just slightly over the 68.3kHz max. Did I get it right (or should I go to stand in the corner)?

"After BW, look at slew rate. Usually it's not much of a factor. With a BW = 31.8kHz, the exponential response of the opamp output is limited to about 1V/us, which is well below the slew limit of 13V/us."

Is there a point in this circuit where slew does come into play?

"The integrator is actually a sampling integrator, so you want a very high input Z opamp, such as a JFET-input opamp. Also, you don't need the speed or power of the preamp."

Jeeze...I feel kinda stupid on that one...the FET thing should have clued me in. I rarely had to worry about high impedence in FX, mostly servo control. Makes sense...

"Also, JPlayer is right, to sample below 15us you need to pay a whole lotta attention to the coil; also the cable, and the MOSFET switch."

I think I have read nearly everything that BBsailor has posted on both this board and Eric Foster's. He is one of (from what I have read) the best at fast coils. As far as cable, for the time being, I will use as short a hunk of s-video cable as poss. I plan to move the MOSFET and preamp circuit down to the coil later anyway. At this point, I'm going to start with the IRF734 for the MOSFET given the coss of 190pf, 450v, 4.9a specs (we'll see, I like that 190pf)

"On the 7660, I clocked it off the master clock to make the noise glitch synchronous with the sampling. Yes, it is less efficient,but should give lower interference. Honestly, I don't think I've seen any difference."

I hadn't thought about the clocking timing with the glitch (though I think you actually mention that in the HH assembly instructions). Deff not a good idea to mess with that, thanks for reminding me to remember what I read...lol.

I think I'm gettin' it.......thanks for all the help!

[Carl-NC]

Is there anyway to infer the overvoltage recovery from the specs that they do publish? Just wondered...

No, you have to test them for this.

So if I understand this.......If I want to reach 7uSec, then I have a tau of 2.33uSec, correct? If that is correct, and my GBW is 55mHz on chip x, then the max BW for the chip should be 68.3kHz, correct? So...If I'm not wrong already....A gain of 800 would result in a BW of 68.7kHz, just slightly over the 68.3kHz max. Did I get it right (or should I go to stand in the corner)?

Yup, perfect. The "3*tau" settling is somewhat arbitrary, you just want to make sure the opamp settling is not a major contributor to the overall settling.

Is there a point in this circuit where slew does come into play?

The highest slew is at the coil & switch, on the order of 100V/us. Which is why you want a really low-C coil, cable, switch, etc. The opamp doesn't see this slew rate and, as I mentioned, response is usually limited by the GBW. Once you set up the opamp for a certain gain, take the resulting tau and the peak output voltage (usually set by the rail voltage) and calculate the maximum BW-limited slew: SR = Vmax/tau. For your example (tau=2.33us) and a 5V supply, SR = 5V/2.33us = 11.67 V/us. This is well below the part's internal SR limit.

I think I have read nearly everything that BBsailor has posted on both this board and Eric Foster's. He is one of (from what I have read) the best at fast coils. As far as cable, for the time being, I will use as short a hunk of s-video cable as poss. I plan to move the MOSFET and preamp circuit down to the coil later anyway. At this point, I'm going to start with the IRF734 for the MOSFET given the coss of 190pf, 450v, 4.9a specs (we'll see, I like that 190pf)

I have doubts about the advantages of moving circuitry to the coil. You will need to provide power, perhaps 2 voltages for the opamp and a high current for the switch. Also clocking, with a fast edge. And then return a signal. And all it buys you is to get rid of the cable capacitance, which can be pretty easily made non-dominant. I think it's a lot of hassle for little gain. I may be wrong.

- Carl

[GT Blocker]

Thanks Carl! (I'm dumb...)

Thanks for clearing up some of my ignorance on SR & BW. Your explanation obviously helped a great deal (or I'd be sitting in the corner..)

I hadn't thought about clocking, i.e., "Also clocking, with a fast edge." The circuitry move didn't seem as nasty until then...in retrospect... more hassle than it's worth, as you mentioned.NM.

I'm glad that I got the gist of the bandwidth/slew thing though. I really want to thank you for making that understandable to me. Hopefully this thread will help clarify some of the mysteries that the 'analog challenged' among us may not understand as yet.

If anyone has any other circuitry related ideas for shorter sample times on the HH (as opposed to coil design), please jump in! I can use all the help I can get.

Last - I want to take a moment to thank everyone that has commented and tried to help with my questions. I have avoided forum boards in the past, due to flames, infighting, etc. It's nice to see civilized (and I would personally say, professional) conduct on a forum board. You folks are a pleasure to 'work' with.

[GT Blocker]

Ummm..say Carl, one more Q (yeah right..)

Carl, in response to an earlier post in this thread, you mentioned,

"After BW, look at slew rate. Usually it's not much of a factor. With a BW = 31.8kHz, the exponential response of the opamp output is limited to about 1V/us, which is well below the slew limit of 13V/us."

Is there a formula to calculate this or is it in the graphs on the chip's specs? It appears not to be much of an issue, given a 13:1 margin, but I was a bit curious as to how to calculate the SR's vs. BW needed.

[Carl-NC]

A BW-limited step response is exponential, with the form v(t) = V*e^(-t/tau). The highest slew rate is right at the beginning, at t=0+, so all you need to do is take the derivative at t=0:

dv/dt = -V/tau*e^(-t/tau)

dv/dt|t=0 = -V/tau

This is for a falling response; a rising response is the same, but positive slope. "V" is usually the supply rail for the opamp. In terms of BW, this becomes:

dv/dt(max) = V*2πf

- Carl

[GT Blocker]

an update

I've been running the National Semi LME49710 in place of the 5534 with no problems so far. Once I can start pushing it past the 'stock' HH (faster coils, quieter environment, etc.) we'll see if it's truly better.

It does act as a drop in replacement, however (at least at this point).
If anyone else tries it out, let me know what you think, if you would.

I also replaced the TL072/082 with a Natsemi LMP7702 (you need a dip carrier for the through hole board though, soic only), and no problems (yet, same caveat).

The BW and slew are MUCH lower on the 7702 although it doesn't seem to matter. I'm hoping the lower noise/rtHz will make a difference later on.

Comments?

Take care, and good hunting, GTB

[rajesh]

I've been running the National Semi LME49710 in place of the 5534 with no problems so far. Once I can start pushing it past the 'stock' HH (faster coils, quieter environment, etc.) we'll see if it's truly better.

It does act as a drop in replacement, however (at least at this point).
If anyone else tries it out, let me know what you think, if you would.

I also replaced the TL072/082 with a Natsemi LMP7702 (you need a dip carrier for the through hole board though, soic only), and no problems (yet, same caveat).

The BW and slew are MUCH lower on the 7702 although it doesn't seem to matter. I'm hoping the lower noise/rtHz will make a difference later on.

Comments?

Take care, and good hunting, GTB

hi you has try ic 6 i mean tl 081 gary used this .........

[learner]

Hi Guys,
can i use a TL 092 in place of the TL 072/082?

[rajesh]

ic no 6 / 8 for h h

[quote=learner;77432]Hi Guys,


dear guys never use but i will try sunday.............

[GT Blocker]

TL092

Hey there learner,

I think the TL092 would be a bad choice, though I bet it will work.

It has nearly twice the noise /sq.rt. hz (34) vs 18nv/sq.rt.hz for the TL072 (I didn't lookup the 082). Each stage amplifies the noise from the previous stage by it's gain (like duh), as was pointed out to me recently {which, btw, wasn't my question in that thread}.

You really would rather not add any more noise 'upstream' than you can avoid, and the integration stage is the middle op-amp and part of that chip.

The LMP7702 I'm testing has a noise factor of only 9nV/sq.rt.hz and seems to work ok (so far) even though the BW is only 2.5mHz and the slew is a measly 1v/uSec.

From what I can tell, the 'back-end' doesn't seem to be too picky about BW and slew, though as I push things a bit, I may eat these words.

As far as the 7702, I'm wondering if the low slew affects the integration, although things seem ok so far.

I don't see why the TL092 wouldn't work (same pinout/series) so if that's what ya got handy...give er a try.