Low Noise Front End for P.I.

[Ferric Toes]

Hi All,

I've had yet another look at alternatives to the NE5534A for the RX front end amplifier. I am trying to find something with a lower noise spec at low frequencies, i.e. 10kHz and downwards. The LT1028ACN8 looked ideal on paper, but in practice, not so good. Had lots of ringing too, on the edges of the TX transients. I tried frequency compensation as per the data sheet, both on the feedback R and between pins 5 & 6. In the latter case, any capacitance made the ringing worse. I don't think it is a layout or supply decoupling problem as much faster amplifiers such as the AD8055, or OPA620 work fine, but the low frequency noise with these, is much too high. I currently use two opamps in the front end; a low noise one with a gain of 10 x, followed by a fast one (8055) with a gain of 50. This gives me speed for short delays, with low noise. That should be the case with the LT1028, but something is amiss.

Has anyone tried any other alternatives? particularly low or ultra low noise devices?

Eric Foster.

[Carl-NC]

Slew rate looks low on the LT1028, might not have enough gas for a PI. I'm still using the AD8031, mostly because it's what I have a lot of, but also it performs a little better than the NE5532. I haven't done thorough testing, though, on any opamps.

I think opamp noise is often spec'd at midband frequencies, and it can be higher at low freq., probably due to 1/f noise.

- Carl

[Ferric Toes]

Hi Carl,

Linear Technology are being helpful and suggested that a LT1007/LT1037 may be better, due to my highish source impedance (2K2). Reg Sniff also suggested a MAX437 which is equivalent. MAXIM data is good because it gives comparisons with the 5534 on noise versus source Z.

With a 5534, I dropped the input resistor to 1K, and the feedback R in proportion, and noticed a drop in noise at the output of the detector. It will be interesting to see if the additional 1nV root Hz that the MAX437 gives will be noticeable.

Eric.

[Carl-NC]

I would think that, since 1K => 4nV/rtHz, we're likely limited more by the resistor, rather than the opamp.

Now that I work for Maxim, I need to get ahold of a variety of Maxim opamps and try them out.

- Carl

[amtech2005]

low noise front end for P.I.

Just an Idea. 10 KHZ or Less miniature torroid coil say 0.01 micro-henry or greater to phase out noise on the opamps. Instead of experimenting with capacitors, Just a crazy Idea to phase out noise. Eugene [email protected]

[Ferric Toes]

Yes, the data sheet shows the 1k resistor noise as 4nV, and the addition of the MAX437 gives a total noise of 5nV. The 5534 ends up at 6nV total. Going from a 2.2k input resistor to a 1k drops the resistor noise by about 1nV rt Hz, and this was evident on a chart recorder plot I made of the detector's output. The Maxim data gives the 437 about 1nV better than the 5534, so hopefully I will see that too.

Incidentally, I typed "resistor noise" into Google, and it came up with a lot of interesting info, like how to reduce it using RC networks around an opamp circuit. I can see hours of "fun" looming up ahead.

Historically, opamp and resistor noise has not been a real problem in PI circuits. However, in some recent designs, particularly those using noise cancelling coils, the performance is not limited by the coil pickup noise, but by circuit noise. That is why I am pursuing this problem yet again.

Eric.

[JC1]

EE

Hi guys,

When adding noise, say op amp and resistor, it is the square root of the sum of the squares.

i.e. 2 nv/rtHz and 3 nv/rtsq sources would add to

sqrt((2^2)+(3^2))= 3.6 nv/rtHz instead of 5 nv/rtHz

a small point but can become important.

also resistor noise is theorical and all resistors produce more noise than calculated. Metal film and "low noise" resistors come close, but carbon can be much greater.

most resistor noise charts are for 25 degrees C and this is fine, but for coil currents above 1 amp the damping resistor can get pretty hot if not physically large enough. I build mine from seperate metal film resistors in || and series to get much more wattage than required.

So Eric down to the circuit noise. Very good indeed!

The correllated double sampling and subtract needs to be able to help lower any low frequency noise as well as 60 Hz.

Also can play around with low noise Fet and build your own front end amp, and follow with, or close loop with op amp.
But you will need a large area Fet and this will require a bit of drain current.

Kinda like John Corbin (not me) did with transistors.

JC1

[Ferric Toes]

Hi JC1,

Can you clear up something that is puzzling me? If I put 10 x 10k resistors in series to get 100k, the noise, at 13nVrtHz per 10k works out to 41.1nVrtHz. (sq.rt of sum of sq's). Is it the same if the resistors are in parallel, i.e. totalling 1k. Or is it nearer the 4.1nVrtHz for a single 1k resistor.

I've come to the conclusion that it is resistor noise that is limiting me at the moment. I've done measurements without the TX connected and the NE5534 gives the lowest noise. This is with the sampling, integrating and dc amplifiers and filters all connected and running. Removing preamp stages and grounding the input to the sampling gates gives a noise level 40db lower than with the preamps in.

Eric.

[_Raz_]

Hi Eric,

Is it the same if the resistors are in parallel, i.e. totalling 1k. Or is it nearer the 4.1nVrtHz for a single 1k resistor.

I may be wrong, but I believe the noise for any series or parallel combination is equivalent to a single resistor with the same resistance as the combination.

the NE5534 gives the lowest noise.

I have to thank you for bringing up this subject. I also have been looking for a replacement for the NE5534 in the frontend of my design, without too much luck (as yet). Devices that look good on paper don't always do so well in the real world. Well, not in the frontend of a PI detector anyway.

How did you find the Max437? I am assuming from your statement above that it did not perform as well as the NE5534.

Mark

[Ferric Toes]

Hi Mark,

Two opamps I have recently tried, are the MAX437 and the LT1208. Neither worked in a PI application. The MAX437 did not like being saturated by the diode clipped back emf spike. It stayed in saturation for over 30uS. The LT1028 exhibited serious ringing on transients, which I was not able to cure by any compensation capacitor; either on the feedback, or between pins 5 and 6. Running the amplifiers without the saturating transients and just looking at the noise, the NE5534A was best, even though the data sheets on the others indicated that there should be an improvement.

By the way, the opamps under test are running at a gain of 10 x to maximise bandwidth with low noise performance. It is followed by a fast AD8055 with a gain of 50, to give an overall gain of 500.

Eric.

[treasurediver]

AD797 ultra low noise

Hi all,

has anybody tried the AD797 opamp for the front end? Specs give a 0.9nv noise figure?
Treasurediver

[Ferric Toes]

Hi Treasurediver,

The AD797 certainly looks like a good one to try. I'll see if I can get one.

Thanks, Eric.

[Carl-NC]

most resistor noise charts are for 25 degrees C and this is fine, but for coil currents above 1 amp the damping resistor can get pretty hot if not physically large enough. I build mine from seperate metal film resistors in || and series to get much more wattage than required.

You can easily calculate the noise... for a resistor the noise energy density is

vn^2 = 4*K*T*R (volts^2 per Hz)

where K = Boltzmann's constant (1.38e-23), T=temperature in Kelvin, and R is the resistor value. So for a 1K resistor at 27C (=300K) the energy density is

vn^2 = 4*1.38e-23*299*1000 = 16.56e-18 v^2/Hz

or vn = 4e-9 v/sqrt(Hz) (i.e., 4nV/rtHz)

A hot 1K resistor -- say, 100C = 373K -- ends up with a vn = 4.54nV/rtHz.

So, since noise is proportional to the square-root of temperature, and we're dealing with the Kelvin scale, temp. has a fairly minimal effect. Still a Good Idea to keep the damping resistor cool.

I may be wrong, but I believe the noise for any series or parallel combination is equivalent to a single resistor with the same resistance as the combination.

Yup, that's correct.

has anybody tried the AD797 opamp for the front end? Specs give a 0.9nv noise figure?

I have some, but my through-hole board is down right now.

- Carl

[JC1]

EE

Hi Guys,

Hey, I'm not going to have the time to go into all the nitty gritty engineering details of some of my statements.

So I will try to provide some link which explains a bit more.
Of course could probably make a career out of resistor/material noise analysis.

As I said before, the theory is the minimum noise.
Only gets worse from there.

Here it is:

Cut and paste.

Resistor noise is made up of three main types: thermal, contact, and shot noise. Thermal noise is mainly dependent on temperature, bandwidth, and resistance, while shot noise is dependent on bandwidth and average DC current, and contact noise is dependent upon average DC current, bandwidth, material geometry and type.

Wirewound resistors are the quietest, having only thermal noise, followed by metal film, metal oxide, carbon film, and lastly, carbon composition. {wirewound can also be very inductive and this can cause other problems}

Since the characteristics of thermal noise have a Gaussian probability density function, and the noise of the two separate sources is uncorrelated white noise, the total noise power is equal to the sum of the individual noise powers. If you model the individual resistors as noise generators, the output noise voltage will be equal to the square root of the sum of the squares of the individual noise sources.

The above equation shows that the noise varies in direct proportion to the square root of the resistance, so if you take two resistors of half the value and square the square root and add them and take the square root of the sum, you end up with the exact same value as you would if you took the square root of a single resistor of twice the value. Therefore, the total noise remains the same.

Contact noise is dependent on both average DC current and resistor material/size. The most significant contributor to noise in guitar amplifiers is the use of low-wattage carbon composition resistors. Since the noise is proportional to resistor size, the use of 2W carbon comp resistors will improve the performance over that of 1/2W resistors. Studies have shown a factor of 3 difference between a 1/2W and a 2W carbon comp resistor operating at the same conditions. {same conditions! not even hot, factor of 3, now heat it up and watch what happens!!!}

The predominant noise in carbon comp, carbon film, metal oxide, and metal film is composed of contact noise, which can be very large at low frequencies because it has a 1/f frequency characteristic. Wirewound resistors do not have this noise, only resistors made of carbon particles or films. This noise is directly proportional to both the current flowing in the resistance and a constant that depends upon the material the resistor is made of.

The material and geometry of the resistor can greatly affect the contact noise. Therefore, if you double the power rating of the resistor, which increases the size and area, you will reduce the contact noise generated by the resistor.

http://www.aikenamps.com/ResistorNoise.htm

Actually getting real numbers on the resistors that are going to be used can be tricky.

The only resistors I have seen that come close to the resistor thermal noise equation were the MOX resistors made by Victoreen {got bought by Ohmite} and they actually do come very close, I measured about 10% more.

http://www.submm.caltech.edu/~cdd/si.../sichrome.html


JC1

[_Raz_]

Hi Eric,

The MAX437 did not like being saturated by the diode clipped back emf spike. It stayed in saturation for over 30uS.

Thanks for the information. The 30uS kind of makes the MAX437 useless for any serious PI front-end. Too bad, as I actually found some in my component collection.

I have looked at a few different ultra low noise op-amps, such as the OPA847 and LMH6624, but their noise characteristics aren't too good as you go below 10Khz or so.

The AD797, as suggested by treasurediver, looks very good on paper. I'll have to track one down too and do some tests. I remember seeing a detector schematic on this forum that used the AD797 in the front-end, a quick search revealed the post below (by Cossaro Fernando).

http://thunting.com/geotech/forums/s...ad.php4?t=8107

the opamps under test are running at a gain of 10 x to maximise bandwidth with low noise performance. It is followed by a fast AD8055 with a gain of 50, to give an overall gain of 500.

Thanks for the tip. I use much less gain in my pre-amp stage (around 50) and don't have any more gain until after the demodulators (4066). I do notice quite a bit of TX pulse noise on the receive signal. I guess I should be thinking of a second pre-amp stage to raise the signal level before the demodulators and probably more filtering on the power inputs of the devices after the demodulators.

Mark