Low Noise Direct Coupled Preamp

They wanted to find _good_ batteries, so why would they even care? I mean, you can always find crappy examples of anything.

"Doctor! Doctor! It hurts when I go like this!"

"So don't go like that."

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

Phil Hobbles about wrote: ====================

** Another massive red herring - only thing the Hobbler f*****ad can ever come up with.
** Totally irrelevant to what was posted.

You have no experience with cells or batteries at all . So can go f*ck yourself .

...... Phil

Reply to
Phil Allison

I don't want to get more noise power. This calculation has no meaning in this application. [...]

I don't see how moving a few wires would change the low frequency response.

I do not want to measure very low frequencies.

The 1/f noise would dominate the measurement and make it useless. I want to start rolling off around 1 Hz, or even a bit higher. Fotunately, the design of this preamp makes it very easy to change the upper and lower cutoff frequencies.

A capacitive coupled preamplifier would suffer extreme transients if the input capacitor value is changed. The preamplifier could be destroyed if the capacitor was suddenly connected to a large voltage source.

Direct coupling as shown here changes the transient into a simple slew rate issue.

I could make a NIST correlator to measure very low frequencies, but that is a different ball game and out of the scope of this preamplifier,

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

My experience is with coaxial cables built using solid teflon dielectric between shield and center conductor.

Such cables show a very large change in electrical length around 20 C, where the teflon crystalline structure undergoes a change in crystal structure. Both dielectric contsant and physical length change, and I don't offhand no which causes the change, perhaps both.

This change is called the "teflon knee" in the literature.

Foamed teflon still shows this effect, but reduced in proportion to the fraction of air to teflon in the foam. Amorphorus teflon does not show this effect. Neither do polyethylene and polypropylene, and their alloys (collectively called "polyolefin" dielectric in the cable literature).

Joe Gwinn

Reply to
Joe Gwinn

You wrote, "The typical 1/f corner frequency of the LT1028 is 3.5 Hz. This limits the usable low frequency response of the preamp to 1Hz or so."

But that isn't true, as I demonstrated. Your circuit is your circuit, and does what you want it to, but it isn't the only one in the world. That's a separate point.

There's been a lot of fuzzy reasoning in this thread, basically along the lines of "my filter cuts off everything past the 3 dB point." When the interfering signal is very strong, as the tempco forcing in biased capacitors is, you have to be way down the skirts of the filter before it becomes negligible.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Why should I measure shit that cannot power my measurement amplifier for ten minutes? I have a few of these 9V blocks for my garage door openers and some CMOS multimeters, but absolutely no incentive to measure their noise.

An 18650 Li pack runs my measurement amplifier for a day and during the night it sits in the charging station and is fresh the next morning. That's how I spell actual use.

You would probably castrate yourself for the fun of it and then complain that you have few children.

Big NiCds or NiMH or Lithium batteries work. The bigger the better. With batteries, too.

Gerhard

Reply to
Gerhard Hoffmann

Yes, it is a single pole filter. The bandwidth is measured at the -3dB points, as is normal for filters. If you know it is a single pole and the

3dB point, you can easily find the response at any point on the curve. The LTspice PLT file gives you that curve.

A single pole filter has a pretty crappy rolloff, but that is not a problem in this application. The measurement time is too fast for the capacitor to change temperature and affect the reading.

I do not need much low frequency response. All I need is to find the corner frequency and the noise floor.

The corner frequency is where the noise starts to rise at about a 45 degree angle. The 1/f noise is pretty much the same for most sources, so once you have the corner frequency, there is not much more to learn.

Measurement speed is also an issue. I may want to sort through dozens or hundreds of LEDS to find the best brand and the quietest ones. If the filter frequency response is too low, I will have to wait longer for the measurement to settle.

The noise floor is much more significant. This separates the noisy sources from the quiet ones. The quiet ones is what I am after, so that is the main emphasis. The noise floor occurs at a much higher frequency than the corner frequency, so low frequency response is not a major concern.

I expect the noise sources to be much greater than the noise floor of the instrument. If I need a lower noise floor, I can always use multiple units in parallel, up to the practical limit set by 1/sqrt(N).

Temperature coefficient of the capacitors is not a concern. The instrument must be in a shielded box to eliminate RFI. This also reduces the random air currents present on the outside. If tempco becomes an issue, I can dump the circuit in a jar of mineral oil to minimize any fluctuations.

Having a higher rolloff frequency also minimizes any effect of capacitor tempco. The cutoff frequency is simply too high for the capacitor to change temperature much and affect the reading. If it still becomes an issue, I can raise the cutoff frequency to allow less time for the capacitor to change value.

All I need is to be able to find the corner frequency, which may be in the tens or hundreds of Hz. Not much is going to happen to the capacitor temperature in that short of time.

I have presented an approach that eliminates the capacitor coupling at the input, that eliminates any danger of destroying the base junction of a bipolar or jfet due to current surge, that settles quickly, is easy to change the rolloff frequencies, has reasonably low noise due to the LT1028, and is simple and inexpensive.

And did I mention - it gets rid of the coupling capacitor at the input.

It was also designed in the Theta state.

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

Are you going to use battery power for the whole system or just as the reference for a voltage regulator, such as the reference DC voltage for a capacitance multiplier ?

In the latter case, two or three silver oxide button cells should do for a long time..

Reply to
upsidedown

Thanks.. When I think about it the vibration stuff is easier to see in the time domain. Or something that sums all the noise . (I measured noise with analog multiplier and low pass filter.)

Yeah the size of the spur will relate to the bin size of your (D)FFT. (bin size in frequency space)

Oh I am suitably impressed. I love all your noise measurements.. good stuff.

George H.

Reply to
George Herold

Yeah well the lab 'dogma' was to change 9V's early and often.... (a few dollar battery is nothing when you are using ~$100's in liquid helium.)

George H.

Reply to
George Herold

Oh I just wanted to add a footnote to the battery discussion. When I first started doing some (serious) noise measurements, I got a little metal box for the opamps and such, and powered it from two 9V batteries. Since I'd had vibration problems in the distant past, I put cap multipliers on each rail. I did look at the supply noise, while talking loudly to the box... nothing. So I moved on to the 'real' experiments.

George H.

Reply to
George Herold

The low ripple power supply is for a GPSDO that needs to be powered on continuously.

The Low Noise Direct Coupled preamp has to be run off batteries. These can be lithium and charged when the system is not in use.

I'm a bit undecided on the source for measuring LEDS and Zeners. The current needs to be adjustable, perhaps up to 10 mA. This could be from a simple resistor connected to the battery, or a current source also connected to the battery. The current source may actually have more noise than a simple resistor. I don't know, and will have to try both to see which is better.

The source for measuring power supply noise will obviously have to be a power supply. Even with Jim Williams 100uV low noise switcher, the ripple will be very high, so all the precautions of measuring in a shielded box and running off batteries may not be totally necessary.

The Low Noise Ripple Filter is expected to bring the noise and ripple down to the low microvolt region, which is much greater than the Direct Coupled preamp noise.

So running off batteries during the measurement may be overkill in this application.

Finally, the output of the preamplifier has to be amplified as necessary to bring the level up into the volts region so it can be sampled with an A/D.

Then it goes to a FFT, which will be a new experience for me. I have never donw a FFT, but I understand there are two main considerations. The high frequency limit is determined by Nyquist, which states the sampling frequency has to be at least twice the highest frequency of interest. For a

1 MHz bandpass, this means the sampler has to be run at or above 2 MHz. So far, so good. But it means the sample clock noise has to be kept out of the preamp, which may be difficult to do even with extensive filtering and shielding.

The second consideration is the sample time has to be equal to the period of the lowest frequency to be measured. For example, to measure down to 1 Hz in the FFT, the sample time has to be 1 second. This has a direct bearing on the drain from the batteries, since the A/D and RAM may draw considerable current during the measurement.

The sample time has a direct bearing on the low frequency corner of the preamp. If we want to measure down to 1 Hz, the rolloff has to be well below that in order to pass the desired frequency without attenuation. In this case, the preamp -3dB point has to be around 0.2 Hz for a single pole RC filter. See the PLT file for the preamp for a more accurate measurement.

However, the 1 second sample time needed for the FFT eliminates many of the boogeymen in low frequency measurements, such as the temperature coefficient of the capacitors. They are not going to change temperature much in 1 second.

Most noise plots only go to 10 Hz, which implies a sample time of 100 milliseconds. This means any temperature change of the capacitors has even less time to affect the measurement.

Finally, we have to thank the pioneers in low noise measurements, such a Fred Walls at NIST, Enrico Rubiola of the Institute Besancon, France, Gerhard Hoffman, and many others for paving the way.

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

Rubiola has a nice paper on cross-correlation techniques:

formatting link

[...]
--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson
[...]

The filter is not for rejecting signals. It is to guarantee the desired signal will be passed without attenuation. In order to pass 1 Hz without attenuation, the filter -3dB point has to be around 0.2Hz. See the PLT file in

formatting link

The output of the preamp will be amplified and sampled in a A/D converter, then fed to FFT conversion software.

As I understand it, the FFT sample time has to be equal to the period of the lowest frequency to be measured. For example, to measure down to 1 Hz, the sample time is 1 second. For a plot to stop at 10 Hz, the sample time is 100 milliseconds.

This does not give the capacitors much time to change temperature and affect the measurement, especially when the preamp has to be placed in a metal shield to minimize RFI. The shield also reduces stray air currents that can affect the capacitor temperature.

So there is less perturbation that can affect temperature, and less time to affect the measurement. My feeling is tempco will not be a problem. It does not seem to be an issue in any of the other approaches to measure low level signals.

I am not interested in measuring very low frequencies, especially as this would drastically increase the settling time. I will let others worry about generating noise plots down to 0.1Hz, or hunting bogies down in very low baseband.

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

Am 27.02.21 um 17:46 schrieb Steve Wilson:

LT3042/3045 is better than you'll need.

Just alone for avoiding ground loops.

I used 14 V from Nicad or NiMH cells and a wire resistor. This near-ccs does not produce excess noise as you can prove by replacing the LED by a 60 Ohm resistor. You get 1nV/rtHz with or without the battery.

The best LED I have seen is HLMP-6000. Surprisingly it was not axed by Avago and Broadcom.

<
formatting link
> and pics left/right

The noise _current_ measurements in the Walls article have to be taken with some grain of salt.

LT4042 after your switcher. If you must admit that you need 100 dB suppression, then you have a pile of dirt on the board close to your precious oscillator.

Fine, but only for measurement.

I could compile FFTW, the Fastest FFT in the West on the Beaglebone Black just so. Download, unzip, type make. But I don't see the need for a FFT other than watching noisy interference to a GPS receiver.

The received signal is below noise so there is not much to see in the FFT. Remember, it looks like noise itself. It takes synchronous averaging in an integrate&dump for the data and Costas loops or similar to track the carrier and the data rate.

BTDT. Back in the 386 time span I've built a combined GPS/Glonass hardware with the Plessey 2010/2020 chip set. It featured 1 bit ADCs. You don't need more. That was actually quite easy, with the chip set registers mapped into PC/AT IO space. Plessey is no longer there.

GPS hardware is easy. The problem is the software. I'd estimate

95% of the effort. That starts when GPS and Glonass do not have the same time scale or you have to remove Faraday by utilizing the delay at different carrier frequencies and and and.

My customer had set up an operation in Moscow for writing the software and even got some key Glonass people, but they tried to use their new western connection for importing text markers and paper folders for quick money. The software made no real progress. The mafia then offered "protection" and my customer simply pulled the plug. Two years later Ashtech had it, the head start was gone.

The sample time must be much higher when your lowest FFT bin is 10 Hz.

Get some uBlox modules, they are 30 years ahead.

Get on the time nuts list at febo.com

Two books that you definitely want to have:

- Robert C. Dixon: Spread spectrum systems with commercial applications I have the 3rd ed. from 6/2000, probably there is a newer one

- Jack K. Holmes: Coherent spread spectrum systems Mine is from 1982, hard to get but good.

Both books from J. Wiley

What an ego boost, in one sentence with the Gods...

Cheers, Gerhard

Reply to
Gerhard Hoffmann

So it is mains powered ?

Why bother with switching mode power supplies ? Just use a 50/60 Hz toroid mains transformer and linear regulators. With these the main problem is the mains frequency and its harmonics requiring carefully PCB layout to avoid "ground loops". Using balanced signals and transformer coupling especially at the GPSDO output etc. will help.

In audio engineering microphone (and pick-up) amplifiers working with millivolt level signals and requiring at least 60-100 dB SNR (i.e. noise below 1 uV) have been used for decades, Does the GPSDO really require better than that environment ?

Do you really need 1 Hz frequency resolution up to 1 MHz ?

Alternatively split the analog signals at 1 KHz and feed the high pass filtered 1 kHz-1 MHz part to one ADC sampled at 2 MHz giving 1 kHz FFT frequency resolution and the low pass filtered (1 Hz - 1 kHz) to an other ADC sampled at a few kHz giving 1 Hz FFT resolution.

Alternatively do it all digital, i.e. run ADC at 2 MHz, use 1024 bin FFT. Digitally decimate the data by dropping the sampling rate (low pass filtering and take only N'th sample) and do the FFT at this lower sample rate. Instead of just two bands, you could do the FFT for each decade in the 1 Hz - 1 MHz frequency range, i.e. 6 FFT's total, each with separate bin width (and noise floor).

Reply to
upsidedown

Am 27.02.21 um 22:21 schrieb Gerhard Hoffmann: < nonsense >

correct: The sample _frequency_ must be much higher when your lowest FFT bin is

10 Hz.
Reply to
Gerhard Hoffmann

I'm not so sure. The ripple reduction graph stops at 10 Hz. The noise is spec'd at 10 KHz and increases dramatically at 10 Hz. I think my numbers are as good or better.

I'm interested in the rejection at 1mHz. I'm not using a band gap reference, so I'm hoping the noise will be lower.

[...]

Good info. Thanks.

Good Info. Thanks.

That was along time ago. I like Rubiola's article from 2018:

formatting link

I was talking about measuring power supply noise. Different problem. [...]

How do you find the corner frequency and noise floor?

I'm not talking about decoding the GPS. I'm talking about locking to the 1 Hz output of the receiver. The time jitter can be quite large, but I have a new phase detector that will significantly reduce it. A good ocxo will be controlled by the pd output, and the low noise power supply will minimize jitter. Witness the excessive jitter caused by power supply noise and ripple on the early Wenzel ULN oscillators, besides the cost and slow delivery:

formatting link

[...]

Please explain more. This is very significant. What is the relationship between the sample time and the FFT resolution? I have searched google but found very little information that made sense.

Have you encountered any problems with capacitor tempco?

Yes, among the best. They do full GNSS but they still have excessive time jitter on the 1 Hz output signal. Long ago I found some articles on systems that would deliver cm accuracy on cellphones but haven't heard anything since. This would require exceptional low noise and jitter GPS receivers, but there isn't much room in a cellphone.

I have a bunch of Rubidiums, but they are very noisy. This is to be expected considering the method they use to lock to the rubidium cell.

Been on it for many years, but haven't been following it lately. I have gigabytes of posts downloaded to my hard drive so I can search topics. There is a huge amount of information and it takes a long time to go through.

AbeBooks $15.81, Amazon $212.03

CAD$41.73 on Amazon, Oct. 1 1981

You do good work. You earned it.

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

Yes, although I plan on using regular 60 Hz transformer power supply instead of a switching unit. That is why I'm interested in the ripple rejection at 120 Hz.

I use a wallwart.

I'm not sure that is related. I'm interested in measuring LED noise corner frequency and noise floor.

Gerhard goes to great lengths to achieve 220pV/root(Hz) for his measurements.

[...]

No, but the problem still remains to meet Nyquist. Once you have done that, the hard part is to keep the sample clock noise out of the preamp.

This might create more switching noise right in the middle of the preamp bandwidth.

[...]
--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

Thanks. What about the sample time? Is that equal to the period of the lowest frequency?

--
The best designs occur in the theta state. - sw
Reply to
Steve Wilson

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.