D. I. Y. Watch Timing Machine.

[gibsonk]

Doh, thank you; I will order up a new one.

Sent from my iPhone using Tapatalk

[Alastair]

Out of curiosity--Why use TL074...?

TL074 is a Quad Op-Amp, and looking at that strip-board Only one of the four is being used.

Be Easier/less confusing and prob less chance of the three Unused amps causing issues with oscillation or latch-ups to use TL071, a Single Op-Amp. There are Many dozens of pin for pin equivalents to this one too, Even the ever popular 741 and its herds of friends!...

 

[svorkoetter]

1 hour ago, Alastair said:

Out of curiosity--Why use TL074...?

TL074 is a Quad Op-Amp, and looking at that strip-board Only one of the four is being used.

 

The circuit uses all four. I'm assuming that gibsonk is testing as he goes along, which I think he has just shown to be a good idea.

[gibsonk]

54 minutes ago, svorkoetter said:

The circuit uses all four. I'm assuming that gibsonk is testing as he goes along, which I think he has just shown to be a good idea.

Not maybe as much as i should have done, finished the circuit but must have a schoolboy somewhere, becasue i got a signal but not from the mic seemed like electrical hum and the software couldn't detect the tick. I swapped out the 0.1 capacitor from the input mic and now it seems to be doing even less. Back to the drawing board later next week when i am back from work. ( i will test more this time) 

 

Its late here so i have parked the amp for tonight, out of curiosity though i tested the mic using a FIIO E17 DAC and Headphone amp, i got better results than i expected. 

 

 

 

 

 

[Alastair]

Ah--OK, Tracked down the schematic now. An amp with bandwidth of 150Hz to 11KHz or therabouts....

A few observations-

I see that you're using one section as the voltage-reference for the Inverting inputs. In my design days, this value was usually chosen as exactly half +B rather than (as in your scheme) offset  to the lower value of 3.6V

--Personally I would have used 4.5V (Half the supply of 9V) and equal value resistors for the voltage-reference divider-pair, R8 and R9 to achieve this.

Any reason why this voltage Isnt half the +B supply?

Offsetting this value as in the scheme will limit the maximum output swing the amp will give before distortion occurs,--You're limiting headroom, which may or may not ever be an issue.--I'll have to build one and find out!

Another thing worth mentioning which can negatively affect noise levels is that this ref voltage supplied by 'Z1D' Should be decoupled close to each gain stage. A 10-20K and 0.47uF would be good on each one-- mounted as close to the Inverting input node as possible.

Good practice for Low Noise would be the addition of 2-3K resistors on each of the inputs of the gain stages, both Inverting and Non-Inverting. Stoppers if you will....

Personally, I would have used Two supplies, Plus and Minus B, this simplifies the amplifier gain stages and reduces noise issues--but does mean you need Two batteries or switching bi-polar PSU if you have to make it single-batt powered.... No free lunch in electronics!

Worth keeping in mind, The Input of your average laptop mic or soundcard is a fairly low impedance--Usually around 4Kish as often they are designed for Electret Mics and are designed as a voltage source for the MOSFET in the electret. I had LOTS of fun years ago with these designing noise-cancelling mics for Voice Recognition software!

 

Edited July 24, 2016 by Alastair

[Alastair]

Thinking about the soundcard I/P issue, an additional Unity-Gain Buffer stage would be a good idea--After the set-output level control-pot, and decoupled by summit like 47uF (This is what I usually used for driving any cables into any computer soundcards.)

Current scheme as is, will alter the output bandpass filters response according to control-pot position due to impedance of soundcard input....

[svorkoetter]

3 hours ago, Alastair said:

Any reason why this voltage Isnt half the +B supply?

 

Yes. The TL074 and similar op-amps can only swing to within 1.5V of the positive rail. So with a 9V supply, that's 7.5V, and 3.6V is close to the centre of that range.

3 hours ago, Alastair said:

Another thing worth mentioning which can negatively affect noise levels is that this ref voltage supplied by 'Z1D' Should be decoupled close to each gain stage. A 10-20K and 0.47uF would be good on each one-- mounted as close to the Inverting input node as possible.

Good practice for Low Noise would be the addition of 2-3K resistors on each of the inputs of the gain stages, both Inverting and Non-Inverting. Stoppers if you will....

I will experiment with that. It's probably especially important in the early stages, as any noise introduced there is amplified by the later stages.

3 hours ago, Alastair said:

Thinking about the soundcard I/P issue, an additional Unity-Gain Buffer stage would be a good idea--After the set-output level control-pot, and decoupled by summit like 47uF (This is what I usually used for driving any cables into any computer soundcards.)

Current scheme as is, will alter the output bandpass filters response according to control-pot position due to impedance of soundcard input....

 

In my experience, the Line In to a sound card usually has around 47k impedance, which will not significantly affect the output high pass (not band pass) filter. As far as decoupling goes, C7 of the output filter already has that effect.

[matabog]

hello Ștefan!

I wanted to use wos to regulate a pendulum clock, but I didn't find 7200bph. I used IT, though, with great results, in waveform view, set to 1/2s. 

[robmack]

Hi Stefan,

I'm currently working on refining the Watch-o-scope amplifier design, taking onto consideration the discussion that has taken place in this thread.  I've been contemplating swapping the 9V battery power source for two CR123A lithium batteries.  These batteries have two forms -  a lithium non-rechargable with 3V and capacity of 1500mAH and a lithium ion rechargable with 3.6V and a capacity of 700mAH.  A pair has the same physical dimensions as a 9V battery with equal or greater capacity.  This change would also eliminate the need for a virtual ground circuit since they would provide up to +/- 3.6V to the TL074.

On 2016-07-24 at 9:10 AM, svorkoetter said:

Yes. The TL074 and similar op-amps can only swing to within 1.5V of the positive rail. So with a 9V supply, that's 7.5V, and 3.6V is close to the centre of that range.

I quoted above because you mentioned that the op amp would only swing within 1.5V of the positive rail.  I'm assuming this is with a single ended supply.  Would this same limitation apply for a dual-ended supply?  If so, that means the available signal using CR123A battery pair would be 7.2V-1.5 = 5.7V.  Although the TL074 is speced for minimum 6V Vdd, with the loss at the output, would this still be enough usable signal (+/- 2.9V) to be picked up by the PC sound card?

- Robert

[wlysenko]

I just complete the Watch-O-Scope microphone and amplifier. I followed the design except that I put the 10k pot on the front panel of the aluminum box and used  a small solder-less protoboard instead of a PCB. It seems to be working but I am having problems getting a reliable signal for the software to do its thing. This may have something to do with noise in the amplifier.

With the gain all the way up and plugging in some headphones, I hear some hiss but no hum. Here is the weird thing: With the input capacitor C1 disconnected, I see noise that is about 3% of full scale on the software's scope mode. But when I connect C1 to the microphone (the normal position) or to ground, the noise increases to about 8%. This doesn't seem normal and I suspect I should be doing something differently. When I put a watch on the microphone, peaks are about 80% of full scale with the valleys (between ticks) going down to about 10% (just above the noise level of the amplifier).

Does anyone have any idea of what I should try to improve this?  I tried putting the grounds to the virtual ground point instead of the battery negative terminal (just move one wire in my setup) but this made no difference.

 

[wlysenko]

Here is a photo of my project. As I said, the problem is that if I short the input to ground, the noise level at the output goes up. I don't think it should be doing that.

[svorkoetter]

On August 31, 2016 at 8:40 AM, matabog said:

hello Ștefan!

I wanted to use wos to regulate a pendulum clock, but I didn't find 7200bph. I used IT, though, with great results, in waveform view, set to 1/2s. 

7200bph isn't one of the listed rates, but you can type any rate you want (between 3600 and 43200 bph) into the rate selection box.

On August 31, 2016 at 3:55 PM, robmack said:

Hi Stefan,

I'm currently working on refining the Watch-o-scope amplifier design, taking onto consideration the discussion that has taken place in this thread.  I've been contemplating swapping the 9V battery power source for two CR123A lithium batteries.  These batteries have two forms -  a lithium non-rechargable with 3V and capacity of 1500mAH and a lithium ion rechargable with 3.6V and a capacity of 700mAH.  A pair has the same physical dimensions as a 9V battery with equal or greater capacity.  This change would also eliminate the need for a virtual ground circuit since they would provide up to +/- 3.6V to the TL074.

I quoted above because you mentioned that the op amp would only swing within 1.5V of the positive rail.  I'm assuming this is with a single ended supply.  Would this same limitation apply for a dual-ended supply?  If so, that means the available signal using CR123A battery pair would be 7.2V-1.5 = 5.7V.  Although the TL074 is speced for minimum 6V Vdd, with the loss at the output, would this still be enough usable signal (+/- 2.9V) to be picked up by the PC sound card?

- Robert

It seems that the op-amp will also only swing within 1.5V of the negative rail as well, so with +/-3.6V, you will have an available signal range of +/-2.1V. This should still be adequate, since the final output is generally no more than +/-1V. However, for a loud watch, you may need a way to attenuate the input, or at least before the final amplifier stage.

I'm not sure why I originally thought that it was only the positive rail that had this limitation. Given that, I should probably slightly modify my design to have virtual ground at the half-way point.

22 hours ago, wlysenko said:

I just complete the Watch-O-Scope microphone and amplifier. I followed the design except that I put the 10k pot on the front panel of the aluminum box and used  a small solder-less protoboard instead of a PCB. It seems to be working but I am having problems getting a reliable signal for the software to do its thing. This may have something to do with noise in the amplifier.

With the gain all the way up and plugging in some headphones, I hear some hiss but no hum. Here is the weird thing: With the input capacitor C1 disconnected, I see noise that is about 3% of full scale on the software's scope mode. But when I connect C1 to the microphone (the normal position) or to ground, the noise increases to about 8%. This doesn't seem normal and I suspect I should be doing something differently. When I put a watch on the microphone, peaks are about 80% of full scale with the valleys (between ticks) going down to about 10% (just above the noise level of the amplifier).

Does anyone have any idea of what I should try to improve this?  I tried putting the grounds to the virtual ground point instead of the battery negative terminal (just move one wire in my setup) but this made no difference.

 

The noise levels you describe sound like they're within the range of normal. The reason you're probably not getting much noise when C1 is disconnected is that there is then no current flowing through R1. All resistors produce noise, and the noise in R1 is amplified the most.

Since you've built the circuit on a breadboard and it's easy to experiment, you might also want to try inserting 10k resistors between virtual ground and each of pins 5, 10, and 12 of the IC, and 0.1uF capacitors between those pins and actual ground. I've been thinking of revising the design to include those as it might reduce noise (as suggested by Alastair in an earlier post).

[robmack]

Thanks Stefan.  I am currently working on a pcb for the amp and including several suggested changes.

* series resistors at each of the non inverting inputs to the filter stages

* decoupling caps on the virtual ground

* use LVM324 rail to rail opamp instead of TL074

* smd components and ground pours for noise reduction

* self contained in a Hammond 1590 case.  Integrated battery holder and power switch

* experimenting with piezo film bonded to the case

* reduce number of 3.5mm Jacks to reduce potential noise sources

I finished the layout this evening and will look at getting pcb prototypes shortly.  I'll post updates to this thread as I progress.

[svorkoetter]

9 hours ago, robmack said:

* decoupling caps on the virtual ground

Just to clarify, those extra caps should go from the op-amp input pins (the ones that are now connected to virtual ground by the new resistors) and actual ground, not from virtual ground to actual ground.

[wlysenko]

Thanks svorkoetter for the comments and suggestions. I put 10k resistors between pins 5, 10, and 12 and virtual ground (instead of the wire jumpers I had previously). Also added 0.1uF capacitors between these pins and actual ground. I saw no difference in behavior (same noise levels).

Since you mentioned noise in R1, I also tried changing R1. With the input grounded, increasing R1 from 33k to 133k reduced the noise level by a factor of about two. Pulling C1 out reduced the noise another factor of two. (Actual numbers are 7%, 4%, and 2% of full scale on the Watch-O-Scope scope.) Does this make sense?

I'll continue to look at this to make sure I did not make a mistake in wiring. Maybe none of this matters but I would like to understand it. Thanks again.

[Alastair]

Is there any way to add 300Hz and 360Hz as timing options so the tuning-fork types can be timed?

[robmack]

On 31/08/2016 at 11:10 PM, wlysenko said:

Does anyone have any idea of what I should try to improve this?  I tried putting the grounds to the virtual ground point instead of the battery negative terminal (just move one wire in my setup) but this made no difference.

Wlysenko,

It's hard to tell from your photo but it looks like you have different types of capacitors in use in your circuit.  There are ceramics,  electrolytic, and possibly tantalum.  If you are using tantalum caps as bypass caps, then they won't do a good job because of their high ESR.  You should probably use ceramics wherever possible.   These types have lower ESR and ESL ratings and might contribute to better noise reduction.  This Intersil app note is helpful for understanding bypass capacitor types ( http://www.intersil.com/content/dam/Intersil/documents/an13/an1325.pdf ).

Also, if you're thinking of keeping this unit intact (i.e. it's not a test bench model), then try to place the bypass caps as close as possible to the pins they are bypassing, and clip the leads as short as possible.  The idea is to reduce the inductance of the circuit as much as possible and long leads go against that objective.

Don't put the virtual ground at the same potential as the circuit's ground.  Virtual ground has to float at +3.6V above the circuit's ground to offer the op amp dual-ended power  source ( + and - ) from a single-ended battery.

By removing/replacing C1, you are disconnecting/connecting the piezo element from the circuit.  If you are getting the same response irrespective of whether C1 is in or out of place, then maybe the IC is not functioning properly.  Can you pull it and test to see that all four elements are functioning?  Also, have you tested the voltage at the output of IC1A (the virtual ground)?  It should be 3.6V DC relative to the negative terminal of the battery.  Test the voltages at the non-inverting inputs to the three filtering stages as well to see that they have the same voltage. 

[wlysenko]

Robmack,

All the bypass capacitors are ceramic.

I understand the business about short leads and keeping the caps near the pins. I will do that when I build the final soldered-up version. But isn't that a concern mainly for high-frequency circuits and not for a kHz device such as this? (I'm not sure about this.)

When I said I temporarily moved the grounds to the virtual ground, I moved the shielded conductors of the two 3.5mm jacks, the connection to the aluminum case, and the CCW end of the pot to the virtual ground point. Nothing else. I did not upset the dual-ended power situation and the virtual ground (before and after this temporary modification) was at 3.6V. Anyway, this did not change the behavior of the circuit.

About C1, let me reiterate: With C1 connected to the piezo element as normal or with the input shorted, I get the same noise level. If I pull C1 out, the noise level goes down. This did not seem reasonable to me (but maybe it is). The amplifier is likely okay. I can hear the watch ticks and they sound normal (I think--never heard them before). My current thinking on the noise problem (if it is one) is that it is normal op-amp noise that can't be avoided or maybe some ground-loop problem (the input and output signal power levels are very different and any interaction could be detrimental).

Thanks for helping me troubleshoot.

 

[wlysenko]

Svorkoetter, you may be right about resistor noise. I worked out the expected noise caused by the resistors in the first stage, assuming an ideal op amp, using the formulas in http://www.ti.com/lit/an/slva043b/slva043b.pdf. I assumed the input is shorted and neglected the effects of C1 and C2 as the reactances are small and large, respectively. Assuming a bandwidth of 5kHz, the rms noise voltage at the input is 1.4uV. The noise gain is (R1+R2)/R1=4, so the noise at the output of stage 1 is 5.8uV.

Changing R1 to 133k yields voltages of 2.2uV and 3.8uV, at the input and output, respectively. The modified/original ratio at the output is 0.7. I measured 0.6 approximately, so the measurement makes sense, contrary to what I first thought.

If anyone is interested, I can post a scan of my handwritten calculation or even do a LaTeX formatted version.

About the bypassing capacitors: Isn't that really required only in digital circuits where we have large binary transitions with high frequency components? This is a purely linear circuit.

[JohnR725]

5 hours ago, wlysenko said:

About the bypassing capacitors: Isn't that really required only in digital circuits where we have large binary transitions with high frequency components? This is a purely linear circuit.

Usually to figure out whether you need a bypass capacitor it helps to look at the tech sheet. For instance For the tl074 found at the link below Sections 10 & 11 Specify type of bypass capacitor and where it should be located.  Then the tech sheets are also handy for circuit board layout for low noise designs.

http://www.ti.com/lit/ds/symlink/tl074.pdf

 

 

 

[wlysenko]

1. I have written up my calculations on resistor noise in the preamp. See the noise article at  wlysenko.blogspot.com.

2. Thanks JohnR725 for the link to the real data sheet for the TL074. I had been looking at an abbreviated version that didn't have all that information.

[Alastair]

For a better option dump the TL074 and rebuild with the NE5534 op-amps (But these are singles, so you'll need to use a couple or three). Has 4nV/Hz as opposed to 18nV Input Noise figures of the TLO series....

TLO series were Good 20 years ago, but even the (Now pretty old) 5534 can better it for this application

But--IMHO, The noise thats causing you issues is probably little to do with the Op-Amp--Or the resistors-- (You're using metal-film right?) but Much more to do with Layout, Screening and Decoupling (Or lack of, or mis-use of)

[rodabod]

10 hours ago, Alastair said:

But--IMHO, The noise thats causing you issues is probably little to do with the Op-Amp--Or the resistors-- (You're using metal-film right?) but Much more to do with Layout, Screening and Decoupling (Or lack of, or mis-use of)

I was thinking the same..... It sounded more like it could be oscillation. I can't see the latest schematic, but it's worth strapping some caps across the +ve and -ve rails liocally to the opamp (on the solder side) and also limiting the bandwidth to whatever you need with some negative feedback in the form of another cap.

 

Regarding the opamps, I agree that the NE5534 is a good choice. It could possibly be bettered by a an OPA604 depending on how high the source impedance of the piezo element is, since the OPA604 has good noise current specs.

Edited September 5, 2016 by rodabod

[Alastair]

A good trick for anti oscillation/noise techniques for op-amps an old designer told me some time ago was to use 2.2K resistors on each of the device's input pins (Just like Grid-Stoppers on Valves), connecting them in series with the node that normally goes to the pins, and the pins themselves .

Certainly decoupling close to the chip bodies with supply decoupling resistors of say, 1K wouldn't hurt, 1-10uF caps would be good..

The reference voltage to the inverting inputs should be decoupled too....

Op-Amps are 'Stupid' Devices,--Meaning They don't know--Or Care-- What the signal is--IF they See an Input Difference between inputs--They Amplify it, correct signal or not, in that--they are faithful!.

Keep all leads as short as possible and route the Input leads well away from supply and output leads. Use Star Grounding techniques, keep all impedances as reasonable as possible avoiding high impedance sources or f/b loops...

[wlysenko]

I measured the gain of the preamp as a function of frequency. The voltage levels were such that, at the max gain frequency, the output rms voltage was 1.08 V. At this point the noise (output voltage with signal generator turned off) was 0.015 V. This value was subtracted from all points plotted. The circuit is as originally designed with the addition of the bypass modifications suggested by svorkoetter but no other changes.

preampgain.pdf