Differential amplifiers ...noise, common mode etc.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've just wound a guitar pickup ...bifilar style....I'm asking on this forum, as readers here are more likely to have experience of similar preamplifier (magnetic type cartridges, balanced amplifiers etc)

For those that might not know, a bifilar wound coil is wound with two wires at the same time, the 'end wire' of one the two wires is soldered to the 'start wire' of the other...you end up with a centre tapped coil.

The pickup has about 800 turns - therefore 400 turns on each leg...the output is quite low, so I'm having to amplify it about 50x to get it up to normal guitar levels.

I'm feeding it into this type of circuit arrangement...

An externally hosted image should be here but it was not working when we last tested it.


The pickup sounds fine but the output is a little noisy (I'm seeing a floor of about 80mv as seen on the scope at the output of that third opamp) ...I'd like to get the noise down as low as possible, but I don't want to pay for crazily expensive opamps!

I'm using an MCP6004 opamp (it's all I had lying around). - it's noise spec @ 28nv/rt(Hz) is poor .

I'm wondering where I can get the most 'win' from a noise performance/bang per buck perspective....what I'm not grasping is how much of the noise is coming from the opamp itself, how much from the circuit (power rails hiss) etc?

Is the VCC/rail noise & hiss induced into the signal by the first two opamps 50x gain stage 'masked' by the third opamp (which only amplifies the differences wrt its its input signals? )
 
Last edited:
Apologies...the circuit isn't clear (I'll try & get it reuploaded later).

A bit more meat - the centre tap of the coil is connected to 1/2 VCC (2.5V in this instance)...therefore out of the first two opamps is a signal sitting on 2.5V.

Is the missing resistor (to the +ve pin to Vref on the third opamp) strictly necessary if the bias is correct coming out of the preceding opamp?

Edit: I guess it is, becuase a quick 'google images' search shows such a resistor present on all instrumentation amplifiers I looked at! (http://freecircuitdiagram.com/wp-content/uploads/2008/10/3_op_amp_instrumentation_amplifier.gif ie R4 on that one)

Ok, I'll put that right tonight, but in the meantime, to my main poiint again...

Is the VCC/rail noise & hiss induced into the signal by the first two opamps' 50x gain stage 'masked out' by the third opamp (becuase it only amplifies the differences wrt its input signals? ). In other words, is the noise created by the (potentially) noisy pair of opamps deemed 'common' & therefore ignored by the thrid opamp?

If so, then surely it doesn't matter about the noise spec of the first two opamps (so long as they're the same). If the thirdopamp doesn't mask the noise created by the preceding 50x gain stages, then I guess using a low noise opamp becoimes really important!
 
Last edited:
the two input op amp's input noise are independent and add vectorially as usual

Vref, power supply noise can be differenced out as common mode - to the degree that your resistors ratio match, and up to the frequency where the op amp loop gain is still large

800 turns is very low for guitar pickups, making it hard to guess what your source impedance looks like

some CMOS op amps are very low noise above 100 Hz - you do have to check supply V - some only handle up to 5 V

http://www.diyaudio.com/forums/solid-state/68793-new-interesting-cmos-op-amps-2.html#post843381
 
th
800 turns is very low for guitar pickups, making it hard to guess what your source impedance looks like

Agreed, but not as low a 'turn count' as this fella's (about 15-20 turns!)...

Pre-Amplifier For Balanced Magnetic Guitar Pickup joebrown.org.uk

At the risk of appearing slow ....just to be clear, any noise generated in the first two (50x gain) opamps by way of opamp generated noise can - to an extent - be eliminated (or highgly reduced) providing close toelerance (1%?) resistors are used?

Is there any win in shelling out for low noise opamps for such a configuration as I'm using? (ie if the opamp noise is 'common' & can be 'commoned out' by the final opamp ...then why bother with low noise opamps?)
 
Last edited:
no, once noise is aded to a signal the only way to remove it is to have an independent measurement of the noise separate from the signal - which you don't have

since you should take a large gain in the front end only the noise of the input op amps should be important - although gain-bandwidth limits on loop gain may mean that gain of ~ 10x is a better compromise for the input op amp gains with higher summing op amp gain to make up the difference

if your source Z at audio frequencies is over a few kOhms then low noise fet or cmos input op amps are probably the best option

some of the low noise cmos op amps are <$2 ea
 
no, once noise is aded to a signal the only way to remove it is to have an independent measurement of the noise separate from the signal - which you don't have


Then I mustn't have grasped what's going on here - just to be clear, I'm not talking about any noise embedded within the signal as presented at the input to the first two opamps - I'm talking about electrical noise induced into the signal within the opamps themselves - isn't such noise 'common' (ie will be the same in both +ve and -ve signals ....& therefore since the noise signature is the same, the third opamp will mask it out?)
 
Last edited:
Noise produced in two different opamps will be uncorrelated, so when you either add or subtract it you get 3dB more noise. If the noise was fully correllated (i.e. from essentially the same source) you would get 6dB more noise from adding and no noise (perfect cancellation) from subtracting. I assume you were hoping for the latter. It doesn't work like that!

Noise is random. Two noise sources can have the same 'signature' (e.g. power level, spectrum) but they are still random so add/subtract to give 3dB extra.
 
Apologies...the circuit isn't clear (I'll try & get it reuploaded later).

A bit more meat - the centre tap of the coil is connected to 1/2 VCC (2.5V in this instance)...therefore out of the first two opamps is a signal sitting on 2.5V.

Is the missing resistor (to the +ve pin to Vref on the third opamp) strictly necessary if the bias is correct coming out of the preceding opamp?

Edit: I guess it is, becuase a quick 'google images' search shows such a resistor present on all instrumentation amplifiers I looked at! (http://freecircuitdiagram.com/wp-content/uploads/2008/10/3_op_amp_instrumentation_amplifier.gif ie R4 on that one)

Ok, I'll put that right tonight, but in the meantime, to my main poiint again...

Is the VCC/rail noise & hiss induced into the signal by the first two opamps' 50x gain stage 'masked out' by the third opamp (becuase it only amplifies the differences wrt its input signals? ). In other words, is the noise created by the (potentially) noisy pair of opamps deemed 'common' & therefore ignored by the thrid opamp?

If so, then surely it doesn't matter about the noise spec of the first two opamps (so long as they're the same). If the thirdopamp doesn't mask the noise created by the preceding 50x gain stages, then I guess using a low noise opamp becoimes really important!


To design a low level preamp correctly one really needs to know the impedance of the 800turn pick up, both DC resistance and the inductance. High inductance will add to high frequency hiss and require a higher impedance input stage for optimum performance.

The third combining op amp can never mask noise from a preceding stage.

Other posts are correct in saying that two equal noise sources, uncorrelated will increase noise by 3dB which is a factor of root of 2.
The reasons for having a bifilar wound coil and a balanced cct design are more important than the slight noise penalty incurred.

You clearly have a rather small signal if a preamp gain of 50 is required and I understand pickups of this ilk have outputs of around the general millivolt level or so. Also I would expect the source resistance to be of the general order of tens or hundreds of ohms.

Designing a cct and choosing an op amp I would highly recommend the good old 5532 or 5534 which have noise of around 3 to 5 nV/rtHz and are very cheap to buy. They also have very low current noise less than 1 pA and with a source impedance of less than a few K ohm that will be fine.

Do not go to the expense of a premium low noise amp such as an AD797 in this application since it has a fairly high current noise and is inappropriate if the pick-up source impedance is over about 1Kohm. Similarly the voltage noise of a fet opamp is too high unless you have a very high signal level and a high source impedance. much higher than is likely with 800 turns.

The best reference work for op amp and transistor, voltage and current noise and matching to source resistance, and also noise figures ( which bit the noise is coming from ) is Horowitz and Hill, The Art of Electronics, pgs 428 onwards. The best textbook I have ever read in over 40 years. Everything is there without overwhelming maths.
 
Last edited:
cheap low noise op amps

I replaced some audibly hissy JRC4558 op amps (1.4 microvolt noise 1000 ohm source impedance) on a high gain magnetic phone cartridge circuit, with ST33078 at $.60 each. I bought them because newark.com (farnell outside US) said they were low noise. Their noise spec is 4.5 nv/sqrtofhz. They don't audibly hiss, now. I have interchanged them in a CS800S amp input (lower gain) with JRC4560, and they sound exactly the same. TI RC4560 (the datasheet I have) is 1.2 microvoltrms with 2000 ohm source impedance. Either might be better than what you've got. The 33078's don't have latching fet inputs like TL07x, and I made a lot of mistakes on my disco mixer with them without blowing any up. So, have fun.
 
Last edited:
I should have mentioned...I'm constrained to a 5V supply (for deciding which opamp to go with)
I had the ST33078's working at +-3.3V for a while until I noticed they were taking more current than the 4558's and decreased the voltage dropping resistors. They sounded fine at 3.3v on mag phono, but on transistor radio input, which hissed (the radio, not the op amps) below 3 VAC out, they didn't have enough headroom. I"ve got them at +-7.8 V now on dual 1n5344 zeners regulating off an 18vdc slot car wall transformer. It sounds so good now with ST33078's on music I put the PAS2 tube preamp (which was legendary even before I put the metal film resistors and plastic caps in), back in the attic.
If you use something as fast as a 4560 or a 33078, make sure you bypass both power supplies at the op amp with disk caps like .01 uf or .1uf, and bypass the feedback resistors with a tiny cap like 20 nf or something. This keeps them from oscillating. The 4558's they replaced couln't go that fast, but the 33078's oscillated at ~1mhz without the fast caps. Made my PA amp fan run really fast.
 
Last edited:
Thanks guys.

If it helps ....the inductance of each leg of the pickup I've wound (ie either side of the centre tap) is 6.9mH & the DC resistance is 86.1 Ohms.

I should have mentioned...I'm constrained to a 5V supply (for deciding which opamp to go with)

Good, the impedance at audio frequencies of that inductance is up to 1K.
Implied is a signal level around a millivolt or two.

The ideal op amp is a standard bipolar one such as 5532 or 5534 with a voltage noise of 5nV/rtHz as I suggested before. Current noise is unlikely to be an issue.

That single 5V supply is a real obstacle, you need some cunning to produce at least +-5V or preferably more.

ALTERNATIVELY. With a single 5V supply a two or three transistor stage is easily designed, quite compact and perform admirably. If you want to go that route I can rustle up a schematic using low noise cheap transistors.
Please post back and I'll see what I can do.
 
Good, the impedance at audio frequencies of that inductance is up to 1K.
Implied is a signal level around a millivolt or two.

The ideal op amp is a standard bipolar one such as 5532 or 5534 with a voltage noise of 5nV/rtHz as I suggested before. Current noise is unlikely to be an issue.

That single 5V supply is a real obstacle, you need some cunning to produce at least +-5V or preferably more.

ALTERNATIVELY. With a single 5V supply a two or three transistor stage is easily designed, quite compact and perform admirably. If you want to go that route I can rustle up a schematic using low noise cheap transistors.
Please post back and I'll see what I can do.

I looked up the datasheets of the 5532 - but it needs +-5V.

I found one contender, a TS974.....

http://www.st.com/stonline/books/pdf/docs/6031.pdf

but I don't see any mention whether its bipolar or not.

a noise figure of just 4 nV/Hz, cheap as can be .....and a 5V single supply is bang in the middle of its supply voltage range.


Thanks for the offer of knocking up a circuit.....but in this instance, I need to stick with opamps (even though discreet will surely perform better!)
 
I looked up the datasheets of the 5532 - but it needs +-5V.

I found one contender, a TS974.....

http://www.st.com/stonline/books/pdf/docs/6031.pdf

but I don't see any mention whether its bipolar or not.

a noise figure of just 4 nV/Hz, cheap as can be .....and a 5V single supply is bang in the middle of its supply voltage range.


Thanks for the offer of knocking up a circuit.....but in this instance, I need to stick with opamps (even though discreet will surely perform better!)

Without research I have no idea what a TS974 is, so good luck. However, there is never a non option, discrete or op-amp. It is all, how much pcb real estate, what cost, how much design, how much verification, how many soldered joints. think of a transistor as a three legged op amp but much much simpler to use, it just needs a couple of extra resistors to do the bis.
 
I looked up the datasheets of the 5532 - but it needs +-5V.

I found one contender, a TS974.....

http://www.st.com/stonline/books/pdf/docs/6031.pdf

but I don't see any mention whether its bipolar or not.

a noise figure of just 4 nV/Hz, cheap as can be .....and a 5V single supply is bang in the middle of its supply voltage range.


Thanks for the offer of knocking up a circuit.....but in this instance, I need to stick with opamps (even though discreet will surely perform better!)


The general type of circuit I am talking about is shown in Fig 4. AN222 National Semiconductors, substituting 2SA1085 and 2SC2547 for the fancy devices shown. A such it will work even better. One can adjust resistor values for supply voltage, but it is a cunning circuit and needs a little care when recalculating.
 
A rail to rail low voltage opamp will always send out more maximum voltage than a "normal" discrete circuit when supplied from the same low voltage. This particularly applies when +-2.5V or 0,+5V is being supplied.

Now try comparing a discrete on 0,+9V to that opamp running on 0,+5V. No contest, the discrete should win out every time.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.