Update on preamp design for piezo pickup

[tempus fugit]

Hey all;

Finally got the pickup in the guitar and had some time to mess around with
preamp designs. A simple voltage follower works very well - clean and quiet,
even with a TL071. However, trying to add any gain at all turns the thing
into frying eggs. There is SO much noise. I also found, though, that any
extra gain isn't really needed; in fact, a gain of more than about 3
distorts the signal when I strum hard. I suppose it would be nice to get a
little more gain (2-3), but not at the expense of the noise that's added. I
also tried the JFET amp posted by J. Legris
(http://www.till.com/articles/GuitarPreamp/) but couldn't get it to work at
all with a J111, which is all I have handy. This was a bit of a
disappointment, because it has fairly small gain (although maybe too much at
10)and should be fairly quiet, as the J111 has a fairly low noise figure.

Just gotta figure out how to make my voltage follower work with a single
supply and I'm set.....


Thanks for the help again guys.

 

[Jamie]

Hey all;

Finally got the pickup in the guitar and had some time to mess around with
preamp designs. A simple voltage follower works very well - clean and quiet,
even with a TL071. However, trying to add any gain at all turns the thing
into frying eggs. There is SO much noise. I also found, though, that any
extra gain isn't really needed; in fact, a gain of more than about 3
distorts the signal when I strum hard. I suppose it would be nice to get a
little more gain (2-3), but not at the expense of the noise that's added. I
also tried the JFET amp posted by J. Legris
(http://www.till.com/articles/GuitarPreamp/) but couldn't get it to work at
all with a J111, which is all I have handy. This was a bit of a
disappointment, because it has fairly small gain (although maybe too much at
10)and should be fairly quiet, as the J111 has a fairly low noise figure.

Just gotta figure out how to make my voltage follower work with a single
supply and I'm set.....


Thanks for the help again guys.

to make it work with a single supply with 0 output, use a Virtual
ground. Get your self a TLE2426CLP or a smt type chip. It's a 3 legged
rail splitter regulator. it will give you a 4.5 volt +/- for your
circuit from a 9V bat.

 

[tempus fugit]

Thanks for the tip Jamie.

Is it possible to get the single supply set up with 2 resistors as a divider
network on a voltage follower? I've tried but haven't had any success yet.
Is there a trick I should no about?

Thanks

 

[Jamie]

Thanks for the tip Jamie.

Is it possible to get the single supply set up with 2 resistors as a divider
network on a voltage follower? I've tried but haven't had any success yet.
Is there a trick I should no about?

Thanks

You can use 2 R's from the power poles of your bat and have the
tie point as your common. The the + & - sides of the bat will
operate the + and - side of the rails to the IC's etc..
all common/grounds tie to the center tie point of these
R's. Now here's the bad part, Poor regulation. you Would need
to select the proper R's that will give you max current that will
ever be requested which will cause drain on your battery of course
when it's not needed. Using Caps from each Rail to the common tie
point will help stabilize the voltage, it still may drift.

If you use the TLE2426 as i suggested, this unit is regulated.
What this means is, at idle, the current will be very low and when
the demand is there, the internal regulator will compensate for your
load and keep it stable.
Now, you don't need to use this component! You can use an single
channel Op-amp where you take the output and use that as your rail
splitter as the common/ground. but that will require some extra
components to bias the + input etc.
The TLE2426 is basically just an OP-Amp circuit with all the external
components already incorporated for the job and is most likely more
precision.

 

[[email protected]]

Finally got the pickup in the guitar and had some time to mess around with
preamp designs. A simple voltage follower works very well - clean and quiet,
even with a TL071. However, trying to add any gain at all turns the thing
into frying eggs. There is SO much noise. I also found, though, that any
extra gain isn't really needed; in fact, a gain of more than about 3
distorts the signal when I strum hard. I suppose it would be nice to get a
little more gain (2-3), but not at the expense of the noise that's added.

We'd need to see your actual circuit to comment properly. There are
lots of ways to add noise besides the opamp. What's more, as I'm sure
you know, the TL071, with en = 18nV at 1kHz, is _not_ among the quiet
JFET opamps. For example, the AD743 has 3.2nV/rt-Hz of noise at
100Hz, whereas the TL071 has 25nV of noise. Hey, that's a lot!

I also tried the JFET amp posted by J. Legris
(http://www.till.com/articles/GuitarPreamp/) but couldn't get it to work at all
with a J111, which is all I have handy. This was a bit of a disappointment,
because it has fairly small gain (although maybe too much at 10) and
should be fairly quiet, as the J111 has a fairly low noise figure.

Actually, J. Legris' J201 choice isn't that quiet either (i.e. 8nV at
500Hz at 100uA), and your J111 isn't much better at 9nV/rt-Hz (500Hz
and 100uA). But then, we don't have your circuit, to know if you're
making mistakes to overshadow these low-noise specs for these parts,
like too much cable capacitance creating en-w-Cin noise, or if perhaps
you're employing some enhancements to improve on the specs (e.g.,
operating at higher currents, keeping critical resistor values under
220 ohms,etc.).

BTW, the really low-noise discrete JFETs get down to well under 1nV.

As we point out in the extended low-noise discussion in AoE, you'll
want to select a JFET with larger area = higher input capacitance.

 

[tempus fugit]

We'd need to see your actual circuit to comment properly. There are

lots of ways to add noise besides the opamp. What's more, as I'm sure
you know, the TL071, with en = 18nV at 1kHz, is _not_ among the quiet
JFET opamps. For example, the AD743 has 3.2nV/rt-Hz of noise at
100Hz, whereas the TL071 has 25nV of noise. Hey, that's a lot!

I could post a schem of the circuits I tried, either here or on ABSE. I had
the TL071 configured as an inverting amp with an Ri of 3.3M and an Rf of
10M. Lowering Ri didn't really affect the noise if at all.

Actually, J. Legris' J201 choice isn't that quiet either (i.e. 8nV at
500Hz at 100uA), and your J111 isn't much better at 9nV/rt-Hz (500Hz
and 100uA). But then, we don't have your circuit, to know if you're
making mistakes to overshadow these low-noise specs for these parts,
like too much cable capacitance creating en-w-Cin noise, or if perhaps
you're employing some enhancements to improve on the specs (e.g.,
operating at higher currents, keeping critical resistor values under
220 ohms,etc.).

BTW, the really low-noise discrete JFETs get down to well under 1nV.

I've seen some of these, but I'd have to order some to try them out. I'm
hoping to be able to get something working with what I've got around here.
Unfortunately, I couldnt get the Tillman pre to work at all (maybe
something's wrong with my J111? - it works fine when set up as a simple
switch), so I can't tell if the noise level will be acceptable.

As we point out in the extended low-noise discussion in AoE, you'll
want to select a JFET with larger area = higher input capacitance.

Sorry, AoE?

Thanks

 

[tempus fugit]

You can use 2 R's from the power poles of your bat and have the
tie point as your common. The the + & - sides of the bat will
operate the + and - side of the rails to the IC's etc..
all common/grounds tie to the center tie point of these
R's. Now here's the bad part, Poor regulation. you Would need
to select the proper R's that will give you max current that will
ever be requested which will cause drain on your battery of course
when it's not needed. Using Caps from each Rail to the common tie
point will help stabilize the voltage, it still may drift.

Thanks Jamie. I tried this setup, but I can't get the thing to work. There
is a tiny amount of signal getting through, but not as much as there should
be. Is it possible to set up a voltage follower with single supply this way?
Oh BTW, battery life is not an issue, as this will be powered by a power
supply.

If you use the TLE2426 as i suggested, this unit is regulated.
What this means is, at idle, the current will be very low and when
the demand is there, the internal regulator will compensate for your
load and keep it stable.
Now, you don't need to use this component! You can use an single
channel Op-amp where you take the output and use that as your rail
splitter as the common/ground. but that will require some extra
components to bias the + input etc.
The TLE2426 is basically just an OP-Amp circuit with all the external
components already incorporated for the job and is most likely more
precision.

I actually have a ICL7660 floating around here somewhere that I could use to
give me a dual (or rail split, for that matter) supply, but I was hoping to
get away with a resistor divider setup if at all possible.

 

[YD]

Late at night, by candle light, "tempus fugit"

Thanks Jamie. I tried this setup, but I can't get the thing to work. There
is a tiny amount of signal getting through, but not as much as there should
be. Is it possible to set up a voltage follower with single supply this way?
Oh BTW, battery life is not an issue, as this will be powered by a power
supply.



I actually have a ICL7660 floating around here somewhere that I could use to
give me a dual (or rail split, for that matter) supply, but I was hoping to
get away with a resistor divider setup if at all possible.

You need to decouple the divider. Any old 1 to 10 uF cap between
divider output and negative rail will do. TI has an appnote on using
opamps in single-supply.

- YD.

 

[tempus fugit]

Late at night, by candle light, "tempus fugit"


You need to decouple the divider. Any old 1 to 10 uF cap between
divider output and negative rail will do. TI has an appnote on using
opamps in single-supply.

Thanks YD.

I put a decoupling cap where you suggested, but still nothing. There is very
little signal at the output.

 

[[email protected]]

tempus fugit wrote:
Winfield wrote...

I could post a schem of the circuits I tried, either here or on ABSE. I had
the TL071 configured as an inverting amp with an Ri of 3.3M and an Rf
of 10M. Lowering Ri didn't really affect the noise if at all.

Whew, the Johnson noise density of a 10M resistor is 400nV/rt-Hz,
no wonder it was excessively noisy! Try a TL071 in a non-inverting
configuration, with say two 4.7k resistors for a gain of about two.
This should have about 230/18 = 13 times less noise. Be sure you've
got a resistor to ground (or a bias voltage) on the opamp's + input.

If that's too much noise you can change to a more quiet opamp, or
revisit the discrete JFET preamp circuit, paying attention to
Vgs vs Id and using an appropriate source-resistor value to bias
the preamp's output DC voltage to 45 to 60% of the 9V battery.
If necessary, you can add a parallel electrolytic plus series
resistor, to increase the gain. Like this (view with a fixed font):

| Guitar-pickup preamp with G = +3dB (x 1.4)
|
| .------------------------+--- +9v
| | |
| 6.8k _|_+
| | 2.2uF --- 10uF
| Vd +----||---+--- out | 25V
| | + | gnd
| |--' 100k
| in ---+---->| |
| | |--, gnd
| 10M | Cx Use the lowest-noise JFET you
| | +--||--, have and choose Rs for Vd = 5V
| gnd | + | separately for each JFET you try.
| Rs Rx If necessary for more gain add
| | | Cx = 2.2uF and choose Rx for a
| gnd gnd modest gain of about 1.5, etc.

 

[Ecnerwal]

Sorry, AoE?

The Art of Electronics. A very, very good book, which Winfield happens
to be the co-author of. Thanks, Win.

 

[tempus fugit]

T

tempus fugit wrote:
Winfield wrote...

Whew, the Johnson noise density of a 10M resistor is 400nV/rt-Hz,
no wonder it was excessively noisy! Try a TL071 in a non-inverting
configuration, with say two 4.7k resistors for a gain of about two.
This should have about 230/18 = 13 times less noise. Be sure you've
got a resistor to ground (or a bias voltage) on the opamp's + input.

If that's too much noise you can change to a more quiet opamp, or
revisit the discrete JFET preamp circuit, paying attention to
Vgs vs Id and using an appropriate source-resistor value to bias
the preamp's output DC voltage to 45 to 60% of the 9V battery.
If necessary, you can add a parallel electrolytic plus series
resistor, to increase the gain. Like this (view with a fixed font):

| Guitar-pickup preamp with G = +3dB (x 1.4)
|
| .------------------------+--- +9v
| | |
| 6.8k _|_+
| | 2.2uF --- 10uF
| Vd +----||---+--- out | 25V
| | + | gnd
| |--' 100k
| in ---+---->| |
| | |--, gnd
| 10M | Cx Use the lowest-noise JFET you
| | +--||--, have and choose Rs for Vd = 5V
| gnd | + | separately for each JFET you try.
| Rs Rx If necessary for more gain add
| | | Cx = 2.2uF and choose Rx for a
| gnd gnd modest gain of about 1.5, etc.

Thanks for the reply. I can't use 4.7K resistors, as the output Z of my
pickup is 1.3Mohms. I will try that JFET idea though.

If I follow the parameters youve listed here, should the J111 preamp work no
matter what? I'm starting to wonder if there's actually something wrong with
my JFET.

Thanks again

 

[Winfield Hill]

Winfield wrote

Thanks for the reply. I can't use 4.7K resistors, as the output Z of my
pickup is 1.3Mohms. I will try that JFET idea though.

No, no, I said "non-inverting" configuration. Look it up, that means
the signal goes to the opamp's + input, with Z = infinity. The 4.7k
resistors are on the other side, isolated from the transducer signal.

If I follow the parameters youve listed here, should the J111 preamp work no
matter what? I'm starting to wonder if there's actually something wrong with
my JFET.

Not at all, read about the huge Vgs variation from part-to-part, and
from type to type, and from manufacturer to manufacturer of the same
type, and from run-to-run of a given manufacturer. This means,
contrary to Donald Tillman's assertion, you often have to select
the self-bias resistor for each individual JFET, as I detailed above.
Certainly you'll get the best results (more output capability, lower
overload distortion, etc.) that way.

 

[tempus fugit]

No, no, I said "non-inverting" configuration. Look it up, that means
the signal goes to the opamp's + input, with Z = infinity. The 4.7k
resistors are on the other side, isolated from the transducer signal.


Not at all, read about the huge Vgs variation from part-to-part, and
from type to type, and from manufacturer to manufacturer of the same
type, and from run-to-run of a given manufacturer. This means,
contrary to Donald Tillman's assertion, you often have to select
the self-bias resistor for each individual JFET, as I detailed above.
Certainly you'll get the best results (more output capability, lower
overload distortion, etc.) that way.

Thanks again for your reply, Win.

Sorry, I misread your original reply - I see what you mean now using the 2
4.7k resistors in a noninverting configuration. I'll definitely give that a
try. I managed to get the JFET working by putting a cap in parallel Rs. The
gain is much increased (maybe a little too much), so I'll have to mess with
that a bit to see if I can get that working better. I was surprised by the
difference in gain with the cap. I haven't really worked much with JFETs -
mostly BJTs, and I've seen where that cap can increase the gain, but holy
cow! The output is actually attenuated without the cap there - it's barely
audible, but comes through loud and clear with the cap in. I'm going to try
some different value resistors for Rd and Rs and see if i can reach a nice
compromise, and put it on the scope as well. The output is pretty noise
free, certainly enough for my purposes. An opamp seems kind of like overkill
when a JFET can do basically the same job, particularly when I need so
little gain.

Oh 1 quick question - the voltage gain is gm*Rd for a JFET - is there an
equation for the gain with the cap? Will the size of the cap make a
difference? Guess that's 2 questions, and here's one more: You said to pick
Rs to make Vd 5v or thereabouts - does this mean to keep Rd at 6.8K and swap
out different resistors for Rs? I couldn't get the circuit to work with
6.8K, so I dropped it to 2.2k. I could try some other values in there too;
2.2K was the 1st thing I tried and it worked so I didn't try any others.

Thanks again.

 

[Winfield Hill]

... I managed to get the JFET working by putting a cap in parallel Rs. The
gain is much increased (maybe a little too much), so I'll have to mess with
that a bit to see if I can get that working better. I was surprised by the
difference in gain with the cap. I haven't really worked much with JFETs -
mostly BJTs, and I've seen where that cap can increase the gain, but holy
cow! The output is actually attenuated without the cap there - it's barely
audible, but comes through loud and clear with the cap in. I'm going to try
some different value resistors for Rd and Rs and see if i can reach a nice
compromise, and put it on the scope as well. The output is pretty noise
free, certainly enough for my purposes. An opamp seems kind of like overkill
when a JFET can do basically the same job, particularly when I need so
little gain.

Oh 1 quick question - the voltage gain is gm*Rd for a JFET - is there an
equation for the gain with the cap? Will the size of the cap make a
difference?

No, it can be expressed as G = Rd/rs, where rs = Rs + 1/gm, so
G = Rd * gm only if Rs = 0, which is what your Rs bypass cap
accomplished. Study my drawing -- that's why I suggested a
resistor Rx in series with the large cap Cx.

Guess that's 2 questions, and here's one more: You said to pick
Rs to make Vd 5v or thereabouts - does this mean to keep Rd at
6.8K and swap out different resistors for Rs?
Yes.

I couldn't get the circuit to work with 6.8K, so I dropped it to 2.2k.
I could try some other values in there too; 2.2K was the 1st thing
I tried and it worked so I didn't try any others.

That's a less desirable approach because lowing Rd means much
higher drain current, instead adjusting the current to the desired
level with the self-bias resistor Rs. You were dealing with Id too
high, which meant too much drop across Rd, saturated the JFET
and destroying the gain.

Alternately, when you fail to optimize the variation of Id due to Vgs
uncertainty, the current might be too low, for a low drop across Rd
and a greatly-reduced signal-output capability.

BTW, I wonder how much gain you really want. Tillman uses only
3dB = x 1.4, and the JFET is really an impedance converter, to
drive the cable, and a soggy distortion limiter for loud signals,
which I understand is sometimes an issue with guitar amplifiers.

 

[tempus fugit]

No, it can be expressed as G = Rd/rs, where rs = Rs + 1/gm, so
G = Rd * gm only if Rs = 0, which is what your Rs bypass cap
accomplished. Study my drawing -- that's why I suggested a
resistor Rx in series with the large cap Cx.


That's a less desirable approach because lowing Rd means much
higher drain current, instead adjusting the current to the desired
level with the self-bias resistor Rs. You were dealing with Id too
high, which meant too much drop across Rd, saturated the JFET
and destroying the gain.

Alternately, when you fail to optimize the variation of Id due to Vgs
uncertainty, the current might be too low, for a low drop across Rd
and a greatly-reduced signal-output capability.

BTW, I wonder how much gain you really want. Tillman uses only
3dB = x 1.4, and the JFET is really an impedance converter, to
drive the cable, and a soggy distortion limiter for loud signals,
which I understand is sometimes an issue with guitar amplifiers.

Hi Win - thanks for the reply again.

I can't get anything out of the 6.8K resistor unless I go up to 22k at Rs,
and even then it is barely audible. Also, this puts Vd at over 6v, so I'm
running out of headroom. Is the Rx Cx trick to lower or increase the gain?

You are right about the gain by the way. I realize I only need a little
gain, and mostly I need a buffer to lower the output Z. I tried the
noninverting opamp you suggested with 5.6K and 3.9K and it sounds great. In
fact, the gain on this may even be a bit too high. I may end up just going
that route. I tried this in a single supply and the sound was much thinner
though, so I'll need to mess with it a bit. Any suggestions there? I used 2
22k resistors to form a voltage divider, and connected Vb/2 to the
noninverting input.
Thanks again.

 

[Phil Allison]

"tempus fugit"

I tried this in a single supply and the sound was much thinner
though, so I'll need to mess with it a bit. Any suggestions there? I used
2
22k resistors to form a voltage divider, and connected Vb/2 to the
noninverting input.

** WRONG !!

You connect your 3.3M input load resistor between Vb/2 and the + input.

The piezo PU connects from + input to ground, as usual.




.......... Phil

 

[tempus fugit]

"tempus fugit"



** WRONG !!

You connect your 3.3M input load resistor between Vb/2 and the + input.

Thanks Phil.

That fixed it up and it sounds just fine now. Odd that that wasn't mentioned
in the textbook that I got the circuit from....