Maker Pro
Maker Pro

Low noise guitar pre-amp / ADC

C

Colin Howarth

Jan 1, 1970
0
Hi,

I'm (trying to) design a dedicated single channel pre-amp / ADC for my
guitar. This is a hobby/learning exercise.

The ADC will probably be the 24bit cirrus logic CS5381. I'd be sampling
at 192 kHz.

A couple of questions:

1. the pickups (Humbucker) have a normal resistance but also a hefty
inductance. For noise calculations (& op-amp choice etc) is it only the
resistance that is relevant? For example, the Linear LT1115 datasheet
refers to source resistance, not source impedance.

2. the guitar output is single ended. Should/can I treat this as what
Analog's SSM2019 datasheet calls pseudo-differential, like this:


tip ------------------------ +
| |
R |
| |
GND C
| |
R |
| |
sleeve ------------------------ -



or do I have to do the dual op-amp single to differential conversion
thing?


Thanks,

colin
 
P

Phil Allison

Jan 1, 1970
0
"Colin Howarth"
1. the pickups (Humbucker) have a normal resistance but also a hefty
inductance. For noise calculations (& op-amp choice etc) is it only the
resistance that is relevant?

** Nope.

The volume pots on your guitar are the dominant sources of circuit noise -
only at full setting will residual noise be from the PU. The source
impedance of such PUs is a function of frequency with a large peak in value
around 5 to 10 kHz - depends a lot on the capacitance of the lead in use.

Imagine the source resistance to be 50kohms and bandwidth to be 7kHz and you
are in the ball park.

For example, the Linear LT1115 datasheet
refers to source resistance, not source impedance.

** That IC should be OK - but the best op-amps for magnetic guitar PUs are
low noise FET types cos the source impedance varies from 10k to 100 kohms.

2. the guitar output is single ended.


** So one connection is ground and PSU common and the other is signal.

Use the basic non-inverting stage.



..... Phil
 
C

Colin Howarth

Jan 1, 1970
0
Phil Allison said:
"Colin Howarth"

** Nope.

The volume pots on your guitar are the dominant sources of circuit noise -
only at full setting will residual noise be from the PU. The source
impedance of such PUs is a function of frequency with a large peak in value
around 5 to 10 kHz - depends a lot on the capacitance of the lead in use.

Imagine the source resistance to be 50kohms and bandwidth to be 7kHz and you
are in the ball park.

Hmmm. Not sure if you (mis)read the original as

"the pickups have normal resistance... is it THE ONLY resistance that is
relevant?"

whereas what I meant was

"the pickups have DC resistance but also a hefty inductance. For noise
calculations is it ONLY THE resistance that is relevant (Johnson noise)?"



As it happens I leave the volume pot on maximum. The resistance is then
8.5 k. Say, do people remove the volume (and tone) pots? They seem to be
useless. (I think the tone control is just a variable (500k) resistor in
series with a 22 nF capacitor to ground ? - it's hard to tell from the
"circuit diagram", http://www.dimarzio.com//media/diagrams/C.pdf , which
looks like something a 6 yr old would draw :)

The cable is coax, with 76 pF/m and 24.4 m‡/m.

** That IC should be OK - but the best op-amps for magnetic guitar PUs are
low noise FET types cos the source impedance varies from 10k to 100 kohms.


The LT1115 has extremely low voltage noise, but NatSemi's LM4562 seems
to have much lower overall THD+N.

By the way, what do they mean by "Differential Input Impedance (30 k‡)
and CM Input impedance (1 G‡)?

** So one connection is ground and PSU common and the other is signal.


OK. One connection is ground and the other is signal. I suppose one
could one day add electronics in the guitar body for true differential
signalling...

Maybe there's no point if one just plays wham, bang, wham, bang, bang,
wham and feeds that into phase distortion and bitcrusher units :)

Use the basic non-inverting stage.


Do you mean the "Single-Ended Input Buffer with Dedicated Reference
Pins" referred to in Cirrus's AppNote, AN241? Or the "Single-Ended to
Differential Input Buffer"? Or something else?

I have to study a bit to prepare my next question (which will be "how
many amplifier/buffer stages should I use?" :)

--colin
 
C

Colin Howarth

Jan 1, 1970
0
MK said:
Look at the Cirrus schematic for the evaluation board for this chip and use
that as a reference design.

I'm looking at the reference design, the evaluation board and some other
circuits...
It is NOT easy to get these converters to work
at anything like the full quoted performance - you will need very careful
attention to board layout, component types etc.

That's why I'm checking here on board layout, component types etc. :)

Some text books pretend that analogue stuff isn't really black magic at
all, if one grasps a couple of basic concepts. But we know better, hey?
:)
I found them to be very sensitive to noise on the VQ pin and that much
larger decoupling caps on that pin were needed than Cirrus suggest.

I'll remember that. Their appnote, AN241 has one topology with an opamp
just for buffering VQ. The whole input buffer "provides proper biasing,
isolation from the switched capacitor currents, low output impedance,
and anti-alias filtering."

Good luck !

Thanks :)
 
P

Paul Keinanen

Jan 1, 1970
0
Hi,

I'm (trying to) design a dedicated single channel pre-amp / ADC for my
guitar. This is a hobby/learning exercise.

The ADC will probably be the 24bit cirrus logic CS5381. I'd be sampling
at 192 kHz.

Is this pre-amp situated within the guitar body or is it in a separate
box ?

If the pre-amp is within the guitar, where is the ADC, is it within
the guitar or as a separate box ?
A couple of questions:

1. the pickups (Humbucker) have a normal resistance but also a hefty
inductance. For noise calculations (& op-amp choice etc) is it only the
resistance that is relevant? For example, the Linear LT1115 datasheet
refers to source resistance, not source impedance.

The wire resistance within the pick-up coil will generate some white
noise voltage, when operated above absolute zero (0 K) temperatures.
The unloaded noise voltage can be easily calculated from temperature
and resistance. The loaded noise voltage depends on the load
resistance (some of the noise may be converted to current noise), but
at room temperatures, the noise power density about -174 dBm/Hz can be
assumed.

The inductive or capacitively reactance does not generate any noise at
the amplifier input terminals, but the amplifier itself may generate
more noise when mismatched.
2. the guitar output is single ended. Should/can I treat this as what
Analog's SSM2019 datasheet calls pseudo-differential, like this:

Quite nice specifications for that chip.

Do you have room for this amplifier and for at least two, preferably
four 9 V batteries within the guitar body ?

Or would it be acceptable to connect the guitar to the ADC using a
CAT5 cable, one pair feeding +/- 15 V to the guitar and the other pair
outputting the amplified single ended signal plus the signal ground ?

In order to avoid noise and hum problems, I would definitively want to
build the amplifier into the guitar, even if the ADC is in a separate
box.

If the volume/tone controls are required at the guitar, first amplify
the weak pick-up signal and then use potentiometers to attenuate the
signals. If the original potentiometers are designed or high impedance
levels, you may have to change these potentiometers to units with
lower resistance.
 
P

Phil Allison

Jan 1, 1970
0
"Colin Howarth"
"Phil Allison"
"Colin Howarth"
Hmmm. Not sure if you (mis)read the original as

"the pickups have normal resistance... is it THE ONLY resistance that is
relevant?"

whereas what I meant was

"the pickups have DC resistance but also a hefty inductance. For noise
calculations is it ONLY THE resistance that is relevant (Johnson noise)?"


** Read my reply again - it is a complete answer.

The winding inductance of magnetic guitar PU's resonates with stray and
cable C in the audio band so the source impedance rises, plateaus and then
falls. This high impedance at high audio frequencies means you need an
op-amp with low current noise - something all FET types have.



The LT1115 has extremely low voltage noise,

** But high current noise which dominates if the source impedance is more
than a few thousand ohms.

Read my reply again.

Do you mean the "Single-Ended Input Buffer with Dedicated Reference
Pins" referred to in Cirrus's AppNote, AN241? Or the "Single-Ended to
Differential Input Buffer"? Or something else?


** I mean use a non-inverting stage.



..... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Paul Keinanen"
In order to avoid noise and hum problems, I would definitively want to
build the amplifier into the guitar,

** Why ?

Don't you believe that a few metres of well shielded co-axial cable can
deliver the signal from a guitar PU without introducing hum or noise ?

It's the guitar's PUs ( and any unshielded internal wiring) that are
sensitive to electric and magnetic hum fields, not the connecting cable.



..... Phil
 
C

Colin Howarth

Jan 1, 1970
0
Paul Keinanen said:
Is this pre-amp situated within the guitar body or is it in a separate
box ?

If the pre-amp is within the guitar, where is the ADC, is it within
the guitar or as a separate box ?


Both were going to be in a separate box, but the more I look at it, the
more I think having the pre-amp as close as possible to the pick-ups and
having it generate a differential signal makes sense. I'd then simply
use microphone cable, XLR connectors and phantom power.

The wire resistance within the pick-up coil will generate some white
noise voltage, when operated above absolute zero (0 K) temperatures.
The unloaded noise voltage can be easily calculated from temperature
and resistance. The loaded noise voltage depends on the load
resistance (some of the noise may be converted to current noise), but
at room temperatures, the noise power density about -174 dBm/Hz can be
assumed.

The inductive or capacitively reactance does not generate any noise at
the amplifier input terminals, but the amplifier itself may generate
more noise when mismatched.


Quite nice specifications for that chip.

Do you have room for this amplifier and for at least two, preferably
four 9 V batteries within the guitar body ?


For a pre-amp circuit - yes. For two 9 V batteries - perhaps, but a bit
impractical. For 4 - no way.

Or would it be acceptable to connect the guitar to the ADC using a
CAT5 cable, one pair feeding +/- 15 V to the guitar and the other pair
outputting the amplified single ended signal plus the signal ground ?


See above. Cat 5 cable isn't really flexible enough anyway.

In order to avoid noise and hum problems, I would definitively want to
build the amplifier into the guitar, even if the ADC is in a separate
box.


The humbuckers are hum cancelling by design.

If the volume/tone controls are required at the guitar, ...


not really

... first amplify
the weak pick-up signal and then use potentiometers to attenuate the
signals. If the original potentiometers are designed or high impedance
levels, you may have to change these potentiometers to units with
lower resistance.



Thanks,

--colin
 
C

Colin Howarth

Jan 1, 1970
0
Phil Allison said:
"Paul Keinanen"


** Why ?

Don't you believe that a few metres of well shielded co-axial cable can
deliver the signal from a guitar PU without introducing hum or noise ?

It's the guitar's PUs ( and any unshielded internal wiring) that are
sensitive to electric and magnetic hum fields, not the connecting cable.


Perhaps Paul (and certainly I) are a victim of that

A little learning is a dangerous thing; drink deep, or taste not the
Pierian spring: there shallow draughts intoxicate the brain, and
drinking largely sobers us again.

thing. :)


I'm surprised no-one has yet said "what's this obsession with low noise?
An E-guitar isn't a Stradivarius!"

The whole issue is (or I'm making it) a bit more complicated than it may
seem.


Pro audio (not that I'm a pro) is mostly digital these days. My digital
audio workstation (aka computer) likes 192 kHz 24bit input. That may
seem like overkill considering that I don't even hear up to 20 kHz
anymore BUT...

Guitar pickups (resistive/inductive) in combination with cable
capacitance have their own resonance, distortion and filtering
characteristics (ie. sound) and, in the old days, these even change
depending on what effects boxes you plug into, due to varying load
impedance.

However, now, ALL the signal modification (including filtering and
distortion) is supposed to be going on in the computer using amp
modelling, equalisation, artificial distortion etc.


If I'm sampling 24 bits, I'd like the input signal to be as clean as
possible. The ADC wants 5.6 Vpp (full scale). That's differential input,
so each signal is supposed to be 2.8 Vpp, ie. around 1 V rms.

The pickups output around 300 mV rms.

1 bit of that is, ummmm, about 18 nV.

-> low noise amps, pre-amp as soon as possible.


--colin
 
P

Phil Allison

Jan 1, 1970
0
"Colin Howarth"
The humbuckers are hum cancelling by design.


** Really ??

Most are nothing of the sort.

What kind do you have ???



..... Phil
 
P

Paul Keinanen

Jan 1, 1970
0
The main problem is the convention of using unbalanced systems with
monophonic 6.35 mm plugs. The pick-ups and microphones are by nature
balanced, but for some strange reason, the electric system in a guitar
is unbalanced.

Typically, all metallic parts, including the strings are connected to
the cold side of the jack. This creates a huge "antenna" i.e. a large
capacitance between the guitar and surrounding electric systems,
including triac controlled stage lights etc. containing 50/60 Hz and
quite high harmonics.

The capacitive reactance will allow some current to flow from stage
lights etc. to the guitar body, through the signal cable shield to the
amplifier power cord ground to the utility company.

This is not a problem in the ideal world with ideal connectors and
zero impedance cable shields. Unfortunately the plug/jack interface
and cable shield (especially in old coiled cords) may have a
significant resistance.

Any capacitively coupled interference current flowing through these
resistances will create a noise voltage drop, which is
_directly_added_ to the audio signal, considerably reducing the SNR.

I guess that everyone using an electric guitar would have experienced
the typical sound of hum with a lot of harmonic, if the plug was not
properly inserted or there was a dry joint in the cable/plug.

Most of these problems can be avoided by using stereophonic jacks on
both the guitar and amplifier and using stereophonic patch cords. With
the pickup coils connected to the L and R poles and the guitar
metallic framework connected to the ring and R and ground connected
together inside the amplifier would avoid any capacitively connected
currents from being added to the signal, even with a bad connection on
the shield.
Perhaps Paul (and certainly I) are a victim of that

A little learning is a dangerous thing; drink deep, or taste not the
Pierian spring: there shallow draughts intoxicate the brain, and
drinking largely sobers us again.

thing. :)


I'm surprised no-one has yet said "what's this obsession with low noise?
An E-guitar isn't a Stradivarius!"

The whole issue is (or I'm making it) a bit more complicated than it may
seem.


Pro audio (not that I'm a pro) is mostly digital these days. My digital
audio workstation (aka computer) likes 192 kHz 24bit input. That may
seem like overkill considering that I don't even hear up to 20 kHz
anymore BUT...

Guitar pickups (resistive/inductive) in combination with cable
capacitance have their own resonance, distortion and filtering
characteristics (ie. sound) and, in the old days, these even change
depending on what effects boxes you plug into, due to varying load
impedance.

I have not been working with electric guitars for a few decades, but I
was a bit surprised that current "high output" pick-ups produce up to
1 Vrms of output. I was used to have something like 100 mV to the
first tube in the amplifier.

If the self resonance frequency with a reasonable cable (200 pF) is in
the order of 10 kHz, that the coil inductance is more than 1 H, is
this really the current situation ?

The traditional design principle has been to try to keep any
mechanical or electrical resonances out of the frequency range of
interest.

What is the point of using "high output" pickups, if this will cause
some nasty resonance peaks within your passband ?

These days an integrated preamplifier is not a problem (e.g. phantom
powered) , thus there is no need to maximize the output voltage and
hence much lower inductances could be used and hence, the self
resonance peak could be moved well above the audio pass band,
especially when the cord capacitance is isolated from the PU.

If the guitar contains multiple pick-ups, each should have an own
preamplifier with possible mixing potentiometer after the
preamplifier.
However, now, ALL the signal modification (including filtering and
distortion) is supposed to be going on in the computer using amp
modelling, equalisation, artificial distortion etc.


If I'm sampling 24 bits, I'd like the input signal to be as clean as
possible. The ADC wants 5.6 Vpp (full scale). That's differential input,
so each signal is supposed to be 2.8 Vpp, ie. around 1 V rms.

While those chips produce "24 bit" data words, the SNR figures are at
best about 120 dB (20 bits).
The pickups output around 300 mV rms.

or 1 Vpp and 120 dB below that is 1 uV.
1 bit of that is, ummmm, about 18 nV.

-> low noise amps, pre-amp as soon as possible.

If you design a guitar amplifier input stage that different kind of
pick-ups can be connected without any gain adjustments, the real 120
dB dynamic range of currently available "24 bit" might barely be
sufficient.
 
C

Colin Howarth

Jan 1, 1970
0
Phil Allison said:
"Colin Howarth"
"Phil Allison"
"Colin Howarth"


** Read my reply again - it is a complete answer.


You mean the total noise is sqrt( e.n^2 + (i.n * R.s)^2 + e.t^2 )
where

e.n is the amp voltage noise
i.n is the amp current noise
R.s is the source IMPEDANCE
e.t is the source RESISTANCE thermal noise

So only the DC RESISTANCE is relevant for Johnson noise whilst the
IMPEDANCE (* current noise) gives another contribution and the voltage
noise is another contribution.

:)

The winding inductance of magnetic guitar PU's resonates with stray and
cable C in the audio band so the source impedance rises, plateaus and then
falls. This high impedance at high audio frequencies means you need an
op-amp with low current noise - something all FET types have.





** But high current noise which dominates if the source impedance is more
than a few thousand ohms.

Read my reply again.


Looks like Analog's ADA4627 is really nice (and slightly expensive)

e.n (@ 1 kHz) = 6.1 nV/rtHz
i.n (@ 100 Hz) = 1.6 fA/rtHz
THD+N (@ 1 kHz, A.v = 1) = 0.000045 %
CMRR = 116 dB
PSRR = 112 dB
SR = 56+ V/µs

Those are comparable values to NatSemi's LM4562 I was thinking of using

e.n (@ 1 kHz) = 2.7 nV/rtHz
i.n (@ 10 Hz, 1000 Hz) = 3.1 , 1.6 pA/rtHz
THD+N = (A.v = 1) = 0.00003 %
CMRR = 120 dB
PSRR = 120 dB
SR = 20 V/µs

except that ADI's i.n is 1000 times better :)


** I mean use a non-inverting stage.


Oh. You mean use a non-inverting stage?

How will using a non-inverting stage (say, with a gain of +3) give me
the differential input that the CS5381 requires?

--colin
 
C

Colin Howarth

Jan 1, 1970
0
Phil Allison said:
"Colin Howarth"


** Really ??

Most are nothing of the sort.

Well, they're supposed to be, aren't they? I thought that was the point?
What kind do you have ???

Some sort of DiMarzio/Ibanez thingies.


--colin
 
C

Colin Howarth

Jan 1, 1970
0
Phil Allison said:
Colin Howarth is a visitor from another planet


Damn! How did you guess?!!!

** Nuff said ???


My calculations are wrong? But to err is human is it not?


All I meant was that (for fun) I'd like the pre-amp to be as low noise
as possible and that means it should be as close as possible to the
pickup, and that fully differential signalling would be nice.

--colin
 
C

Colin Howarth

Jan 1, 1970
0
Paul Keinanen said:
The main problem is the convention of using unbalanced systems with
monophonic 6.35 mm plugs. The pick-ups and microphones are by nature
balanced, but for some strange reason, the electric system in a guitar
is unbalanced.

Typically, all metallic parts, including the strings are connected to
the cold side of the jack. This creates a huge "antenna" i.e. a large
capacitance between the guitar and surrounding electric systems,
including triac controlled stage lights etc. containing 50/60 Hz and
quite high harmonics.

The capacitive reactance will allow some current to flow from stage
lights etc. to the guitar body, through the signal cable shield to the
amplifier power cord ground to the utility company.

This is not a problem in the ideal world with ideal connectors and
zero impedance cable shields. Unfortunately the plug/jack interface
and cable shield (especially in old coiled cords) may have a
significant resistance.

Any capacitively coupled interference current flowing through these
resistances will create a noise voltage drop, which is
_directly_added_ to the audio signal, considerably reducing the SNR.

I guess that everyone using an electric guitar would have experienced
the typical sound of hum with a lot of harmonic, if the plug was not
properly inserted or there was a dry joint in the cable/plug.

Most of these problems can be avoided by using stereophonic jacks on
both the guitar and amplifier and using stereophonic patch cords. With
the pickup coils connected to the L and R poles and the guitar
metallic framework connected to the ring and R and ground connected
together inside the amplifier would avoid any capacitively connected
currents from being added to the signal, even with a bad connection on
the shield.


I have not been working with electric guitars for a few decades, but I
was a bit surprised that current "high output" pick-ups produce up to
1 Vrms of output. I was used to have something like 100 mV to the
first tube in the amplifier.

If the self resonance frequency with a reasonable cable (200 pF) is in
the order of 10 kHz, that the coil inductance is more than 1 H, is
this really the current situation ?


A quick google shows that Fender pickups are between 2.4 and 4 H which
would put the resonance at around 5 kHz. Don't know about my particular
pickups.

The traditional design principle has been to try to keep any
mechanical or electrical resonances out of the frequency range of
interest.

What is the point of using "high output" pickups, if this will cause
some nasty resonance peaks within your passband ?

These days an integrated preamplifier is not a problem (e.g. phantom
powered) , thus there is no need to maximize the output voltage and
hence much lower inductances could be used and hence, the self
resonance peak could be moved well above the audio pass band,
especially when the cord capacitance is isolated from the PU.


.... which it would be with a pre-amp. Then high output would give a
better SNR?

If the guitar contains multiple pick-ups, each should have an own
preamplifier with possible mixing potentiometer after the
preamplifier.


While those chips produce "24 bit" data words, the SNR figures are at
best about 120 dB (20 bits).

How's that SNR arrived at? Is that an "in practice" figure?
or 1 Vpp and 120 dB below that is 1 uV.


If you design a guitar amplifier input stage that different kind of
pick-ups can be connected without any gain adjustments, the real 120
dB dynamic range of currently available "24 bit" might barely be
sufficient.

It's all just for fun. And for my pickups.

Thanks for the detailed post.

--colin
 
P

Phil Allison

Jan 1, 1970
0
Colin Howarth is a visitor from another planet
Damn! How did you guess?!!!


** Having two heads is a dead give away.


My calculations are wrong?


** Plain stupid.

All I meant was that (for fun) I'd like the pre-amp to be as low noise
as possible and that means it should be as close as possible to the
pickup,


** Yawnnnnnnnn....



..... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Paul Keinanen lives in a crater Mars "

Phil Allison:

" Don't you believe that a few metres of well shielded co-axial cable can
deliver the signal from a guitar PU without introducing hum or noise ?

It's the guitar's PUs ( and any unshielded internal wiring) that are
sensitive to electric and magnetic hum fields, not the connecting cable. "

Typically, all metallic parts, including the strings are connected to
the cold side of the jack. This creates a huge "antenna" i.e. a large
capacitance between the guitar and surrounding electric systems,
including triac controlled stage lights etc. containing 50/60 Hz and
quite high harmonics.

The capacitive reactance will allow some current to flow from stage
lights etc. to the guitar body, through the signal cable shield to the
amplifier power cord ground to the utility company.

This is not a problem in the ideal world with ideal connectors and
zero impedance cable shields. Unfortunately the plug/jack interface
and cable shield (especially in old coiled cords) may have a
significant resistance.

Any capacitively coupled interference current flowing through these
resistances will create a noise voltage drop, which is
_directly_added_ to the audio signal, considerably reducing the SNR.

I guess that everyone using an electric guitar would have experienced
the typical sound of hum with a lot of harmonic, if the plug was not
properly inserted or there was a dry joint in the cable/plug.

Most of these problems can be avoided by using stereophonic jacks on
both the guitar and amplifier and using stereophonic patch cords. With
the pickup coils connected to the L and R poles and the guitar
metallic framework connected to the ring and R and ground connected
together inside the amplifier would avoid any capacitively connected
currents from being added to the signal, even with a bad connection on
the shield.


** What a load of absolute and utter TWADDLE !!!!

Not one single word of is fact.



..... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Colin Howarth is a very sick man "
Well, they're supposed to be, aren't they? I thought that was the point?



Some sort of DiMarzio/Ibanez thingies.


** Why don't you just **** off and die ??



..... Phil
 
P

Paul Keinanen

Jan 1, 1970
0
How's that SNR arrived at? Is that an "in practice" figure?

I was just trying to point out that 24 bits would theoretically be
24*6=144 dB. In practice the best claims in manufacturer data sheets
for ADCs and DACs is about 120 dB, typically with 192 kHz sampling
rate but measurement bandwidth to something like 10 Hz - 20 kHz (so
there is some noise shaping :).

As others have pointed out, the layout is critical, probably requiring
a multilayer board, perhaps one or two contiguous ground plane layers
etc.
 
Top