elektroda.net NewsGroups Forum Index - Electronics Design - Current divider mirror
John Larkin
Guest
Guest
Tue Feb 09, 2010 3:44 am
On Mon, 8 Feb 2010 18:08:12 -0600, "Tim Williams"
<tmoranwms_at_charter.net> wrote:
Quote:
"Fred Bartoli" <" "> wrote in message
news:4b70a0fd$0$21600$426a74cc_at_news.free.fr...
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
news:4b70a0fd$0$21600$426a74cc_at_news.free.fr...
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
Reference designators would help!
You could eliminate the dual diode and just let the upper PNP kill the
+ side current source, with suitable shuffling. Saves some parts and
power.
The behavior of Q1 versus pot rotation is interesting.
John
Spehro Pefhany
Guest
Guest
Tue Feb 09, 2010 4:51 am
On Mon, 08 Feb 2010 14:23:44 -0800, the renowned John Larkin
<jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
Quote:
On Mon, 08 Feb 2010 17:12:19 -0500, Spehro Pefhany
speffSNIP_at_interlogDOTyou.knowwhat> wrote:
On Mon, 08 Feb 2010 13:51:28 -0800, John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to +
is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
If you select the transistors carefully, or get lucky.
Putting two or three or seventeen transistors in parallel in a circuit
like this doesn't scale the mirror ratios, it just sort of half-assed
averages out the betas. If there is a 2:1 ratio, it's because the NPNs
have an average beta twice that of the PNPs.
A 2N4403 has a min specified beta of 60 at 1 mA, typ about 250, no
specified max.
speffSNIP_at_interlogDOTyou.knowwhat> wrote:
On Mon, 08 Feb 2010 13:51:28 -0800, John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to +
is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
If you select the transistors carefully, or get lucky.
Putting two or three or seventeen transistors in parallel in a circuit
like this doesn't scale the mirror ratios, it just sort of half-assed
averages out the betas. If there is a 2:1 ratio, it's because the NPNs
have an average beta twice that of the PNPs.
A 2N4403 has a min specified beta of 60 at 1 mA, typ about 250, no
specified max.
Beta only enters this as an error due to the base currents, which is
of the order of 1/beta. In the case of the -2*I current sink there are
3 base currents effectively subtracted from the input current, so for
a beta of 250 there will be about 1.2% error = 3/250 in the output
current.
Quote:
I got a crap mark in Uni for a circuit much like this one, because the
TA didn't understand it. 8-(
Or because he did?
TA didn't understand it. 8-(
Or because he did?
Didn't. It was a bit different because it was a linear IC design so
the emitter resistors were not required to gobble up potential Vbe
mismatch, and I shut off the |2*I| current source with a saturated
transistor across the Vbe junction. It worked fine when built on a
breadboard with monolithic transistor arrays.
Quote:
Next problem?
John
John
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff_at_interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Tim Williams
Guest
Guest
Tue Feb 09, 2010 5:05 am
"Spehro Pefhany" <speffSNIP_at_interlogDOTyou.knowwhat> wrote in message
news:shl1n5hhubvi18ucbr23s3vl17599n8nvl_at_4ax.com...
Quote:
Didn't. It was a bit different because it was a linear IC design so
the emitter resistors were not required to gobble up potential Vbe
mismatch, and I shut off the |2*I| current source with a saturated
transistor across the Vbe junction. It worked fine when built on a
breadboard with monolithic transistor arrays.
the emitter resistors were not required to gobble up potential Vbe
mismatch, and I shut off the |2*I| current source with a saturated
transistor across the Vbe junction. It worked fine when built on a
breadboard with monolithic transistor arrays.
I did that in an even earlier drawing (which I don't think I ever
breadboarded, actually),
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Function_Generator.gif
but turning off transistors slows things down. Nice part about diode gates,
they are basically instantaneous.
Tim
--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
John Larkin
Guest
Guest
Tue Feb 09, 2010 5:10 am
On Mon, 08 Feb 2010 22:51:36 -0500, Spehro Pefhany
<speffSNIP_at_interlogDOTyou.knowwhat> wrote:
Quote:
On Mon, 08 Feb 2010 14:23:44 -0800, the renowned John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 08 Feb 2010 17:12:19 -0500, Spehro Pefhany
speffSNIP_at_interlogDOTyou.knowwhat> wrote:
On Mon, 08 Feb 2010 13:51:28 -0800, John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to + is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
If you select the transistors carefully, or get lucky.
Putting two or three or seventeen transistors in parallel in a circuit
like this doesn't scale the mirror ratios, it just sort of half-assed
averages out the betas. If there is a 2:1 ratio, it's because the NPNs
have an average beta twice that of the PNPs.
A 2N4403 has a min specified beta of 60 at 1 mA, typ about 250, no
specified max.
Beta only enters this as an error due to the base currents, which is
of the order of 1/beta. In the case of the -2*I current sink there are
3 base currents effectively subtracted from the input current, so for
a beta of 250 there will be about 1.2% error = 3/250 in the output
current.
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 08 Feb 2010 17:12:19 -0500, Spehro Pefhany
speffSNIP_at_interlogDOTyou.knowwhat> wrote:
On Mon, 08 Feb 2010 13:51:28 -0800, John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to +
is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
If you select the transistors carefully, or get lucky.
Putting two or three or seventeen transistors in parallel in a circuit
like this doesn't scale the mirror ratios, it just sort of half-assed
averages out the betas. If there is a 2:1 ratio, it's because the NPNs
have an average beta twice that of the PNPs.
A 2N4403 has a min specified beta of 60 at 1 mA, typ about 250, no
specified max.
Beta only enters this as an error due to the base currents, which is
of the order of 1/beta. In the case of the -2*I current sink there are
3 base currents effectively subtracted from the input current, so for
a beta of 250 there will be about 1.2% error = 3/250 in the output
current.
Right. I missed a couple of lines so it looked beta limited to me.
Need new eyeballs, maybe later this year. Besides, I'm busy saving the
F16 fleet today.
Quote:
I got a crap mark in Uni for a circuit much like this one, because the
TA didn't understand it. 8-(
Or because he did?
Didn't. It was a bit different because it was a linear IC design so
the emitter resistors were not required to gobble up potential Vbe
mismatch, and I shut off the |2*I| current source with a saturated
transistor across the Vbe junction. It worked fine when built on a
breadboard with monolithic transistor arrays.
We had understandings with most of our TAs. They left early, we left
right after, we faked all the reports, everybody was happy.
Shorting the upper current-source Vbe junction is actually a better
way to wigwag the triangle than the diode steering thing.
John
JosephKK
Guest
Guest
Tue Feb 09, 2010 6:29 am
On Mon, 08 Feb 2010 17:12:19 -0500, Spehro Pefhany <speffSNIP_at_interlogDOTyou.knowwhat> wrote:
Quote:
On Mon, 08 Feb 2010 13:51:28 -0800, John Larkin
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to + is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
I got a crap mark in Uni for a circuit much like this one, because the
TA didn't understand it. 8-(
jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote:
On Mon, 8 Feb 2010 15:26:41 -0600, "Tim Williams"
tmoranwms_at_charter.net> wrote:
"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:3ju0n51fhfrj2pukuet0obtibm2p7lnmu3_at_4ax.com...
http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/Triangle.gif
That circuit is so astoundingly bad that it's worth saving.
Bad for what?
It has an unusually wide range, as discrete circuits go. On 680pF, I got
from 1Hz to 4.5MHz in a single range. Six decades isn't bad for a 2N440x.
Smaller RF BJT's would go lower, or you could do range switching or more
exotic approaches (leaky photodiode?).
Besides the wide range, what impressed me is the diffamp got the 2N4403
switching in about 20ns, for a balls-out maximum frequency around 35MHz.
Faster transistors would also take this arbitrarily high (the PHEMTs you're
so fond of would probably make a proper RC oscillator in the GHz). Say, do
they even make P type stupidfast transistors, SiGe or otherwise? I remember
they don't bother with P type GaAs or InP since they suck for holes.
I don't remember if, when set at ~10Hz or so, the frequency drifted by a
decade or so when I touched the leftmost transistor. It ought to. I do
remember seeing it change in steps, since I was using a wirewound pot.
Tim
Let's see how many people can each find one problem.
The most obvious one is the the charge/discharge currents depend on
the betas of the upper and lower mirror transistors. That's probably
why there are two PNPs and three NPNs: they were selected to work.
John
? The top mirror gives +I, the bottom mirror gives -2*I, where I is
the current through the NPN emitter follower.
When the 2N4403 (emitter connected to +
is "on" you get +I (cause it
conducts the -2I away through the diode), when it is off you get a net
of -I.
I got a crap mark in Uni for a circuit much like this one, because the
TA didn't understand it. 8-(
Protest time.
langwadt@fonz.dk
Guest
Guest
Tue Feb 09, 2010 6:36 pm
On 9 Feb., 01:08, "Tim Williams" <tmoran...@charter.net> wrote:
Quote:
"Fred Bartoli" <" "> wrote in messagenews:4b70a0fd$0$21600$426a74cc_at_news.free.fr...
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
--
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
--
posting a schematic here follows a predictable pattern :)
all the old guys laugh at how naive and horrible wrong it is, ...
old guys find glasses, and agree it might not be so bad....
20 post later they agree its a great design, they all did exactly
the same 40 years ago except they used a few secret tricks
that made it even more awesome
;)
--Lasse
John Larkin
Guest
Guest
Tue Feb 09, 2010 6:41 pm
On Tue, 9 Feb 2010 08:36:42 -0800 (PST), "langwadt_at_fonz.dk"
<langwadt_at_fonz.dk> wrote:
Quote:
On 9 Feb., 01:08, "Tim Williams" <tmoran...@charter.net> wrote:
"Fred Bartoli" <" "> wrote in messagenews:4b70a0fd$0$21600$426a74cc_at_news.free.fr...
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
--
posting a schematic here follows a predictable pattern :)
all the old guys laugh at how naive and horrible wrong it is, ...
old guys find glasses, and agree it might not be so bad....
20 post later they agree its a great design, they all did exactly
the same 40 years ago except they used a few secret tricks
that made it even more awesome
"Fred Bartoli" <" "> wrote in messagenews:4b70a0fd$0$21600$426a74cc_at_news.free.fr...
??? Not even, but the schmidt trigger has wrong feedback (take it from the
other 3904 collector).
Indeed! Then the 4.7k positive feedback is actually positive.
That's ONE error. Come on, can't you so-called professionals do better than
this? ;-D
Tim
--
posting a schematic here follows a predictable pattern :)
all the old guys laugh at how naive and horrible wrong it is, ...
old guys find glasses, and agree it might not be so bad....
20 post later they agree its a great design, they all did exactly
the same 40 years ago except they used a few secret tricks
that made it even more awesome
Except that this is not a great design. It's an interesting
all-discrete breadboard that would benefit greatly from a few changes
and a little math.
It does use the current mirror that I posted previously
ftp://jjlarkin.lmi.net/Mirror1.JPG
and for some reason mis-read as drawn. Luckily, usenet isn't life.
John
whit3rd
Guest
Guest
Wed Feb 10, 2010 1:49 am
I hate to get back to the original question, but my brain
itched last night...
There's a common circuit design that fills the original
requirement handily: a transconductance multiplier.
Instead of a current mirror, Q1 and Q2 with the emitters
tied together to a negative supply, add Q3 and Q4, and tie
their emitters to that supply instead. base+collector of Q3
is fed a current (call it I_gain), and base + collector of Q4
is fed another current (call this I_ref).
Then you use two voltage-follower op amps -- buffer
the potential on Q3's collector+base, and connect the emitter
of Q1 to that output, and connect the emitter of Q2
to the potential on Q4's collector+base.
The balance equation now reads:
Vbe_Q1 + Vbe_Q3 = Vbe_Q2 + Vbe_Q4
Multiply both sides by Q_e/kT, exponentiate both sides
of the equation, and multiply both sides of the result by
I_sat... then recognize the equation to have four
terms which are all Ebers-Moll transistor emitter
currents,
I_Q1 * I_gain / I_ref = I_Q2
Those are emitter currents, so it has another factor of alpha
(the collector/emitter current ratio of Q2, which is nearly one).
I_output = alpha * I_Q1 * I_gain/I_ref
so it's actually a lot like a current mirror with adjustable
transistor areas.
The transistors all must be at the same temperature (close and
on the same heatsink, with small self-heating), and all have
the same I_sat, so matched quad transistors are required.
Mat-04 or THAT300 are candidates. The voltage-follower
op amps hold the gain setting, so they haven't any slew rate
requirements to speak of.
Usually, this kind of circuit is inconvenient because of the current-
in
and current-out configuration.
elektroda.net NewsGroups Forum Index - Electronics Design - Current divider mirror
