Driver to drive?

[whit3rd]

On Tuesday, November 18, 2014 3:45:50 PM UTC-8, Joerg wrote:

rickman wrote:
On 11/18/2014 10:57 AM, Joerg wrote:
rickman wrote:

[...]

I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful...

Coupling coils and a steep transformation ratio can do that. ... where ultra-low power
consumption and bandwidth requirements clash... best solved
with discretes.
Yes, if transformers are used then power gain can provide voltage gain.
But I think that is a bit overkill for this design.

Actually the ones for power supplies are good candidates.

Stagger-tuned IF transformers are better ones when some bandpass filter is
the next requirement. Power efficiency, small size, are plusses; "nonstocked
items" issues are the minus side.

 

[rickman]

On 11/18/2014 10:27 AM, RobertMacy wrote:

On Mon, 17 Nov 2014 13:06:30 -0700, rickman <gnuarm@gmail.com> wrote:

...snip...

Your coil though is not electrostatically shielded. The reading I
have done indicates the noise sources generate a more significant E
field which can be shielded. So my starting point was a shielded
antenna. Boosting the signal is just one part of the problem. Digging
the signal out of the noise is the other. I supposed it could be
shielded easily enough. If I go with a small coil with more turns I
would use a ferrite core. With a two inch diameter that would be a
lot of ferrite.

Actually you can use a 'hollow' core. envision a 'paper towel' roll tube
made of magnetic material. The center section often does not add as much
as one would think.

Shielding is something as simple as one turn, left open, of copper tape,
wrapped around the outside of the coil. The gap prevents shorting the
turn and the broad conductive are provides electrostatic shielding.
often best to GND the tape.

...snip...
Why is a few uV "oodles of margin"?

input noise in the range of less than 10-20nV/rtHz, so lots of S/N

What input noise?

Not looking for beaucoup volts. I'm looking for a low power
consumption and enough volts. How would a follower give voltage
gain? I have only seen followers in a current amplification stage
with no voltage gain.


I thought a follower would provide the necessary burffering and eat less
power, probably wrong. Probably get the 'gain' for free.

What gain for free? A follower has no voltage gain. Why is a follower
less power?

What? How do you get 50 uV into an accumulator? Using an ADC at 240
kSPS would totally blow the power budget.

You only care about a SINGLE frequency, so why are you proposing an ADC?
The DFT is performed by simply a synchronous detection average shoved
into an accumulator. You only get one frequency bin, but there in lies
all the information you seek.

I'm not proposing an ADC. I don't get what you are talking about. How
do you plan to get the signal into the digital accumulator?

My concept is to get the signal level high enough that it can make a
statistical difference in the state of the input. With the huge
processing gain from a very narrow bandpass filter I expect to see
several 10's of dB improvement in SNR. Rather like the way they
recover a direct sequence spread spectrum signal or more accurately
very much like a lock-in amplifier.

But I have started to doubt the ability of this circuit to work behind
the digital input on the chip. If the signal and noise are too small
to overcome any latent hysteresis in the input circuit there might be
no signal to improve. I was just trying to explore the possibility of
adding a low power consumption amplifier. As usual many people take
the discussion on many tangents. That is how I learn.


Glad you appreciate the tangents. Very much appreciate the openness to
enable the fun of contributing.

The amp was just something I was exploring to make me more comfortable
having a fallback position, although I still don't have a design. The
best part I've found now that suits the design will give 40 dB of gain
in one stage, I can't seem to cascade two stages effectively. In the
simulation the capacitive loading appears to be *much* more than the 10
pF the spec sheet claims.

Still, I think it is time to move on and do some testing of the digital
circuit to see what it can be made to do. I wish I could find my
prototyping socket board. I haven't used it in years and it seems to
have gone MIA. This is the point where I need it.

--

Rick

 

[Jamie M]

On 11/17/2014 3:53 PM, sroberts6328@gmail.com wrote:
> On the systems I saw, beam positioning was galvanometer scanner based with a F-Theta lens for focal plane correction. The object to be made sat

in a tank of argon, and a metal dust tornado" was swirled around the
object to be sintered.
> The lab I was in was making turbine blade prototypes from high temperature materials. As well as other rapid prototypes. These were later annealed

in some fashion to become "single Crystal" metals. After SLS, they
went into a conventional oven to consolidate the sintering before
further treatment.

No oxygen is allowed in, for obvious reasons.

I worked in a university laser lab at the time. We were given their castoffs as worn but usable gear when they upgraded to better lasers. It kept

us running another two years. Most fun I ever had loading a truck. As
a treat(ment) I was given a tour of the SLS labs on site.

Syd,
Flame spraying is used for very course deposition, and usually a plasma torch running inert gas is the source these days. Flam spraying is usually

for placing a hardening layer on a structural object or for building
back up worn spots on a shaft. It is very crude, resolution wise, and
the target
>

Hi,

Sounds pretty interesting, also I was wondering if it would be possible
to couple a CO2 laser (200Watt to 1kW) to an optical fiber bundle and
then a focusing lens?

cheers,
Jamie

 

[jeroen Belleman]

On 18/11/14 22:03, rickman wrote:
[Snip!]

The amp was just something I was exploring to make me more comfortable
having a fallback position, although I still don't have a design. The
best part I've found now that suits the design will give 40 dB of gain
in one stage, I can't seem to cascade two stages effectively. In the
simulation the capacitive loading appears to be *much* more than the 10
pF the spec sheet claims.
[...]

Someone must have already said so, but how can you possibly
expect the input capacitance to stay 10pF in a simple grounded
source stage with 40dB gain? The datasheet doesn't give the
test circuit, but I wouldn't be at all surprised if Ciss was
measured with both source *and* drain at AC ground.

Hint: Miller effect. Solution: Cascodes, and perhaps bootstraps.

Jeroen Belleman

 

[rickman]

On 11/18/2014 4:28 PM, jeroen Belleman wrote:

On 18/11/14 22:03, rickman wrote:
[Snip!]

The amp was just something I was exploring to make me more comfortable
having a fallback position, although I still don't have a design. The
best part I've found now that suits the design will give 40 dB of gain
in one stage, I can't seem to cascade two stages effectively. In the
simulation the capacitive loading appears to be *much* more than the 10
pF the spec sheet claims.
[...]

Someone must have already said so, but how can you possibly
expect the input capacitance to stay 10pF in a simple grounded
source stage with 40dB gain? The datasheet doesn't give the
test circuit, but I wouldn't be at all surprised if Ciss was
measured with both source *and* drain at AC ground.

Hint: Miller effect. Solution: Cascodes, and perhaps bootstraps.

Yes, those words have been mentioned, but I wasn't aware that the miller
effect boosted the input capacitance. That explains and lot and I can
see that now.

Interestingly enough, Joerg provided a circuit with a Cascode second
stage but it doesn't provide more gain! In my original circuit was just
playing with the bypass capacitor in the drain leg of the second stage
and found it would affect the gain of the first stage without
appreciably affecting the gain of the overall circuit! I suppose by
reducing the gain of the second stage it reduces the Miller effect and
the apparent input capacitance raising the gain of the first stage. I
expect if I were to write out the equations for this it would show
little impact on the overall gain because of the two compensating effects.

Thanks for helping me understand this.

Now I need to figure out why the Cascode design doesn't deliver a higher
gain. It should, right? Ah, I just found it. He left the output on
the drain to emitter leg and the output is now on the collector! That's
much better giving some 75 dB at 60 kHz now.

Thanks, I think I learned a few things. Now if I can bias the circuit
properly.

--

Rick

 

[Joerg]

rickman wrote:

On 11/18/2014 4:28 PM, jeroen Belleman wrote:
On 18/11/14 22:03, rickman wrote:
[Snip!]

The amp was just something I was exploring to make me more comfortable
having a fallback position, although I still don't have a design. The
best part I've found now that suits the design will give 40 dB of gain
in one stage, I can't seem to cascade two stages effectively. In the
simulation the capacitive loading appears to be *much* more than the 10
pF the spec sheet claims.
[...]

Someone must have already said so, but how can you possibly
expect the input capacitance to stay 10pF in a simple grounded
source stage with 40dB gain? The datasheet doesn't give the
test circuit, but I wouldn't be at all surprised if Ciss was
measured with both source *and* drain at AC ground.

Hint: Miller effect. Solution: Cascodes, and perhaps bootstraps.

Yes, those words have been mentioned, but I wasn't aware that the miller
effect boosted the input capacitance. That explains and lot and I can
see that now.

Interestingly enough, Joerg provided a circuit with a Cascode second
stage but it doesn't provide more gain! In my original circuit was just
playing with the bypass capacitor in the drain leg of the second stage
and found it would affect the gain of the first stage without
appreciably affecting the gain of the overall circuit! I suppose by
reducing the gain of the second stage it reduces the Miller effect and
the apparent input capacitance raising the gain of the first stage. I
expect if I were to write out the equations for this it would show
little impact on the overall gain because of the two compensating effects.

Thanks for helping me understand this.

Now I need to figure out why the Cascode design doesn't deliver a higher
gain. It should, right? Ah, I just found it. He left the output on
the drain to emitter leg and the output is now on the collector! That's
much better giving some 75 dB at 60 kHz now.

Yup. Sorry about that, I probed at the collector but forgot to note it
in the schematic:


Version 4
SHEET 1 1340 680
WIRE 1008 -416 1008 -448
WIRE 1008 -256 1008 -336
WIRE 1152 -256 1008 -256
WIRE 1264 -256 1152 -256
WIRE 1008 -176 1008 -256
WIRE 32 -128 -16 -128
WIRE 128 -128 32 -128
WIRE 368 -128 368 -160
WIRE 944 -128 848 -128
WIRE 128 -112 128 -128
WIRE -16 -96 -16 -128
WIRE 848 -96 848 -128
WIRE -16 0 -16 -16
WIRE 128 0 128 -48
WIRE 368 0 368 -48
WIRE 416 0 368 0
WIRE 448 0 416 0
WIRE 608 0 512 0
WIRE 768 0 608 0
WIRE 848 0 848 -16
WIRE 368 32 368 0
WIRE 1008 32 1008 -80
WIRE 1264 48 1264 -256
WIRE 240 96 -16 96
WIRE 320 96 240 96
WIRE 768 96 768 0
WIRE 832 96 768 96
WIRE 960 96 832 96
WIRE 240 144 240 96
WIRE 368 144 368 128
WIRE 448 144 368 144
WIRE 496 144 448 144
WIRE 1008 144 1008 128
WIRE 1088 144 1008 144
WIRE 1136 144 1088 144
WIRE -16 160 -16 96
WIRE 768 160 768 96
WIRE 368 176 368 144
WIRE 1008 176 1008 144
WIRE 496 192 496 144
WIRE 1136 192 1136 144
WIRE 1264 224 1264 112
WIRE 240 256 240 224
WIRE -16 288 -16 240
WIRE 368 288 368 256
WIRE 496 288 496 256
WIRE 496 288 368 288
WIRE 1008 288 1008 256
WIRE 1136 288 1136 256
WIRE 1136 288 1008 288
WIRE 368 336 368 288
WIRE 768 336 768 240
WIRE 1008 336 1008 288
FLAG 368 336 0
FLAG -16 0 0
FLAG 32 -128 V2.2
FLAG -16 96 Vin
FLAG 240 256 0
FLAG -16 288 0
FLAG 128 0 0
FLAG 368 -160 V2.2
FLAG 448 144 Vs
FLAG 1008 336 0
FLAG 1152 -256 Vout
FLAG 1008 -448 V2.2
FLAG 1088 144 Vs2
FLAG 768 336 0
FLAG 416 0 G1
FLAG 608 0 Vin2
FLAG 832 96 Vin3
FLAG 1264 224 0
FLAG 848 0 0
SYMBOL voltage -16 -112 R0
WINDOW 123 0 0 Left 2
WINDOW 39 24 124 Left 2
SYMATTR SpiceLine Rser=1
SYMATTR InstName V1
SYMATTR Value 2.2v
SYMBOL voltage -16 144 R0
WINDOW 123 24 152 Left 2
WINDOW 39 24 124 Left 2
SYMATTR Value2 AC 1
SYMATTR SpiceLine Rser=10
SYMATTR InstName V2
SYMATTR Value SINE(0 50uV 60K)
SYMBOL res 224 128 R0
SYMATTR InstName R1
SYMATTR Value 10Meg
SYMBOL cap 112 -112 R0
SYMATTR InstName C5
SYMATTR Value 100ľF
SYMBOL res 352 -144 R0
SYMATTR InstName R3
SYMATTR Value 100k
SYMBOL njf 320 32 R0
SYMATTR InstName T1
SYMATTR Value JBF862
SYMBOL res 352 160 R0
SYMATTR InstName R2
SYMATTR Value 100k
SYMBOL cap 480 192 R0
SYMATTR InstName C1
SYMATTR Value 10ľF
SYMBOL res 992 -432 R0
SYMATTR InstName R6
SYMATTR Value 100k
SYMBOL njf 960 32 R0
SYMATTR InstName T2
SYMATTR Value JBF862
SYMBOL res 992 160 R0
SYMATTR InstName R5
SYMATTR Value 100k
SYMBOL cap 1120 192 R0
SYMATTR InstName C3
SYMATTR Value 1000nf
SYMBOL cap 448 16 R270
WINDOW 0 32 32 VTop 2
WINDOW 3 0 32 VBottom 2
SYMATTR InstName C2
SYMATTR Value 0.01ľ
SYMBOL res 752 144 R0
SYMATTR InstName R4
SYMATTR Value 10Meg
SYMBOL cap 1248 48 R0
SYMATTR InstName C4
SYMATTR Value 10pF
SYMBOL npn 944 -176 R0
SYMATTR InstName Q1
SYMATTR Value 2N3904
SYMBOL voltage 848 -112 R0
WINDOW 123 0 0 Left 2
WINDOW 39 24 124 Left 2
SYMATTR SpiceLine Rser=1
SYMATTR InstName V3
SYMATTR Value 1.6v
TEXT 504 -200 Left 2 !.ac dec 10 0.1 10Meg
TEXT -24 400 Left 2 !.lib spice_BF862.prm

Thanks, I think I learned a few things. Now if I can bias the circuit
properly.

Just use a resistive divider from 2.2V and a bypass cap on the base.
Something like 2V or a little less would be fine.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[rickman]

On 11/18/2014 10:52 AM, Joerg wrote:

rickman wrote:
On 11/16/2014 8:00 PM, Joerg wrote:
rickman wrote:
On 11/16/2014 6:49 PM, Joerg wrote:
rickman wrote:
On 11/16/2014 7:49 AM, RobertMacy wrote:

[...]


Why resonant antenna? You can gain significantly voltage coming in.
Envision the generated voltage creating N*2piFB*Area will be
small, but
put in a tank with Q=100 and the signals start flying around 100X
bigger, higher impedance but this is low frequency voltage.
Tapping off
the parallel resonance you just gained 100X in the incoming voltage.
Now
you need more like a buffer than a gain stage.

Note, I'm an Analog Designer from waaaay back, but sometimes going
straight to digital has its advantages.

That was what I had hoped to show with this project. I envisioned it
two years ago I believe, but put it aside and have only given it a
little thought since then.

The comment was about a tuned circuit in the drain leg of an amplifier
stage. The antenna is already tuned and the calculated Q is 90.


A word of caution here. If you rely on a Q of 90 for the input side
which low your circuit will still take in all kinds of crud. The noise
comes from myriad switch mode supplies and with an Iq of just a few
microamps that can quickly swamp your amplifier. Once that happens
everything mixes with everything else in there and the WWVB signal
might
drown.

Which switch mode supplies, in the same device or in other devices? That
is one reason why I prefer not to have an amplifier. I am investigating
what is possible with low power.


I meant power supplies in the vicinity. And not just those, there's also
tons of <100kHz noise from CFL, LED lighting, induction ranges and so on.

Low power and high unity gain bandwidth bite each other. Same for low
power and intermodulation performance.


Really good WWVB clocks had two crystals in the front end for
filtering.
"Modern" ones don't and the result is as expected. Here in the
California Sierra they don't work well anymore.

I considered a crystal, but I haven't found any good info on how to
design that and it makes the device work for only one frequency.


Designing crystal filter means bench time, simulating crystals at narrow
bandwidth is very tedious and time consuming. And yes, it only works at
one frequency.

It also means a lot of book time. I tried to look into that once but
just didn't find my info on design, including from the xtal makers. They
are all about oscillators though.


I believe you will have no other choice. Especially since you said in an
answer to Robert that you want to drive an FPGA pin with hysteresis
which is a really crummy "detector". With a Q of just 90 man-made noise
can swamp you. Unless you live in a rural neigborhood with no overhead
wiring.

I'm not trying to use hysteresis on the digital input pins and it is not
supposed to have any. But the input is a differential digital input and
likely is not nearly as good as a proper comparator. So I am assuming
it will have some "issues" including some amount of hysteresis that will
need to be overcome by the signal. The hysteresis may not be an issue
since there will likely be more noise than signal.

... With a
tuned antenna you can adjust it for multiple frequencies of operation.
There are a number of time signals broadcast in the range of 50 to 80
kHz which could be tuned using one configuration.


You could switch crystals but that does get old when you have to cover
almost all time signal transmitters in the world.

This is not a product, but just something I'm pushing around and would
like to build. The first step is just to make it work with low power.


That's good, and gets you involved in analog. BTW, the problem in your
circuit is Miller in T2. If you reduce R6 to 100ohms T1 shows sufficient
bandwidth. So cascode is the name of the game. Got to run, our Labradors
want their morning walk but I quickly put in a BJT to show the effect.
Replace the .lib statement with yours (didn't work on my PC). V3 would
normally be replaced by a resistive divider with a bypass cap:

Ok, thanks. I was not aware that the Miller effect was about the input
capacitance. I was thinking it had to do with feedback interfering with
the frequency response. I can see how it would also add to the input
capacitance.

I was confused for a bit as you had left the output on the drain of the
JFET so the gain was less than the original circuit. But once it was
moved to the collector all is good!

I replaced your voltage source with a diode from the 2.2 volt rail, a
cap in parallel and a resistor to ground. I think this is good enough
to prototype. Supply current is just 12 uA.

--

Rick

 

[rickman]

On 11/18/2014 6:06 AM, Adrian Tuddenham wrote:

rickman <gnuarm@gmail.com> wrote:


I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful for positive feedback. But maybe I'm just not
getting it.

Think power, not voltage.

I understand the difference, but an opamp works off of voltages.

--

Rick

 

[rickman]

On 11/18/2014 10:57 AM, Joerg wrote:

rickman wrote:
On 11/17/2014 4:56 PM, whit3rd wrote:
On Sunday, November 16, 2014 5:01:35 PM UTC-8, rickman wrote:
X-Received: by 10.66.165.166 with SMTP id
yz6mr69274036pab.7.1416261376137;
Mon, 17 Nov 2014 13:56:16 -0800 (PST)
X-Received: by 10.50.66.144 with SMTP id
f16mr325984igt.17.1416261376022; Mon,
17 Nov 2014 13:56:16 -0800 (PST)
Path:
eternal-september.org!mx02.eternal-september.org!feeder.eternal-september.org!news.glorb.com!uq10no642402igb.0!news-out.google.com!c9ni14081igv.0!nntp.google.com!h15no2246572igd.0!postnews.google.com!glegroupsg2000goo.googlegroups.com!not-for-mail

Newsgroups: sci.electronics.design
Date: Mon, 17 Nov 2014 13:56:15 -0800 (PST)
In-Reply-To: <m4bhcv$l0n$1@dont-email.me
Complaints-To: groups-abuse@google.com
Injection-Info: glegroupsg2000goo.googlegroups.com; posting-host
9.221.140.126;
posting-account=Qm_QoAAADOaDCYsqOFDAW8NJ8sFHoE
NNTP-Posting-Host: 209.221.140.126
References: <m3bnhn$82n$1@dont-email.me> <m49519$nvf$1@dont-email.me
5acc3bb1-059d-4f6f-9c4c-b94ab5790b41@googlegroups.com
m4bhcv$l0n$1@dont-email.me
User-Agent: G2/1.0
MIME-Version: 1.0
Message-ID: <4072a29d-913d-4646-afa4-4ebcb87a63b2@googlegroups.com
Subject: Re: Very Low Power Preamp
From: whit3rd <whit3rd@gmail.com
Injection-Date: Mon, 17 Nov 2014 21:56:16 +0000
Content-Type: text/plain; charset=O-8859-1
Content-Transfer-Encoding: quoted-printable
Xref: mx02.eternal-september.org sci.electronics.design:348164

On Sunday, November 16, 2014 5:01:35 PM UTC-8, rickman wrote:
On 11/16/2014 5:47 PM, whit3rd wrote:
On Saturday, November 15, 2014 7:18:17 PM UTC-8, rickman wrote:
On 11/4/2014 6:29 PM, rickman wrote:
I am working on a project for receiving a very narrow bandwidth
signal
at 60 kHz [and] keep the power consumption to
an absolute minimum

How about a programmable-bias-current op amp? LM4250 is easily
available, there used to be lots of others. OTAs like LM13700 have
similar
low-power specifications, with the additional feature that output loads
don't change the power requirement.

The LM4250 has too low a GBP to work at 60 kHz. The max GBW is about
300 kHz and the setting I would need brings it closer to 60,000 kHz
where it would have a gain of about 1.

Not completely relevant; remember, an amplifier performs three functions,
and voltage gain is only one. The other two, power gain (because the
output
impedance is lower than input impedance) and input/output isolation,
are enough to make this chip useful. There have been suggestions
already to use positive-feedback circuitry, and isolation REALLY helps
there.

I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful for positive feedback. But maybe I'm just not
getting it.


Coupling coils and a steep transformation ratio can do that. But you are
already on the right track. Such problems where ultra-low power
consumption and bandwidth requirements clash are usually best solved
with discretes. That also provides the best learning environment in the
arts of analog design. I've met many younger EEs who work in analog but
can only do IC design and work with opamps and other IC. A multi-stage
transistor-level amp throw them for a loop.

Yes, if transformers are used then power gain can provide voltage gain.
But I think that is a bit overkill for this design. Transformers are
a bit of a PITA to use, but that is likely because I am not very
familiar with them other than for power supplies, lol.

--

Rick

 

[Tim Wescott]

On Mon, 17 Nov 2014 14:34:15 -0500, rickman wrote:

On 11/17/2014 1:39 PM, sroberts6328@gmail.com wrote:
Well, My NDA expired on the subject from 6 years ago. 400-1000 Watt Co2
in a inert atmosphere. Fiber lasers can be used with some metals.

The trick is not the laser, its the specialized lensing.

I haven't looked at the video. I assume the beam is invisible as CO2
lasers work in the infrared range.

There's bright visible light, but it's not clear to me whether it's the
laser, the incandescence of the metal, or if the laser has a spotter light
on it.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

 

[Joerg]

rickman wrote:

On 11/18/2014 10:52 AM, Joerg wrote:
rickman wrote:
On 11/16/2014 8:00 PM, Joerg wrote:
rickman wrote:
On 11/16/2014 6:49 PM, Joerg wrote:

[...]

Really good WWVB clocks had two crystals in the front end for
filtering.
"Modern" ones don't and the result is as expected. Here in the
California Sierra they don't work well anymore.

I considered a crystal, but I haven't found any good info on how to
design that and it makes the device work for only one frequency.


Designing crystal filter means bench time, simulating crystals at
narrow
bandwidth is very tedious and time consuming. And yes, it only works at
one frequency.

It also means a lot of book time. I tried to look into that once but
just didn't find my info on design, including from the xtal makers. They
are all about oscillators though.


I believe you will have no other choice. Especially since you said in an
answer to Robert that you want to drive an FPGA pin with hysteresis
which is a really crummy "detector". With a Q of just 90 man-made noise
can swamp you. Unless you live in a rural neigborhood with no overhead
wiring.

I'm not trying to use hysteresis on the digital input pins and it is not
supposed to have any. But the input is a differential digital input and
likely is not nearly as good as a proper comparator. So I am assuming
it will have some "issues" including some amount of hysteresis that will
need to be overcome by the signal. The hysteresis may not be an issue
since there will likely be more noise than signal.

You might also see a great deal of noise on the threshold level. The
only way to overcome it is more amplitude but that's tough with a
low-power circuit when you must maintain good intermodulation specs at
the same time. If you really want to use a digital input I'd consider an
external comparator in front of that. They are cheap and come in
SOT23-5, SC70-5 or even smaller.

Some hysteresis is needed because in the absence of signals you could
otherwise have a wildly switching input.

... With a
tuned antenna you can adjust it for multiple frequencies of operation.
There are a number of time signals broadcast in the range of 50 to 80
kHz which could be tuned using one configuration.


You could switch crystals but that does get old when you have to cover
almost all time signal transmitters in the world.

This is not a product, but just something I'm pushing around and would
like to build. The first step is just to make it work with low power.


That's good, and gets you involved in analog. BTW, the problem in your
circuit is Miller in T2. If you reduce R6 to 100ohms T1 shows sufficient
bandwidth. So cascode is the name of the game. Got to run, our Labradors
want their morning walk but I quickly put in a BJT to show the effect.
Replace the .lib statement with yours (didn't work on my PC). V3 would
normally be replaced by a resistive divider with a bypass cap:

Ok, thanks. I was not aware that the Miller effect was about the input
capacitance. I was thinking it had to do with feedback interfering with
the frequency response. I can see how it would also add to the input
capacitance.

Miller acts like a multiplier of Cgd. Cgd is smaller than Cds but if
your stage gain is 40dB the gate will see 100*Cds. It can swamp
everything. That is also one reason why designing power converters for
hundreds of volts is tough.

I was confused for a bit as you had left the output on the drain of the
JFET so the gain was less than the original circuit. But once it was
moved to the collector all is good!

Yup, sorry, the collector is where I probed and then I forgot to move
the output label and 10pF cap there.

I replaced your voltage source with a diode from the 2.2 volt rail, a
cap in parallel and a resistor to ground. I think this is good enough
to prototype. Supply current is just 12 uA.

That should work well.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[rickman]

On 11/18/2014 6:09 PM, Joerg wrote:

rickman wrote:
On 11/18/2014 10:52 AM, Joerg wrote:
rickman wrote:
On 11/16/2014 8:00 PM, Joerg wrote:
rickman wrote:
On 11/16/2014 6:49 PM, Joerg wrote:

[...]

Really good WWVB clocks had two crystals in the front end for
filtering.
"Modern" ones don't and the result is as expected. Here in the
California Sierra they don't work well anymore.

I considered a crystal, but I haven't found any good info on how to
design that and it makes the device work for only one frequency.


Designing crystal filter means bench time, simulating crystals at
narrow
bandwidth is very tedious and time consuming. And yes, it only works at
one frequency.

It also means a lot of book time. I tried to look into that once but
just didn't find my info on design, including from the xtal makers. They
are all about oscillators though.


I believe you will have no other choice. Especially since you said in an
answer to Robert that you want to drive an FPGA pin with hysteresis
which is a really crummy "detector". With a Q of just 90 man-made noise
can swamp you. Unless you live in a rural neigborhood with no overhead
wiring.

I'm not trying to use hysteresis on the digital input pins and it is not
supposed to have any. But the input is a differential digital input and
likely is not nearly as good as a proper comparator. So I am assuming
it will have some "issues" including some amount of hysteresis that will
need to be overcome by the signal. The hysteresis may not be an issue
since there will likely be more noise than signal.


You might also see a great deal of noise on the threshold level. The
only way to overcome it is more amplitude but that's tough with a
low-power circuit when you must maintain good intermodulation specs at
the same time. If you really want to use a digital input I'd consider an
external comparator in front of that. They are cheap and come in
SOT23-5, SC70-5 or even smaller.

Some hysteresis is needed because in the absence of signals you could
otherwise have a wildly switching input.

In the absence of an input I would expect a wildly switching input. In
fact, I am going to provide feedback to assure it!

Thanks again for the help. I had gotten off on the dual gate FET
tangent and was finding no traction because they seem to want more than
the 2 or 3 volts I have to work with. I like this a lot better but I
might try simulating one of the dual gate FETs just for fun.

The Cascode circuit is actually pretty simple. It is easier for me to
understand if I picture it as a common source followed by a common
gate/base circuit. Then I can visualize the operation with things I am
familiar with.

Just out of curiosity, is this ever done with a common drain as the
first FET? I expect that would certainly minimize the Miller effect and
provide the current gain needed to drive the common gate/base transistor.

--

Rick

 

[Joerg]

rickman wrote:

On 11/18/2014 6:09 PM, Joerg wrote:
rickman wrote:
On 11/18/2014 10:52 AM, Joerg wrote:
rickman wrote:
On 11/16/2014 8:00 PM, Joerg wrote:
rickman wrote:
On 11/16/2014 6:49 PM, Joerg wrote:

[...]

Really good WWVB clocks had two crystals in the front end for
filtering.
"Modern" ones don't and the result is as expected. Here in the
California Sierra they don't work well anymore.

I considered a crystal, but I haven't found any good info on how to
design that and it makes the device work for only one frequency.


Designing crystal filter means bench time, simulating crystals at
narrow
bandwidth is very tedious and time consuming. And yes, it only
works at
one frequency.

It also means a lot of book time. I tried to look into that once but
just didn't find my info on design, including from the xtal makers.
They
are all about oscillators though.


I believe you will have no other choice. Especially since you said
in an
answer to Robert that you want to drive an FPGA pin with hysteresis
which is a really crummy "detector". With a Q of just 90 man-made noise
can swamp you. Unless you live in a rural neigborhood with no overhead
wiring.

I'm not trying to use hysteresis on the digital input pins and it is not
supposed to have any. But the input is a differential digital input and
likely is not nearly as good as a proper comparator. So I am assuming
it will have some "issues" including some amount of hysteresis that will
need to be overcome by the signal. The hysteresis may not be an issue
since there will likely be more noise than signal.


You might also see a great deal of noise on the threshold level. The
only way to overcome it is more amplitude but that's tough with a
low-power circuit when you must maintain good intermodulation specs at
the same time. If you really want to use a digital input I'd consider an
external comparator in front of that. They are cheap and come in
SOT23-5, SC70-5 or even smaller.

Some hysteresis is needed because in the absence of signals you could
otherwise have a wildly switching input.

In the absence of an input I would expect a wildly switching input. In
fact, I am going to provide feedback to assure it!

Just watch out that the FPGA doesn't let off an evil hiss

Thanks again for the help. I had gotten off on the dual gate FET
tangent and was finding no traction because they seem to want more than
the 2 or 3 volts I have to work with. I like this a lot better but I
might try simulating one of the dual gate FETs just for fun.

The dual-gate is indeed not so great here because of the low supply voltage.

The Cascode circuit is actually pretty simple. It is easier for me to
understand if I picture it as a common source followed by a common
gate/base circuit. Then I can visualize the operation with things I am
familiar with.

Just out of curiosity, is this ever done with a common drain as the
first FET? I expect that would certainly minimize the Miller effect and
provide the current gain needed to drive the common gate/base transistor.

Never seen it and it would not make much sense because then the first
stage wouldn't contribute any gain. The cascode you have now virtually
eliminates the Miller effect by holding the emitter at a constant
potential of base voltage minus a diode drop.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

rickman wrote:

On 11/18/2014 10:57 AM, Joerg wrote:
rickman wrote:

[...]

I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful for positive feedback. But maybe I'm just not
getting it.


Coupling coils and a steep transformation ratio can do that. But you are
already on the right track. Such problems where ultra-low power
consumption and bandwidth requirements clash are usually best solved
with discretes. That also provides the best learning environment in the
arts of analog design. I've met many younger EEs who work in analog but
can only do IC design and work with opamps and other IC. A multi-stage
transistor-level amp throw them for a loop.

Yes, if transformers are used then power gain can provide voltage gain.
But I think that is a bit overkill for this design. Transformers are a
bit of a PITA to use, but that is likely because I am not very familiar
with them other than for power supplies, lol.

Actually the ones for power supplies are good candidates. You'd just
have to pick one with the right turns ratio and no air gap. Meaning not
one for flyback converters. They used to be a pain 20 years ago because
you had to wind it yourself but nowadays even Digikey has lots of them.
Mini-Circuits has good RF transformers but those can be pricey and you
don't really need that for a 60kHz application.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Lasse Langwadt Christense]

Den onsdag den 19. november 2014 09.55.40 UTC+1 skrev rickman:

On 11/19/2014 3:19 AM, Jamie M wrote:
On 11/18/2014 2:44 PM, sroberts6328@gmail.com wrote:
other hand goes through high power silica fibers.

Jamie, Start with Jeff Hecht's book "The Laser Guidebook" and
Silfvast's Book "Laser Fundamentals" before you go down this beam path.

The learning curve is extremely steep an


Hi,

Thanks, I recall hearing from previous discussions on here that glass
is opaque to CO2 laser wavelengths. I was also wondering about a
relatively cheap CO2 40Watt laser tube like this one:

http://www.ebay.com/itm/151058818267

Could the output laser beam from something like that be focused down
to a small enough spot that it would melt metal powder? If so I guess
as was mentioned already the bonding to the layer below might be
the reason a cheap 3D metal printer can't be made with these tubes,
otherwise it could be good for really high resolution small metal
parts. Maybe the concept of a "heated bed" common in plastic filament
3D printers could be used to keep the small (ie 1cm x 1cm) work area
hot enough to make the layers bond easier.

Your lens would need to be quartz or perhaps a plastic that is
transparent to IR. In chemistry we used quartz cuvettes to hold samples
for spectrophotometry... although I think that was for UV. I guess the
opaqueness of glass depends on the frequency of IR as we did use glass
test tubes for some IR work... I think. Heck, that was some 40+ years
ago.

lenses for CO2 laser are afaict made of Zinc selenide

I can't see how a 40W CO2 laser would be nearly enough for welding, that
is what is used at the lower end in laser cutters that can only just cut something like 6mm plastic or wood

-Lasse

 

[Ken]

On 19-Nov-14 3:07 PM, Grant wrote:

Hi,

Would you ever design such a wiring loom as this?

http://media.englishrussia.com/newpictures/Russian-military-factory-Sheglovski-Val-or-Sheglovs-Earthwork/1623286_900.jpg

or <http://goo.gl/D571Qv

Yeah, fault-finding would be so easy...

Nope.

Grant.

It's an MFR. They take about 3 weeks to assemble.

 

[Ken]

On 19-Nov-14 3:49 PM, Ken wrote:

On 19-Nov-14 3:07 PM, Grant wrote:
Hi,

Would you ever design such a wiring loom as this?

http://media.englishrussia.com/newpictures/Russian-military-factory-Sheglovski-Val-or-Sheglovs-Earthwork/1623286_900.jpg


or <http://goo.gl/D571Qv

Yeah, fault-finding would be so easy...

Nope.

Grant.




It's an MFR. They take about 3 weeks to assemble.

From what I think is the same facility;

http://ic.pics.livejournal.com/bmpd/38024980/1622553/1622553_900.jpg


Why is the ceiling fully reflective? Save on lighting?

 

[Jamie M]

On 11/18/2014 2:44 PM, sroberts6328@gmail.com wrote:

other hand goes through high power silica fibers.

Jamie, Start with Jeff Hecht's book "The Laser Guidebook" and Silfvast's Book "Laser Fundamentals" before you go down this beam path.

The learning curve is extremely steep an

Hi,

Thanks, I recall hearing from previous discussions on here that glass
is opaque to CO2 laser wavelengths. I was also wondering about a
relatively cheap CO2 40Watt laser tube like this one:

http://www.ebay.com/itm/151058818267

Could the output laser beam from something like that be focused down
to a small enough spot that it would melt metal powder? If so I guess
as was mentioned already the bonding to the layer below might be
the reason a cheap 3D metal printer can't be made with these tubes,
otherwise it could be good for really high resolution small metal
parts. Maybe the concept of a "heated bed" common in plastic filament
3D printers could be used to keep the small (ie 1cm x 1cm) work area
hot enough to make the layers bond easier.

cheers,
Jamie

 

[Adrian Tuddenham]

rickman <gnuarm@gmail.com> wrote:

On 11/18/2014 6:06 AM, Adrian Tuddenham wrote:
rickman <gnuarm@gmail.com> wrote:


I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful for positive feedback. But maybe I'm just not
getting it.

Think power, not voltage.

I understand the difference, but an opamp works off of voltages.

If you have a stage which has a voltage gain of 1 but delivers more
current, you can then obtain 'voltage gain' from passive components such
as a transformer or a resonant circuit. As your appication appears to
need some tuned circuits, this might be something worth thinking about.

I have even seen a circuit which gives 'voltage gain' using cascaded
resistor and capacitor networks but no inductors (it was in Wireless
World some time in the 1950s - sorry I can't be more precise).


--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
www.poppyrecords.co.uk

 

[rickman]

On 11/19/2014 3:37 AM, Adrian Tuddenham wrote:

rickman <gnuarm@gmail.com> wrote:

On 11/18/2014 6:06 AM, Adrian Tuddenham wrote:
rickman <gnuarm@gmail.com> wrote:


I usually can figure out what people mean when they post something new
to me. But I can't grasp how an amp with a gain of 1 would be able to
do anything useful for positive feedback. But maybe I'm just not
getting it.

Think power, not voltage.

I understand the difference, but an opamp works off of voltages.

If you have a stage which has a voltage gain of 1 but delivers more
current, you can then obtain 'voltage gain' from passive components such
as a transformer or a resonant circuit. As your appication appears to
need some tuned circuits, this might be something worth thinking about.

I have even seen a circuit which gives 'voltage gain' using cascaded
resistor and capacitor networks but no inductors (it was in Wireless
World some time in the 1950s - sorry I can't be more precise).

It is something to consider.

--

Rick