Favourite parts with off-label uses?

[mpm]

On Monday, April 6, 2020 at 9:43:33 PM UTC-4, jla...@highlandsniptechnology..com wrote:

On Mon, 6 Apr 2020 14:50:46 -0700 (PDT), mpm wrote:

While not a component...

Early in my career, it was demonstrated to me that you could make a reasonably accurate RF millivoltmeter out of an analog Simpson 260 voltmeter.

(Without modifying it, of course.)

Mo found this at some neighborhood junk sale and bought it for me. $3
or something.

https://www.dropbox.com/s/tz0panr4f1nlqer/RF_Ammeter.JPG?raw=1

It's probably a thermocouple.

How did that Simpson thing work?

I have some Spice models of an RF detector using an SMS7621
low-barrier schottly. It works around 50 mV RMS.

Wow John, that is a nice looking vintage analog meter!!
Nice find.

From an old, faded, memory...
The trick with the Simpson 260 was you connect the red lead to the uAmp input, and put the black lead in the Common (-). But, you don't connect the black lead to your circuit. What you do instead is loosely drape is across the unit under test, and then probe with just the red lead. Essentially, the meter is reading rectified RF.

The above description about just leaving one lead disconnected but loosely coiled-up in close proximity to the circuit under test is correct. But I might have the meter connections wrong. At this point, I would probably have to have a Simpson 260 in front of me to figure it out -- but it's basically along those lines. The meter's not designed to do it, but it will. And it more accurate than you would think, for such a weird, oddball approach. But hey, if you need a relative RF millivoltmeter and don't have one handy.....

You can also fix a Quintron QBT-250 paging transmitter with a 9-iron golf club, and certain old Ford pickup trucks with garlic bread. (Maybe I shared those stories here before?, but they're a little off-topic in any case.)

You got a use in mind for that meter? (Just wondering)

 

[Michael Terrell]

On Tuesday, April 7, 2020 at 6:09:48 PM UTC-4, Carl wrote:

"Phil Hobbs" wrote in message
news:bc1dc43e-b9a1-7e91-15dc-5f851bc46c7b@electrooptical.net...

On 2020-04-07 09:15, George Herold wrote:
On Monday, April 6, 2020 at 5:43:54 PM UTC-4, Phil Hobbs wrote:
On 2020-04-05 13:39, jlarkin wrote:
On Sun, 5 Apr 2020 12:17:03 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-04-05 11:42, jlarkin wrote:
On Sun, 5 Apr 2020 10:52:29 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:
big snip

We've started putting little Sensirion T/H sensors in a lot of things.
When using TECs, it's awfully nice to be able to compute the dew
point,
for instance, and in outdoor applications (e.g. our fire sensors for
harvesters) it's good to be able to predict when the window is liable
to
fog up on the inside.

We use IP67+ enclosures with bags of 5A molecular sieve inside, which
is
super cheap and will absorb 50% of its own mass in water. Simon has
had
to learn a whole lot about enclosures and mechanical design generally.
Turns out that you have to put an air vent on the enclosure to prevent
pumping water inside due to atmospheric pressure differences. That
leads to working of the O-ring seals, which wears them out.

We considered using a bellows, but atmospheric pressure varies +-7% or
so, which makes for a pretty big, floppy bellows. Hermetic
construction
is possible but very expensive, and relies on glass or ceramic
insulated
connectors. The glass would have had to be brazed or indium-soldered
to
the lid, which leads to CTE mismatch problems.

A balloon inside?

Doesn't help--you need something to reduce the pressure drop across the
seals. An aluminized rubber diaphragm inside the box, with one side
vented, might work.

I made a calculation that 50 grams of 5A molecular sieve would keep the
inside dry for about 10 years.

Phil, I'm not sure I understand the problem. I wonder if you could say
a bit more?
(I do dream of someday maybe using a sealed box...)
You've got a sealed box.
But pressure differences can cause the seals to fail.
So you put a vent in it.
but now that will pump water into your box.
(Because you have molecular sieve in there and it will
reduce the vapor pressure of water in the box.)
So you need a long vent such that the diffusion time for
H2O is long....

Is that close to stating the problem correctly?

Some sort of long thin tube?


Air pressure varies by about +-7% (~1 psi) over time scales of a day or
two. Our enclosures have a volume of about 300 ml, and will be mounted on
top of a harvester. It will be painted a famous dark green colour, so its
albedo will be low. Thus on sunny days, it will experience temperature
fluctuations of as much as 70C over time scales of hours. That will lead to
fluctuations on the order of +-2 psi in a sealed enclosure.

This isn't a whole lot of force, but there are all sorts of white papers
from Parker and other companies that say that pumping liquid water past
O-rings is a significant problem in outdoor equipment. One reason is that
the pressure variations basically centre on zero, and O-ring seals really
really like a single sign of pressure difference, which causes the O-rings
to stay seated.

In a vented 300 ml enclosure with 50g of 5A molecular sieve, we can get
about 24000 cycles of 20% air exchange at a dew point of 20C before the
sieve saturates, so that's many years.

Liquid getting in is much worse.

Cheers

Phil Hobbs

Rube Goldberg (or someone bored and tired of staring at the walls ) would
suggest a spring-loaded piston with a solenoid release. Compress spring and
"arm" solenoid release, seal in box with oring seals and hermetic wire
feedthroughs, trigger solenoid so piston releases and raises the pressure in
the box, and voila, happy orings . If you have to open the box for
service, just rearm and fire the piston again once the box is resealed.

Something like this? <https://www.ebay.com/c/838420942>

I had a piece of test equipment that I couldn't open the shipping container, I had to wait for a really hot day to put it outside, until the pressure equalized. I bought one of these, so it was no longer temperature sensitive storage.

 

[John Larkin]

On Tue, 7 Apr 2020 15:16:32 -0700 (PDT), mpm <mpmillard@aol.com>
wrote:

On Monday, April 6, 2020 at 9:43:33 PM UTC-4, jla...@highlandsniptechnology.com wrote:
On Mon, 6 Apr 2020 14:50:46 -0700 (PDT), mpm wrote:

While not a component...

Early in my career, it was demonstrated to me that you could make a reasonably accurate RF millivoltmeter out of an analog Simpson 260 voltmeter.

(Without modifying it, of course.)

Mo found this at some neighborhood junk sale and bought it for me. $3
or something.

https://www.dropbox.com/s/tz0panr4f1nlqer/RF_Ammeter.JPG?raw=1

It's probably a thermocouple.

How did that Simpson thing work?

I have some Spice models of an RF detector using an SMS7621
low-barrier schottly. It works around 50 mV RMS.


Wow John, that is a nice looking vintage analog meter!!
Nice find.

From an old, faded, memory...
The trick with the Simpson 260 was you connect the red lead to the uAmp input, and put the black lead in the Common (-). But, you don't connect the black lead to your circuit. What you do instead is loosely drape is across the unit under test, and then probe with just the red lead. Essentially, the meter is reading rectified RF.

The old VOMs used copper oxide rectifiers! Before silicon was
discovered.

The above description about just leaving one lead disconnected but loosely coiled-up in close proximity to the circuit under test is correct. But I might have the meter connections wrong. At this point, I would probably have to have a Simpson 260 in front of me to figure it out -- but it's basically along those lines. The meter's not designed to do it, but it will. And it more accurate than you would think, for such a weird, oddball approach. But hey, if you need a relative RF millivoltmeter and don't have one handy.....

You can also fix a Quintron QBT-250 paging transmitter with a 9-iron golf club, and certain old Ford pickup trucks with garlic bread. (Maybe I shared those stories here before?, but they're a little off-topic in any case.)

It's a sin to waste garlic bread.

You got a use in mind for that meter? (Just wondering)

No, it's just on a shelf for admiration.


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[Clifford Heath]

On 7/4/20 11:03 pm, George Herold wrote:

On Monday, April 6, 2020 at 7:32:00 PM UTC-4, Clifford Heath wrote:
George, I'm trying to understand this. What do you need 20 zeners for?

FWIW, I'm currently making a broadband noise source for testing filters
up to 1.5GHz. Still scratching around for the best source to put before
a string of ERA-3 MMIC amplifiers.

The BFR93A data sheet says abs max Vbe is 2V, but it doesn't zener at
5V. Although avalanche zeners produce much more noise, I'd rather not
boost my 5V supply (though I might need to). I wonder how much reverse
current a microwave Shottky diodes (say HSMS-286) would survive. Abs max
peak reverse voltage is 4V, so it might withstand 5V anyway...

Any better suggestions for a device I might have in the drawer already?

Clifford Heath.

Arghh! typo. sorry a 20V zener! Run near the knee you get these big
avalanche spikes, with ~1us rise/ fall times.

1 GHz noise sounds hard.

It doesn't look hard, a lot of people have done it. I've built the
amplifier, I just need a noise diode that will give me me 20db ENR or
better to feed into it. The only zeners I can easily get are 1W and
above, so have big capacitance, only good for audio. I need a physically
small device that will have low capacitance. That's why I started with a
GHz transistor.

I've thought about trying an HSMS286 series microwave Schottky, which
break down at about 4V (if the data sheet is to be believed). The
literature says that devices with a sharp breakdown curve last longer -
a soft curve indicates partial breakdown at the edge of the die, which
leads to early failure. So I'll need to evaluate the sharpness of the
breakdown to see if it's likely to last a while.

But first I'm going to crank the Vbe on this BFR93A to see where it
really breaks down. If I can get it to zener on w 12v supply, that'll
probably do.

I'd prefer not to need a higher supply voltage, but I might have to
compromise on that. Pure zener noise (low voltage) is much lower
amplitude than higher-voltage devices which have avalanche multiplication.

> What about a spectrum analyzer and tracking generator?

The only generators I have are several home made ones, and I don't have
my 100-4000MHz generator complete yet. The closest thing I have to a
spectrum analyser is a couple of SDRs.

Hmm, I should finish that generator

Clifford Heath

 

[Gerhard Hoffmann]

Am 08.04.20 um 03:05 schrieb Clifford Heath:

On 7/4/20 11:03 pm, George Herold wrote:
On Monday, April 6, 2020 at 7:32:00 PM UTC-4, Clifford Heath wrote:
George, I'm trying to understand this. What do you need 20 zeners for?

FWIW, I'm currently making a broadband noise source for testing filters
up to 1.5GHz. Still scratching around for the best source to put before
a string of ERA-3 MMIC amplifiers.

The BFR93A data sheet says abs max Vbe is 2V, but it doesn't zener at
5V. Although avalanche zeners produce much more noise, I'd rather not
boost my 5V supply (though I might need to). I wonder how much reverse
current a microwave Shottky diodes (say HSMS-286) would survive. Abs max
peak reverse voltage is 4V, so it might withstand 5V anyway...

Any better suggestions for a device I might have in the drawer already?

Clifford Heath.

Arghh!  typo. sorry a 20V zener!  Run near the knee you get these big
avalanche spikes, with ~1us rise/ fall times.

1 GHz noise sounds hard.

It doesn't look hard, a lot of people have done it. I've built the
amplifier, I just need a noise diode that will give me me 20db ENR or
better to feed into it. The only zeners I can easily get are 1W and
above, so have big capacitance, only good for audio. I need a physically
small device that will have low capacitance. That's why I started with a
GHz transistor.

I've thought about trying an HSMS286 series microwave Schottky, which
break down at about 4V (if the data sheet is to be believed). The
literature says that devices with a sharp breakdown curve last longer -
a soft curve indicates partial breakdown at the edge of the die, which
leads to early failure. So I'll need to evaluate the sharpness of the
breakdown to see if it's likely to last a while.

But first I'm going to crank the Vbe on this BFR93A to see where it
really breaks down. If I can get it to zener on w 12v supply, that'll
probably do.

I'd prefer not to need a higher supply voltage, but I might have to
compromise on that. Pure zener noise (low voltage) is much lower
amplitude than higher-voltage devices which have avalanche multiplication.

I have done that thing with the BFR93A many years ago and at the
Weinheim VHF ham meeting there was someone who measured these
homebrew things. I got 22 dB noise or sth. like that and it was
flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz. I used a CCS just like the charger
of my HP-35

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor. Yes, they
have a noise figure of a few dB but you know it and it's flat.
More flat than a BFR93A breakdown, and much better than any
Z-diode with its huge capacitance. 1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

The ERAs may also need more than 5V if you include the bias resistor?

What about a spectrum analyzer and tracking generator?

The only generators I have are several home made ones, and I don't have
my 100-4000MHz generator complete yet. The closest thing I have to a
spectrum analyser is a couple of SDRs.

Hmm, I should finish that generator

good idea!

cheers, Gerhard

 

[Clifford Heath]

On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:

Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?

> The ERAs may also need more than 5V if you include the bias resistor?

I figure they want 40mA, and based on data sheet values for other
voltages (from 7V up) it looks like they settle around 3.3V, so I
calculated a bias resistor for 5V and was pretty close. I guess it will
be more temperature sensitive than it would be with a higher supply voltage.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Anyhow, I'll raise it to 12V and try that before adding more ERA-3s and
switching to a resistor.

Many thanks for your advice,

Clifford Heath.

 

[Clifford Heath]

On 8/4/20 1:51 pm, Clifford Heath wrote:

On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?

Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
<https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0>

I guess I need to reduce gain somewhat?

I haven't investigated the spectrum yet.

>> The ERAs may also need more than 5V if you include the bias resistor?

They're now biassed as recommended from 12V, with 240R.

Replacing the BFR93A with 100R drops the noise to under 1mV - still
visible, but will need more amplification (no oscillation visible). One
more stage might be enough gain.

--
Clifford Heath

I figure they want 40mA, and based on data sheet values for other
voltages (from 7V up) it looks like they settle around 3.3V, so I
calculated a bias resistor for 5V and was pretty close. I guess it will
be more temperature sensitive than it would be with a higher supply
voltage.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Anyhow, I'll raise it to 12V and try that before adding more ERA-3s and
switching to a resistor.

Many thanks for your advice,

Clifford Heath.

 

[Lasse Langwadt Christense]

onsdag den 8. april 2020 kl. 10.31.41 UTC+2 skrev Clifford Heath:

On 8/4/20 1:51 pm, Clifford Heath wrote:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?


Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0

I seem to remember something about using two zeners and subtracting
them to get symmetric output

 

[Gerhard Hoffmann]

Am 08.04.20 um 05:51 schrieb Clifford Heath:

On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?

No idea. But a customer of mine had some problems with the
autocorrelation of pseudo-noise for ranging purposes. The measured
S-curve did not look so triangle-like as it was supposed to be.

The ERAs may also need more than 5V if you include the bias resistor?

I figure they want 40mA, and based on data sheet values for other
voltages (from 7V up) it looks like they settle around 3.3V, so I
calculated a bias resistor for 5V and was pretty close. I guess it will
be more temperature sensitive than it would be with a higher supply
voltage.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Pure resistance from high VCC is probably easier for the beginning.
Use smaller R in series, it's a matter of C over the resistor vs. C from
resistor to GND. You get predictability and spend VCC.

I had quite good success with home etching, it helps during the lock-down.
This here is from the same double-sided Eurocard (100 * 160 mm); the
foil has been made with a Kyocera color laser on Bungard pre-sensitized
material. I can etch what I can print. The A3 OKI color laser I used to
have previously was clearly better, but the machine was a royal pain.
The bottom side is GND and unetched. No ado with alignment.

If it needs to be really precise, i.e. microwave filters, I go to a
nearby print shop and have them make an offset film from my .pdf file.
That increases the wall time from .pdf to soldering from 1h:30 to 3h and
costs €8.

<
https://www.flickr.com/photos/137684711@N07/49748806273/in/dateposted-public/
<
https://www.flickr.com/photos/137684711@N07/49749272951/in/dateposted-public/
>

Both boards were made on one Eurocard.

Cheers, Gerhard

 

[George Herold]

On Wednesday, April 8, 2020 at 8:31:51 AM UTC-4, Lasse Langwadt Christensen wrote:

onsdag den 8. april 2020 kl. 10.31.41 UTC+2 skrev Clifford Heath:
On 8/4/20 1:51 pm, Clifford Heath wrote:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?


Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0


I seem to remember something about using two zeners and subtracting
them to get symmetric output

Yeah I've done that in the audio. It helps some, but unless you match the
zeners... Well symmetrical noise is over rated... no one cares much about
low frequencies.

George H.

 

[George Herold]

On Wednesday, April 8, 2020 at 4:31:41 AM UTC-4, Clifford Heath wrote:

On 8/4/20 1:51 pm, Clifford Heath wrote:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?


Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0

I guess I need to reduce gain somewhat?

Hi Cliff, I'm not at all a HF noise guy.
(So listen to Gerhard and not me .:^)
But the non-symmetry is typical. You might want to look at the
signal as a function of the bias current. That is your 'best' knob.

I haven't investigated the spectrum yet.

How fast a 'scope do you have? Have you tried triggering on the
noise (pulses) With a DSO you can average the pulses/ noise peaks
and get a pseudo-spectrum... at least a guess at the max freq.

Good luck,
George H.

The ERAs may also need more than 5V if you include the bias resistor?

They're now biassed as recommended from 12V, with 240R.

Replacing the BFR93A with 100R drops the noise to under 1mV - still
visible, but will need more amplification (no oscillation visible). One
more stage might be enough gain.

--
Clifford Heath

I figure they want 40mA, and based on data sheet values for other
voltages (from 7V up) it looks like they settle around 3.3V, so I
calculated a bias resistor for 5V and was pretty close. I guess it will
be more temperature sensitive than it would be with a higher supply
voltage.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Anyhow, I'll raise it to 12V and try that before adding more ERA-3s and
switching to a resistor.

Many thanks for your advice,

Clifford Heath.

 

On Wed, 8 Apr 2020 05:31:45 -0700 (PDT), Lasse Langwadt Christensen
<langwadt@fonz.dk> wrote:

onsdag den 8. april 2020 kl. 10.31.41 UTC+2 skrev Clifford Heath:
On 8/4/20 1:51 pm, Clifford Heath wrote:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 03:05 schrieb Clifford Heath:
I have done that thing with the BFR93A many years ago

Oh, thanks, nice to know I'm on a right track. I basically just guessed
that the BFR93A should work, not that I saw it elsewhere.

flat enough for tuning filters. Not good enough to measure
noise figure, both for amplitude and source impedance.
The breakdown voltage of the BFR93A was 5 or 6V IIRC.
I think it went to ~ 2 GHz.

That would be good enough for my first try!

I used a CCS just like the charger
of my HP-35

I just thought I'd use a resistor, since I have a known supply voltage.
Not a good idea?

Today I would not use a "Zener" source. If you have them, spend
2 more ERAs and amplify the noise of a 50 Ohm resistor.

Well now it's interesting, but I have exactly the right amount of room
left on this proto-PCB for two more, and I have 20 ERA-3's.

It's on a grid-punched board with assorted top-side copper strips. I
used sticky-backed copper foil to make a ground plane on the back. It
doesn't seem such a bad way to prototype RF circuits, given that my
toner-transfer process is in recess (the new printer creates porous
blacks that don't etch well).

Yes, they
have a noise figure of a few dB but you know it and it's flat.

Assuming the ERA-3 or the layout doesn't filter it, it should stay flat.

1K may produce more voltage noise,
but @ 50 Ohms you know that the ERAs will behave.

Well I could try 220R and just check it's not oscillating I guess.
Otherwise 50.

The problem with noise filter measurements is that the bandwidth
of the noise is so large and if you want to see the stop band
you need lots of power. Thus, you will have large signal conditions
at the output of your noise generator. Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.

Right. Not sure what that will do to the spectrum though?


Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0


I seem to remember something about using two zeners and subtracting
them to get symmetric output

That was also done with gas discharge noise tubes, essentially
thyratrons. There was also a trick of attaching a magnet to one tube
to make the noise symmetric. I haven't a clue why that would work.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

On Tue, 7 Apr 2020 06:25:39 -0700 (PDT), George Herold
<ggherold@gmail.com> wrote:

On Monday, April 6, 2020 at 2:09:59 PM UTC-4, dagmarg...@yahoo.com wrote:
On Saturday, April 4, 2020 at 8:49:22 PM UTC-4, pcdh...@gmail.com wrote:
Following up on blocher's sterling work,(*)

Many of us use parts off-label, often very successfully. A few examples:

SAV-551+ pHEMTs make very good wideband bootstraps. Their f_max is around 12 GHz, but they're amazingly stable.

74HC4352s make good flying-capacitor diff amp front ends.

TMUX1511s make very nice analogue lock-ins--their Coff*Ron FOM is almost in a class with relays, but 1E5 times faster.

Zero-ohm jumpers have about the right resistance to stabilize LDO regulators with ceramic output caps. (It's good to be able to disconnect the supplies during bring-up, and putting the jumper between the reg and the output cap has this additional benefit.

Your faves?

Cheers

Phil Hobbs
(*) who may be bulegoge's good twin, given the similarity of their emails

o CMOS gates for power supplies, precision voltage switching,
class-C r.f. power amps.

o Various semiconductor junctions as varicaps.

o I'm still searching for a way to use crappy ceramic caps as
varactors. Tuning a WWVB loop antenna might finally be that chance.

Hi James, I spent part of the weekend trying to think about
crappy ceramic caps as part of a parametric oscillator.
(After re-reading Pippard's "Physics of Vibration"
the oscillator seems like an easier first step.)
But I have no idea how to couple two different signals
into/ through the capacitor. Well maybe a bridge.

Do you have any clever ideas?

George H.


o LEDs as detectors.

Cheers,
James Arthur

I have thought about making an oscillator from a ceramic cap and
passives, but if it could be done probably someone would have done it.
It could certainly be done with a pump oscillator and some diodes and
stuff.

One use for nonlinear ceramic caps is to make high voltage shock
lines. Put a slow pulse into a shock line and it gets faster as it
propagates.





--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

On Tuesday, April 7, 2020 at 9:25:45 AM UTC-4, George Herold wrote:

On Monday, April 6, 2020 at 2:09:59 PM UTC-4, dagmarg...@yahoo.com wrote:
On Saturday, April 4, 2020 at 8:49:22 PM UTC-4, pcdh...@gmail.com wrote:
Following up on blocher's sterling work,(*)

Many of us use parts off-label, often very successfully. A few examples:

SAV-551+ pHEMTs make very good wideband bootstraps. Their f_max is around 12 GHz, but they're amazingly stable.

74HC4352s make good flying-capacitor diff amp front ends.

TMUX1511s make very nice analogue lock-ins--their Coff*Ron FOM is almost in a class with relays, but 1E5 times faster.

Zero-ohm jumpers have about the right resistance to stabilize LDO regulators with ceramic output caps. (It's good to be able to disconnect the supplies during bring-up, and putting the jumper between the reg and the output cap has this additional benefit.

Your faves?

Cheers

Phil Hobbs
(*) who may be bulegoge's good twin, given the similarity of their emails

o CMOS gates for power supplies, precision voltage switching,
class-C r.f. power amps.

o Various semiconductor junctions as varicaps.

o I'm still searching for a way to use crappy ceramic caps as
varactors. Tuning a WWVB loop antenna might finally be that chance.

Hi James, I spent part of the weekend trying to think about
crappy ceramic caps as part of a parametric oscillator.
(After re-reading Pippard's "Physics of Vibration"
the oscillator seems like an easier first step.)
But I have no idea how to couple two different signals
into/ through the capacitor. Well maybe a bridge.

Do you have any clever ideas?

George H.

Sorry George, parametric amplifiers only momentarily grabbed my
attention ages ago (when I was trying to generate 1GHz from HCMOS).

I've always found that, except maybe at microwave frequencies that keep
getting higher and higher, active circuits won out.

Cheers,
James Arthur

 

20 ma LED current source chips make nice MMIC current limiters in place of resistors.

20 ma Led current source chips plus optoisolators make nice protected interlock inputs to 48 or 90VDC depending on the chip, for a whopping 63 cents.

Steve

 

[John Larkin]

On Wed, 8 Apr 2020 10:26:57 -0700 (PDT), sroberts6328@gmail.com wrote:

>20 ma LED current source chips make nice MMIC current limiters in place of resistors.

Newer MMICs have internal current control loops, so can be biased from
a voltage source through an inductor. The control loops have various
weird low frequency effects. My new GHz o/e converter circuit will
only work with old-style MMICs.

20 ma Led current source chips plus optoisolators make nice protected interlock inputs to 48 or 90VDC depending on the chip, for a whopping 63 cents.

The LED current limiters are cool, but most need a lot of voltage
drop.

There are some self-protecting SSRs that have low drop, basically pure
ohmic up to some hundreds of mA.

Steve
--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[George Herold]

On Wednesday, April 8, 2020 at 12:02:54 PM UTC-4, dagmarg...@yahoo.com wrote:

On Tuesday, April 7, 2020 at 9:25:45 AM UTC-4, George Herold wrote:
On Monday, April 6, 2020 at 2:09:59 PM UTC-4, dagmarg...@yahoo.com wrote:
On Saturday, April 4, 2020 at 8:49:22 PM UTC-4, pcdh...@gmail.com wrote:
Following up on blocher's sterling work,(*)

Many of us use parts off-label, often very successfully. A few examples:

SAV-551+ pHEMTs make very good wideband bootstraps. Their f_max is around 12 GHz, but they're amazingly stable.

74HC4352s make good flying-capacitor diff amp front ends.

TMUX1511s make very nice analogue lock-ins--their Coff*Ron FOM is almost in a class with relays, but 1E5 times faster.

Zero-ohm jumpers have about the right resistance to stabilize LDO regulators with ceramic output caps. (It's good to be able to disconnect the supplies during bring-up, and putting the jumper between the reg and the output cap has this additional benefit.

Your faves?

Cheers

Phil Hobbs
(*) who may be bulegoge's good twin, given the similarity of their emails

o CMOS gates for power supplies, precision voltage switching,
class-C r.f. power amps.

o Various semiconductor junctions as varicaps.

o I'm still searching for a way to use crappy ceramic caps as
varactors. Tuning a WWVB loop antenna might finally be that chance..

Hi James, I spent part of the weekend trying to think about
crappy ceramic caps as part of a parametric oscillator.
(After re-reading Pippard's "Physics of Vibration"
the oscillator seems like an easier first step.)
But I have no idea how to couple two different signals
into/ through the capacitor. Well maybe a bridge.

Do you have any clever ideas?

George H.

Sorry George, parametric amplifiers only momentarily grabbed my
attention ages ago (when I was trying to generate 1GHz from HCMOS).

I've always found that, except maybe at microwave frequencies that keep
getting higher and higher, active circuits won out.

No worries, I'm mostly interested for pedantic reasons.
The only 'good' physics demo's of this are mechanical, AFAIK.
(oh and mag amps...)

At some point I suffer from not understanding transformers well enough.
(like where the LF and HF corners are.)

George H.

Cheers,
James Arthur

 

[Clifford Heath]

On 9/4/20 12:34 am, George Herold wrote:

On Wednesday, April 8, 2020 at 4:31:41 AM UTC-4, Clifford Heath wrote:
On 8/4/20 1:51 pm, Clifford Heath wrote:
Ok with 12V, 2.6mA through the BFR93A's b-e, it zeners at 5.8V.

Two stages of amplification is possibly too much - the noise is
assymetrical around 0V, see this scope photo:
https://www.dropbox.com/s/bmjnhm988160zjc/NoiseScope%20BFR93A%202.5mA%202xERA3.jpg?dl=0

I guess I need to reduce gain somewhat?
Hi Cliff, I'm not at all a HF noise guy.
(So listen to Gerhard and not me .:^)
But the non-symmetry is typical. You might want to look at the
signal as a function of the bias current. That is your 'best' knob.

Oh, good to know, it didn't seem right that the ERA-3's could be
producing this on overload. it makes sens that it's typical zener behaviour.

I haven't investigated the spectrum yet.
How fast a 'scope do you have? Have you tried triggering on the
noise (pulses) With a DSO you can average the pulses/ noise peaks
and get a pseudo-spectrum... at least a guess at the max freq.

500MHz Tek 7904, with all the nice plug-in's . No, I haven't tried
triggering on it. But the noise looks almost the same regardless of the
timebase, so that's a good sign.

But I also have a HackRF and I'm not ashamed to use it as a spectrum
analyser.

Clifford Heath.

 

On Wednesday, April 8, 2020 at 7:20:56 PM UTC-4, Clifford Heath wrote:

On 8/4/20 10:22 pm, Gerhard Hoffmann wrote:
Am 08.04.20 um 05:51 schrieb Clifford Heath:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.
Right. Not sure what that will do to the spectrum though?

No idea. But a customer of mine had some problems with the
autocorrelation of pseudo-noise for ranging purposes. The measured > S-curve did not look so triangle-like as it was supposed to be.

We'll see how the resistor noise looks. I'm just worried that much
amplification will show the noise of the first ERA3, or power supply or
ambient noise that I haven't filtered or screened out.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Pure resistance from high VCC is probably easier for the beginning.
Use smaller R in series, it's a matter of C over the resistor vs. C from
resistor to GND. You get predictability and spend VCC.

I could easily put a tiny wire loop inductor in series after the R, at
risk of coupling to the next stage and singing. That should help at GHz.

I had quite good success with home etching, it helps during the lock-down.

I had excellent success with toner transfer, handling 0.6mm pin-pitch
parts easily, when I was using my ancient HP LJ6L. That finally died,
and the replacement HP2055DN produces a porous checker-plate that's
visible only under magnification. I haven't worked out how to stop it
doing that, but the etchant gets through the tiny gaps, even if I use
the toner transfer foil. Perhaps I'll have another go at it, I've only
tried etching once (printing many times, but only one seemed good enough
to etch.

I haven't been able to get toner transfer working since I changed
toner carts. But I've dabbled with two promising variations.

1) If you lacquer-coat the PCB *then* transfer the toner, the
lacquer fills in the toner's pores. The lacquer itself, meanwhile,
can be easily removed with alcohol, but the toner's pores seem to
stay filled.

The toner image is printed on the paper you peel off the back of
adhesive labels, or in my case, backing peeled off adhesive
shelf-liner paper from the one-dollar store.

2) A variation of the above is to spray the lacquer, then transfer
the toner almost immediately, before the lacquer dries. The lacquer
itself grabs the toner.
https://www.youtube.com/watch?v=0dpCi9kgpuw

3) A third method is to zap the toner with acetone/alcohol mix to
make the toner tacky, then stick it to the board. That always
smears the traces when I try it. "Cold Toner Transfer"
e.g. https://www.youtube.com/watch?v=cVhSCEPINpM

Cheers,
James Arthur

 

[Clifford Heath]

On 8/4/20 10:22 pm, Gerhard Hoffmann wrote:

Am 08.04.20 um 05:51 schrieb Clifford Heath:
On 8/4/20 12:39 pm, Gerhard Hoffmann wrote:
Getting closer to saturation
than 20 dB will damage the crest ratio / noise statistics.
Right. Not sure what that will do to the spectrum though?

No idea. But a customer of mine had some problems with the
autocorrelation of pseudo-noise for ranging purposes. The measured > S-curve did not look so triangle-like as it was supposed to be.

We'll see how the resistor noise looks. I'm just worried that much
amplification will show the noise of the first ERA3, or power supply or
ambient noise that I haven't filtered or screened out.

I'm using 0805 resistors but not series inductors, for now. Will that
affect flatness (rather than just lose some gain)?

Pure resistance from high VCC is probably easier for the beginning.
Use smaller R in series, it's a matter of C over the resistor vs. C from
resistor to GND. You get predictability and spend VCC.

I could easily put a tiny wire loop inductor in series after the R, at
risk of coupling to the next stage and singing. That should help at GHz.

> I had quite good success with home etching, it helps during the lock-down.

I had excellent success with toner transfer, handling 0.6mm pin-pitch
parts easily, when I was using my ancient HP LJ6L. That finally died,
and the replacement HP2055DN produces a porous checker-plate that's
visible only under magnification. I haven't worked out how to stop it
doing that, but the etchant gets through the tiny gaps, even if I use
the toner transfer foil. Perhaps I'll have another go at it, I've only
tried etching once (printing many times, but only one seemed good enough
to etch.

I leave the back side as solid copper, just fitting jumper wires if I
need a trace. If I can't route mostly single-sided, I don't make it myself.

This here is from the same double-sided Eurocard (100 * 160 mm); the
foil has been made with a Kyocera color laser on Bungard pre-sensitized
material. I can etch what I can print. The A3 OKI color laser I used to
have previously was clearly better, but the machine was a royal pain.

> The bottom side is GND and unetched. No ado with alignment.