As Adrian and Bob pointed out, W. J. Riley's site has all the information
needed to make the higher cost TPLL version he did including his PCBs.

The bigger mystery seems to be how easy a TPLL can be built without loosing
performance.
One version of the TPLL tester only needs 8 circuit parts plus + Power
supplies etc,
These are all clearly specified on the bottom block diagram, dated June 7,
2010 at
http://www.ke5fx.com/tpll.htm
And even that complete working, simple version, is good enough that the
performance is still mostly limited by the HP10811 Reference Osc and not the
TPLL circuit.

BOM for "Extra simple TPLL"
Nothing is critical, performance is determined by the Ref Osc.

1. Phase detector  =  SYPD-1
2. 100KHz  LPF = two each 220 ohm in series and two 0.0047uf caps to gnd
3. Amp = OP-27 Pin 3 is input, pin 6 is output
4. Amp feedback gain = + 300 set using a 30K feedback and 100 Ohm resistor
   to gnd
5. 20 Hz LPF =  8K ohm series R and 1uf to gnd
6. Ref Osc = HP10811
7. A slow mv DVM and/or a 16 bit ADC sampling at 100Hz or more.
8. misc connectors, PS, S/W, etc.

The configuration of the TPLL 1.0 that John tested, uses a 3dB and 5dB pad
for isolation, (no real need for Osc buffers),
and has a higher closed loop PLL bandwidth using an op37 with more gain, (a
tighter TPLL)
This can be seen by clicking on the underlined "Here" of John's report next
to Fig 1.7 at
"Warren's annotated block diagram can be seen HERE."

ws

Hi Tom,

Bill has actually published detailed schematics etc here:
http://www.stable32.com/A%2010%20MHz%20OCVCXO%20and%20PLL%20Module.pdf

Btw. what do you think about his small DMTD system?
http://www.wriley.com/A%20Small%20DMTD%20System.pdf

Adrian

Tom Van Baak schrieb:

On Wed, Jan 30, 2013 at 6:16 PM, Richard (Rick) Karlquist
richard@karlquist.com wrote:

For once the "best" is also cheap:  Batteries.

But not all batteries are the same.  You want one with low internal
resistance, so a lead acid flooded battery will be the best.

Chris Albertson
Redondo Beach, California

On 1/30/13 8:28 PM, Chris Albertson wrote:

Most NiCd have very low internal resistance.. much lower than lead acid.

But aside from batteries, Rick's question is interesting, and I'd turn
it around a bit.. What "off the shelf catalog product" has the lowest
noise?  Do you go get a linear regulated supply from someone like Acopian?

And even more interesting.. if you're concerned about efficiency and
want to use a DC/DC switcher to convert some ratty DC supply (12V
vehicle power, or a solar panel, for instance) to something really
quiet, what's the best strategy, using off the shelf bricks and modules.

That is, I'm sure someone could do a fabulous job with a box full of Ls
and Cs and discrete components and a couple months to design, prototype,
and build.. but if someone came to you and said, we want a mobile
microwave system to study propagation in 3 weeks because they just got
permission to go up on some mountain.  Something like a CW carrier at
tens of GHz multiplied up from your very quiet oscillator, and they're
going to look at turbulence and scintillation in the path, so time
scales of milliseconds to 1000s of seconds are important.

What would you order from Newark, Allied, mouser, etc. (assuming you
have your OCXOs and such sitting around).

Lester Veenstra wrote:

Depends on the variety of 723 some are noisier than others.
Some use an internal zener reference, some use a bandgap reference.
The original used a zener reference.

Bruce

On Wed, Jan 30, 2013 at 9:05 PM, Jim Lux jimlux@earthlink.net wrote:

Really?  I'm talking about flooded cells not gel.  I think large
automotive engine start battery might only have 20 milliohms
resistance.  But the only NiCd I know about are the AA sized ones.
Not fair to compare.  Perhaps a size size NiCd would be even better.
I don't know.

Got a reference, Google did not turn up much.  I'd expect that some
place there might be a table.

I've powered resonably large transmitters with a bank of golf cart
batteries.  Not because I needed the low noise but because this was on
a sail boat in the ocean.
Chris Albertson
Redondo Beach, California

Hi Rick;
Starting with a hp supply this finesse regulator will really clean up the output. http://www.wenzel.com/documents/finesse.html
Best Wishes;
Thomas Knox

Content-Type: text/plain; charset=ISO-8859-1

In message BAY162-W38A3AC22812AC79A0FBFC6DF1D0@phx.gbl, Tom Knox writes:

I've played with that, and it can do impressive things.

However, to big caveats:

1. It's very temperature sensitive so on average you only get tens
   of dB, not hundred dB damping.

2. Noise which comes in from the load gets amplified.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Poul-Henning Kamp wrote:

With some care its possible to make the emitter current of the shunt
transistor approximately PTAT so that, at least for small signals the
temperature dependence of the rejection is reduced significantly.

Its also possible to build a feedback style shunt regulator that has
considerably higher supply rejection than the Wenzel circuit using an
opamp and a shunt transistor together with a small resistance in series
with one of the supply leads.

Bruce

On 1/31/2013 12:20 AM, Bruce Griffiths wrote:

Sorry, what does PTAT mean? I'm not familiar with that term.

The Wenzel article gives three circuit topologies. The third shows an
opamp feeding a 2N4401 across 0.05 ohm in the power path. Sounds like
what you are describing. Did you miss that one or are you describing
something different?

Lester wrote:

If you are willing to design your own regulator using a 723, you may
as well use a few more parts to get a much better result.  Neither
the internal reference nor the internal error amp in a 723 is
anywhere near state of the art today with respect to noise, tempco,
or speed.  Using readily available buried zener references and
low-noise, high-speed op amps (or even a few discrete transistors),
you can do several orders of magnitude better than a 723 in all
respects.  The web is overflowing with designs (though not all of the
circuits you find perform as advertised, so evaluate them with a
critical eye and use your own sound judgment).

But wasn't the original question what is available off-the-shelf?

Best regards,

Charles

Am 31.01.2013 03:16, schrieb Richard (Rick) Karlquist:

The E5052B signal source analyzer has low noise power
and control voltage supplies for just that purpose.

E5052 are not that exotic any more. At a customer's
I have seen 5 of them being sent on one day for calibration
and that did not create any bottleneck. :-)

regards, Gerhard

Am 31.01.2013 05:28, schrieb Chris Albertson:

Fred Walls & Co have done tests on batteries. The article is
somewhere on the NIST time/freq server.

regards, Gerhard

Hi

To answer the original question - Power Design makes some pretty quiet bench
supplies. If you are doing low noise testing, batteries often will let you
get rid of one more ground loop. Even well built power supplies are not as
well line isolated as a battery.

Bob

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Charles P. Steinmetz
Sent: Thursday, January 31, 2013 5:33 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Low noise power supplies?

Lester wrote:

If you are willing to design your own regulator using a 723, you may
as well use a few more parts to get a much better result.  Neither
the internal reference nor the internal error amp in a 723 is
anywhere near state of the art today with respect to noise, tempco,
or speed.  Using readily available buried zener references and
low-noise, high-speed op amps (or even a few discrete transistors),
you can do several orders of magnitude better than a 723 in all
respects.  The web is overflowing with designs (though not all of the
circuits you find perform as advertised, so evaluate them with a
critical eye and use your own sound judgment).

But wasn't the original question what is available off-the-shelf?

Best regards,

Charles

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

Rick,

the E3610A is really clean.

If that isn't good enough, there are some audio-related circuits like
the ALWSR.
http://www.andrewweekes.talktalk.net/Manuals/ALWSR_rev2.9_Iss005s.pdf
It's based on a Walt Jung design (Analog Devices).

Here's the man himself with tons of valuable information about ultra low
PS design:
http://waltjung.org/Library.html

Adrian

Richard (Rick) Karlquist schrieb:

Ah, sorry, I didn't see that 2nd PDF! I see the design is significantly different than the TPLL that you made. Any idea why?

We've see http://www.thegleam.com/ke5fx/tpll/TightPLL_BB.gif before and that's partly why I wrote what I wrote. It's high level and pretty far from a schematic/BOM. However, your comments 1 - 8 begin to address that. Thanks much.

You've also talked about TPLL 1.0 and 2.0, as if we should all know what they are; is there a comprehensive description of those; something detailed enough that a random time nut could build one?

Thanks,
/tvb

Hi Rick,

I'll answer the question with anther question -- how does one properly measure power supply noise? Does it boil down to a single number, a couple of key numbers, or is it a plot, or several plots?

I ask because without some sort of standard test and reporting method it just becomes a word game. Like, "use batteries because they are better";  "use my design because it is quiet"; "this has lower noise than that". RMS AC ripple or nV/rtHz sounds like a good start, but I'm wondering if there's something more complete.

I was never quite satisfied with the outcome of comparing a half dozen power supplies this way:
http://www.leapsecond.com/pages/tbolt/noise.htm

We have rigorous ways to compare and report oscillator performance; both as numbers and as plots. Is there something equivalent for power supplies?

Thanks,
/tvb

On 31 January 2013 04:28, Chris Albertson albertson.chris@gmail.com wrote:

My first thought was that a battery was best. Obviously batteries will
have a long term drop in voltage as the battery gets more depleted,
but is there any short term noise from a battery? There is going to be
thermal noise, but I wonder if some types of batteries are better from
a noise perspective than others?

Perhaps a number of batteries having the same nomilal voltage,
capacity and internal resistance are all equal, but perhaps some are
more equal than others!

Dave

Dave

Hi

I think the comparison of PSD on a power supply to phase noise and phase
noise plots is a pretty good one in this case.

For most applications nV/sqrt(Hz) is a pretty good way to check things out
on a supply or regulator. It's not quite the same thing as dbc / sqrt(Hz)
but it conveys the same sort of information. Unless you have a very
sensitive part, anything below 10 nV/ sqrt(Hz) is likely to be "a very quiet
supply". The main limit you hit is in the sub 100 Hz region where you likely
see things like popcorn noise.

AC ripple is no different than spurs in phase noise testing. You sometimes
see people who ignore them when plotting phase noise. I'd suggest that they
are an important part of characterizing a power supply.

As with phase noise, frequency ranges are going to be application dependant.
I may not care about 0.1 to 10 Hz phase noise for project A. It may be the
only thing I care about for project B. Same thing with power supplies.

With a good enough voltmeter you could carry the analogy one step further
and compute an ADEV like number on the output voltage. I suspect that's
carrying things a bit far.

Bob

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Tom Van Baak
Sent: Thursday, January 31, 2013 9:43 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Low noise power supplies?

Hi Rick,

I'll answer the question with anther question -- how does one properly
measure power supply noise? Does it boil down to a single number, a couple
of key numbers, or is it a plot, or several plots?

I ask because without some sort of standard test and reporting method it
just becomes a word game. Like, "use batteries because they are better";
"use my design because it is quiet"; "this has lower noise than that". RMS
AC ripple or nV/rtHz sounds like a good start, but I'm wondering if there's
something more complete.

I was never quite satisfied with the outcome of comparing a half dozen power
supplies this way:
http://www.leapsecond.com/pages/tbolt/noise.htm

We have rigorous ways to compare and report oscillator performance; both as
numbers and as plots. Is there something equivalent for power supplies?

Thanks,
/tvb

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

And this (very interesting) thread brings up the question of measurement
methods.  Some time ago I searched around and didn't find much on a
standard way to measure noise on low voltage DC supplies.

John

On 1/31/2013 11:36 AM, Bob Camp wrote:

On 31.01.2013 09:05, Adrian wrote:

The discussion is interesting, but the original question did ask about
off the shelf so I have been watching for any mention of HP/Agilent
precision power supplies.  The E3610A specification says 200 uVrms, 2mv
pp ripple and noise over the range 20Hz to 20 MHz.  The now obsolete
HP6114A spec is much better at 40uVrms, 100uV pp up to 20MHz.  The 6114A
is no longer available (except on eBay), but there should be something
equivalent or better.

I discovered the 6114A one day when I needed a nice stable voltage
source to check some data acquisition equipment.  My first pick was a
gel-cel,  but with a 34401 DMM you can watch those self-discharge.  A
regular bench supply kind of wanders around.  But after a few minute
warm-up the 6114A that I found sitting on a storage shelf was nice and
stable.  I mentioned the specification for noise because I have observed
its stability, but have not made any measurements on its noise.

Gary

On 31 January 2013 17:11, John Ackermann N8UR jra@febo.com wrote:

Did you try the volt-nuts mailing list?

Dave

There are at least two methods.

First, the broadband or time domain method that uses a low noise
amplifier and oscilloscope, and specifies noise voltage within a certain
bandwidth. For details check application notes from LTC and Analog Devices.

The narrowband or frequency domain method specifies noise voltage within
a normalized bandwidth (nV/sqrt(Hz)) at a specific frequency.
A typical test setup is included in the HP 3048A phase noise measurement
system as so-called base-band noise measurement. For higher noise
voltages, the 3561A dynamic signal analyzer alone is used with a DC
blocking capacitor. For lower noise voltages, the signal is first
amplified by the LNA that is part of the 11848A phase noise test set.
For details please see the HP 3048A manual that can be downloaded from
http://www.hparchive.com/hp_equipment.htm
Note the Opt K23 DC blocking filter. There is an insertion loss of 6 dB
to be entered in the 3048A software when using the DC block in
conjunction with the 11848A.
Obviously, any FFT analyzer plus LNA and coupling cap can be used for
that method. The major advantage of the 3048A system is the automatic
generation of plots.

Adrian

John Ackermann N8UR schrieb:

Content-Type: text/plain; charset=ISO-8859-1

In message C7259FFFCD3E42C2815B939BA386A0A8@pc52, "Tom Van Baak" writes:

Phase Noise measured on a 0Hz carrier ?

If you want the dynamic behaviour, it gets much more tricky, because
then you have both the spectrum of the load-changes and the
supply-changes, resulting in a spectrum output from the PSU.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

tvb wrote:

There are a number of "standard" ways, some of which have been
mentioned by others, none of which is all that helpful IMO.

What I find most useful is a plot of noise density vs. frequency
from, say, 0.1 Hz to as high as you require.  The data should be
taken and processed with sufficient frequency resolution to show any
spurs in the band of interest.  It is often helpful to have several
plots, each covering part of the band of interest, to improve the
displayed resolution of spurs.

NOTE:  Designing a preamp for collecting the data is far from
trivial.  Articles have been written about it (see, for example,
Linear Technology Application Note 124 by Jim Williams
<www.linear.com/docs/28585>).

Best regards,

Charles

Charles,

thanks for posting. That is the LTC application note I had in mind.

And here is Bruce's contribution to low noise PS design and measurement:

http://www.ko4bb.com/~bruce/LowNoisePowerSupplies.html

Don't miss to scroll fully down. There is a link to an IEEE paper
discussing chemical battery noise measurement.

Adrian

Charles P. Steinmetz schrieb:

Hi

If you are driving a spectrum analyzer, the 10,000X mentioned in the app note simply is not needed. A gain of 10X or less will get you to below 2 nv / sqrt(Hz)  at 100 Hz and beyond. A 10 Hz blocking cap does not need to be a "24 hours to stabilize" device.  An AD 797, a couple of metal film resistors, and a fairly large (say 47 uf) plastic cap work pretty well.

Bob

On Jan 31, 2013, at 7:20 PM, Charles P. Steinmetz charles_steinmetz@lavabit.com wrote:

Bob wrote:

The band from 10 Hz down to 0.1 or 0.01 Hz is generally important
when testing oscillators.  To keep the 797 input noise density below
a few nV per root Hz, the terminations must have very low
resistance.  With such low resistance, a 47 uF cap won't even get you
to 10 Hz, much less 0.1 or 0.01 Hz.

One more thought: Many oscillators have internal regulators that are
not nearly as good as what you can build.  No sense using an external
supply with 5 nV per root Hz noise density if it will be re-regulated
inside the oscillator by a circuit that has a noise density of 250 nV
per root Hz.

Best regards,

Charles

Good point.

Just to clarify my original request, my colleagues are currently using
batteries, and they don't have internal regulators.  Their problem is
not that the batteries don't work, but they want to travel around and
give demos and the batteries obviously are a PITA.

BTW, one of papers cited had a reference to Fred Walls paper on
battery noise.  Fred knocks it out the park as usual.  He was
like a rock star at FCS in those days.  A rare example of
government productivity.

Rick Karlquist N6RK

Awhile back I ran some baseband plots of various supplies with an HP 3048A
(image attached).  In my experience measuring actual OCXOs, an LM317T or
LM338K is quiet enough to avoid influencing oscillator PN.  With these
variable-voltage parts, you can bypass the reference pin for some additional
improvement, but I don't believe I did that for these plots.

It's easy to spot the difference between a 7812/7815 and an LM317T (see red
versus green/white traces).  As a lazy approach, try measuring the
oscillator with both a 78XX and an LM317T.  Because the 78XX is about 10 dB
noisier across most of the spectrum, If you don't see a difference, you can
assume that further optimization is pointless.  Near 1 Hz this call may be
questionable.

If you don't need an LDO, don't use one.  If you do, use the quietest part
you can find.  The best LDOs seem to be about as quiet as an ordinary
LM317T.

I've mentioned before that you need to be careful with large LC filters
downstream of the regulator.  A good power source will exhibit a low
impedance at ALL offsets of interest.

You sometimes see NIST circuits where the power is conditioned by a
Darlington emitter follower whose base is fed with an RC-filtered Zener
diode.  The purple and orange traces are pretty informative with regard to
that approach.  On the orange trace, where the only filtering is the RC
network between the Zener and the base, notice how the noise becomes worse
than all of the other sources below 10 Hz.  Here, the RC filter on the Zener
becomes less effective and the Darlington pair obligingly amplifies the
diode noise.

An additional LC filter after the regulator may have the effect of herding
the entire noise spectrum into a high-Q peak, even though the LC corner
frequency is much higher than the RC filter in the base circuit (violet
trace).  Depending on your OCXO's supply rejection characteristics this
could be a good thing or a bad thing.

Finally, make sure the OCXO has good RF bypassing where its power supply pin
enters the case.  If in doubt, solder a 0.1 uF ceramic right at the point of
entry.  I've seen $2000 Wenzels that didn't bother doing this.  I'm sure
they looked good in a screen room.

-- john
Miles Design LLC

Are 4 and 5 regulated? Or is the zener picked to compensate for the VBE
drops?

Any of these circuits where the output transistor is in an emitter
follower configuration will have its noise effected by the load current,
since that current directly effects the output transistor transconductance.

On 1/31/2013 10:03 PM, John Miles wrote:

transconductance.

The load in all cases was a couple hundred mA worth of resistance at the
voltage in question, as I recall.  It's been a few years since I captured
these plots.

The Zeners were just whatever parts came to hand, 1N474x parts basically.  I
didn't care about the 1.4V drop across the Darlington, just the resulting
noise.

-- john
Miles Design LLC

Bob Camp, Thursday, January 31 11:36 AM (Local NY time):
((...snip...))

No, I disagree. That's not carrying things nearly far enough.

I really think we should figure out a reliable methodology for this sort
of thing. This is time nuts after all, so perhaps some time/frequency
applications might end up needing sub-nanovolt regulation (and there are
those whom might simply "need" it because they're nuts)

Tom Van Baak, Thursday, January 31, 2013 9:43 AM (Local NY time):
((...snip...))

Thanks Tom.

I found that writeup about various power supplies rather useful. I'm
doing some research and development for power supplies right now, and
hope to come up with something completely nuts by the time I'm ready to
go from learning how to work with thunderbolts / GPS Disciplined OCXO to
something exotic which, today, might be far over my head.

--Sarah

Hi

The circuit I described, is (as stated) quiet down to 100 Hz. It's 3 db bandwidth is well below 10 Hz with the 47 uF cap. If you need it quiet down to 10 Hz or 0.000000001 Hz, you will need to buy a few more caps. It's still not rocket science.

For most OCXO or atomic standard testing applications out there, 10 Hz is low enough.

Bob

On Jan 31, 2013, at 9:41 PM, Charles P. Steinmetz charles_steinmetz@lavabit.com wrote:

Linear Technology application note 83

"Performance Verification of Low Noise, Low Dropout Regulators"

holds some interesting information. I am however unsure whether what was low
noise in 2000 is still low noise in 2013

Best regards
Ulrich Bangert

Ok, but open loop as I described?

I bicmos design, there are two common "junk" buffers. The junkiest (sp?) is going up a PNP and down a NPN. No feedback. You live with the vbe mismatch. Next up the food chain is the long tail pair (diff amp) with emitter follower. With one gain stage, it is reasonable stable.

-----Original Message-----
From: "John Miles" jmiles@pop.net
Sender: time-nuts-bounces@febo.com
Date: Fri, 1 Feb 2013 01:30:24
To: 'Discussion of precise time and frequency measurement'time-nuts@febo.com
Reply-To: Discussion of precise time and frequency measurement
time-nuts@febo.com
Subject: Re: [time-nuts] Low noise power supplies?

transconductance.

The load in all cases was a couple hundred mA worth of resistance at the
voltage in question, as I recall.  It's been a few years since I captured
these plots.

The Zeners were just whatever parts came to hand, 1N474x parts basically.  I
didn't care about the 1.4V drop across the Darlington, just the resulting
noise.

-- john
Miles Design LLC

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

This is a keeper.  Note the strong peak at 60 Hz for the unfiltered
darlington.

Regards.

Max.  K 4 O DS.

Email: max@maxsmusicplace.com

Transistor site http://www.funwithtransistors.net
Vacuum tube site: http://www.funwithtubes.net
Woodworking site
http://www.angelfire.com/electronic/funwithtubes/Woodworking/wwindex.html
Music site: http://www.maxsmusicplace.com

To subscribe to the fun with transistors group send an email to.
funwithtransistors-subscribe@yahoogroups.com

To subscribe to the fun with tubes group send an email to,
funwithtubes-subscribe@yahoogroups.com

To subscribe to the fun with wood group send a blank email to
funwithwood-subscribe@yahoogroups.com

----- Original Message -----
From: "John Miles" jmiles@pop.net
To: "'Discussion of precise time and frequency measurement'"
time-nuts@febo.com
Sent: Friday, February 01, 2013 12:03 AM
Subject: Re: [time-nuts] Low noise power supplies?