[time-nuts] Fury Interface Board: How about TI OPA277? -- Very Long Reply
xaos
xaos at darksmile.net
Sat Nov 3 09:13:56 EDT 2007
Hello everyone,
Needless to say, but I have also been following this discussion closely.
First, I must confess, I started this project without making clear
what my requirements were.
Personally, I knew what I wanted. However, when I asked the members of this
list for their feedback, I did not properly explain the details.
My apologies, to everyone here, for any mix-ups.
The idea was this:
What I have is this:
1. A Fury GPSDO.
2. Quite a few HP OCXO's. Some are 10544A's and others are 1081-xxxx types.
Over the years I've collected these from 5328A's or 5370A/B's.
3. I have some Crystek CO27VS12DE-02-10.00, OCXO's.
4. A bunch of different power supplies. Lambda's are really nice.
5. OrCAD Capture/Layout/PSPICE
6. Tools and test equipment.
6. Some free time.
What I'd like to get:
A MTI OCXO. The 230 series types.
What I am trying to do:
1. Experiment with different types of 10MHz OCXO's and find one, with best
frequency stability and accuracy.
2. Use the Fury to discipline the OCXO and generate a 10 MHz signal.
I need this to feed all my test equipment.
3. Design a proper system and build a board.
4. The board must handle the different OCXO's I have. It must take the EFC
signal, from the Fury, and translate it to the EFC input of the
different OCXO's.
The translation circuit must handle 0-(+5VDC) input and have the option
of producing:
[0-5V] or [0-10V] or [-10 - +10V]
This will, not only, handle my current OCXO's but leave room for
future expansion.
Between these 3 voltage ranges, I think I have most industry OCXO's
covered.
4. While I am making a board for myself, make a few more for
anyone here who would be interested.
5. Have fun.
Why:
My primary, secondary, tertiary and quaternary frequency standards are
the HP Z3801A's. They are not getting any younger. I need something new
here.
That's it.
As expected, the members of time-nuts have been just great with their
feedback.
However, the discussion has become heated at times.
This was not a surprise. We are all techies and we do this for a living
and we'll be damned if we give up a point without a fight.
I do not want this to be a deterrent to the discussion. I want to hear
from every person who has something to offer.
If I mean, if anyone is getting beat up here, it's me... right? My
original design
is pretty much scrapped.
The thing is, I expected this to happen. It happens all the time in the
real world
and this is how really good systems are made. By consensus.
Anyway, moving on to more technical details...
Over the course of this discussion the following have become clear.
a. The power supply must have 3 outputs and they must be isolated from
each other.
+24V for the HP OCXO oven.
+12V for the OCXO supply
+/- 12V for other circuits.
b. The +/-12V, for the EFC translation circuit, might need to get +/-15V
from the
power supply. If the output is to swing to [-10V] - [+10V] then this
might be too close to the 12V rails.
Does anyone know of a quiet voltage regulator for that range?
If I add more filter caps to fix this, would this be enough? Might be
bulky.
c. In order to handle the different voltages required, the only option
is a 24V power supply
capable of producing + and - 24V.
c. The 3 grounds must be of a star configuration (from the power supply
ground) to avoid noise.
d. The EFC translation circuit is going to be a challenge. The right
OPAMP + reference selection
is extremely critical. I am currently working on the simulation of
the circuit that Bruce has provided.
I know what Said is saying about, "how much EFC voltage deviation is
really required".
However, I don't want to worry about that.
If I have a Crystek, today, I might have brand-X tomorrow. I don't
want to play
games with the OCXO EFC range. I must be able to handle the entire range.
e. The 10MHz output from the OCXO must be connected, via an isolation
transformer, back to the Fury.
This will isolate the boards. I am thinking of the Mini-Circuits
TMO-1-1+ or TMO-1-1T or T1-6-X65.
Questions that I still have:
1. Besides creating a star configuration, for all 3 grounds, what else
should I do to minimize noise
and maximize isolation between them?
2. In the case of the HP 10544A an isolation amplifier might be needed
to connect to the Fury OCXO 10MHz
input. The other OCXO's can connect to a 5Ohm load. Is this true?
If this is true then shall I just feed the output of all OCXO's
through the isolation amplifier?
In other words, shall I also do this for the OCXO's that can handle
50 Ohm loads?
I don't think I have an option here. If I add a Mini-Circuits 1:1
transformer to connect back
to the Fury, then some sort of isolation amplifier will be needed.
3. Said said, that a thermistor should be placed between the Fury OCXO
VCC and ground. Since I will be
breaking out these signals to the interface board, I assume I can
connect the thermistor there.
Am I missing something there?
4. Should I be remote sensing the +12V, (from Fury OCXO connection) in
order to track my +12V power supply?
I get the impression that the Fury controls this voltage as well. Am
I correct in this?
Bruce suggested this and it sounds like a good idea.
Well, that's all I can think of now. I've been up all night working on
this and I need some sleep.
Before I go, let me just say a big thank you to Bruce, Said, Kevin, John
Ackermann and all
the members of time-nuts for your help.
You guys are great.
Good Night,
-George
Bruce Griffiths wrote:
> ); SAEximRunCond expanded to false
> Errors-To: time-nuts-bounces+xaos=darksmile.net at febo.com RETRY
>
> SAIDJACK at aol.com wrote:
>
>>
>> In a message dated 11/2/2007 17:51:33 Pacific Daylight Time,
>> kevin-usenet at horizon.com writes:
>>
>>
>>
>>>> H'm... and if you really want the full +/-10V range, fitting a regulator
>>>> and op-amp into the 2V of available headroom requires an LDO and a
>>>> rail-to-rail output op-amp.
>>>>
>>>>
>>
>> Hi guys,
>>
>> I've been following the discussion about increasing the Fury EFC range from
>> 0-5V to -10V to +10V with great interest!
>>
>> I just had an idea on how to avoid all the issues potentially introduced by
>> using an Opamp circuit.
>>
>> Let's take a step back and see how much EFC voltage deviation is really
>> required:
>>
>> 1) let's assume we use an HP 10811, so temperature stability is very good
>> and certainly requires less than +/-2.5V range to compensate for (on the MTI
>> double oven units we typically see less than 100uV deviation on the EFC due to
>> temperature!).
>>
>> 2) Now let's assume an aging of 5E-08 per year - certainly good OCXO's will
>> be better than this. 5E-08 per year at 10MHz is about 0.00137Hz aging per day.
>>
>> 3) For 10811's I have measured a range of 4Hz for a 5V EFC change, so let's
>> assume it's EFC gain is 0.8Hz/Volt. This into 0.00137Hz/day means a voltage
>> change of 0.00171V per day.
>>
>> This means a -2.5V to +2.5V EFC range would be enough to compensate for
>> about 8 years of aging on our well-aged theoretical OCXO, so going to +/-10V is
>> probably much more than needed.
>>
>> 4) So why don't we just run the OCXO ground at +2.5V instead of 0V, and run
>> the Fury ground at 0V?
>>
>> This means the Fury's EFC output (0V to 5V) looks like a -2.5V to +2.5V
>> range to the OCXO due to the OCXO's ground being offset by 2.5V.
>>
>> The 10MHz output of the OCXO can be easily transformer-coupled into Fury as
>> someone has said earlier, so no problem here.
>>
>> Offsetting the OCXO ground by 2.5V should be possible by adding a -2.5V low
>> noise regulator to the system. EFC current is very low, so a low noise
>> negative voltage reference may be used to generate the -2.5V.
>>
>> No need for opamps, complex bipolar voltage regulators, etc. Of course any
>> noise or drift in the -2.5V regulator would show up in the EFC voltage as an
>> error.
>>
>> What do you think?
>>
>> bye,
>> Said
>>
>>
>>
>>
> Said
>
> Another consideration is that for 10544A's and similar oscillators which
> are only intended to drive high impedance loads( >= 1K) a simple 2
> transistor cascode buffer (or maybe a 10811A style common emitter stage
> with series feedback in the emitter circuit) may be required to allow
> them to drive a 50 ohm load or an RF transformer satisfactorily. A
> cascode buffer has higher isolation than a common emitter stage.
>
> Bruce
More information about the time-nuts
mailing list