50 kHz VCO w/sine output

On Sep 28, 9:56=A0am, me0...@yahoo.com wrote:
> On Sep 28, 6:21=A0am, George Herold <gher...@teachspin.com> wrote:
>
>
>
>
>
> > On Sep 27, 5:00=A0pm, Joerg <inva...@invalid.invalid> wrote:
>
> > > Tim Wescott wrote:
> > > > On Tue, 27 Sep 2011 13:18:06 -0700, George Herold wrote:
>
> > > >> On Sep 27, 3:40 pm, whit3rd <whit...@gmail.com> wrote:
> > > >>> On Tuesday, September 27, 2011 9:54:36 AM UTC-7, George Herold wr=
ote:
> > > >>>> 30kHz to 100kHz VCO with sine wave output. I was also thinking I
> > > >>>> could do this with a ~20MHz varactor controlled VCO as a variabl=
e
> > > >>>> clock into a DSG chip. =A0This looks easier to my unsophisticate=
d
> > > >>>> digital mind.
> > > >>> Workable, of course, but not exactly a VCO anymore... and there's=
 the
> > > >>> Nyquist filtering to do afterward.
>
> > > >>>> Finally I offer the following, =A0a Wien bridge oscillator =A0wi=
th
> > > >>>> varactor diodes as the capacitors.
> > > >>> Not a great choice, because the Wein bridge requires MATCHED
> > > >>> capacitors, tuned together, for best performance. =A0 It can be d=
one with
> > > >>> a few trimmers (probably easiest to trim the offset + gain of the
> > > >>> control voltages). =A0Then, there's the level translation problem=
 (the
> > > >>> capacitors in a Wien bridge don't have one end grounded).
> > > >> Hi Whit3rd,
> > > >> Did you look at my schematic? =A0I seem to have it running in LTsp=
ice.
> > > >>> Good sinewave LC VCOs in narrow ranges are easier, and a 4.0 MHz =
sine
> > > >>> source and 4.030 to 4.100 MHz sine VCO can be mixed down to get y=
ou
> > > >>> what you want.
> > > >> Ah, good. =A0This was my first suggestion on how to get what was w=
anted.
> > > >> It was rejected. =A0But perhaps I should push it a bit more!
>
> > > > It's what I'd push, if I couldn't convince people to use a DDS or i=
f
> > > > there were some overriding systems reason why a DDS wouldn't do. ..=
.
>
> > > I'll second that. Maybe we should make some signs and start a virtual
> > > picket line. Need a nice chant though. "Rah-rah-rah, os-ci-llate,
> > > down-con-vert, rah-rah-rah". Oh, and drums, of course :-)
>
> > Can I get some scantily clad drum majorettes too? =A0:^)
>
> > Seriously walking about last evening, I think this is the way to go!
> > We'll do heterodyning and VCO side band generation. =A0I'll order some
> > varactor diodes and build some LC tank ciruit. =A0Got a favorite
> > oscillator circuit?
> > Or should I get something ready to run?
>
> > Now I just have to convince the 'powers that be'.
>
> > George H.
>
> > > [...]
>
> > > --
> > > Regards, Joerg
>
> > >http://www.analogconsultants.com/-Hidequoted text -
>
> > > - Show quoted text -- Hide quoted text -
>
> > - Show quoted text -- Hide quoted text -
>
> > - Show quoted text -
>
> A couple of words of caution if you decide to go the down-converter
> route.
>
> I missed your final amplitude requirements, but I would caution you if
> you decide to use the 602 part and require an output approaching 0dBm
> or better. =A0This really is a low power part with a weak IIP3.

I didn't post any drive requirements. :^)   But I was picturing it
running into a mini-circits mixer... 7dBm on the LO (IIRC) and I
assume something similar from the VCO, but at a few MHz I can
certainly add some gain.
>
> Even though you have a single tone at the input, over driving parts
> such as these can cause funny distortion effects.
>
> Re VCO: =A0I doubt very seriously that you could use a free running
> VCO. =A0The damn thing will just be doing to much of the =93wautsi=94. =
=A0So
> plan on lock=92d VCO.

Oh that sounds more complicated.  I've only done a little RF stuff and
most of that was 'cut and try', without really understanding what the
'F' I was doing.

>
> I would recommend, as others have suggested, a DDS system.

Hmm and around I go again.

Thanks for the advice.
>
> You might consider getting an eval board from AD. =A0They use to pass
> those things out like candy. =A0But in this economy, who knows. =A0The bi=
g
> advantage of the eval board is that they come with control software
> that can be run off the pc.
>
> Hope this helps.- Hide quoted text -
>
> - Show quoted text -

George Herold wrote:
> On Sep 27, 5:00 pm, Joerg <inva...@invalid.invalid> wrote:
>> Tim Wescott wrote:
>>> On Tue, 27 Sep 2011 13:18:06 -0700, George Herold wrote:
>>>> On Sep 27, 3:40 pm, whit3rd <whit...@gmail.com> wrote:
>>>>> On Tuesday, September 27, 2011 9:54:36 AM UTC-7, George Herold wrote:
>>>>>> 30kHz to 100kHz VCO with sine wave output. I was also thinking I
>>>>>> could do this with a ~20MHz varactor controlled VCO as a variable
>>>>>> clock into a DSG chip.  This looks easier to my unsophisticated
>>>>>> digital mind.
>>>>> Workable, of course, but not exactly a VCO anymore... and there's the
>>>>> Nyquist filtering to do afterward.
>>>>>> Finally I offer the following,  a Wien bridge oscillator  with
>>>>>> varactor diodes as the capacitors.
>>>>> Not a great choice, because the Wein bridge requires MATCHED
>>>>> capacitors, tuned together, for best performance.   It can be done with
>>>>> a few trimmers (probably easiest to trim the offset + gain of the
>>>>> control voltages).  Then, there's the level translation problem (the
>>>>> capacitors in a Wien bridge don't have one end grounded).
>>>> Hi Whit3rd,
>>>> Did you look at my schematic?  I seem to have it running in LTspice.
>>>>> Good sinewave LC VCOs in narrow ranges are easier, and a 4.0 MHz sine
>>>>> source and 4.030 to 4.100 MHz sine VCO can be mixed down to get you
>>>>> what you want.
>>>> Ah, good.  This was my first suggestion on how to get what was wanted.
>>>> It was rejected.  But perhaps I should push it a bit more!
>>> It's what I'd push, if I couldn't convince people to use a DDS or if
>>> there were some overriding systems reason why a DDS wouldn't do. ...
>> I'll second that. Maybe we should make some signs and start a virtual
>> picket line. Need a nice chant though. "Rah-rah-rah, os-ci-llate,
>> down-con-vert, rah-rah-rah". Oh, and drums, of course :-)
> 
> Can I get some scantily clad drum majorettes too?  :^)
> 

Your wife might object :-)


> Seriously walking about last evening, I think this is the way to go!
> We'll do heterodyning and VCO side band generation.  I'll order some
> varactor diodes and build some LC tank ciruit.  Got a favorite
> oscillator circuit?
> Or should I get something ready to run?
> 

You could just use the SA612, figure 7 shows how to hook up LC to pins 6
and 9. You can do Colpitts and Hartley, the transistor would be behind
those two pins with its base at pin 6.

As the next poster (me0223) wrote, this is a fairly wimpy chip. So don't
push it. If you need more amplitude hang an opamp behind it. Or a small
audio amplifier chip if you also need a low impedance output, many of
those easily go up to 100kHz. Then you could generate several watts and
scare the bats out of the rafters :-)


> Now I just have to convince the 'powers that be'.
> 

Been there ...

-- 
Regards, Joerg

http://www.analogconsultants.com/

On 09/28/2011 09:49 AM, George Herold wrote:
> On Sep 27, 10:06 pm, Phil Hobbs
> <pcdhSpamMeSensel...@electrooptical.net>  wrote:
>> On 09/27/2011 02:17 PM, Tim Wescott wrote:
>>
>>
>>
>>> On Tue, 27 Sep 2011 09:54:36 -0700, George Herold wrote:
>>
>>>> 30kHz to 100kHz VCO with sine wave output.
>>
>>>> I know this has been covered before in previous threads.
>>
>>>> The obvious way to do this is a digital signal generator (DSG) IC with
>>>> maybe an A-D to measure the voltage.  I&#4294967295;ve not done much digital stuff
>>>> in a while, so this has a bit of a learning curve.  But if anyone has
>>>> some good chips to look at or app notes then please do share.  I've been
>>>> looking at IC's on Analogs web site, but there are a lot to choose from!
>>
>>>> I was also thinking I could do this with a ~20MHz varactor controlled
>>>> VCO as a variable clock into a DSG chip.  This looks easier to my
>>>> unsophisticated digital mind.
>>
>>>> Finally I offer the following,  a Wien bridge oscillator  with varactor
>>>> diodes as the capacitors.    I was also thinking I could use crappy Z5U
>>>> ceramic caps instead of the varactors.
>>
>>> It'll be rough to get the high impedances and low voltage swings you need
>>> for the Wien bridge oscillator to work either with varactors or with a
>>> suggestion that I've made previously of using JFETs as variable resistors.
>>
>>> So: a high frequency VCO to a DDS chip, or a fixed-frequency reference to
>>> a DDS that you write a frequency to with a micro, or (if you are for some
>>> reason insisting on "analog only") a VCO feeding a mixer, beating against
>>> a crystal oscillator.
>>
>>> I'd look to the DDS solution, or see if
>>
>> you can make a suitable block
>>
>>> with a PIC or similar, a good DAC, and some code.
>>
>> You can still get low-nanofarad Y5V caps in small sizes, and those plus
>> a boost converter are about the best things going right now for high
>> kilohertz stuff.  All the MVAM-style varactors are long gone.
>
> Yeah searching for varactor's lots of them were/ are OBSELETE.
> sigh.
>
> PIN photdiodes have C's that change by factors of 4 or 5.  But that
> seems like a spendy solution.
>
>
>>
>> I'd probably be wanting to use a tri-wave oscillator with a sine shaper.
>>    (You don't have to use the crappy diode ones--you can use a tanh to
>> roll over the peaks and then subtract a bit of the original sine, as we
>> discussed a year or so back.  I posted a suggestion athttp://electrooptical.net/www/sed/sed.html, but it's a much older
>> technique.
>
> Hmmm, Thanks Phil, I'd forgotten about that.  (I had no need for it at
> the time.)  How do I get a tanh function out of a diff pair.  (Yeah,
> I'll try google too.)
>
> George H.

Diff pairs do tanh automatically--it's getting them to do anything else 
that takes work.

You just shove the tri wave between the bases, and subtract the 
collector currents, either with a current mirror or an op amp.  If you 
don't mind an offset current of I_e/2, just connect one collector to the 
supply and come out the other one.

Cheers

Phil Hobbs

On Wed, 28 Sep 2011 11:32:24 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>On 09/28/2011 09:49 AM, George Herold wrote:
>> On Sep 27, 10:06 pm, Phil Hobbs
>> <pcdhSpamMeSensel...@electrooptical.net>  wrote:
>>> On 09/27/2011 02:17 PM, Tim Wescott wrote:
>>>
>>>
>>>
>>>> On Tue, 27 Sep 2011 09:54:36 -0700, George Herold wrote:
>>>
>>>>> 30kHz to 100kHz VCO with sine wave output.
>>>
>>>>> I know this has been covered before in previous threads.
>>>
>>>>> The obvious way to do this is a digital signal generator (DSG) IC with
>>>>> maybe an A-D to measure the voltage.  I&#4294967295;ve not done much digital stuff
>>>>> in a while, so this has a bit of a learning curve.  But if anyone has
>>>>> some good chips to look at or app notes then please do share.  I've been
>>>>> looking at IC's on Analogs web site, but there are a lot to choose from!
>>>
>>>>> I was also thinking I could do this with a ~20MHz varactor controlled
>>>>> VCO as a variable clock into a DSG chip.  This looks easier to my
>>>>> unsophisticated digital mind.
>>>
>>>>> Finally I offer the following,  a Wien bridge oscillator  with varactor
>>>>> diodes as the capacitors.    I was also thinking I could use crappy Z5U
>>>>> ceramic caps instead of the varactors.
>>>
>>>> It'll be rough to get the high impedances and low voltage swings you need
>>>> for the Wien bridge oscillator to work either with varactors or with a
>>>> suggestion that I've made previously of using JFETs as variable resistors.
>>>
>>>> So: a high frequency VCO to a DDS chip, or a fixed-frequency reference to
>>>> a DDS that you write a frequency to with a micro, or (if you are for some
>>>> reason insisting on "analog only") a VCO feeding a mixer, beating against
>>>> a crystal oscillator.
>>>
>>>> I'd look to the DDS solution, or see if
>>>
>>> you can make a suitable block
>>>
>>>> with a PIC or similar, a good DAC, and some code.
>>>
>>> You can still get low-nanofarad Y5V caps in small sizes, and those plus
>>> a boost converter are about the best things going right now for high
>>> kilohertz stuff.  All the MVAM-style varactors are long gone.
>>
>> Yeah searching for varactor's lots of them were/ are OBSELETE.
>> sigh.
>>
>> PIN photdiodes have C's that change by factors of 4 or 5.  But that
>> seems like a spendy solution.
>>
>>
>>>
>>> I'd probably be wanting to use a tri-wave oscillator with a sine shaper.
>>>    (You don't have to use the crappy diode ones--you can use a tanh to
>>> roll over the peaks and then subtract a bit of the original sine, as we
>>> discussed a year or so back.  I posted a suggestion athttp://electrooptical.net/www/sed/sed.html, but it's a much older
>>> technique.
>>
>> Hmmm, Thanks Phil, I'd forgotten about that.  (I had no need for it at
>> the time.)  How do I get a tanh function out of a diff pair.  (Yeah,
>> I'll try google too.)
>>
>> George H.
>
>Diff pairs do tanh automatically--it's getting them to do anything else 
>that takes work.
>
>You just shove the tri wave between the bases, and subtract the 
>collector currents, either with a current mirror or an op amp.  If you 
>don't mind an offset current of I_e/2, just connect one collector to the 
>supply and come out the other one.
>
>Cheers
>
>Phil Hobbs


Another way to smooth a triangle onto a sine is with linear segment
breakpoints. One opamp and a mess of resistors and diodes can do that,
but there are several ways.

Two breaks works pretty well. It can be done at high level, which is
convenient.

I wonder how you'd find the optimum break formula. Or, in general, how
one best approximates a function, over some range, with N linear
segments.

John

On Wed, 28 Sep 2011 06:21:38 -0700, George Herold wrote:

> On Sep 27, 5:00&nbsp;pm, Joerg <inva...@invalid.invalid> wrote:
>> Tim Wescott wrote:
>> > On Tue, 27 Sep 2011 13:18:06 -0700, George Herold wrote:
>>
>> >> On Sep 27, 3:40 pm, whit3rd <whit...@gmail.com> wrote:
>> >>> On Tuesday, September 27, 2011 9:54:36 AM UTC-7, George Herold
>> >>> wrote:
>> >>>> 30kHz to 100kHz VCO with sine wave output. I was also thinking I
>> >>>> could do this with a ~20MHz varactor controlled VCO as a variable
>> >>>> clock into a DSG chip. &nbsp;This looks easier to my unsophisticated
>> >>>> digital mind.
>> >>> Workable, of course, but not exactly a VCO anymore... and there's
>> >>> the Nyquist filtering to do afterward.
>>
>> >>>> Finally I offer the following, &nbsp;a Wien bridge oscillator &nbsp;with
>> >>>> varactor diodes as the capacitors.
>> >>> Not a great choice, because the Wein bridge requires MATCHED
>> >>> capacitors, tuned together, for best performance. &nbsp; It can be done
>> >>> with a few trimmers (probably easiest to trim the offset + gain of
>> >>> the control voltages). &nbsp;Then, there's the level translation problem
>> >>> (the capacitors in a Wien bridge don't have one end grounded).
>> >> Hi Whit3rd,
>> >> Did you look at my schematic? &nbsp;I seem to have it running in LTspice.
>> >>> Good sinewave LC VCOs in narrow ranges are easier, and a 4.0 MHz
>> >>> sine source and 4.030 to 4.100 MHz sine VCO can be mixed down to
>> >>> get you what you want.
>> >> Ah, good. &nbsp;This was my first suggestion on how to get what was
>> >> wanted. It was rejected. &nbsp;But perhaps I should push it a bit more!
>>
>> > It's what I'd push, if I couldn't convince people to use a DDS or if
>> > there were some overriding systems reason why a DDS wouldn't do. ...
>>
>> I'll second that. Maybe we should make some signs and start a virtual
>> picket line. Need a nice chant though. "Rah-rah-rah, os-ci-llate,
>> down-con-vert, rah-rah-rah". Oh, and drums, of course :-)
> 
> Can I get some scantily clad drum majorettes too?  :^)
> 
> Seriously walking about last evening, I think this is the way to go!
> We'll do heterodyning and VCO side band generation.  I'll order some
> varactor diodes and build some LC tank ciruit.  Got a favorite
> oscillator circuit?
> Or should I get something ready to run?
> 
> Now I just have to convince the 'powers that be'.

If this is a low production volume system then you'll probably spend less 
$$ in the end to just buy a VCO from MiniCircuits.  If it's high volume 
then you can spend more time engineering the VCO yourself and the parts 
will cost less.

What do the 'powers that be' expect?  I spent years working at a company 
that was very "transistor averse" -- they were much happier with a square 
inch of board space and a $5.00 super-zoot op amp than they were with a 
quarter square inch occupied by a $0.25 transistor and a few resistors.  
One learned to do what was going to get through the design review in the 
end...

-- 
www.wescottdesign.com

George Herold wrote:
> 
> Yeah DDS from a micro looks like the 'best' path.  It's just outside
> my present comfort zone and so hard to predict how long it will
> take.


   You can get a "AD9850 Module DDS Signal Generator with Circuit
Diagram" in ebay for about $13 and demo software on the Analog Devices
website.  There is also VB source code & 80*51 source code availible for
free.

   I just bought a couple for a project I'm working on.


-- 
You can't have a sense of humor, if you have no sense.

George Herold wrote:
> 
> Can I get some scantily clad drum majorettes too?  :^)


   I don't know. CAN you? :)


-- 
You can't have a sense of humor, if you have no sense.

On Wed, 28 Sep 2011 08:57:52 -0700, John Larkin
<jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

>
>
>Another way to smooth a triangle onto a sine is with linear segment
>breakpoints. One opamp and a mess of resistors and diodes can do that,
>but there are several ways.
>
>Two breaks works pretty well. It can be done at high level, which is
>convenient.
>
>I wonder how you'd find the optimum break formula. Or, in general, how
>one best approximates a function, over some range, with N linear
>segments.
>
>John

Many years ago I wrote an optimization program for doing this for a
nonlinear function with 'n' breakpoints. It would slide around the
points from reasonable initial positions and optimize whatever error
function you wanted (absolute error, error squared or whatever). These
days you could probably do it with Excel's solver. I used it for
fitting thermocouple curves. 

On Wed, 28 Sep 2011 14:40:01 -0400, Spehro Pefhany
<speffSNIP@interlogDOTyou.knowwhat> wrote:

>On Wed, 28 Sep 2011 08:57:52 -0700, John Larkin
><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>>
>>
>>Another way to smooth a triangle onto a sine is with linear segment
>>breakpoints. One opamp and a mess of resistors and diodes can do that,
>>but there are several ways.
>>
>>Two breaks works pretty well. It can be done at high level, which is
>>convenient.
>>
>>I wonder how you'd find the optimum break formula. Or, in general, how
>>one best approximates a function, over some range, with N linear
>>segments.
>>
>>John
>
>Many years ago I wrote an optimization program for doing this for a
>nonlinear function with 'n' breakpoints. It would slide around the
>points from reasonable initial positions and optimize whatever error
>function you wanted (absolute error, error squared or whatever). These
>days you could probably do it with Excel's solver. I used it for
>fitting thermocouple curves. 

Did it always converge to a sensible result? I can imagine all sorts
of silly or chaotic behavior. This seems like an interesting problem.

I usually just eyeball it, for soft curves at least. I guess you want
shorter segments on the curvier parts. For software fits, we just do
some insane number of equally-spaced segments, like 64 or so.

For shaping a triangle into a sine, what should be optimized is the
harmonic distortion, not the RMS error, I guess. Fiddling in Spice is
probably as good a way as any.

John

On Wed, 28 Sep 2011 11:57:41 -0700, John Larkin
<jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

>On Wed, 28 Sep 2011 14:40:01 -0400, Spehro Pefhany
><speffSNIP@interlogDOTyou.knowwhat> wrote:
>
>>On Wed, 28 Sep 2011 08:57:52 -0700, John Larkin
>><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>>
>>>
>>>
>>>Another way to smooth a triangle onto a sine is with linear segment
>>>breakpoints. One opamp and a mess of resistors and diodes can do that,
>>>but there are several ways.
>>>
>>>Two breaks works pretty well. It can be done at high level, which is
>>>convenient.
>>>
>>>I wonder how you'd find the optimum break formula. Or, in general, how
>>>one best approximates a function, over some range, with N linear
>>>segments.
>>>
>>>John
>>
>>Many years ago I wrote an optimization program for doing this for a
>>nonlinear function with 'n' breakpoints. It would slide around the
>>points from reasonable initial positions and optimize whatever error
>>function you wanted (absolute error, error squared or whatever). These
>>days you could probably do it with Excel's solver. I used it for
>>fitting thermocouple curves. 
>
>Did it always converge to a sensible result? 

It was a long time ago, but provided you started with a sensible start
position I think it did. 

>I can imagine all sorts
>of silly or chaotic behavior. This seems like an interesting problem.

I think it would be rather difficult to do in general, and I think I
spent a bit of time looking at the general problem, but really I just
wanted to calculate optimal precision resistor values for each input
type so I could send the design to production and go have a beer. 
Plotting the nonlinearity and picking points manually was good enough
to meet the spec. 

>I usually just eyeball it, for soft curves at least. I guess you want
>shorter segments on the curvier parts. For software fits, we just do
>some insane number of equally-spaced segments, like 64 or so.

The BFI method works fine. 

>For shaping a triangle into a sine, what should be optimized is the
>harmonic distortion, not the RMS error, I guess. Fiddling in Spice is
>probably as good a way as any.
>
>John
>