RF Automatic Level Control

[piglet]

On 07/01/2020 09:36, Cursitor Doom wrote:

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!

The zeners CR3 and CR6 in the signal path of the ALC diff amp is a bit
quirky but why bother improving an already good-enough circuit?

piglet

 

[piglet]

On 07/01/2020 09:32, Clifford Heath wrote:

On 7/1/20 5:05 am, Cursitor Doom wrote:
Greetings gentlemen,

I'm trying to improve on this 40 year old design by HP. The ALC
section is supposed to maintain the RF source output level across a
range of 4Mhz to 1300Mhz constant to within 1dB. Unfortunately, due to
a combination of component ageing and a restricted design budget at
that time, it doesn't meet this spec; not even close. There is
"considerable room for improvement" as they say.
I'm inclined to believe there are better solutions available today
with better frequency and temperature compensation than that afforded
by the handful of discretes shown; possibly something at least partly
monolithic?
Suggestions welcome...

Here's the schematic; you can see the relevant signal path is shown by
dashed lines:

https://tinyurl.com/yjzxtgn4

It looks like a perfectly acceptable circuit. The generator makes
4MHz->1.3GHz by mixing two VCO outputs (looks like 3.6-4.5 and
3.6-3.8GHz) so only the low product survives (no deliberate LPF needed.

The ALC uses a diode to derive a DC level to measure the output
amplitude, which is immediately RC filtered. It won't steal much energy
or distort the signal much (the RC charges up until the diode stops
stealing energy), and it works the same way across the whole frequency
range.

That level signal drives an error amp that feeds a two-diode modulator
which robs energy from the 3.6-3.8GHz signal before the low pass filter.
The process flat-tops symmetrically because of the two diodes, and the
2nd harmonic distortion is discarded by the following filter.

I reckon that either your error amp is faulty or one of the two
modulator diodes is busted.

Clifford Heath.

Those two diodes in the amplitude modulator are possibly PIN devices
(conceptually can be viewed as current controlled variable resistances)
rather than clippers?

piglet

 

[Cursitor Doom]

On Tue, 7 Jan 2020 20:16:19 +1100, Clifford Heath <no.spam@please.net>
wrote:

On 7/1/20 10:18 am, Cursitor Doom wrote:
On Mon, 06 Jan 2020 12:28:07 -0800, John Larkin
jlarkin@highland_atwork_technology.com> wrote:

That Yandex site is too picky about showing the image.

Dropbox is free.

Sigh....
Try this one: https://imgur.com/a/MzbIf8O

The original Yandex image is still here, trivially accessible (as you
said) to anyone with a modicum of nous:

https://avatars.mds.yandex.net/get-pdb/2846114/394be906-be9c-474a-b262-f63b8da1d8bf/s1200

Thanks, Clifford; thought I was in some alternative universe there for
a while. ;-)
--

"The BEST Deal is NO DEAL"

 

[Cursitor Doom]

On Tue, 7 Jan 2020 11:45:32 +0000, piglet <erichpwagner@hotmail.com>
wrote:

On 07/01/2020 09:36, Cursitor Doom wrote:

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!


The zeners CR3 and CR6 in the signal path of the ALC diff amp is a bit
quirky but why bother improving an already good-enough circuit?

Neither of those are in the signal path, though, Piglet!
--

"The BEST Deal is NO DEAL"

 

[piglet]

On 07/01/2020 11:58, Cursitor Doom wrote:

On Tue, 7 Jan 2020 11:45:32 +0000, piglet <erichpwagner@hotmail.com
wrote:

On 07/01/2020 09:36, Cursitor Doom wrote:

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!


The zeners CR3 and CR6 in the signal path of the ALC diff amp is a bit
quirky but why bother improving an already good-enough circuit?

Neither of those are in the signal path, though, Piglet!

They are not in the *RF* signal path but they are in the amplitude error
and demand signal paths. The amplitude leveling signal is thus subject
to the vagaries and drift of those zeners - I haven't perused the
circuit but first glance suggests they are intended to approximately
cancel each out?

piglet

 

[Cursitor Doom]

On Tue, 7 Jan 2020 12:11:41 +0000, piglet <erichpwagner@hotmail.com>
wrote:

On 07/01/2020 11:58, Cursitor Doom wrote:
On Tue, 7 Jan 2020 11:45:32 +0000, piglet <erichpwagner@hotmail.com
wrote:

On 07/01/2020 09:36, Cursitor Doom wrote:

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!


The zeners CR3 and CR6 in the signal path of the ALC diff amp is a bit
quirky but why bother improving an already good-enough circuit?

Neither of those are in the signal path, though, Piglet!


They are not in the *RF* signal path but they are in the amplitude error
and demand signal paths. The amplitude leveling signal is thus subject
to the vagaries and drift of those zeners - I haven't perused the
circuit but first glance suggests they are intended to approximately
cancel each out?

I *think* they constitute some kind of elementary temperature
compensation arrangement, but I'm not sure. CR6 is nowhere near the
signal, it forms part of the fixed reference voltage network to the
non-inverting input. I think you're correct about everything else
though - as usual!
--

"The BEST Deal is NO DEAL"

 

[Bill Sloman]

On Tuesday, January 7, 2020 at 8:36:21 PM UTC+11, Cursitor Doom wrote:

On Mon, 6 Jan 2020 20:17:26 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

For ALC, a diode isn't a bad idea in 2020. All you care about is that
the response be flat with frequency, so a very quick Schottky in a
low-inductance layout is the bee's knees.

Are you referring to the detector diode? The manual doesn't state what
kind of diode it is but I'm sure they'll have used something flat with
a fast recovery. It's not that aspect that's bothering me.

You appear to be a bit confused about the op amp idea--an error amp
isn't a comparator, it's a linear thing, just the sort of application op
amps are best at. Given the speed requirement, an LM358A would probably
be just ducky.

The op amp is only dealing with relatively slowly undulating DC, so
I'd have thought just about any jellybean device would suffice.

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!

Analog Devices does wideband product detectors which are a whole lot better.

This has been mentioned in previous threads.

You can roll your own at up to 500MHz by using an AD834 analog multiplier as a squarer, but AD has special purpose parts that go faster.

--
Bill Sloman, Sydney

 

[Cursitor Doom]

On Tue, 7 Jan 2020 20:39:56 +1100, Clifford Heath <no.spam@please.net>
wrote:

Actually forget that. Your level detector diode has been blown by a
pulse coming in the output, and so the ALC is running open loop.

That diode seems fine. I've scoped it at E10 and I'm getting exactly
the expected DC readings as shown on that node when I turn the RF
output control pot (not shown) between 0 and 10dBm. However, it's not
fluctuating at all as I expected given it's detecting the RF out which
most certainly is.
<baffled>
--

"The BEST Deal is NO DEAL"

 

[Phil Hobbs]

On 2020-01-07 04:36, Cursitor Doom wrote:

On Mon, 6 Jan 2020 20:17:26 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

For ALC, a diode isn't a bad idea in 2020. All you care about is that
the response be flat with frequency, so a very quick Schottky in a
low-inductance layout is the bee's knees.

Are you referring to the detector diode? The manual doesn't state what
kind of diode it is but I'm sure they'll have used something flat with
a fast recovery. It's not that aspect that's bothering me.

You appear to be a bit confused about the op amp idea--an error amp
isn't a comparator, it's a linear thing, just the sort of application op
amps are best at. Given the speed requirement, an LM358A would probably
be just ducky.

The op amp is only dealing with relatively slowly undulating DC, so
I'd have thought just about any jellybean device would suffice.

But - and here's the thing - no one has thus far suggested how the
posted design (from February 1979!!) could be improved!

The ALC portion is perfectly reasonable in 2020. There are chips to do
it, but if all you want is to keep the level stable, you don't have to
pay $$$--a 15-cent BAT15 is much better than good enough.

BITD HP sold very good RF diodes--I remember their P/N 5082-2835 fondly,
and still have a few.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Jeff Liebermann]

On Mon, 06 Jan 2020 22:19:45 +0000, Cursitor Doom <cd@notformail.com>
wrote:

On Mon, 06 Jan 2020 11:09:35 -0800, Jeff Liebermann <jeffl@cruzio.com
wrote:

1. Any good reason for not bothering to disclose the model number of
the HP signal generator? It just might be interesting to lookup the
specs for the generator.

8754A and it's not a sig gen; it's a network analyser.

Thanks. Please note that it's really difficult to build a network
analyzer without a signal source.

HP8754A user manual:
<http://hparchive.com/Manuals/HP-8754A-Manual.pdf>
Pg 2-3 Block Diagram and description showing ALC loop.

HP8754A user and service manual:
https://literature.cdn.keysight.com/litweb/pdf/08754-90015.pdf (27MB)
Pg 8-23 RF Source (A17) theory of operation
Pg 8-27/8-28 RF Source (A7) schematic
Pg 7-9/7-10 Source (A7), component locations
Pg 6-13/6/14 A7 components

I could not find the reference designator or part number for the ALC
diode. The parts list does show 12 diodes, but without a reference
designator, I can't tell which diode is the ALC diode. Anything
printed on the PCB?

More:
<http://jvgavila.com/hp8754a.htm>

2. How did you determine that the RF signal level varies across the
frequency range? Translation: What instrument do you have that can
accurately measure to within +/-1dBm from 0 to 1300MHz?

I have a spectrum analyser (2 actually) which both go up to 22Ghz.
However, on this occasion I had my 350Mhz Tek scope right next on the
bench so used that for the sake of convenience. It was readily
apparent the amplitude variations were probably north of 4dB, even
within the first 150Mhz.

Amazing. Since the detector diode is an envelope detector, the
harmonic and spurious signal content of the signal will have an effect
on the ALC circuit operation. When you looked at a low frequency
signal on your 350MHz scope (model number omitted again), did it look
like a sine wave, or did it exhibit some distortion? I'll be really
impressed if you can see and measure any level of distortion with an
analog oscilloscope.

3. If you want it stable and/or accurate over an unspecified
temperature range, are you prepared to test it in an environmental
oven?

I don't have one.

In that case, you will have some difficulties improving on the
circuit. It's easy to design something that works at one specific
temperature. It's not so easy to design something that works over a
specified temperature range. It's also possible to improve on
something that works as the working model will act as a reference to
determine if anything has actually been improved. You can't do that
if the reference model is non-functional.

I'm snipping the rest of your post as it relates to
signal generators which this isn't.

Actually, it has quite a bit to do with the problem. Quoting myself:

4. If the output leveling is really that bad, perhaps
something is fried in either the amplifier/detector section,
or in whatever you're using for an output attenuator.

I'm inclined to agree with the others, who suggest that something is
probably fried in the circuit. I can't add my guess(tm) to the
others, because you seem reluctant to provide any clue as to what
you're viewing on your unspecified model 350MHz scope over at low freq
part of the frequency range, or any other measurements. The quality
of the answers you receive is proportional to the information you
provide.

Also:
>>> My favorite mistake is transmitting into the signal generator.

You probably didn't transmit into your network analyzer, but
discharging a BFC (big fat capacitor) might have zapped the detector
diode.

Lastly:
>>> 6. Good luck.

I changed my mind. No amount of luck will help you redesign an
instrument that only needs some troubleshooting and repair.


--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

 

[Cursitor Doom]

On Tue, 7 Jan 2020 19:35:29 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

The ALC portion is perfectly reasonable in 2020. There are chips to do
it, but if all you want is to keep the level stable, you don't have to
pay $$$--a 15-cent BAT15 is much better than good enough.

Thanks for that bill of clean health, Phil. I only wondered because
that particular model was designed to appeal to small start-up
companies that couldn't afford the other VNAs that were produced by HP
and others BITD. Consequently they nickle 'n' dimed it where they
could to keep the cost down. It still came to as much as three new
family cars back in '79, but I guess that was a worthwhile improvement
on six.
--

"The BEST Deal is NO DEAL"

 

[Cursitor Doom]

On Mon, 06 Jan 2020 11:09:35 -0800, Jeff Liebermann <jeffl@cruzio.com>
wrote:

[snip]

Okay, Jeff. You've very generously devoted a lot of time to your
carefully-considered remarks and they deserve a proper response which
owing to the lateness of the our here I'm not able to provide right
now, having wasted time earlier in pointless arguing with one of our
resident trolls who shall remain nameless. I'll answer in full
tomorrow and all will be revealed......


--

"The BEST Deal is NO DEAL"

 

[Bill Sloman]

On Wednesday, January 8, 2020 at 1:06:42 PM UTC+11, Cursitor Doom wrote:

On Mon, 06 Jan 2020 11:09:35 -0800, Jeff Liebermann <jeffl@cruzio.com
wrote:

[snip]

Okay, Jeff. You've very generously devoted a lot of time to your
carefully-considered remarks and they deserve a proper response which
owing to the lateness of the our here I'm not able to provide right
now, having wasted time earlier in pointless arguing with one of our
resident trolls who shall remain nameless. I'll answer in full
tomorrow and all will be revealed......

Or all that Cursitor Doom understands about the problem, which isn't likely to be all that useful.

--
Bill Sloman, Sydney

 

[Clifford Heath]

On 7/1/20 10:51 pm, piglet wrote:

On 07/01/2020 09:32, Clifford Heath wrote:
On 7/1/20 5:05 am, Cursitor Doom wrote:
Greetings gentlemen,

I'm trying to improve on this 40 year old design by HP. The ALC
section is supposed to maintain the RF source output level across a
range of 4Mhz to 1300Mhz constant to within 1dB. Unfortunately, due to
a combination of component ageing and a restricted design budget at
that time, it doesn't meet this spec; not even close. There is
"considerable room for improvement" as they say.
I'm inclined to believe there are better solutions available today
with better frequency and temperature compensation than that afforded
by the handful of discretes shown; possibly something at least partly
monolithic?
Suggestions welcome...

Here's the schematic; you can see the relevant signal path is shown by
dashed lines:

https://tinyurl.com/yjzxtgn4

It looks like a perfectly acceptable circuit. The generator makes
4MHz->1.3GHz by mixing two VCO outputs (looks like 3.6-4.5 and
3.6-3.8GHz) so only the low product survives (no deliberate LPF needed.

The ALC uses a diode to derive a DC level to measure the output
amplitude, which is immediately RC filtered. It won't steal much
energy or distort the signal much (the RC charges up until the diode
stops stealing energy), and it works the same way across the whole
frequency range.

That level signal drives an error amp that feeds a two-diode modulator
which robs energy from the 3.6-3.8GHz signal before the low pass
filter. The process flat-tops symmetrically because of the two diodes,
and the 2nd harmonic distortion is discarded by the following filter.

I reckon that either your error amp is faulty or one of the two
modulator diodes is busted.

Clifford Heath.

Those two diodes in the amplitude modulator are possibly PIN devices
(conceptually can be viewed as current controlled variable resistances)
rather than clippers?

Quite possibly but not necessarily, they are operating in the knee
region, with the control current pushing them closer to full-on (which
means full attenuation). But the net effect of their non-linearity
(assuming they're matched) is progressive clipping leading to harmonic
distortion - which at >7GHz is easy to lose/filter. If not matched,
mostly the same anyhow, just assymetric. A poor man's variable
attenuator, but quite good enough for the task.

Thanks to the OP, I know that these broad-range generators used this
dual-VCO topology, but it's really good to see more detail.

CH.

 

On Wednesday, 8 January 2020 06:23:39 UTC, Clifford Heath wrote:

On 7/1/20 10:51 pm, piglet wrote:
On 07/01/2020 09:32, Clifford Heath wrote:

It looks like a perfectly acceptable circuit. The generator makes
4MHz->1.3GHz by mixing two VCO outputs (looks like 3.6-4.5 and
3.6-3.8GHz) so only the low product survives (no deliberate LPF needed.

The ALC uses a diode to derive a DC level to measure the output
amplitude, which is immediately RC filtered. It won't steal much
energy or distort the signal much (the RC charges up until the diode
stops stealing energy), and it works the same way across the whole
frequency range.

That level signal drives an error amp that feeds a two-diode modulator
which robs energy from the 3.6-3.8GHz signal before the low pass
filter. The process flat-tops symmetrically because of the two diodes,
and the 2nd harmonic distortion is discarded by the following filter.

I reckon that either your error amp is faulty or one of the two
modulator diodes is busted.

Clifford Heath.

Those two diodes in the amplitude modulator are possibly PIN devices
(conceptually can be viewed as current controlled variable resistances)
rather than clippers?

Quite possibly but not necessarily, they are operating in the knee
region, with the control current pushing them closer to full-on (which
means full attenuation). But the net effect of their non-linearity
(assuming they're matched) is progressive clipping leading to harmonic
distortion - which at >7GHz is easy to lose/filter. If not matched,
mostly the same anyhow, just assymetric. A poor man's variable
attenuator, but quite good enough for the task.

Thanks to the OP, I know that these broad-range generators used this
dual-VCO topology, but it's really good to see more detail.

CH.

I thought the whole point of PIN diode attenuators is that at
sufficiently high frequency the PIN diode really does look like a
resistor not a diode and therefore doesn't clip. The charge carrier
lifetime in the intrinsic region needs to be many times longer than
the period of the lowest frequency to be attenuated for this to work.
Having dual VCOs and attenuating one of them before mixing neatly
makes this condition true.

John

 

[Phil Hobbs]

On 2020-01-08 06:25, jrwalliker@gmail.com wrote:

On Wednesday, 8 January 2020 06:23:39 UTC, Clifford Heath wrote:
On 7/1/20 10:51 pm, piglet wrote:
On 07/01/2020 09:32, Clifford Heath wrote:

It looks like a perfectly acceptable circuit. The generator makes
4MHz->1.3GHz by mixing two VCO outputs (looks like 3.6-4.5 and
3.6-3.8GHz) so only the low product survives (no deliberate LPF needed.

The ALC uses a diode to derive a DC level to measure the output
amplitude, which is immediately RC filtered. It won't steal much
energy or distort the signal much (the RC charges up until the diode
stops stealing energy), and it works the same way across the whole
frequency range.

That level signal drives an error amp that feeds a two-diode modulator
which robs energy from the 3.6-3.8GHz signal before the low pass
filter. The process flat-tops symmetrically because of the two diodes,
and the 2nd harmonic distortion is discarded by the following filter.

I reckon that either your error amp is faulty or one of the two
modulator diodes is busted.

Clifford Heath.

Those two diodes in the amplitude modulator are possibly PIN devices
(conceptually can be viewed as current controlled variable resistances)
rather than clippers?

Quite possibly but not necessarily, they are operating in the knee
region, with the control current pushing them closer to full-on (which
means full attenuation). But the net effect of their non-linearity
(assuming they're matched) is progressive clipping leading to harmonic
distortion - which at >7GHz is easy to lose/filter. If not matched,
mostly the same anyhow, just assymetric. A poor man's variable
attenuator, but quite good enough for the task.

Thanks to the OP, I know that these broad-range generators used this
dual-VCO topology, but it's really good to see more detail.

CH.

I thought the whole point of PIN diode attenuators is that at
sufficiently high frequency the PIN diode really does look like a
resistor not a diode and therefore doesn't clip. The charge carrier
lifetime in the intrinsic region needs to be many times longer than
the period of the lowest frequency to be attenuated for this to work.
Having dual VCOs and attenuating one of them before mixing neatly
makes this condition true.

The carrier lifetime is part of the story but not all of it. The
recombination rate goes as 1/(doping density), but it's hard to get that
below about a nanosecond in silicon, even at 1E21 doping. (Gold-doped
diodes such as 1N914s have reverse recovery times of a few nanoseconds.)

Pure (intrinsic) silicon has a minority carrier lifetime of a bit below
a millisecond at room temperature, iirc (dim memory says 250 us).
However, you can make good RF diodes out of material with long carrier
lifetime--you just make the junction thin enough that all the charge
gets swept out in a small fraction of a cycle.

A PIN diode has a thick region of low doping in the junction, so it
stores a lot of charge there. That's what makes the series resistance
nearly constant over a cycle, which gives PIN diode attenuators their
lowish distortion.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Cursitor Doom]

On Tue, 07 Jan 2020 16:54:20 -0800, Jeff Liebermann <jeffl@cruzio.com>
wrote:

[snip references for which many thanks]

I could not find the reference designator or part number for the ALC
diode. The parts list does show 12 diodes, but without a reference
designator, I can't tell which diode is the ALC diode. Anything
printed on the PCB?

I'm not sure which one you're referring to here, Jeff. If it's the
signal pick-off one, the detector at the RF out stage, then it's
internal to that amplifier module which is monolithic and not
accessible without a can-opener. Consequently it wouldn't appear on
the parts list AFAIAA.

Amazing. Since the detector diode is an envelope detector, the
harmonic and spurious signal content of the signal will have an effect
on the ALC circuit operation. When you looked at a low frequency
signal on your 350MHz scope (model number omitted again),

Tek 2465A (is that really relevant?) Oh, you mean does it have a 50
ohm input? Yes, it does but you have to alter the input mode otherwise
it's just 15pf||1Meg IIRC. Come to think of it I'd best just check I
did have that on the right setting; not sure I did that.

did it look
like a sine wave, or did it exhibit some distortion? I'll be really
impressed if you can see and measure any level of distortion with an
analog oscilloscope.

I wasn't looking for distortion but certainly there was none visible,
no. No clipping either.

In that case, you will have some difficulties improving on the
circuit. It's easy to design something that works at one specific
temperature. It's not so easy to design something that works over a
specified temperature range. It's also possible to improve on
something that works as the working model will act as a reference to
determine if anything has actually been improved. You can't do that
if the reference model is non-functional.

OKay, I think we've established there's nothing to be gained from a
re-design; I'll just focus now on fixing it.

I'm inclined to agree with the others, who suggest that something is
probably fried in the circuit. I can't add my guess(tm) to the
others, because you seem reluctant to provide any clue as to what
you're viewing on your unspecified model 350MHz scope over at low freq
part of the frequency range, or any other measurements. The quality
of the answers you receive is proportional to the information you
provide.

It's kind of hard to provide that sort of detail on a text only
bulletin board, though.

You probably didn't transmit into your network analyzer, but
discharging a BFC (big fat capacitor) might have zapped the detector
diode.

Nope. No BFCs involved and proper static protection precautions
observed, too.

I changed my mind. No amount of luck will help you redesign an
instrument that only needs some troubleshooting and repair.

Yup, that's the goal now.
--

"The BEST Deal is NO DEAL"