开云体育

So, for fun I thought I'd make my own version of the 3GHz 53132-68003 Channel 3 option for my 53132A Universal Counter.
I am able to source divide-by-128 prescaler chips here in Japan (I'm using a UPB1506GV by NEC) and the rest of the parts are available on Digikey and Mouser etc.


I went with a coplanar microstrip at almost 50 ohms for the entirety of the signal path, and instead of the obsolete original MSA-0986 amplifiers, I am using two BGA420 amplifiers for a total of roughly 26dB of gain at 1.8GHz, down to 18dB at 3GHz. I wonder if I should add another amp stage?



I mostly replicated the HP design, but the original diodes are also obsolete so I went with RB876WTL, they have a slightly higher Vr than the originals used by HP (HSMS-8202) at 5V rather than 4V, but their Ct is 0.58pF compared to the 0.26pF of the originals.
How much of an effect will this capacitance have and is there a better diode to use here? Also what sort of maximum capacitance should I aim for for a working frequency just above 3GHz? (I'm still learning these design details...)



Here are links to the schematics and board layout so far, also 3D renders of the front and rear of the board too:





For reference, here is the relevant pages from the HP 53132A CLIP:


Any help, advice, pointers, heckling, etc would be greatly appreciated.
I'll release the design files once I'm done, and as I am making 10 of these, I'll have 9 to sell once the design is all characterised and tested for anyone interested.



Thanks!
Jared

Looks like it should work OK.
On my board, I connected pin 4 (Channel 3 Detect Enable) to pins 9 and 10 (CH3CODE0 and CH3CODE1).
Pin 4 is an open collector drive so it should not matter if it is grounded. I do not think it will matter to the operation of the board.
The only other issue I had was too much gain, it caused spurious counting when no input was attached.
All was fine if any signal in range (frequency and amplitude) was there.
Working on an updated version now that has an on board shield to see if that helps..

ed

One other thought, The two amps and the divider could take up to about 50mA (UPB1506 @ 19mA and BGA420 @ 15mA).
I am not sure if that accounts for output currents.
With a nominal 12V input, the regulator will see a 7V drop across it times the current for 0.343W
With the MSOP at 160C/W junction to ambient per watt it could get quite hot, but not outside its specs.
Unfortunately I did not see a junction to case number for the part.
The L78M05CDT is better at 100C/W junction to ambient and has a junction to case of 8C/W.
With the case bonded to the ground plane, you should be relatively close to the 8C/W number.
Also, 5V is available on pin 7 of the connector if you install R434 on the main board with a 0 ohm resistor so that could be an option.

ed

I'll double check the regulator to see if it will be ok. I have a big ground plane which should help, but I'll run more calculations to make sure it won't burn itself up.
It would have been nice if HP would have put that resistor in for the 5V by default for all instruments....

I wonder if a couple meg or tens of meg resistor or so at the input down to ground would help prevent spurious counts without affecting the actual operation too much?



Thanks!

Ok, I decided to use the cheaper Diodes Incorporated for the Vreg.

At a max ambient temperature of 55 deg C (From the 53132A Specifications), the max input voltage allowable by the vreg is 13.64V before hitting thermal shutdown.
So with a big ol' ground plane to sink heat into, I reckon it should be ok.



Jared

Still think that an SOT-23 package is a bit small for 50mA with a 7V drop.
For a small increase in $$ an SOT-89 or SOT-223 package regulator would be a better choice.
You can not get much heat out of the SOT-23 or SOT-23-5 regulators due to the small leads.
That may be why they just provide the junction to ambient number in the data sheet.
The larger tab on the SOT-89 and SOT-223 gives them a better heat transfer to the plane, and you want one that has a ground tab, not a input or output tab like some regulators..
I am not sure the low noise aspect of the AP2210 would make much difference here.
Maybe its just my aerospace background taking over here, we usually did not have much room for temperature rise...

As far as the input noise is concerned, I tried a 3dB attenuator at the input and that did not help, so I would assume a high value resistor would not help either.
I found that with at least one nylon standoff the input was stable with no signal..
The board was OK with three of the ground connected standoffs but the 4th one initiated the noise problem.
It was quiet on the bench.

ed

I agree with Ed and think it's best to use a heavier regulator, or maybe throw some of the power into a series dropping resistor or Zener diode.

Regarding spurious noise and oscillations, here's something I wrote in time-nuts a few months back:

"It's a bit tricky to use prescalers for arbitrary input signals. As others have mentioned, they self-oscillate when there isn't a valid input signal. This is not only a possible, but nearly certain characteristic due to the (typically) ECL input amplifier being biased right at the logic threshold, followed by lots of gain, then divided by flip-flops. These were typically used in PLL synthesizers over limited frequency ranges, with known, adequate power levels.

If you are counting strong signals at well defined DC levels, you can use ECL directly with proper biasing. For a general purpose counter, you want lots of sensitivity and wide frequency range, so the usual prescaler with self-biasing and AC coupling is a simple way to go. You just have to be aware of the limitations. You can eliminate the oscillation tendency by giving up some sensitivity, but it's easiest to just ignore it unless it causes trouble."

Ed

I remember someone mentioned trying a 3 dB attenuator in front, with no effect on oscillation. The oscillation is in the prescaler itself, and it doesn't matter how much gain is ahead of it.

The usual way to inhibit oscillation is to throw the input DC bias out of wack a little bit, with some resistance to Vcc, which is +5V for PECL operation, or ground for regular (N)ECL. This shifts the amplifier chain away from the "ideal" (Vbb) DC bias level, reducing the gain. It can be done experimentally, determined by the particular device input characteristics. A starting R value in the 1k to 10k range should be noticeable. You can monitor the input DC level with a high resistance like 100k tacked to the node, connected to a DVM. The loss in sensitivity can be made up for by more gain in the other amplifiers, if needed.

One thing that might help somewhat, is to use a much larger coupling cap into the input of the prescaler, so the Z there will be dominated by the driving amplifier's characteristics. It could also aggravate it, or just change the frequency - it's another experimental kind of deal. Bigger signal coupling caps throughout can also reduce the lower frequency cutoff, if that's desirable. Most prescalers can count as low as you want if the signal and cap are big enough, and edges fast enough.

Ed.

Just remembered one more thing from something I noticed before, but didn't look closely - had to go back to the original HP circuit to be sure. Take a look at the rather elaborate decoupling circuit they used on the other input (pin 8) to the prescaler. On close inspection, you can see the extra damper R10/C23, and R6 to ground, apparently a factory selected value - that's to shift the DC bias just like I was talking about, except using the opposing input. You can picture it as a differential input. They used ground though, to move it, which works the same way, while I'm more of a purist and go with Vcc, by habit. And of course, R6 had to be determined and selected somehow, to go with the particular IC U5.

Ed

Ed's absolutely right. At 25C ambient, that's already right at the limit for that package, and you shouldn't assume a 25C ambient for the innards of an instrument.

Use a version in a beefier package.

-- Cheers
Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

Alrighty, I've swapped to a SOT-89 package for the vreg, I chose an which has a bit more thermal headroom.
It also does away with the tantalum caps and uses two 1uF ceramic caps instead, so nice for a little BOM and cost reduction.

So two more questions remain.. That biasing of the prescaler and also I have yet to nail down the protection diode selection.

I was wondering about that fancy stuff going on in the inverted input of the prescaler, I thought it must have had a reason like the biasing, but datasheets don't make much mention of it...
How would one go about getting into the ballpark for the design of that network of parts before empirical testing to nail the values?
I have a handful of the prescaler chips here, so I could rig something up pretty easily.

And secondly, the diode selection.. I see that the (obsolete) diodes that HP chose have a capacitance of 26pF, but a reverse voltage rating of 4V.
How can I figure out an upper limit to the diode capacitance? higher reverse breakdown voltage means higher capacitance, so it seems to be a trade-off between durability against user abuse and performance. :D


Thanks!
Jared

You don't have to care about the diode reverse voltage ratings - when anti-parallel, each can only see forward voltages, unless the current rating is exceeded. What matters is that (indicates the RF power it can clamp at the input), and the zero-bias capacitance (for speed).

The prescaler inputs are symmetric. You can treat it as a differential amplifier, and put the input signal to either one, while the other is bypassed - it will toggle one opposite edges if swapped. The ECL DC bias (Vbb) is usually developed by some sort of internal regulator circuitry, and each input connects to it through some medium resistance maybe in the hundreds of ohms range. You can put the DC bias into either one, just like the signal. What you want to do is upset the bias level a little bit to reduce the gain. HP used a selected R value for pulling the bias, and you can do the same if you can figure out how much is right. If you're designing for "production," it may be better to use a fixed resistor in series with a pot, to effectively select a value that's right for the particular part.

You can get some good ideas by studying the old ECL parts and documents. My favorite info source for this is the Motorola from the 1980s - I think the data book was called "MECL-III" family. Later generations like 10H and up started improving the performance, always tending faster, and with temperature-compensated Vbb,, lower operating voltage, and more function variety. A good proxy for a prescaler input section is the 116 or 216, for instance. These basic models extend from MC10216, to 10H216, to very high speed into the microwave 100E series and such. The 116 and 216 are triple line receivers, among the most highly used types, for interfacing small signals to ECL. Picture a single part with the three stages cascaded for high gain, so you can toggle the following flip-flops with mV input signals - if you have it biased right. The line receiver type parts almost always include access to the Vbb for setting up these scenarios with external resistors. The prescalers have the same setup with internal resistors, and don't provide access to Vbb. But, you can measure it - it's roughly the DC voltage present on the input pins.

Ed

Hi Jared,

I assume that a decimal point went AWOL and that you meant 0.26pF. I have never seen a 26pF diode with a 4V breakdown -- not because it's impossible, but because there is no market for one.

A good guide for how much is too much is to calculate the capacitance reactance at the highest frequency, and compare it to what shunts it in that circuit (50 ohms, e.g.). At 1GHz, say, a 1pF cap presents about a 160 ohm reactance, which is larger, but not very much so, than 50 ohms. At 10GHz, it's 16 ohms, which begins to look more like a short compared to 50 ohms. You get the idea.

-- Cheers
Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

On Tue, Oct 10, 2023 at 11:10 AM, Ed Breya wrote:

You don't have to care about the diode reverse voltage ratings - when anti-parallel, each can only see forward voltages, unless the current rating is exceeded. What matters is that (indicates the RF power it can clamp at the input), and the zero-bias capacitance (for speed).

Of course.... That makes perfect sense.

The prescaler inputs are symmetric. You can treat it as a differential amplifier, and put the input signal to either one, while the other is bypassed - it will toggle one opposite edges if swapped. The ECL DC bias (Vbb) is usually developed by some sort of internal regulator circuitry, and each input connects to it through some medium resistance maybe in the hundreds of ohms range. You can put the DC bias into either one, just like the signal. What you want to do is upset the bias level a little bit to reduce the gain. HP used a selected R value for pulling the bias, and you can do the same if you can figure out how much is right. If you're designing for "production," it may be better to use a fixed resistor in series with a pot, to effectively select a value that's right for the particular part.

Yeah, this application doesn't care if the signal is inverted so we are free to do what we want.
This one will be just a small run to cover my initial costs so a selected resistor would be fine, and then the full design files will be released, so maybe I'll make provision in v2.0 for a selected resistor and a trimpot depending on the builders preference.

You can get some good ideas by studying the old ECL parts and documents. My favorite info source for this is the Motorola from the 1980s - I think the data book was called "MECL-III" family. Later generations like 10H and up started improving the performance, always tending faster, and with temperature-compensated Vbb,, lower operating voltage, and more function variety. A good proxy for a prescaler input section is the 116 or 216, for instance. These basic models extend from MC10216, to 10H216, to very high speed into the microwave 100E series and such. The 116 and 216 are triple line receivers, among the most highly used types, for interfacing small signals to ECL. Picture a single part with the three stages cascaded for high gain, so you can toggle the following flip-flops with mV input signals - if you have it biased right. The line receiver type parts almost always include access to the Vbb for setting up these scenarios with external resistors. The prescalers have the same setup with internal resistors, and don't provide access to Vbb. But, you can measure it - it's roughly the DC voltage present on the input pins.

Ed

I'll check that out, thanks!



On Tue, Oct 10, 2023 at 12:13 PM, Tom Lee wrote:

Hi Jared,

I assume that a decimal point went AWOL and that you meant 0.26pF. I
have never seen a 26pF diode with a 4V breakdown -- not because it's
impossible, but because there is no market for one.

Ah yup, a typo there, 0.26pF is what I meant. :D

A good guide for how much is too much is to calculate the capacitance
reactance at the highest frequency, and compare it to what shunts it in
that circuit (50 ohms, e.g.). At 1GHz, say, a 1pF cap presents about a
160 ohm reactance, which is larger, but not very much so, than 50 ohms.
At 10GHz, it's 16 ohms, which begins to look more like a short compared
to 50 ohms. You get the idea.

-- Cheers
Tom

It makes sense. I'm still learning this black magic high frequency stuff, so I may have a few more basic-level questions to come. Bear with me.... :)



Thanks!
Jared

No worries -- this was all new stuff to all of us at one time or another.

As for black magic, it's got that rep, sure, but after a while you realize that it's all pretty straightforward. All you need is a wizard's hat and a chicken. Adjust the latter as necessary.

-- Cheers,
Tom

--
Prof. Thomas H. Lee
Faculty Co-Director, SystemX Alliance
Director, Stanford-Samsung Research Initiative

Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

If it doesn't work, then change the incantation, go have lunch, and start with a new chicken.

Harvey

Re the L78M05CDT you could also try a Traco Power TSR 2-2450.

This is a switchmode drop in replacement for LM7805 with a 2mA idle current, rated for 2A. It does not need heatsinking. I've used them when repairing the power supply in multimeters which have closed cases and no active cooling (Fluke 8000 series bench meters).

Warren

There are a number of such chips, one particularly nice one is an isolated DC supply 5 volts to 5 volts. TEA1-0505 (Traco), this is rated for 200 ma with an isolation of 1500 volts.? I was somewhat surprised to find these chips (and there are bigger ones that are adjustable and at more current, no TO-220 form factor, though).

Harvey

I would not use a switching supply on the prescaler board.
I think at about 50mA, a DPACK or SOT-223 regulator should do the job and stay relatively cool..
Even the SOT-89 should be OK as the most of the heat will go to the ground plane through the tab.
And on that note, the tab should be embedded in the ground plane, no relief connection.
That is assuming that the tab is the ground pin on the regulator.

The SOT-89 package regulator I chose should do the job from my calculations.
It has a decent tab that I have connected directly to the top ground plane, and a bunch of vias feeding through to the bottom ground plane too.

I also settled on BAT150-04R diodes for the protection diodes.
0.27pF capacitance, 110mA current rating and 100mW power dissapation should be adequate I think?

Next I just have to work out that biasing on the invert input pin on the prescaler chip and then I think I'll be ready to spin some PCB's up for v1.0