Replacement for the PM9610 Prescaler for Philips PM6654 High precision Counter

[SaabFAN]

Since I have bought a PM6654C on ebay, I have had the idea of adding the Prescaler-module to it to give it the ability to detect frequencies up to 1.5Ghz or even higher.

The reasons behind this: My cheap 2.4GHz counter from ebay doesn't really fit into my setup on my bench and adding the functionality to the PM6654C would save one device. Also i don't like the fact that my device has the connector and the silk-screen for the RF-Option on the Front-Panel but no RF-Option installed

Fortunately, Philips provided full schematics for the options in the Service Manual, so I started remodeling the circuit in Circuit Maker with slightly different components:

- I'm using a BGA2869 MMIC from NXP instead of a 6-stage amplifier made out of discrete transistors.
- Instead of using one 1:2 Prescaler and one 1:8 prescaler, I use 2 MC12095D Prescalers that perform division by 4 and offer a shutdown-mode.
- Instead of the NE5512, I use a NE5535 Dual OpAmp for the Shutoff-Control (turns off the last prescaler-Chip when the input-signal is too weak.
- I am also using standard SMA-Connectors, so the connector on the Coax-Cable that is attached to the BNC-Connector in the Front-Panel needs to be replaced by an SMA-Connector.

The budget for this thing is less than 40€, which includes 10 PCBs from china.

What I need though, are the correct measurements for the board, as well as where to connect the little connector called BU2. The mainboard has no cutouts for it and there's nothing on the Input-Board either.
Could anyone with that counter and the PM9610 installed take some pictures? Google isn't too helpful in this regard unfortunately.

[SaabFAN]

Schematic for the Original Unit and the new Version with more modern components.

[edavid]

I can't help with the photo, but I would love to buy a couple of boards when you have them.

[SaabFAN]

I have finished the layout of the PCB in CircuitMaker.

I have tried to avoid bends in the traces that carry high frequency-stuff and chose a Coplanar Waveguide for the 50Ohm trace that runs from the Input to the amplifier and into the Prescaler.
The board-size is 15cm X 4cm - the unit is rather small in terms of height.
I haven't put anything of that sort in the board yet, but I think that at least everything before the MMIC should be shielded - The Philips Engineers apparently didn't think so when they made the second generation of the board, despite the fact that it is right next to the Display, which does consume a bit of current

Btw. on the 3D-View the board has staight connectors - That's wrong. The Connector has to be angled, but was too lazy to search for a fitting model in the Circuitmaker-Database.

Ideas / Suggestions are welcome. I haven't calculated the signal-levels before and after the amplifier yet, but NXP offers quite a broad range of MMICs in the small 6-pad SMD-Package. So there is the possibility of using a different type if the one I have specified now doesn't work.

[SaabFAN]

I have just checked the signal-levels: The MMIC has a maximum rating of 10dBm Input-Power, which should be enough for the automatic attenuator, which should kick in once the voltage rises above 0,5V.
In the other direction the MC12095 has a toggle-sensitivity of 200mVpp, which is just above the signal-level if one puts 10Vrms into the Input-Connector.

[Robert763]

Hi,
That was very timely, I just "found" the PM6654 I bought on ebay 5 years ago that did not work this evening. I was looking for something else in the attic. I'd forgotten all about it. Anyway having fixed it this evening, I thought it would be nice to fit a prescaler, googed PM6654 and this popped up.

I would be very keen to have a PCB if you get a batch made.

[Fraser]

I also own this model of counter 

If a prescaler PCB, unpopulated or populated, becomes available to purchase, I would love one for my unit.

Fraser

[SaabFAN]

Just to be on the safe side, I have added two Minimelf 1N4148 Diodes at the Input of the first Prescaler. I doubt that the MMIC will be able to dump too much voltage into the Prescaler and blow it up, but with RF and possible impedance missmatches, anything is possible

Anyway, if you guys think that there's nothing fundamentally wrong with the circuit I drew up, I'll check if it fit well into the counter (with a paper model^^) and then order the boards. I already have all the components, so once I order them it will be about 2 to 3 weeks until I can show some success, or some smoky stuff

[SaabFAN]

I just added the necessary Jumper-Links to make the use of MC12093 instead of the MC12095 possible because the latter one has been discontinued. The MC12093 ist still active and available on Digikey for quite expensive 4,50€. The MC12095 can be found on ebay for about 1€ in signle quantities or 5€ for 10 pieces.

This updates the design to REV. C.

I haven't checked how well the board fits yet and I'm not sure if I really want to or if I'll just roll with it now and later just buy some larger metal-pieces to form the retainers

[SaabFAN]

REV. D:

- Replaced the two LL4148 with one BAV99.
- Replaced the Bourns 3313S-1-102 Trim Pot by a Bourns 3296W-1-102LF - I have quite a few of these Trim-Pots here and I guess everyone else has at least a few of them too
   - Complete rearrangement of the board to make the pot more accessible if the one with the top-screw has to be used instead of the one with the screw mounted at the side.
- Made it more clear that there is an inductor connected in parallel to C18 in the schematic.

I have also checked inside the Counter if it all fits and to verify that just shorting Pins 5 and 6 on the connector activates Channel C - It does

I'm now going to order the boards. I've checked the Footprints of all components (the Footprint of the Trim-Pot was wrong: The Wiper was on Pin 3 instead of 2, like it is specified in the datasheet - The database of Circuitmaker sometimes isn't too reliable in this regard...) and have made sure that the trim-pot is easily reachable. If there's no GPIB-Module installed, it can be reached rather easily with a screwdriver. If the pot with the side-mounted screw is used, there's no problem in that regard of course.

Once I have the boards here, I need something to verify their performance. Unfortunately, the fastest thingie I have here is a Si5153 on an Adafruit Board, which goes up to about 160MHz. I could try to build a fast enough oscillator, but given my history with VCOs and hand-made inductors, I see some problems ahead

Btw. once I have the boards, where would be the best place to put them up for sale? I'll order 10 and only need one of course. eBay?

EDIT: 3D-View of the Board and BOM added.

[Fraser]

Hiya,

As previously stated, I would love to buy a board from you. You could place then in the Buy and Sell section of this forum for our fellow members. Set your price to that with which YOU are comfortable. I.e. I do not expect you to sell them at cost. Once you have fed the forum members with their needs you will get feedback on performance and any issues. You then have the option to sell PCB's and/or complete kits on eBay with Bona Fide quotes from happy members here who have built it. Such feedback is invaluable for sales. You can also reference this forum thread as part of the eBay auction.

I would buy a complete kit of parts if it were available so please consider me a very interested party.

Best Wishes

Fraser

[ON5Bi]

Hi

If a prescaler PCB, or PCB comes available to purchase, I would love one to.

Marc

[SaabFAN]

Short update:
- Boards are still listed as "Producing" on the ShenZhen2U Website
- If someone is interested, here is the Link to the CircuitMaker project: http://circuitmaker.com/Projects/Details/Thorsten-Schroeder/Philips-PM9610-Replacement Maybe if there are enough "thumbs up" for it it gets to the main page with the features projects 

[SaabFAN]

The boards are here and my curse strikes again: They kinda work, but not too well -.-

As you can see on the attached pictures, there's alot of noise and oscillation in the signal after the first prescaler and it gets worse after the second prescaler.
As you can imagine, the counter doesn't really know what to do with such an unclean signal and displays higher frequencies than the one that is put in. With no input the counter counts about 1,21GHz, which is probably the noise that gets amplified and lowpass-filtered by the DC Block-Capacitors.

At the moment I have the following culprits:
- The Prescalers are capable of too high frequencies and are toggling on a harmonic as well as the fundamental frequency - They are capable of 2.5GHz and per datasheet their useful operation area starts at 500MHz. This would mean I need other prescalers.
- Noise gets amplified enough for the prescalers to toggle on them, which I would say is unlikely because I have to go down to 10mV/div on my scope to see noise on the supply-lines, which is probably already inside the noise-floor of the scope. This would mean bigger or more caps.
- The prescaler cannot be connected directly to each other and require DC-Blocks to work properly. This would mean I have to hack the board to get a DC-Block in there.
- The prescalers need larger output capacitances to suppress parasitic oscillations and harmonics. This could be fixed the easiest way: Just add a bunch of caps

Any other reasons for this behavior?

[artag]

I had this problem some years ago when I made a prescaler for another Philips counter. I didn't have the original schematic so I just adapted a prescaler from another design (an Elektor project that used an MB512).

In my case, it was OK when given a reasonable signal, but tended to read non-zero when disconnected. The problem is self-oscillation. If you google prescaler self-oscillation you'll find it's a common problem with most home made designs. I think one way of dealing with it on commercial designs is to block the divided signal if it's below the level the input can work reliably with.

[mos6502]

If you're looking for a different prescaler, check out the Fujitsu MB506. I've had very good results with it. On perfboard even.

[SaabFAN]

Well, I got it mostly to work now.
It took some cutting and component-acrobatics, but I finally managed to get a decent waveform at the output that the counter can use. I also discovered that the biggest problem wasn't the board itself, but the Si5351 Adafruit board I used as a signal generator. Using it at 120MHz caused the output to contain considerable jitter which made the output of the prescaler-board look a lot worse than it actually was.
Self-Oscillation and a lot of harmonics is riding on top of the output waveform, but at as there is a circuit on the board that shuts off the output if there is no input-signal, this is no problem as long as the input-signal is handled correctly.

Which brings me to the biggest problem I'm having at the moment: It does not handle the signal correctly! The second Prescaler performs the correct /4 division, but the first one doesn't.
I already changed the chip because I suspected a faulty one from the 5 piece lot I bought on ebay, but the problem remains: The first prescaler only does division by 2. Both ICs have basically the same connection with both SW-Pins grounded.
Btw. the datasheet specifies the SW-Pin to be open for a low-condition, but if I leave the pin unconnected the voltage at the pin rises to about 3,5V and it still keeps dividing just by 2.
The picture with the counter displaying 100MHz was taken with the math function active which added a multiplier of 0.5 to the counter-result.

When searching for other prescaler-chips I also found some interesting replacements (pin-compatible).
A suitable set of chips would be the MC12026 (both control Pins set to high) together with the MC12083 (/2 divider with standby). The MC12026 could also be used standalone with some board-hacking, but that would defeat the standby-control as this chip has no standby-input. This chip is compatible with the boards I have here now.

About building the board: I've specified BAT54JFilm and BAP50-03 diodes in several places, but with a cut off leg regular BAT54C or schottky diodes in a similar package will also fit.
I also used 0402 components in several places so components ranging from 0805 down to 0402-package fit on the footprints.
With board revision D some cutting of traces and mounting resistors and capacitors on top of each other is necessary.
I don't know yet if I'm going to order another batch, as this board seems to work except for the problem with the first prescaler mentioned earlier (anyone have an idea about that?).

EDIT: Was a bit late and forgot to add the picture of the counter with the prescaler in it

[TheSteve]

Is there any chance your input signal isn't clean enough and the first prescaler is triggering on an undesired harmonic? Are you able to scope the input and output to see it really only dividing by 2? My only other thought is that the input signal is so noisy going into the first prescaler that it sees voltage on pin 6 even when it is grounded.

[SaabFAN]

The sine-wave at the input is pretty clean.
But as you can see in the attached pictures, there's excessive jitter on it at the output of the first prescaler.

When I look at the output signal of the first prescaler more closely, it seems as if there's some frequency-mixing going on...

[SaabFAN]

By the way, this is the signal from the Si5351. It is not exactly clean, but I don't have any better signal source upwards of 10MHz here at the moment.

[SaabFAN]

I have looked at the difference between the connection of the first and second prescaler:
The only difference is that the first one has a trace running under it that is connected to the +5V supply.
Could that be enough to cause oscillation if there's some noise on this particular trace?

[TheSteve]

I have looked at the difference between the connection of the first and second prescaler:
The only difference is that the first one has a trace running under it that is connected to the +5V supply.
Could that be enough to cause oscillation if there's some noise on this particular trace?

Hard to say for sure but having a solid ground plane under the IC is important and standard practice for the frequencies involved.

[Performa01]

I’ve had a brief look at your circuit and there are a few issues striking me immediately.

D2 and D3 are specified as BAT54? These are totally inadequate, with a capacitance of 10pF @ 1V, compared to the original BAT17, which is 1pF @ 0V.

This means that the input of the MMIC is loaded with ~20pF and the corner frequency is 318MHz, thus making the circuit very insensitive at high frequencies.
Furthermore, at higher frequencies like 1GHz, the input impedance seen by the MMIC is very low and almost purely capacitive, which might lead to stability issues.

Apart from the capacitive loading, these diodes appear to be part of an RF detector which produces the control voltage for the pin diodes. Therefore these diodes need to work up to several GHz, whereas a slow BAT54 will not be good for more than maybe 100MHz.

Then the output of the MMIC is loaded with 25 ohms, which means a major mismatch (C13, R11 in parallel with C14, R20).
BTW. What is that shorted C18? Even if it’s not shorted, it will still act like a short at 1GHz or above anyway. This again will not aid stability and clean signals.

So I recommend solving these issues first before looking any further.
BAT14 is still available, but of course any other UHF mixer diode will do the job, as long as the capacitance is <1pF and the max. forward current can be at least 30mA.

[SaabFAN]

I’ve had a brief look at your circuit and there are a few issues striking me immediately.

D2 and D3 are specified as BAT54? These are totally inadequate, with a capacitance of 10pF @ 1V, compared to the original BAT17, which is 1pF @ 0V.

This means that the input of the MMIC is loaded with ~20pF and the corner frequency is 318MHz, thus making the circuit very insensitive at high frequencies.
Furthermore, at higher frequencies like 1GHz, the input impedance seen by the MMIC is very low and almost purely capacitive, which might lead to stability issues.

Apart from the capacitive loading, these diodes appear to be part of an RF detector which produces the control voltage for the pin diodes. Therefore these diodes need to work up to several GHz, whereas a slow BAT54 will not be good for more than maybe 100MHz.

Then the output of the MMIC is loaded with 25 ohms, which means a major mismatch (C13, R11 in parallel with C14, R20).
BTW. What is that shorted C18? Even if it’s not shorted, it will still act like a short at 1GHz or above anyway. This again will not aid stability and clean signals.

So I recommend solving these issues first before looking any further.
BAT14 is still available, but of course any other UHF mixer diode will do the job, as long as the capacitance is <1pF and the max. forward current can be at least 30mA.

Thanks for pointing out the problems.
I completely missed the parallel connection of the termination resistors
I just replaced the resistors with 100Ohms each, which improved the sensitivity a bit.

The Diodes will have to stay the same for the time being, as I don't have any better ones, but I will try to get better ones on ebay (digikey requires 65€ order and I just ordered something about a week ago^^).

C18 is connected in parallel with an tiny microstrip-inductor on the board, forming a notch-filter at about 1,55GHz. My best guess is that it is supposed to stop RF-Power from reaching the detector if the input-frequency exceeds 1,5GHz and therefore the Prescaler is disabled - I haven't paid too much attention to that part though and just copied it from the original schematic

The biggest success I just had by inserting some copper foil underneath the first prescaler that I connected to ground. At least when connected on the bench, I now have the correct division ratio at the output. It seems to require some "warm-up" time, but after about 30 to 60seconds it shows the correct frequency (this might be a problem with the Si5351 though).

[Performa01]

So you’re making progress … 

Don’t you have a ground plane under the prescaler? If so, the copper foil shouldn’t make a difference. Sorry if this sounds daft, but I haven’t really studied your PCB design.

As for the diodes – for the time being, even a humble “computer diode” like 1N914 or 1N4148 would surpass the BAT54 by quite a margin. Even better would be 1N4151 or at least 1N4448.

Yes, these are no Schottky, but that doesn’t really matter for now.

Try to replace the BAT54 with the fastest Si diode (lowest capacitance) you have, and then compare the sensitivity at 1GHz or above – you should see a substantial improvement, even though quite obviously no Si diode can ever match a dedicated RF detector/mixer Schottky diode.

EDIT: oh, forgot to mention the notch filter at 1.55GHz - isn't your design meant to go higher than that?