By Wolfgang Sauerwein DL 8 JT

In this article, DL8JT describes his home made Z3801A distribution amplifier using commercially available Maxim evaluation kits and parts from surplus PC Ethernet boards.

He has also found an interesting note in an old HP "Bench-brief" about ground-loops.

Most of us have more than one piece of equipment to be synchronized at 10 MHz to the Z3801A "in house standard". Commercially available 10 MHz-distribution-amplifiers are too expensive (even on Ebay). Building it with many outputs with discrete components is a frustrating job. Fortunately, Maxim-IC  has an IC (MAX 4135 with a gain of 2 or MAX4136 with variable gain) as a video distribution amplifier. Each device has 6 analog outputs with a bandwidth from DC to 185 MHz!

Unfortunately, these chips are available only as "hard-to-handle" SMD versions and Maxim recommends GHz proof layout for the board (no prototype-boards) !

But now, best of all, Maxim offers an evaluation kit for a very reasonable price (about US-$ 65. in Europe), which includes a completely assembled and tested board. It is designed for 75 Ohm impedance and is equipped with the fixed-gain MAX4135, but you can get a free sample of the MAX 4136 (variable gain) with your order. The board requires + and - 5 Volts only.

I don't understand why they don't use the MAX4136, which would make it easier to adjust the output level, especially because the board is designed for the MAX4136 as well; you just need two SMD-resistors or a pot. (Nevertheless, I will not try to replace the 24 pin SMD chip due to missing SMD experience after 40 years as an electronic engineer !)

The gain of two will just compensate the loss in the serial output resistor of 75 Ohms with the termination-resistor of 75 Ohms in a coaxial video systems. Input and the six outputs are gold-plated SMA-connectors.

If you download the data sheets, use caution with the size of the board layouts on the MAX4135 Kit data sheet, they are scaled, although they appear to be 1:1 for quick reproduction and I was fooled by that !

While I was waiting for my shipment from Maxim I started to build a prototype board (100x160mm) with transformers and isolated BNC output connectors, assuming the size (74x59mm) of the layout on the Maxim data sheet, but now the actual boards are larger (95x76mm) and I have some trouble sandwiching the two boards as intended !

I planned to sandwich the two boards to save 6 each coaxial cables with SMA connectors by making a direct vertical connection between the SMA center pins and the transformers on the connector/transformer-board. This will save 6 costly SMA cables.

This photo shows how I originally intended to mount the Maxim kit on my prototype board. I will probably cut my Maxim boards between the SMA input connector and the channel enable DIP switch. Then, solder the unnecessary enabling lines to ground and drill two additional mounting holes, or just let the Maxim board extend outward.

Caution: When you receive your kit, all outputs will be disabled; set all DIP's to "ON" to enable! For this extreme accurate 10 MHz reference, I decided to isolate each output from the others to prevent ground loops (like on your stereo equipment). This requires isolated BNC jacks and 10 MHz RF transformers (i.e. Minicircuits T 1-1). These components cost about US-$ 10.- per channel, but I found a very cheap solution!

Try to get obsolete 10 Mbit Ethernet PC boards from any source (i.e. your company, any company near your location or Ebay). I was able to purchase 19 unused, brand new boards for a total of US $ 5 only on Ebay, including a BNC-T connector and a diskette, because 10 MHz only is absolutely obsolete. Modern 10/100 MHz boards are available now between 5-10 dollars, but these do not have the BNC output any more!

By Ethernet specs, all outputs have to be isolated by transformers and almost all designs are more or less identical. The BNC coax connection uses a floating IC (typically a 83C92P) which is connected to the rest of the logic by an IC like chip containing 3 each 1:1 10 MHz wideband-transformers (One for transmit, receive and collision signals, similar to the Minicircuits T1-1). Data sheets from various manufacturers are available on the internet.

I have found that that some, but very few of the Ethernet transformers contain internal termination resistors, especially those for the RJ-45 jack. This will degrade the output of the 10 MHz signal drastically. Check your component specifications via the Internet. The only missing spec is the maximum power of these transformers, but it should be adequate for our applications. Two of these triple transformer chips can isolate the six outputs of the MAX 4135/6. Most of the boards contain many more "goodies" !

A necessary power source for the isolated IC is a DC/DC converter. This is typically the largest module on the board and is a DC-101 from YCL. It converts 5 V to 9V DC at about 280 mA. If your boards have an additional RJ-45 twisted pair connector, you will find another chip with 2 each RF-transformers with integrated baluns and a 2/3-pole 17 MHz low pass filters for receive and transmit (typically YCL 20F001). These can be used instead of the triple transformers if you want some low pass filtering of your output or if you need the 10 Mhz signal at a remote site via twisted pair copper cable.

The isolated BNC-Connectors are typically from AMP and most boards have a 20.000 MHz crystal plus other components like capacitors and SRAM's usable for other applications. For a total of 12 outputs to feed all of my equipment, I will use two sets of boards in a sandwich style construction.

Due to the uncritical low frequency of 10 MHz I do not care about impedance differences between 50 Ohm and 75 Ohm of the MAX-4135-EV-kit. In the past we used silver-plated cavities at frequencies above 144 MHz; today PC's are clocked over 2 GHz on epoxy boards designed by digital oriented engineers!

Most of the commercial equipment like counters and synthesizers specify an input impedance of about 1 kOhm for external 10 MHz sources, but even with a specified 50 ohm everything should work well. If you run out of ports, several units should be able to share one channel in parallel. However, this would be against the ground loop isolation requirements!

My installation which will be housed in an "cannabilized" HP cabinet. I will connect the inputs of two distribution amplifiers in parallel to the output of the Z3801A. The twelve output connectors will be on the back. On the front I will have an additional 10 MHz and an 1 pps output derived from the differential (quasi ECL signals) available at the RS 422 / RS 232 connector. This DB-25 connector will be extended to another DB-9 or DB-25 connector at the rear-panel of the cabinet.

To prevent overloading the input of the MAX4135 with the output of the Z3801A, I will replace the input-termination-resistor R7 of the kit with a miniature variable CERMET-variable-resistor of 100 Ohms, (don't use wire-wound versions !) which has about the distance between it's outer wires as the SMD-termination-resistor R7; cut the trace between the printing "R7" and connect the input (Pin20) to the wiper of the pot.

This will allow me to adjust the input until clipping of the output occurs. Unfortunately I have no schematic drawing software to draw complete diagrams for you, but the information above should be understandable by most anyone involved in electronics.

I have not finished my project yet, and additional ideas from other users are highly welcome!

© Wolfgang Sauerwein DL 8 JT