High Speed Sine to Square Wave converter based on LT1016 Comparator

Several medium to high speed ADC’s which I would like to test required a low jitter square wave input for optimum performance. In addition, I wanted a >20MHz square wave signal to test my active oscilloscope probe and simulate high speed digital signals in a simple manner.

Since my function generator is only able to produce square waves up to 5MHz I decided to build a simple comparator based sine to square converter. For a first try I selected the LT1016 high speed comparator from Linear:


LT1016 Pros:

  • Single 5V supply Rail produces ~3.3V Output swing
  • Complimentary TTL / CMOS output
  • 3000V/s small signal gain
  • Very high bandwidth product (50GHz)
  • 5$ / psc

LT1016 Cons:

  • Maximum ~10mA output current – Not able to drive 50 Ohm loads directly to 3.3V CMOS / TTL level


The working principle is very simple. The sine wave input gets AC coupled and biased to half the supply voltage. Which gets compared against half the supply voltage on the second comparator input. The LT1016 produces from this a complimentary output which can be very handy as well.

The first test board was wired up with a big continuous ground plane underneath and plenty of decoupling capacity on the supplies. Series output resistance is 470Ohms resulting in a 330mVpp Square wave into a 50Ohms load (measured 420mVpp).

Measrment of the prototype with a 20MHz sine wave input.

This worked pretty good but I really wanted a more permanent solution. Schematic of the new design is as simple as the prototype:

My prototype PCB was ordered. The schematic stayed the same but the PCB board offers better routing options with power and signal layers directly over the ground plane, via stitching and 50 Ohm impedance matching for the signals. Sine input is via a W.FL2 micro coaxial, output via two SMA edge mount connectors. I choose the W.FL2 connector since the maximum frequency as well as the input power for the sine wave will stay low and since it allows for a more flexible mounting into a (eg) Hammond aluminum case.

The finished PCB is shown below.

The results were slightly better than the prototype. Raise/fall time were faster compared to the first prototype (3.5/3.5nS vs 4.9/3.7nS) – Probably because the lower series resistance. I was, however, only able to test to 20MHz sine input today and will certainly try with another signal generator. Note that series output resistance was chosen to be 200Ohms resulting in a 628mVpp square wave into a 50 Ohm load (500mV measured).

One improvement I might make is to add a potentiometer to the second comparator input to adjust the duty cycle to exactly 50% if desired. Also, a 5V input regulator would make the device a bit more flexible.


Schematic and Gerber files as well as the measurement data in form of cvs data can be downloaded from my Github account.



Agilent / HP6632A Part I

I just received a HP6632A System Power Supply which can be have for about 100-200$ from Ebay and which should make a quiet nice Lab supply. Lets open it and see what`s inside 🙂


Lots of dust everywhere. As to be expected from a fan cooled device with lots of airflow. The unit should look a lot better after some basic cleaning. More worryingly: One of the input diodes seems to have fallen apart! Fortunately I found one with close enough specs on one of my scrap boards.

Since I had to remove the PCB anyway I took the unit completely apart and cleaned everything out. I also tried to save and restore the 60x60mm 12V Papst fan which sounded like a turbine. But even after disassembly, cleaning, lubricating and balancing the fan unit the noise is still unacceptable. Sounds like the ball bearings have gone. In addition to that, the fan blades are quiet small and the fan runs at very high RPM`s so I will replace it with some slower running fan with larger blades.

Another option would be a speed controlled Fan. But the three pin connectors “FAN” pinout is simply GND-12V-GND with no PWM signal output. The schematics also indicate that the unit only does over temperature detection and shutdown without actual temperature measurements. That`s annoying, but maybe I`ll hack something together in the future.


In order to remove just the PCB and repair the diode, quiet a lot has to be taken apart. First open the case, then remove the aluminium middle strut and unscrew the transformer. Plug out the transformer lead out on the board side – this way it`s much easier to correctly reconnect them later. No two power connectors have the same pin count. Make sure to disconnect everything and clearly lable the four front connectors, also try not to drop the transformer on the board or on the floor!!!


A quiet nice way to mount the PCB are the little standoffs, on which the hole board can slide out – after removing just three screws!!!


Tilt the board forward on the heatsink and carefully take it out of the casing. It`s a bit tricky but it should just fit.


Now you can unscrew the front panel (beware of scratches) by removing the two screws under the fake leather just behind the front cover on the sides. Now cleaning with pressured air and some mild soap is in order. Here`s a look at the solid, clean main case unit. BTW, as always, do not use alcohol, isopropanol or similar at any plastic or rubbery surfaces which could be damaged by the aggressive cleaning.


A look at the dusted off board. The main heat sink on the right of the board has been dusted off with air pressure and is quiet clean on the inside now.


The construction and layout is, as it is to be expected from such an expensive device – very nice. Some minor points are the excessive use of discrete diodes and transistors with individual heat sinks and the somewhat lose main heat sink when the fan plate is removed -The hole aluminium sink is then split in two and lose. Also, the capacitors seem to be only 85°C rated which is a bit disappointing because they are so close to warm or even hot components and might dry out at some point.


Since the “020” front panel connection option is almost never installed, but HP has provided the necessary connectors on the front of the PCB as well as holes trough the plastic part of the front panel unit.  I got some rather nice 5mm banana plug connectors which fit nicely into a 9mm hole. It`s very important for the connectors to be isolated on their mounting point since the front panel is made out of conductive aluminium and the outputs will short otherwise!!!


In order to drill the holes I would  recommend starting with a 3mm drill from the back and then use a 5mm and after that the 9mm drill from the front. Also I would suggest to take the front panel off and protect the keypad and the electronics with some tape from the aluminium dust which could block & short out things later on. Fix the front panel down securely and use a vertical drill. Also, clean off the holes and remove any rough aluminum around the edges!



The third, smaller hole is intended for a 5mm acyl light guide and indicator LED (more in part II).

I used approx. 15cm long 15mm2  wire and crimped connectors (blue color code) on both sides. You could also solder them on to the connector but I like everything removable. Resistance check with 6.5digit multimeter and 4 wire test setup reveals about 0.01 Ohms between the original connectors on the back and the newly installed front plugs.









Don`t forget to install the sense wires (Sense+ to Out+ and Sense- to Out-) on the back. You could also connect cables on the backside sense ports and connect them to the front output ports, but since the resistance is that low I don`t think it is worth the trouble. If you worry about the lead resistance, I would rather connect a shielded & twisted pair cable and connect it directly at the sense terminals and your load as suggested by the original HP user manual to the 6632A.


Will soon continue with part II.

Teardown of a D-Link dir815 Wireless N-Router

Small, cheapish 2 Channel 2.5/5.8Ghz Router with 4 Port 10/100 Switch. Follow the links to get to the datasheets for the IC`s of the device. Full size pictures of the teardown at the end of the page.


Programing interface:

There is a Serial (UART) port for the RT3662 chip on the bottom left side of the pcb. The pinout is:

  • RXD
  • 3.3V
  • GND
  • TXD

Serial protocoll: 57600 baud, 8N1 (8 bit, no parity, 1 stop bit).


LAN Driver / Switch

  • IC+ IP175D 5 port 10/100 Ethernet switch

Wireless SOC (System on chip)

  • 5.8GHz band, Media Raillink RT3662 dual band N channel, 300MBit/s
  • 2.4Ghz dual band, Raillink RT3092 PCI wireless driver with RF stage

Rf Amplifier Stage:

  • 2x 5GHz siGe 2537L 30dB power amplifier
  • 2x 2.4GHz SST 12P15A 32dB power amplifier
  • Additional lumped & active (?) filters and other small RF stuff



Teardown pictures (Click for full size Download).

Analog stage of the Dlink DIR815
Backside of the DIR815 PCB
Dir 815 board overview
Overview with shield removed.
RT366 Wlan SOC of the DIR815
RT366 Wlan SOC of the DIR815