Archive for March, 2018

Audio Adapter for the Si5351 Signal Generator

I had just finished building a three port RF signal generator when the January 2018 QST arrived in the mail. January is the annual Do It Yourself issue and there was an intriguing article by Keith Kunde, K8KK titled “A Low Distortion Digital Audio Oscillator”. The heart of Keith’s circuit is a MAX294 filter chip, a sophisticated 8th order switched capacitor filter capable of rounding off a square wave into a pretty good sine wave. The ‘294 is fed with a square wave at the desired frequency and also a clock signal 100 times the desired frequency. Keith generates his clock with a 555 timer and then feeds a divide by 128 chip to develop the base output frequency. My new Si5351 RF generator could be a crystal controlled clock source and all I would only need the divider and filter circuits to add audio capability. The MAX294 costs about six bucks each.

Goals for the Audio Adapter were:

  1. 20 to 20000 Hz capability
  2. Sine or Square wave output
  3. Line level output, one volt peak to peak into 600 ohms
  4. Be as portable as the signal generator, meaning battery operation in an Altoids tin
  5. Two channels
  6. Direct frequency readout from the signal generator display

That sixth requirement means decimal dividers throughout. If I add one additional decade to the ‘294 chip’s clock divide by 100 requirement, the total division would be 1000. When the signal generator display shows 25 megahertz, the output would be 25 kilohertz. No math necessary.

The MAX294 has an uncommitted on board op amp but it is intended as a pre or post processing filter and has weak output specs. To get line level output I decided to buffer the ‘294 output with an LM358 op amp connected as a voltage follower. Later on, this proved to be the most troublesome part of the project.

Internet searching did not turn up many choices for decade frequency dividers. I decided on the 74hct390 chip which I have used before. It’s a dual decade divider, each side has a divide by two, and a divide by five stage. Cascading these gives a divide by ten with BCD output of the counter state. For this two channel build I would need a total of ten divide by ten stages, achievable with five 74hct390s. This diagram shows the basic filter wiring.

Basic MAX294 Application

Basic MAX294 Application

The last divider stage is a divide by two so the ‘294 filter sees a symmetrical square wave. 100k and 39k resistors form a voltage divider network to get the TTL level square wave down to an amplitude the ‘294 can handle without clipping.

 

I decided the project was indeed viable and ordered parts from Digikey.

Software in the signal generator was tweaked to have a lower range limit of 100 Kilohertz. The Si5351 should go all the way down to 10 Khz with appropriate software support but rather than press my luck, I added two additional decade divider stages to the design. A three position switch selects a frequency tap from the chain so the frequency ranges are:

  • Divide by 10**3  100 to 25,000 Hz
  • Divide by 10**4  10 to 2,500 Hz
  • Divide by 10**5  1 to 250 Hz

With the additional ranges, I can still read the LCD display directly but have to mentally move the decimal point. The first requirement is met. Also added is an SPDT switch to connect the op amp input to either the filter input (square wave) or the filter output (sine wave). That meets the second requirement.

Digikey parts finally in hand, I bread boarded the circuit. The dividers are on the left, the op amp is on the right.

Audio Adapter Breadboard

Audio Adapter Breadboard

This is the final schematic of the filter circuits. Eagle made the drawing and I did go through the exercise of generating a PC board  just to see if there really was room for the parts.  Since the project is a one-off, I built everything on perf board, using 30 gauge wire wrapped around IC pins with my standard wire wrap tool – a ballpoint pen refill. It would have been a lot faster if I had a real PC board.

Analog Board Schematic

Analog Board Schematic

Analog Board

Analog Board

A 2000 mAh portable USB power pack was sacrificed to obtain an 18650 battery, it’s associated charger, and 5 volt up converter. That would satisfy the portability requirement. I stuffed the battery, charger board, four RCA jacks and three slide switches into an Altoids tin while waiting for Digikey.

I used Eagle to generate a schematic and proposed PC board for the higher level digital divider circuits. This board had to fit between the 18650 battery and the range switches, and I had to nip off the corners to work around the battery charger circuit. This board is mounted on #4 screws soldered to the bottom of the Altoids tin. I built the digital divider board first, it is much simpler than the analog boards.

Digital Board Schematic

Digital Board Schematic

Digital Divider Board

Digital Divider Board

I used the analog portion of the bread board to verify the digital dividers were working properly. I found the 74hct390 chips would accept a signal as high as 90 MHz. They will be loafing at 25 MHz. This photo shows the digital board seated in the bottom of the Altoids tin. Only six wires will connect to the analog board in the tin lid, 5 volt power, ground, two divided outputs from the range switches, and two audio output leads. Audio connects to 1000 ufd  capacitors nestled on either side of the battery.

Digital Divider Board and Charger Installed

Digital Divider Board and Charger Installed

Now came extensive testing on the breadboard. Keith Kunde’s QST article discusses the level and DC offset considerations of the MAX294 filter when used with a single ended supply. I used a pair of 10k resistors to create a half supply voltage virtual ground and bypassed that rail with 22 uF capacitors.  I’m using DC coupling between the dividers, the filter chip, and the op amp so had to do some fussing with the voltage divider parameters Keith talks about to get a filter output with no clipping. My final divider has 100k resistor in series with the TTL level divider output, and a 36k resistor to virtual ground. I also found that connecting a VOM in series with the 5 volt supply to measure current seriously upset the DC balance for some reason. It appears the final circuit will draw about 25-30 milliamps from the battery. At that rate, the adapter should run 60 hours on a charge.

But the output waveform from the LM358 buffer amp was horrible! To make a long story short, Google “LM358 Crossover Distortion”.  The ‘358 has a class B push pull output stage and when the signal crosses over from one side to the other it takes a short nap. After a good bit of troubleshooting and research, the fix is simple. Add a 1000 ohm resistor from amp output to ground. This forces the output stage into a class A region and I was able to get 2 volts peak to peak out with no glitches and it will drive a 100 ohm load with no clipping or distortion.  I believe 2200 ohms to ground will also work and draw less current through the amp.

Also the ‘358 has a slow slew rate which causes the sides of the square wave to lean noticeably.   If I knew then what I know now I would have ordered an MC34072 amp which has better specs and no crossover problem, still less than a dollar.

The following photos are the analog board top, bottom and trial fit into the lid. The board is supported in the Altoids tin by the pot mounting nuts at the front and a single L shaped bent paper clip soldered to the lid at the back. I’ve also added an LED in the lid to remind me to turn the power off.

Top of Analog Board

Top of Analog Board

Bottom of Analog Board

Bottom of Analog Board

Connecting Digital Divider Board to Analog Board

Connecting Digital Divider Board to Analog Board

I took a couple of weeks part time to wire and debug the analog circuits but I was finally satisfied and mounted the board in the Altoids tin lid. It JUST clears the battery and the inductor on the battery charger. This photo is as finally assembled. Note a small bit of blue clay on the charger inductor, that’s how I checked the clearance.

Interior of Finished Audio Adapter

Interior of Finished Audio Adapter

These are the money shots of the fully assembled Audio Adapter.

Dual Channel Audio Adaptor

Dual Channel Audio Adaptor

Dual Channel Audio Adaptor

Dual Channel Audio Adaptor

But does it work?

This shot has sine wave selected on one channel and square wave selected on the other.

MAX294 Filter Out and In

MAX294 Filter Out and In

A 10 KHz sine wave. As K8KK noted in QST, there is some clock noise on the wave form at 100x the output frequency.

10000 Hz Sine Wave

10000 Hz Sine Wave

This is a 1 (ONE!) Hz sine wave. Notice the sweep speed setting on the right. It took some creative manipulation of room lights and the camera shutter speed to get this to show properly. You can definitely see the clock noise on this trace.  It looks a lot like the synthesized sine waves I experimented with a couple of years ago.

One Hertz Sine Wave

One Hertz Sine Wave

The next two photos show the waveform when the two channels are combined in an external resistor network. 1500 ohms from either side to a junction, then 1500 ohms to ground to load the signal.

Combined Waveform: 10000 + 10500 Hz

Combined Waveform: 10000 + 10500 Hz

Combined Waveform: 1000 Hz + 1100 Hz

Combined Waveform: 1000 Hz + 1100 Hz

Finally, here is the family connected together to the resistive combining network.

Set Up for Combined Waveform Test

Set Up for Combined Waveform Test

This was an enjoyable but sometimes frustrating project. It’s definitely the densest thing I ever built and would be much simpler if I wasn’t too cheap to order a PC board. I learned that 0603 resistors are not a good choice for hand soldering. I learned that LM358 op amps suck. It is working great now though, and even though I have two other audio oscillators, this is a welcome addition to my test equipment stable.

de WB8NBS/9

 

Addendum

April 27 2018:  I acquired a Brother P-touch labeler. Labels help my feeble memory. The P-touch came with a short reel of 12 mm black on white background – these labels were made in two line mode which fits well on the edge of the lid. I will probably replace them with black on clear sometime as that would better keep the Altoids theme.

PTouch Labels Added

PTouch Labels Added

Advertisements
Advertisements