This project takes the popular Heathkit IG-18 (and 5218) audio generator to a whole new level of performance, replacing the main board with a design based-on the
HP-339A. The write-up is in the form of what I call construction notes, which is brief.

New Developments

PCB Created for the Project!

We are pleased to present contributions from reader Larry Burk, who created an excellent layout for the IG-18SL in Eagle CAE software. After doing the layout, he fabricated his own, single-sided PCB and successfully built the working generator. Performance results seem at least as good as the ones we obtained. In fact, noting that our distortion results were limited by the HP8903A analyzer we used, Larry was able to measure far lower distortion at midband frequencies, using special techniques. His project, with the PCB, is shown below.

Tips on Fabricating the PCB

For readers who want to fabricate the PCB, Larry provides the following tips (edited and augmented):

• The board is intended for single-sided, DIY fabrication. As such, the topside traces are to guide topside hand wiring and the silkscreen was for guidance, rather than application to the board. While some silkscreen marks may overlap pads, PCB vendors typically mask out the silkscreen over the pads, anyway.
• The overall dimensions of the board are 5.4 x 4.0-inches.
• The PCB provides for a few enhancements, over the IG-18SL schematic:
• Pot R54 (1kohm) is added, in series with R51. If the pot is used, reduce R51 to 2.2K. Adjust the pot for minimum (second) harmonic distortion.
• Two 0.1uF bypass caps (C49,50) are added at U4.
• The power supply seems quieter with two 10uF tantalum caps added, across C5 and C6. Those can be tack soldered across C5 and C6, under the PCB.
• Since Newark no longer has HA2625’s Dick Moore indicates that a LT1468 works very well for U1. It is a direct replacement, except that the compensation components are left out. Those would be C44,45,48, CR7,8, R27,28. Dick also left out C32,46 and R29, though he and Larry think that these last three might be useful. LT1468CN8 is available from Digikey or from Linear Tech directly.
• Eagle version-6 is required to open the .brd file.
• Eagle generates errors if an error check is run on the board. That’s because the red topside traces were only intended to indicate where jumper wires are used, not for actual PCB fabrication. As-is, the traces overlap. After inspection just approve all of them and it will be okay.
• In this build, the high voltage supplying the neon lamp caused noise/distortion which was hard to locate. It would be worthwhile during checkout to disconnect it at transformer end to see if it is adding anything, especially on the high value range resistances. [Larry’s knob inserts had been modified, requiring that the frequency vernier pot be oriented with its leads near the lamp. Normally, the leads will be positioned on the opposite side from the lamp —ed.]
• The Electroswitch 6P4T rotary switch for S1, is getting expensive; about $27 now. Mouser has an Alpha rotary switch (SR2921F-0304-19R0B-E9-S-W) for $5.16 that would work but a perf board for most of the larger caps would need to be made and mounted where the old power supply is located, similar to what Dick Moore did. It’s not as nice a switch.
• For the square wave generator, the CD40106B was substituted for the equivalent 74C14B, since the other chip was handy. R5 needed to be increased to 56K, to obtain a 50% duty cycle. Other specifications were not tested but the chips have similar speed ratings.
• The Heathkit attenuator resistors were changed to the HP style shown at the bottom of the page on Dick Moore’s site. It was a pretty easy job and is accurate.
• Gerber files for PCB fabrication are provided below. If instead, you wish to generate your own output files from the .brd Eagle file, the topside metal file should be discarded and not sent to board house, because of the overlaps mentioned above. The silkscreen can still be used, though I didn’t, since the PCB was made at home. If you wish, you can view the Gerber files using the free application, GC-Prevue. Note that it will show errors (which are harmless), due to the fact that informational parts of the layout extend past the Eagle zero-reference. That was needed to use free Eagle to create the board.
• Warning: Larry has kindly generated Gerber files to allow PCBs to be fabricated at commercial board vendors. However, those files have not at this point been used to make a board. Please proceed at your own risk. We would very much like to hear, your results, if you have boards made with the files —ed.

PCB Files for Download

The PCB files package includes:

• Readme IG-18SL Osc.txt - a list of the files.
• IG-18SL Osc.brd - Complete Eagle v.6 layout file.
• IG-18SL Osc.sts - Bottom solder mask
• IG-18SL Osc.sol - Bottom copper
• IG-18SL Osc.plc - Silkscreen
• IG-18SL Osc.drd - Excellon drill data
• IG-18SL Osc.dri - Readable ASCII summary of drill information.
• IG-18SL Osc.gpi - Readable ASCII summary of photoplotter information.

Formats: txt, dri, gpi are ASCII. The drd file is excellon drill data. Others are Gerber. Note that this layout is intended for a single sided board. The topside traces shown in the BRD file are only to guide topside jumper installation. They are not suitable for fabricating a topside copper layer on the PCB.

Intriguing New Developments from Dick Moore

In addition to creating the PCB for the IG-18SL, Larry Burk also contacted Dick Moore, who has a fascinating website here. Dick has done extensive work on various approaches to enhancing the IG-18. With one of Larry’s PCB’s Dick constructed the circuit and did some advanced measurements which showed that the distortion at lower frequencies could be far lower than what I was able to measure, with the limitations of the HP8903A analyzer. His write up on it is here. I must say that the distortion figures seem almost too good to be true (under 0.0001% at 1kHz, 0.0003% at 10kHz). However, his measurement methods, while indirect, certainly make sense. Note that Dick calls this build the “IG-339A.” As mentioned above, Dick also found that an LT1468 opamp works very well, in place of the hard-to-find HA3-2625.

He also experimented with using the original Heathkit tuning network with the new PCB circuitry, calling that build, the “IG-339B.” His write up on that version is here. The Heathkit tuning network uses a 10:1 ratio of capacitors in the bridged-T, versus HP’s 100:1 ratio. It originally seemed likely that the higher ratio was responsible for the low distortion performance of the HP-339 oscillator. However, he reports measured distortion figures under 0.0005% over 100Hz to 10kHz. While I find this very surprising, Dick obviously has done a lot of excellent work on this, so I would hardly question his results. I really wish that I had time right now to get back into this project but alas, there are so many others that I’m juggling, now. (Maybe you know the feeling—so many toys, so little time :)

Reader Comments

Posted by Steve L.

June 12, 2011 - 07:44 am

Hi Larry, Thank you for the kind comment. I did not notice much hum and the 20-1000Hz distortion measurement was at the floor of the analyzer. It's possible that the folks who dealt with the transformer were actually chasing a ground loop problem. The new board layout has the power supply section comfortably spaced from the audio circuits. I was also very careful about ground connections, as is reflected in the layout diagrams and schematic. So I didn't find a reason to change the transformer situation. Let me know how it goes!

Posted by Larry

June 11, 2011 - 11:02 pm

Thanks for link Steve. Regarding the power tran, I note you did not relocate,shield or change to toroid as some others have indicated. Did you see no improvement in doing so? I have a HP339A so I know what this circuit is capable of doing. Good work.

Posted by Steve L.

June 11, 2011 - 10:15 pm

Hi Larry, The HA3-2625-5 is available from Newark electronics here: Newark link The listing is a little confusing but it IS in the DIP package and they have 400 in stock and they will sell quantity-1 for $8.55. Note that after stock is depleted, they will only be available in multiples of 500. I did not try any other opamps, since I felt confident that this one would work well there (per the original HP design). However, I imagine that there are some high-performance opamps which could match it. This one has 100MHz gain-bandwidth, 35V/us slew rate and works from ±15V power supplies. I will be happy to help if you have any other questions. Steve

Posted by Larry

June 11, 2011 - 09:52 pm

Have you a source for the HA3-2625-5 IC? If not have you tried another low distortion opamp?

Posted by Steve L.

February 13, 2010 - 08:11 pm

Please feel free to add any comments or questions you might have. No sign-up is needed.

Add your comments here...

Copyright © 2010-2012, Stephen H. Lafferty. All rights reserved.