Thanks to everyone who's sent manuals or helped.
I certainly understand about real work Dick, and have enjoyed reading your pages very much!
I've just started testing my unit with ARTA and an M-Audio Transit that does 96K sampling at best. I've settled on 9KHz as the test frequency so that I can see the 5th harmonic. I saw about .09% after the recap and cleaning with the only mod being the 100uF cap in place of C5.
Removed the square wave load by lifting R15, and R5 which throws off the output balance and saw the distortion reduced to .068%, but then later it was around .075%, not sure if this was due to warm up.
This unit was built in 1973 and I think it already has the grounding mod.
I certainly understand about real work Dick, and have enjoyed reading your pages very much!
I've just started testing my unit with ARTA and an M-Audio Transit that does 96K sampling at best. I've settled on 9KHz as the test frequency so that I can see the 5th harmonic. I saw about .09% after the recap and cleaning with the only mod being the 100uF cap in place of C5.
Removed the square wave load by lifting R15, and R5 which throws off the output balance and saw the distortion reduced to .068%, but then later it was around .075%, not sure if this was due to warm up.
This unit was built in 1973 and I think it already has the grounding mod.
Well, there are grounding mods and grounding mods -- they really matter with this old dog.
I found the use of the constant current FET for Q3 to be the most helpful thing I did, once I put in really good matched transistors for Q1 & 2. The second most helpful was raising the value of the emitter resistance on Q3 and then raising the collector load on Q2.
Please keep us up to date with any other mods you make.
I found the use of the constant current FET for Q3 to be the most helpful thing I did, once I put in really good matched transistors for Q1 & 2. The second most helpful was raising the value of the emitter resistance on Q3 and then raising the collector load on Q2.
Please keep us up to date with any other mods you make.
I have been thinking along the same lines but I'm going to also bias the output well into class A.
Yes, raising the Q3 emitter resistance is one way to balance the diff pair currents. I'd also bypass it with a large cap for maximum open loop gain. We might want to try a small Cdom cap and see if it lowers the distortion at 50 - 100 KHz.
I think they used 1.5K for the Q3 emitter resistor, then bypassed it with a big cap - see the link to pictures:
http://cgi.ebay.com/Heathkit-IG18-V...Electrical_Test_Equipment&hash=item2ea68eb193
Q3 current source load makes a lot of sense to maximize the open loop gain.
I will keep you up to date, but I'm looking around for a better distortion analyzer, or perhaps a low cost USB sound device that will do 192KHz with low noise and distortion. An EMU 0202 or 0404 perhaps. I want something that will work with a laptop but those are discontinued unfortunately.
Yes, raising the Q3 emitter resistance is one way to balance the diff pair currents. I'd also bypass it with a large cap for maximum open loop gain. We might want to try a small Cdom cap and see if it lowers the distortion at 50 - 100 KHz.
I think they used 1.5K for the Q3 emitter resistor, then bypassed it with a big cap - see the link to pictures:
http://cgi.ebay.com/Heathkit-IG18-V...Electrical_Test_Equipment&hash=item2ea68eb193
Q3 current source load makes a lot of sense to maximize the open loop gain.
I will keep you up to date, but I'm looking around for a better distortion analyzer, or perhaps a low cost USB sound device that will do 192KHz with low noise and distortion. An EMU 0202 or 0404 perhaps. I want something that will work with a laptop but those are discontinued unfortunately.
Last edited:
Feedback Pot
It looks to me as if pins 1 and 2 on the feedback pot should be reversed. The lamp would then act as the series feedback element with the full pot resistance as the shunt load to AC ground. Then the wiper just picks off the correct amount of feedback as a voltage divider.
I believe that the feedback pot could be overloaded (burnt out) with the osc putting out 10V and the pot at the bottom of the range, perhaps it stops oscillating before this can happen. Reversing pins 1 and 2 makes more sense to me.
It looks to me as if pins 1 and 2 on the feedback pot should be reversed. The lamp would then act as the series feedback element with the full pot resistance as the shunt load to AC ground. Then the wiper just picks off the correct amount of feedback as a voltage divider.
I believe that the feedback pot could be overloaded (burnt out) with the osc putting out 10V and the pot at the bottom of the range, perhaps it stops oscillating before this can happen. Reversing pins 1 and 2 makes more sense to me.
Bypassing the Q3 emitter resistor cause me no end of grief at any frequency above a few kHz -- serious instability that I couldn't correct; when you think about the equivalent GBW product of very high open loop gain and operation at 100kHz, you'd need a whole rack of RF amps to do the job.... So open loop gain has to be high enough, but not too high, and the rest comes down to great linearity everywhere else.
As to the feedback pot, the lamp runs at between 800 and 1200 ohms in normal operation, and when the output is fully across the lamp, the current is limited by the lamp, so I dont' think there's a possibility of damaging the pot. After that, I don't think it matters which way you wire it up... I don't see a performance difference being likely.
As to the feedback pot, the lamp runs at between 800 and 1200 ohms in normal operation, and when the output is fully across the lamp, the current is limited by the lamp, so I dont' think there's a possibility of damaging the pot. After that, I don't think it matters which way you wire it up... I don't see a performance difference being likely.
It is quite difficult to get a USB sound device to sample at the higher rates due to limits to USB and computer latency. That said I did get an EMU 0202, but I have to turn off all other software and my wireless ethernet to get low distortion readings at high sampling rates, eg 96 kHz or above. There is some information on the web about doing this. Sorry, but I have lost the links.
Whenever USB 3 comes out things will probably be fine.
Jim
I have read this thread with interest, looking to resurrect the IG-18 I have here. I want to make the thing work stock before doing any upgrades. However, the unit I have is substantially different from the schematics I find. The available schematics show 8 transistors in the wave generator board, and one transistor in the power supply. The IG-18 I have has 3 transistors in the power supply and 18 transistors in the generator circuit. In the corner of each board on the foil side is WM-1 and WM-2 for the generator and power supply boards respectively. It sure looks stock.
I'd like to know what I have here, before I trace it out by hand.
I'd like to know what I have here, before I trace it out by hand.
You have Morrey's super oscillator, a 1975 Audio Amateur modification. I think the schematic and article are available on the web. If you can't find them, I can send them to you.
I've increased the emitter R for Q3 to balance the diff pair currents, and bypassed it with 100 uF and this unit is working fine but I do not have a current source load for Q3 yet so perhaps that would be an issue with the increased gain. You might have reached a point where the amp was no longer unity gain stable with the current source load - not sure.
The feedback pot is very touchy in my unit and I suspect that it is damaged, not sure. The board is also very touchy in that area, just taping around that area. Anyway, it is simply better to wire the pot correctly. I'll check the pot with an ohm meter.
Thanks Jim, I had a feeling that USB might be an issue but I don't mind disabling things to have a portable system. I'll look for some tips on the web if I ever get a better card.
I'll probably try a current source load for Q3 next. Was also thinking about the beta multiplier on the VAS and wanted to point out that Reg's load resistor of 220K is way too high, it needs to provide the turn off current for Q3 and 1 to 10K is more reasonable. A good example of a high performance OP amp is the Jensen 990 where 2K is used in that position:
http://www.johnhardyco.com/pdf/990-2007.pdf
http://www.diyaudio.com/forums/solid-state/107404-simulation-je-990-op-amp-deane-jensen.html
The VAS is biased at about 6 mA, and the outputs even higher. It is designed to easily drive 600 ohm loads which lines up well with this design since it has to drive potentially a 600 ohm load in parallel with the low Z feedback network.
http://www.johnhardyco.com/pdf/990-2007.pdf
http://www.diyaudio.com/forums/solid-state/107404-simulation-je-990-op-amp-deane-jensen.html
The VAS is biased at about 6 mA, and the outputs even higher. It is designed to easily drive 600 ohm loads which lines up well with this design since it has to drive potentially a 600 ohm load in parallel with the low Z feedback network.
Yes, I measured the lamp here, with 10V RMS out, at about 900 ohms by putting a 100 ohm resistor in series with it and measuring the current then voltage across the bulb.
The pot measured OK, but I tried swapping pins 1 and 2 and the adjustment seems to work better. Perhaps it is just the wiper making and breaking contact.
I agree, I do not expect much if any difference in performance. There was the possibility of output loading on the amp, since with the pins swapped the lamp is always in series with the full 750 ohms of the pot for minimal loading.
Oscillations
I'm starting to see HF oscillations and was wondering if anything in the grounding helps to stop this.
What I have found partly by accident, is that the oscillations seem to be very sensitive to the switch settings and wiring. I tested this theory by adding a series RC of 1K and 10 pF from H to K on the PC board to give the negative feedback path a low inductance local shunt path. This seemed to help a lot but it is still not completely gone. I missed, at the time that I tried this, the fact that Cx shunts this path and it would be easy to put say 30 pF local to the PC board and make the X1000 500 pF cap 470 pF instead. The 30 pF introduces a very small error with the other larger caps.
The Cy return to point J on the PC board is essentially going to AC ground and since there is no DC component it could go directly to actual ground. Has anyone tried this or have any thoughts concerning the change? If point J is strongly decoupled to ground such as with a good quality 100 uF then it is not an issue but with the .68 or even 6.8 uF it just seems wrong - directly to ground would be better.
I'm starting to see HF oscillations and was wondering if anything in the grounding helps to stop this.
What I have found partly by accident, is that the oscillations seem to be very sensitive to the switch settings and wiring. I tested this theory by adding a series RC of 1K and 10 pF from H to K on the PC board to give the negative feedback path a low inductance local shunt path. This seemed to help a lot but it is still not completely gone. I missed, at the time that I tried this, the fact that Cx shunts this path and it would be easy to put say 30 pF local to the PC board and make the X1000 500 pF cap 470 pF instead. The 30 pF introduces a very small error with the other larger caps.
The Cy return to point J on the PC board is essentially going to AC ground and since there is no DC component it could go directly to actual ground. Has anyone tried this or have any thoughts concerning the change? If point J is strongly decoupled to ground such as with a good quality 100 uF then it is not an issue but with the .68 or even 6.8 uF it just seems wrong - directly to ground would be better.
I forgot to mention that, of course, the theory is that there is some inductance in the wiring from the switches to the PC board and while it is probably not significant even at 100 kHz, the amp is flat out to 1-3 MHz in this unity gain configuration where it does make a difference. The cap range switch is wired to the Tens switch and it would be better to star wire all the switches to the PC board to minimize the inductance for each run. The caps have the greatest potential for problems since they can form another resonant circuit with the lead inductance so that switch should certainly go straight to the board.
There are two ways to get 20 kHz:
The Tens set to 20, Units 0, Decimal 0, range X1000
Tens set to 10, Units 10, Decimal 0, range X1000
These use the same caps and the swtich settings provide the same resistance, with different resistors and wiring, and the distortion is different between the two settings .0045% second HD in one case and .0065% with the Tens set to 20.
I expect that this is due to wiring inductance and might be cleaned up with star wiring for the switches - I'll give it a try. Push button switches with short traces to the amp would be best as was done in the HP.
I'm getting a low of .0024% THD at 9 kHz but it is quite touchy at this point.
I'm thinking that less than .01% THD from 20 to 20 kHz is a reasonable goal without changing transistors or any drastic mods, but I'll try for less than .005%.
There are two ways to get 20 kHz:
The Tens set to 20, Units 0, Decimal 0, range X1000
Tens set to 10, Units 10, Decimal 0, range X1000
These use the same caps and the swtich settings provide the same resistance, with different resistors and wiring, and the distortion is different between the two settings .0045% second HD in one case and .0065% with the Tens set to 20.
I expect that this is due to wiring inductance and might be cleaned up with star wiring for the switches - I'll give it a try. Push button switches with short traces to the amp would be best as was done in the HP.
I'm getting a low of .0024% THD at 9 kHz but it is quite touchy at this point.
I'm thinking that less than .01% THD from 20 to 20 kHz is a reasonable goal without changing transistors or any drastic mods, but I'll try for less than .005%.
Thanks for the updates. I would be happy with 0.005 for most of my testing and 0.002 would be great. Like a typical Canadian, everything else is on hold while I enjoy the summer weather, mainly hiking, swimming, fishing and the occasional BBQ and beer, but will return to the soldering iron during the fall. Getting back to this will be one of my first projects, so any progress you make will be welcome.
Thanks
Jim
Thanks
Jim
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.