Re-greening the Heathkit IG-18

[PB2]

OK Jim sounds good!

[PB2]

I've mentioned that the amp is unity gain, that is obviously around the loop that provides oscillation, however the bridged-T notch filter provides 15.6 dB of notch depth giving the amp 15.6 dB of gain and then the positive feedback path should have 15.6 dB of attenuation to provide proper feedback for oscillation. Away from the notch the negative path has no attenuation and the amp is then unity gain at frequencies where it might spuriously oscillate. Here is a link to Dick's discussion on the IG-18 with excellent coverage of the bridged-T filter:
IG-18 #1

I noticed that the design is very sensitive to any Cdom capacitance and this is due to the diff pair not having enough available current to charge the capacitance, it slew rate limits at higher frequencies with even 10 pF of Cdom capacitance. This means that the design will be very sensitive to the internal capacitances of the VAS transistor.

I've attached an LtSpice simulation of the stock IG-5218 (R3 = 56K and C5 = 100uF) where I removed the positive feedback loop so that we can look at the performance of the amp in the 15.6 dB gain configuration. I simplified the negative feedback loop for flat response as an amp rather than an oscillator. It shows all of the DC bias issues that I've mentioned. Note the imbalance in Q2 and Q3 Ic currents, and then how Q2 current clips when the experimental Cdom network is connected and the output shows signs of slew rate limiting.
Also note the currents in R8, R10, and R20.

[PB2]

Here is the sim file showing the frequency response with the negative feedback network set up normally for 100 kHz oscillation and the positive path open and driven as an amplifier. The roughly 15.6 dB of gain is clearly seen at 100 kHz and there is some peaking in the response up around 10 MHz.

The circuit can be observed as an oscillator by changing R100 to 1000 Meg and tying in the 900 ohm resistor simulating the lamp to the feedback trim center point. Change the simulation type to transient. The oscillation builds and then clips due to the fact that the resistor does not provide amplitude control as a real lamp would.
I just noticed that if you connect the Cdom network it does not clip.

[PB2]

Originally Posted by PB2

Note the imbalance in Q2 and Q3 Ic currents, and then how Q2 current clips when the experimental Cdom network is connected and the output shows signs of slew rate limiting.
Also note the currents in R8, R10, and R20.

Forgot to update this after I fixed the reference designators in the simulation to match the original schematic. Here it is corrected:
Note the imbalance in Q1 and Q2 Ic currents, and then how Q1 current clips when the experimental Cdom network is connected and the output shows signs of slew rate limiting.
Also note the currents in R13, R14, and R215.

[davada]

In the absence of a lamp model anywhere on the net that I could find I made one.

The lamp models the VI behavior of the SYL-90 lamp used in the Heath IG-18 models.

The model emulates the lamp converting the VI curve to resistance.
A resistors behavior is modified to follow the resistance curve of the SLY-90 as a function of the voltage across the resistor, AKA lamp.

Dick provided the VI data from his IG-18 lamp.

Unzip the file and install the .sub and .sym files in the LT spice library sub and sym file folders.
Open the .sub file in LT spice to view the sub circuit.

If anyone is interested in how this was done then I will document the procedure.

Cheers,

David.

[PB2]

Wow! Nice work gentlemen! Thank you David and Dick.

Are either of you on the Yahoo LtSpice group and if so, do you know if anyone has done an oscillator example like this there? I'm a member but have not looked carefully at the uploaded files there.

[davada]

Originally Posted by PB2

Wow! Nice work gentlemen! Thank you David and Dick.

Are either of you on the Yahoo LtSpice group and if so, do you know if anyone has done an oscillator example like this there? I'm a member but have not looked carefully at the uploaded files there.

Thanks Pete.

I never thought of looking there.
I found another lamp model in the group files.
It's done completely differently than this one and has a much nicer lamp symbol.

Have a look.

I don't know where one could get such detailed information on a lamp to customize
a lamp for this model.

At least mine is DIY friendly. Measurements taken from the lamp.

[davada]

Hi Pete,

Here is the instruction and explanation of how the lamp was modeled.

Any lamp model can be fit to the sub circuit by replacing the coefficients.

Cheers,

David.

[PB2]

Thanks David, once again nice work!

[PB2]

I've mentioned several times that this unit is very touchy. There are large disturbances in the output if I tap the circuit board and especially the lamp. Has anyone else observed this, is it normal for this unit? I've looked for cold solder joints and cracked lands in the PC board and have not found any. Is the lamp "microphonic"? Perhaps I need to replace the lamp.
I've purchased a few more lamps so that I can prototype another unit with new parts - I might just use an OP amp since that seems to be the logical way to go. It is easier than building up the amp to a Jensen 990 level for example and obviously the HP 239A has outstanding performance with an OP amp as Dick has shown:
IG-18 #3

There are some challenges in this design. The 10V RMS output really pushes many OP amps to their slew rate limits and certainly to higher distortion especially at 110 kHz and into a 600 ohm load. I believe that this can be solved by having the OP amp based oscillator put out something like 1 to 2 V RMS and then use inverting OP amp buffers to increase the output to 5 or 10 V RMS. OP amps are inexpensive so the amplification could be done in 2 inverting stages. Might make sense to put one stage in front of the attenuator and another after in order to make the output less sensitive to differences in loading.

The lamp is a limitation in low frequency distortion due to the thermal time constant and the obvious solution is to use a FET as is done in the HP 239A.

I had not planned to spend too much time on this, especially for the basic mod, so I might just abandon simplistic upgrades to what I view as a somewhat flawed design.