Hi Dick,
I've read over your pages and find them very interesting, nice work!
I was wondering if you've seen "Low Distortion Oscillator" by Linsley Hood
that was published in Wireless World 1977? He uses a beta multiplier on
the VAS, in fact it is a MPSA14 darlington and a current source load.
I wondered if you tried a current source or bootstrap for the VAS load
in the IG-18.
Also wondering if you've looked at composite OP amps to use in place
of the Harris for your #3 mod. Here's a Burr Brown app note about it:
focus.ti.com.cn/cn/lit/an/sboa002/sboa002.pdf
I've read over your pages and find them very interesting, nice work!
I was wondering if you've seen "Low Distortion Oscillator" by Linsley Hood
that was published in Wireless World 1977? He uses a beta multiplier on
the VAS, in fact it is a MPSA14 darlington and a current source load.
I wondered if you tried a current source or bootstrap for the VAS load
in the IG-18.
Also wondering if you've looked at composite OP amps to use in place
of the Harris for your #3 mod. Here's a Burr Brown app note about it:
focus.ti.com.cn/cn/lit/an/sboa002/sboa002.pdf
Last edited:
@PB2 -- Yes I actually have a copy of that article. In IG-18 #1, I do use a FET for a constant current source load on Q3, and it helped, but beta multiplying caused me problems at higher frequencies, especially near 100kHz; Miller C is a problem. I haven't looked at the Linsley-Hood paper for a while, but I think that he intended that design for maybe 10kHz, tops.
As to composite opamps, I haven't -- thanks for the link, I'll go have a look. One of the appeals of the 239A design is the sparse number of parts. I'm hopeful about the OPA1641....
As to composite opamps, I haven't -- thanks for the link, I'll go have a look. One of the appeals of the 239A design is the sparse number of parts. I'm hopeful about the OPA1641....
Yes the 239A does look quite simple, very nice, and being a production unit it hopefully has all the potential bugs worked out. It seems to do a bit better than the Linear AN-43 Fig. 47 design, yet the Fig. 47 design uses another op amp for common mode suppression for a virtual ground at the midpoint of the bridge. It also uses a composite made up of an LT1115 with an LT1010 output buffer. Looks like Figs. 47 and 48 are swapped in that app note.
Just noticed that you measured far better than spec for the 239A and I wonder if they are all that good.
Just noticed that you measured far better than spec for the 239A and I wonder if they are all that good.
Last edited:
Those 44H11s are 10A 50W devices, wow that is nearly a power
amp - the large resistors in the emitter leg certainly limit the current.
I'd think that those outputs would have excessive capacitance being
so large. Drivers are only 100mA max, I'd think it would be better
to use .5 A devices such as MPSA92 and comp.
The output stage is the same topology as the Tiger amps with
output protection.
R31 seems to be missing on the other side.
Last edited:
@PB2 -- The 239A and 339A oscillators are nearly identical -- I have a 239 and a 339, and they are indeed nearly identical in every way, except the 239 has the advantage of a switched ground for floating AC coupled ground or direct chassis ground, which is a big advantage, depending on the other gear being used. I have a friend with a 339, and he gets approximately the same results I do, so I would say "they're all like that."
As to the KH balanced output amp, I have a hard time understanding why that level of overkill was felt to be necessary, but then I don't know what the max output level is on the 1120, nor do i know how these things were actually used in military or industrial applications....
As to the KH balanced output amp, I have a hard time understanding why that level of overkill was felt to be necessary, but then I don't know what the max output level is on the 1120, nor do i know how these things were actually used in military or industrial applications....
Did you consider that all that's needed to run op amps on a single supply is to DC bias the inputs at about half the supply? That could be done in the IG-18 just by leaving in the bias pot circuitry. Essentially just replace all the transistors with say for example the LT1115 and LT1010 in Figure 47 of that App note, or perhaps leave in the emitter follower to buffer the OP amp. Maybe I'll take a look at something like this.
Wondering why the output was biased at 29V and not half of the supply, perhaps to heat the bulb to the proper DC operating point?
Wondering why the output was biased at 29V and not half of the supply, perhaps to heat the bulb to the proper DC operating point?
Well, that still means throwing away a lot of power to get the supply down to where the opamps are happy; I personally like using split supplies for opamps because in my experience they are quieter, and biasing uses more parts. Just a preference.
An easy way to use the Heath transformer is to wire the primary for 240V, which gets the secondary output down where using a FW bridge is a happy thing -- I never did this IG-18 #2 because I used a lower noise toroid transformer for my rebuild, but I will try this with the 239A mod I'm planning for IGT-18 #3 and see if the Heath transformer is quiet enough to let the oscillator do really well. It's an open question at the moment.
In the original, I've found you can't trust Heath's schematic voltages; don't know why. Anyway, the bias pot setting changes the bias point....
An easy way to use the Heath transformer is to wire the primary for 240V, which gets the secondary output down where using a FW bridge is a happy thing -- I never did this IG-18 #2 because I used a lower noise toroid transformer for my rebuild, but I will try this with the 239A mod I'm planning for IGT-18 #3 and see if the Heath transformer is quiet enough to let the oscillator do really well. It's an open question at the moment.
In the original, I've found you can't trust Heath's schematic voltages; don't know why. Anyway, the bias pot setting changes the bias point....
Interesting but looks like a lot of work:
Modifications: Heathkit IG-18 Audio Generator
Source for Harris/Intersil HA3-2625:
http://www.newark.com/intersil/ha3-2625-5/ic-op-amp-100mhz-35v-µs-dip-8/dp/06F5228
Modifications: Heathkit IG-18 Audio Generator
Source for Harris/Intersil HA3-2625:
http://www.newark.com/intersil/ha3-2625-5/ic-op-amp-100mhz-35v-µs-dip-8/dp/06F5228
I dusted off my old IG-18 and recapped it, the only mod being the 100uF upgrade for C5.
When I first looked at the schematics, years ago, a few things jumped out at me.
First the output stage seems to be class B or very low bias class AB.
Second the feedback network, and DC bias network have a very low impedance. This did not make sense at first but then I realized that the lamp has a low impecance and probably needs a good amount of current to warm it up. It seems to me that the output stage should be biased for class A to get the output devices into their best range for beta, Ft, and for the best linearity.
I've always been curious to measure the output stage idle current by way of the drop across R13 and R14. The bias on this unit is set per the manual with 22V at the emitter of Q4. I set the feedback for 10V RMS with the output set full, then put a .1uF cap from the wiper of the feedback pot to ground to stop the oscillation in order to measure the DC operating points. This worked as long as the frequency was set to 10 - 100 KHz. Oddly enough, I measured 62mV across R13 for 4.13 mA and 0.000 V across R14 indicating that the amp is clearly in class B. But the 4.13 mA has to be going somewhere and looking closer at the schematics it can be seen that the square wave generator is DC coupled with a path to ground. The voltage across R15 indicated that it was taking 1.74 mA and the majority of the remaining current must be going through the lamp, feedback pot, and to the bias pot since the base of Q1 and Q2 only require microamps to provide their bias point.
While the unit works, something does not seem right. Looking closer we can see that the base of Q2 is biased from the top of the feedback pot - pin 1 and the base of Q1 from the other side of the lamp L1 by way of RX, RY, and RZ. Ideally, for balance they should be biased to the same DC voltage but here they are not since there is some DC drop through the lamp.
I also measured the collector currents in Q1 and Q2 and they are far off balance.
My gut feeling from looking at these schematics years ago was that while it might work, it looks to be seriously flawed and I think this is confirmation. I don't think that there is a miswire or failed component in this unit.
Perhaps I should start a new thread.
When I first looked at the schematics, years ago, a few things jumped out at me.
First the output stage seems to be class B or very low bias class AB.
Second the feedback network, and DC bias network have a very low impedance. This did not make sense at first but then I realized that the lamp has a low impecance and probably needs a good amount of current to warm it up. It seems to me that the output stage should be biased for class A to get the output devices into their best range for beta, Ft, and for the best linearity.
I've always been curious to measure the output stage idle current by way of the drop across R13 and R14. The bias on this unit is set per the manual with 22V at the emitter of Q4. I set the feedback for 10V RMS with the output set full, then put a .1uF cap from the wiper of the feedback pot to ground to stop the oscillation in order to measure the DC operating points. This worked as long as the frequency was set to 10 - 100 KHz. Oddly enough, I measured 62mV across R13 for 4.13 mA and 0.000 V across R14 indicating that the amp is clearly in class B. But the 4.13 mA has to be going somewhere and looking closer at the schematics it can be seen that the square wave generator is DC coupled with a path to ground. The voltage across R15 indicated that it was taking 1.74 mA and the majority of the remaining current must be going through the lamp, feedback pot, and to the bias pot since the base of Q1 and Q2 only require microamps to provide their bias point.
While the unit works, something does not seem right. Looking closer we can see that the base of Q2 is biased from the top of the feedback pot - pin 1 and the base of Q1 from the other side of the lamp L1 by way of RX, RY, and RZ. Ideally, for balance they should be biased to the same DC voltage but here they are not since there is some DC drop through the lamp.
I also measured the collector currents in Q1 and Q2 and they are far off balance.
My gut feeling from looking at these schematics years ago was that while it might work, it looks to be seriously flawed and I think this is confirmation. I don't think that there is a miswire or failed component in this unit.
Perhaps I should start a new thread.
Hi Pete -- I'm doubtful that going full class-A on the output stage will make very much difference, but it certainly is worth a try if the current can be found and the devices can do it.
As designed, the output stage is definitely running class AB -- there is absolutely no sign of crossover distortion that I can see when the bias and feedback are set optimally. One can easily add some resistance to the base-spreading diodes to increase the output stage standing current -- 10mA seems reasonable; and uncoupling the square-wave generator makes sense all the way around.
I considered using a current-mirror for the first stage devices, as I mention on my page, but that's pretty far from a "simple" modification, and I just never got around to it. Richard Andresen has come up with a circuit for this that needs to be tested, and I haven't gotten around to that yet, either.
As to another thread, I'm OK with pursuing all of this here -- it's not hijacking.
Best,
Dick
As designed, the output stage is definitely running class AB -- there is absolutely no sign of crossover distortion that I can see when the bias and feedback are set optimally. One can easily add some resistance to the base-spreading diodes to increase the output stage standing current -- 10mA seems reasonable; and uncoupling the square-wave generator makes sense all the way around.
I considered using a current-mirror for the first stage devices, as I mention on my page, but that's pretty far from a "simple" modification, and I just never got around to it. Richard Andresen has come up with a circuit for this that needs to be tested, and I haven't gotten around to that yet, either.
As to another thread, I'm OK with pursuing all of this here -- it's not hijacking.
Best,
Dick
Thanks Dick, OK then I will continue here.
I did a rough simulation of the DC conditions in LtSpice and the behavior is very similar to what I'm seeing. With the square wave generator load, R5, and R3 disconnected there is about 360 uA Ie in Q4 and 100 uA in Q5 it really should idle at 10 mA or more for the best HF performance and linearity. When I put R3 in circuit then there is about 1.1 mA in Q4 and nearly zero in Q5, so it is R3 that throws off the DC balance. R5 throws it off even more, 220K makes it worse than what I'm seeing in real hardware, 1 Meg is a close match for what I'm seeing probably due to higher beta transistors in simulation.
If I add the square wave load then there is about 4mA in Q4 and nearly zero in Q5 - this clearly is not good. I consider this and R3 and R5 to be a design errors.
Also, I'm seeing 290 uA Ic in Q1 and 100 uA in Q2 which is very close to what I measured and is clearly not a well balanced diff pair.
It is clear that that the diff pair is not balanced and that one output device is starved for current, even with ideally matched transistors in simulation.
I believe that an early prototype probably used lower beta diff pair transistors that required much more Ib and therefore R5 was added. To maintain the balance a resistor should also be added to Q1 and perhaps R3 was used thinking that it would help maintain a common Vb - not a very good solution. R5 and R3 completely throw off the balance with higher beta transistors that were probably used in production and here in simulation.
I think I'm going to remove R15, R5 and R3.
I did a rough simulation of the DC conditions in LtSpice and the behavior is very similar to what I'm seeing. With the square wave generator load, R5, and R3 disconnected there is about 360 uA Ie in Q4 and 100 uA in Q5 it really should idle at 10 mA or more for the best HF performance and linearity. When I put R3 in circuit then there is about 1.1 mA in Q4 and nearly zero in Q5, so it is R3 that throws off the DC balance. R5 throws it off even more, 220K makes it worse than what I'm seeing in real hardware, 1 Meg is a close match for what I'm seeing probably due to higher beta transistors in simulation.
If I add the square wave load then there is about 4mA in Q4 and nearly zero in Q5 - this clearly is not good. I consider this and R3 and R5 to be a design errors.
Also, I'm seeing 290 uA Ic in Q1 and 100 uA in Q2 which is very close to what I measured and is clearly not a well balanced diff pair.
It is clear that that the diff pair is not balanced and that one output device is starved for current, even with ideally matched transistors in simulation.
I believe that an early prototype probably used lower beta diff pair transistors that required much more Ib and therefore R5 was added. To maintain the balance a resistor should also be added to Q1 and perhaps R3 was used thinking that it would help maintain a common Vb - not a very good solution. R5 and R3 completely throw off the balance with higher beta transistors that were probably used in production and here in simulation.
I think I'm going to remove R15, R5 and R3.
Last edited:
Here it is, it was you:
http://www.diyaudio.com/forums/equipment-tools/168718-heath-ig-18-ig-5218-a-2.html#post2297243
I'll try 250 ohms in the simulation but I don't expect it to change the results since they do match real hardware fairly well.
Dick wrote this on the other thread:
"The lamp itself is marked SYL 90V. I was able to cross-ref to a Sylvania 90MB, which is a T2 bayonet-mount bulb with the right envelope shape, 90 volts at 30mA, giving a hot resistance of 3k ohms. I pulled one lead of the lamp and its cold resistance is around 280 ohms. I say around because even the small current from the ohmmeter raised it to 290 ohms while I was measuring. This all seems to be in the right ballpark to me, even though the cold R seems low. "
We'd never see 3K ohms since it never sees the full 90V obviously.
Have you modded your unit Jim? I noticed you mentioned working on yours in the other thread.
http://www.diyaudio.com/forums/equipment-tools/168718-heath-ig-18-ig-5218-a-2.html#post2297243
I'll try 250 ohms in the simulation but I don't expect it to change the results since they do match real hardware fairly well.
Dick wrote this on the other thread:
"The lamp itself is marked SYL 90V. I was able to cross-ref to a Sylvania 90MB, which is a T2 bayonet-mount bulb with the right envelope shape, 90 volts at 30mA, giving a hot resistance of 3k ohms. I pulled one lead of the lamp and its cold resistance is around 280 ohms. I say around because even the small current from the ohmmeter raised it to 290 ohms while I was measuring. This all seems to be in the right ballpark to me, even though the cold R seems low. "
We'd never see 3K ohms since it never sees the full 90V obviously.
Have you modded your unit Jim? I noticed you mentioned working on yours in the other thread.
Last edited:
I have to say that I never liked this design, the schematic looked odd to me in many ways and the trim adjustments are touchy, there are things that just do not make any sense.
I noticed that it is possible to set the frequency to zero and thought that perhaps this would be a good way to measure the DC operating point with the oscillation stopped. Tried it and took a measurement, the output goes to the positive rail - time to look at the schematics. Rx, Ry, and Rz go to infinity when the frequency is set to zero and thus there is no DC negative feedback path to stabilize the "OP" amp's DC output. In fact, the only bias provided to the base of Q1 is through R3 and it seems that the designer was trying to hold the diff pair bases at the same voltage which would work for an ideally balanced OP amp but unfortunately this is not what we have here and is not really possible with thermal considerations - some negative feedback is needed.
I'm sure most here know that with OP amps that draw a small but significant Ib at the inputs it is important to have the same DC path resistance to the bias supply for each input. This would be the Thevenin source impedance seen looking out of the base of Q1 and Q2. The best solution would have been to use a FET input or any OP amp with ultra low input Ib requirements, then the source resistance balance is not critical. This is what was done in the HP239A where a Harris HA-2625 OP amp is used.
R6 (10K) is a highly compromised value for Q2 since the path for Q1 is completely dependent on the frequency control settings. About 1K ohm at the lowest settings and 1Meg at the highest when the tens and units are set to zero. I don't have the full manual and I'm not sure if there are any warnings about not setting the tens and units to zero, it obviously doesn't work well if at all. In fact, this unit has highly reduced amplitude with the tens and tenths set to zero and the units set to 1, clearly it is better to set it to 10 and drop down by one on the range switch. Then the combined resistance is 10K and a match for R6.
One has to wonder if R5 was an attempt to partially null the Ib of Q1 and if it was incorrectly put on the base of Q2.
Does anyone know if the full manual is available online?
I noticed that it is possible to set the frequency to zero and thought that perhaps this would be a good way to measure the DC operating point with the oscillation stopped. Tried it and took a measurement, the output goes to the positive rail - time to look at the schematics. Rx, Ry, and Rz go to infinity when the frequency is set to zero and thus there is no DC negative feedback path to stabilize the "OP" amp's DC output. In fact, the only bias provided to the base of Q1 is through R3 and it seems that the designer was trying to hold the diff pair bases at the same voltage which would work for an ideally balanced OP amp but unfortunately this is not what we have here and is not really possible with thermal considerations - some negative feedback is needed.
I'm sure most here know that with OP amps that draw a small but significant Ib at the inputs it is important to have the same DC path resistance to the bias supply for each input. This would be the Thevenin source impedance seen looking out of the base of Q1 and Q2. The best solution would have been to use a FET input or any OP amp with ultra low input Ib requirements, then the source resistance balance is not critical. This is what was done in the HP239A where a Harris HA-2625 OP amp is used.
R6 (10K) is a highly compromised value for Q2 since the path for Q1 is completely dependent on the frequency control settings. About 1K ohm at the lowest settings and 1Meg at the highest when the tens and units are set to zero. I don't have the full manual and I'm not sure if there are any warnings about not setting the tens and units to zero, it obviously doesn't work well if at all. In fact, this unit has highly reduced amplitude with the tens and tenths set to zero and the units set to 1, clearly it is better to set it to 10 and drop down by one on the range switch. Then the combined resistance is 10K and a match for R6.
One has to wonder if R5 was an attempt to partially null the Ib of Q1 and if it was incorrectly put on the base of Q2.
Does anyone know if the full manual is available online?
Last edited:
I have not modded mine, just recapped it and cleaned the contacts. I had big plans but richiem, after a great effort, did not get the kind of improvement I was hoping for. My simulation efforts did not lead me to an easy fix. I have been away from this for some time, but recall that I had to set the lamp resistance much higher than is reasonable (5k) to get the simulated circuit to act reasonably. This was probably due to a bad spice model.
Jim
Last edited:
Just a quick note to say that at some point, I'm going to try a current mirror for the collectors of Q1 & 2 and see about cleaning up the various imponderables here. I've been doing some gratis repair work on PCR machines for a microbiologist in New Orleans, and also playing with some ESI and BECO impedance bridges, and have had a problem getting back to the IG-18 and my HP 334 project. All in good time, I guess....
@PB2 -- I have an SG-5218 schematic for you -- due to size, it's in several pieces. PM me to get a copy.
@PB2 -- I have an SG-5218 schematic for you -- due to size, it's in several pieces. PM me to get a copy.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.