CD4069 vs 40106

[Rob Strand]

Can you estimate the load on the 62V rail?

You can work it out by measuring the voltage *across* each plate resistor and calculating the plate current,
Ip1 = VRp1 / Rp1
Ip2 = VRp2 / Rp2
...
Total current = Ip1 + Ip2 ...

My guess for a couple of tubes at that voltage you might be looking at 500uA to 1mA max.  Could even be less.

Assuming the article is correct and the 2k7 + 1n parts gives indicated 100kHz oscillation frequency.
I'm assuming the 40106 switches are about 200 ohms based on the datasheet typicals at Vdd=10V.

The difference between 22n caps and 47n caps is pretty small, perhaps 1V.   So beefing up the caps might not give much.

At 1mA load I'd expect to see the voltage drop to about 55V to 56V.

If the total B+ current is higher then obviously you expect to see more voltage drop but it needs to be quite a lot of current to drop the down to 45V.

I've assumed 200 ohms switches for the 40106 which seems reasonable to me.   However, if your switches are higher resistance for some reason then that will make the sag worse.

[This guy verified the impedances,
https://2n3904blog.com/cd40106-output-drive-characteristics/
They are about 140 ohms for low and 220 ohms for high.
So the 200 ohms  is a good assumption.]

Perhaps you could check the clock frequency is about 100kHz.  A low clock frequency will stuff things up.
Based on the article, I'm also assuming you are using 47nF caps? and 1N4148/1N914 diodes?

[PRR]

... why the output voltage of a 40106 (around 62V) would drop down to about 45V when you add a 6V voltage regulator to the power supply?....

The usual 'regulators' are rated 35V or 40V MAX input. Which implies they may act as '45V Zeners' when over-volted.

[Rob Strand]

The usual 'regulators' are rated 35V or 40V MAX input. Which implies they may act as '45V Zeners' when over-volted.

6V regulator branches off the 9V supply for the tube filaments

He's got the regulator off the 9V DC.  So no problems there.

I still think it's worth going through the motions of things to check in Reply #20.

All the stuff below is crap.  Ignore it.   

I had another look this morning to see what was going on and realized I stuffed up the gate model.
The sad thing is I checked the gate model before I started and didn't see the blatant error. 

My apologies. 


I did some more digging and I found something but it may not be the cause of your problems.   

if you check out the schematic,
http://zwz.cz/Obrazky/110V%20voltage%20multiplier%20with%2040106.png

Notice that each inverters are in cascade that means if the gates were ideal the output on every second gate is the exact same signal.

In practice the gates have a delay and also an output impedance.

The delay means the signals get slowly out of whack as we go from left to right on the schematic.   The heavy cap loading on the output probably won't help the cause. If we wired the input of every second inverter together that would synchronize all the outputs.   It's not that simple because the first gate is used as an oscillator and we can't connect all the input there.   Anyway ignore that detail for this thought experiment.

My previous 55V result didn't have the cascaded inverters it was all parallel (to simplify the simulation).  Later I thought about what the delays would do.

To cut a very long story short.    Many configurations with delays didn't work well.  I tried wiring caps in reverse order on the gates and that didn't help.   I even tried parallel connections and that didn't work well either.

The root cause of the problem is creating a complementary using an invert.   Not actually the cumulative delays.  So you have the clock from the oscillator then you create the complementary signal by passing that clock through an inverter it's stuffs things up.  It seems a benign process but that is main thing that stuffs things up.   If I created the complementary output synchronously, for example the Q and /Q output from a flip flop then everything works fine.  And how things get connected after that seem to produce only minor differences in performance.
[As a caveat I deliberately had spice gate models which act a bit dodgy so I could see the effects. But it's possible I went too far and that has exaggerated this issue.]

Making a synchronous complementary output will mean adding flip flops.  We don't want to do that if we avoid it.

The synchronous clock issue doesn't occur with the Co-ckcroft-Walton multiplier.

Anyway, the converter in this circuit might be the answer, since it doesn't need a complementary clock,
http://recordinghacks.com/images/mic_extras/cad/E300-schematic.png

[Paul Marossy]

40106 has its own 9V supply and is measuring 9.2V on Pin 14. 6V regulator branches off the 9V supply for the tube filaments, just like on the schematic. 9V goes into the 7806 and 5.8V comes out. That's why I am perplexed. Can't understand why the voltage drop on the output of the 40106 when that is added to the power supply.

Try pulling the tubes out.

When the tube heaters are off they pull no plate current.  But now when the heaters are powered the tubes start pulling a plate current.  That plate current loads down the B+ rail and it sags down.   The output impedance of those converters is fairly high so the voltage has a tendency to sag under load.   If you can see ripple on B+ rail under load then beefing up the caps on the converter cap can prevent sag as the ripple causes the average DC voltage to drop [20V drop would need 40V ripple].   It can turn out the ripple is low and the voltage still sags in these cases the larger caps don't help much.   What's happening is the on resistance of the switches (in the 40106 gates) have unavoidable voltage drops and that "loses" voltage to the caps.    You can lower on-resistance by paralleling inverters.   Paralleling some of the inverters earlier in the converter might have more effect than paralleling later ones - I haven't analysed the details.

Yeah I expect a little bit of sag but not 15-20 volts. I've built this thing twice now with two different PCBs I designed and same results. I get the big voltage drop on the output of the 40106 and I have other weird problems... I can't use a regulated power supply because after 3-4 minutes the 7806 drops down to half the output voltage. If I use a wall wart, that doesn't happen but I get major noise, like a very aggressive 60 cycle sounding hum. I tried to get around that with with a hum buster circuit but then it won't power on at all. WTF?! I don't get it. I feel like a complete idiot... with 20 years of guitar effect building experience and it's kicking my ass like this. 

Are all 40106s "created equal" or are there buffered/unbuffered versions like the CD4049? In the case of the CD4049 if you use the buffered version in the Three Legged Dog you get a squealing pig, but with the unbuffered chip it works as intended. I'm wondering if I have something like that going on here. I suspect there's some kind of oscillation going on, and maybe that explains the angry sounding noise with the wall wart? In any case, the PCB I ordered online from Germany has arrived, so if I build that and still this problem persists then it has to be the regulator or the 40106 that is the problem. Right now the two don't want to play together. 

[Rob Strand]

Yeah I expect a little bit of sag but not 15-20 volts.

It's normal to get a lot sag on those things.  How much to expect depends on a lot of details.   Spice simuations are a good guide about what to expect.  I'm seeing 54V with about 1mA.   Without knowing the real load that's still a guess.  If the load is more then 1mA then I'd expect more drop.

I can't use a regulated power supply because after 3-4 minutes the 7806 drops down to half the output voltage.

That could be too much power dissipation in the 7806.   When it gets hot the over temp shuts it down.  As to why, you would need to measure the output current from the 7806 and work out if the part is getting too hot.

I don't think you should be regulating the voltage to the DC converter (40106).   For hum, you can try RC filters at the output of the 60V rail.   it could also be heater related.

One thing about those converters is you need to put a large cap across the input supply.  So right across the 40106 power pins.  The first diode needs to connect to close to the 40106 power rail and the ground of the last cap needs to connect near the 40106 ground.    With the input supply cap you can get a whole heap of noise problems.  The noise could easily be misinterpreted as oscillations.   The lack of supply cap can show up as droop on the output.  Start with 100uF.   It really should be a low ESR cap, normal caps could be hit and miss.    You should also verify the frequency of the oscillator of the first stage.  If it's >= 300kHz it might create problems and it puts more demands on the quality of the 100uF cap.  Down around 100kHz things are more manageable.   Too low a frequency and the 47nF caps in the converter will start to produce too much ripple.

You don't want any of the grounds around the 40106 converter to pass be shared with audio.    The 100uF input caps helps prevent current pulse from the convert get into the audio but it can only do that if the whole converter layout is tight.

Are all 40106s "created equal" or are there buffered/unbuffered versions like the CD4049? In the case of the CD4049 if you use the buffered version in the Three Legged Dog you get a squealing pig, but with the unbuffered chip it works as intended.

The buffered vs unbuffered this is a big issue because the buffered version has more gain and that promotes oscillations.   On the Three Legged Dog the gates are being used as linear amplifiers.     On the DC converter the 40106 is operating as a switch so you won't get the same types of problems.   Where you can get problems is parts tolerances on the 40106 affecting the oscillator frequency.   You can also get variations in the switch resistances which could promote sag.   Dodgy ebay parts might not even meet the spec and they could have high resistances and a lot of sag.

[Ben N]

PAiA used hex invertors as charge-pumps for several of their tube-based projects.  Here, a 4049 is used to produce 50V for a phantom-powered mic preamp.

So that one uses three inverter RC oscillator and only has three inverters in parallel for the boosted output.

12ma.

[Rob Strand]

PAiA used hex invertors as charge-pumps for several of their tube-based projects.  Here, a 4049 is used to produce 50V for a phantom-powered mic preamp.

So that one uses three inverter RC oscillator and only has three inverters in parallel for the boosted output.

12ma.

I think 12mA would be the input current on DC rail to the multiplier circuit?F

For 12mA in on the DC corresponds I estimate to about 1.7mA load at the output.   
The voltage drops from 62V down to 50V.

Is there two tubes?  1.7mA would be need three or more tubes.

On the simulator I'm seeing,
IL,         Vo,        Iin
1uA,      61.6V,   near zero DCin
1mA,      53.7V,  7.1mA DCin  (100mVpp ripple)
1.7mA,   50.0V,  12.0mA DCin

Edited: corrected values and removed the false 2.6mA offset due to the simple gate model I'm using.

[Paul Marossy]

Update: I built the PCB I received in mail and the strange power supply issues persist but is working as expected otherwise. I can't power it with my benchtop LM-317 based power supply, because after a few minutes the 7806 drops down to 50% the voltage it is supposed to output. If I use straight 9V wallwart I get massive aggressive hum. If I use same wall wart with a different LM-317 regulator box I made, the hum is significantly reduced but not eliminated. If I try to use the Anderton hum buster circuit with it, it will not power on at all. I don't get why I should be having these problems. I forgot to measure the output of the 40106 but I expect it's higher than on my previous build(s). I measured around 36V on one of the plates, and I think the other was around 21V or something like that. I will measure when I get it boxed up. Maybe I kept making same mistake but I couldn't find any mistakes. 

BTW Rob Strand, this circuit uses one 12AU7 or 12AT7. 12AX7 is too noisy.