Dual Polarity Voltage Multiplier...

[rogs]

I wanted to try out a dual polarity voltage multiplier, for a new version of my OPIC multi pattern mic project.
I've always had good results using multipliers based on circuits in THIS PAPER which Jakob uploaded a while back.
(No oscillator inductor values or orientation to worry about! )
But creating a dual polarity version of that circuit needs quite a lot of components - and makes for a difficult layout (especially on stripboard ! )
So I thought I'd try a slightly different variation on the same idea, based on the Dickson Charge Pump concept.
Preliminary schematic here:


Simple to construct - and easier to extend without introducing any more active stages.... I've even managed to get a prototpye stripboard layout to fit into a BM800 body!
I've used a 4093 as a quad inverter - simply because I had some thru hole ones lying round - but any Schmitt triggered 4000 CMOS inverters will do ... Like the 40106 for example......
Because there is no current drawn from the output, you could actually just manage with a square wave oscillator and just one inverter in a minimum component configuration.
Seems to work quite well....

EDIT: I've fitted a stripboard version - together with an OPA1641 impedance converter - into a Chinese 'Neumann' style body.
It's also fitted with one of Ari's 'flat K.47' capsules..........I think it's the best sounding mic I've done so far...

 

[ccaudle]

I was glad you mentioned the term "Dickson charge pump" in that previous thread, I was not familiar with that name, and it gave me a good search term to spend some time with.

I have not built up my own proto yet, but I think it should only need two buffers, since the output current is essentially 0 once all the capacitors and the capsule are charged up. First buffer drives half the diode/cap stages on both the positive and negative sides, and forms the oscillator, second buffer drives the second half of the diode/cap stages.

I was hoping that one of those really small packages that just have one or two gates would work. I think they will, but at the cost of needing a second zener regulated supply and additional diode stages, since all the new small package devices are 5V max supply, and it does not seem reasonable to run the op-amp capsule buffer from 5V.

I am interested in how compact this circuit could be made, so I will have to think about whether the extra zener and diode/cap stages to generate from 5V is worth the small gate size, or perhaps a small H-bridge device containing two P-channel/N-channel transistor pairs to make discrete transistor inverters would be better.
And if I order parts to build one (as opposed to using whatever I can find in my parts drawers), I think Schottky diodes to reduce the voltage drop through each stage.
Actually, now that I think about it, normal switching diodes might work just as well because the reverse leakage of Schottky might be on the same order as the leakage across the devices in the output, and with such low forward current the drop of a standard switching diode will probably not be the 0.6V you usually think of for PN diodes. The advantage of the lower forward voltage of Schottky devices may be lower than I initially thought.

 

[mrgrooves666]

HiRogs!. I'm not tech savy, but very interested following up on this new version. I'm also about to embark in a build of a pair of multipattern mics with your Opic design, with Ari's F47 capsules, hopefully as soon as she has them available. Already have all the parts for the build, and practically eating my nails here In the meantime. I do suspect they will be terrific workhorse mics!
What are the advantages of this design VS your original design based on the Holmes paper?

 

[rogs]

What are the advantages of this design VS your original design based on the Holmes paper?

The design based on the Holmes paper uses a separate inverter for each stage.
That makes it more complex to lay out than the simpler charge pump circuitry, which actually only requires a square wave clock and a single inverter as the active components to function.
For this task the voltage multiplier doesn't need to supply any current (or virtually none!) so sharing the oscillator output - and its inverse - alternately between the stages makes for a simpler circuit layout (at least for stripboard prototypes it does! )
Here, I've used the other inverters in the IC to share the load, simply because they're in the IC and any unused inputs have to be terminated anyway!
It also helps here with the stripboard layout.

The original OPIC multi pattern version includes 3 op-amps and several resistors in the signal path.
Using a dual output voltage multiplier allows for the single op-amp layout in the OPIC LDC version to be used for a multi pattern version.
This version will have a lower noise floor - by around 3dB in theory.
In practice the ambient noise in the 'real world' tends to swamp the self noise generated by either version!

The original version also includes an adjustment to compensate for differences which can sometimes be found between the output levels of dual capsules. This control can adjust the output level of the 'cardioid' side of the capsule by up to 6dB,
That should allow the figure of 8 configuration to be calibrated fairly accurately.
Ari's 'flat 47' capsules are matched within c.1dB, so the adjustment option is not really required for those capsules.

I'm a great fan of 'simple'...... I like to try and minimise the component count in the internal impedance converter circuitry within the mic itself, and play around with things like EQ externally.
This was really just another experiment to try out .... just a small part in the fascinating world of hobby mics!

Here's a copy of my experimental dual 4093VM prototype 'stripboard' layout, if it's of any interest?.... (It will fit into a BM800 body!)

 

[rogs]

The stripboard layout constraints for this dual VM project got me thinking again about the original single output VM I used for my OPIC project.
Although that circuit was based on a series of 6 serially linked Hex inverters, as suggested in Rory Holmes' article HERE , it actually only requires 2 clocks of opposite polarity - a Dickson Charge Pump - to perform this task.
Although in this case I have actually retained the serial gate structure, I have found a simpler way to lay out the stripboard to suit the inverter IC pinouts.
There might be other options of course, especially as there is actually only a requirement for a single ultrasonic oscillator - and it's inverse - as the active components?
I've created a new single output voltage multiplier, which is simpler to construct onto stripboard for experimental purposes, has fewer components, and includes a variable resistor to allow for variable output voltages in the range 50v to 85v using a 15v Zener diode - or with different ranges from 40 to 100v using alternative Zeners.

Schematic and prototype photos here:



(Full project notes: HERE )

 

[mrgrooves666]

Hi Rogs, with the original OPIC multiplier, what value of R5 are you using with Ari's Flat 47?

 

[Khron]

Hi Rogs, with the original OPIC multiplier, what value of R5 are you using with Ari's Flat 47?

You mean that last resistor in series with the output to the capsule? Why would you think the absolute value is (quite that) crucial?

 

[rogs]

Hi Rogs, with the original OPIC multiplier, what value of R5 are you using with Ari's Flat 47?

I have one of Ari's Flat 47's fitted into a multi-pattern OPIC. That is fitted with the original VM, and I have R5 fitted as 1k.
There are 3 op-amps in that circuit, and the circuitry draws c.7mA in total from the phantom power supply.
That provides a DC supply to the op-amps of c.16v.

R5 is fitted as 1k, providing an output polarisation voltage of 70 volts - measured at the cathode of D7 - so that the actual voltage applied across the capsule is 70v, minus the op-amp 'half rail' bias voltage - which is 8 volts.

So the capsule actually has 63v across it.

I think the 'R5' Khron is describing refers to the output series resistor in the newer 'adjustable' version, which is 1M and, as he suggests, is not a critical value.

One final point. I have specified a CD40106 as the VM project hex inverter.
It is of course possible to use alternative types of 40106 - like the HEF40106 - but the thresholds are a bit different, so the output voltages may vary a little from the values listed in this circuit. For example, you would probably need to fit R5 as 2k2 to get a 72v output from a VM fitted with a an HEF40106 intead of a CD40106.
For the newer 'variable' type, this is not a concern of course....

 

[mrgrooves666]

You mean that last resistor in series with the output to the capsule? Why would you think the absolute value is (quite that) crucial?

Nope not that one, but thanks. I thought I meant R5 in original Opic (with the first VM), wich determines polarization voltage...
Thanks Rogs!

 

[Emmathom]

Here we are!

Yes I've read Jules website too... and JLI are far too expensive on their shipping rates to Europe (you & I are concerned)
About these OPA's they are VSSOP-8 and need adaptater to DIP-8 (not easy to solder!) Mouser sells them and so OPA1642 jfet's one...

and Mic Mic Sharf is very interesting !

 

[rogs]

Here we are

Your comment from the other thread :
Yes I've read Jules website too... and JLI are far too expensive on their shipping rates to Europe (you & I are concerned)
About these OPA's they are VSSOP-8 and need adaptater to DIP-8 (not easy to solder!) Mouser sells them and so OPA1642 jfet's one

There are two types of packages for the OPA164* series -- SOIC and VSSOP. SOIC is much easier to hand solder.

The attached image shows how I use the 1641 single amp. A tiny dab of Blue Tack on the underside of the device holds it in place before you solder.
Notice I don't fit pin 3 in the adaptor. That allows me to keep the high very impedance (1 Gohm) resistor away from the stripboard...

 

[Emmathom]

Your comment from the other thread :
Yes I've read Jules website too... and JLI are far too expensive on their shipping rates to Europe (you & I are concerned)
About these OPA's they are VSSOP-8 and need adaptater to DIP-8 (not easy to solder!) Mouser sells them and so OPA1642 jfet's one

There are two types of packages for the OPA164* series -- SOIC and VSSOP. SOIC is much easier to hand solder.

The attached image shows how I use the 1641 single amp. A tiny dab of Blue Tack on the underside of the device holds it in place before you solder.
Notice I don't fit pin 3 in the adaptor. That allows me to keep the high very impedance (1 Gohm) resistor away from the stripboard...

View attachment 122942

Great job ! so many way to investigate to build his own mic ! as I said I'm a real beginner and my first projetcs are about to start (as soon as I get the last ordered components).
* repair 2 sE Electronics X1S by throwing away the pcb (full of cms) and made a polar+pre ciruit on a stripboard using Thet schematic
* build a pair of U47Fet > my first intent was to stay with cardio only but as Thet advised me to solder the whole thing (so I won't have to dismantle the mic later in case of I would use it as multi-pattern) I changed my mind and go now for a "U47Fet multi" (?) based on Thet's polarization & pre circuit (wich I have received a couple of days ago)
and because we are here online now, may I ask you for a nice body for my U47 project ? Studio 939 sells good ones for $225 (!) or Ali Express has some too...
Thanx for your help Rogs

 

[Emmathom]

I just looked at Mic Mic Sharf's <OPA "Alice" Microphone (round PCB)> and ask myself if it could fit the sE X1S body... ? it could an OPA alternative to "rebuild" those mics...

 

[Emmathom]

to get back to U47 Arienne caps this is how I intent to connect the capsule :

 

[jp8]

I was glad you mentioned the term "Dickson charge pump" in that previous thread, I was not familiar with that name, and it gave me a good search term to spend some time with.

I have not built up my own proto yet, but I think it should only need two buffers, since the output current is essentially 0 once all the capacitors and the capsule are charged up. First buffer drives half the diode/cap stages on both the positive and negative sides, and forms the oscillator, second buffer drives the second half of the diode/cap stages.

I was hoping that one of those really small packages that just have one or two gates would work. I think they will, but at the cost of needing a second zener regulated supply and additional diode stages, since all the new small package devices are 5V max supply, and it does not seem reasonable to run the op-amp capsule buffer from 5V.

I am interested in how compact this circuit could be made, so I will have to think about whether the extra zener and diode/cap stages to generate from 5V is worth the small gate size, or perhaps a small H-bridge device containing two P-channel/N-channel transistor pairs to make discrete transistor inverters would be better.
And if I order parts to build one (as opposed to using whatever I can find in my parts drawers), I think Schottky diodes to reduce the voltage drop through each stage.
Actually, now that I think about it, normal switching diodes might work just as well because the reverse leakage of Schottky might be on the same order as the leakage across the devices in the output, and with such low forward current the drop of a standard switching diode will probably not be the 0.6V you usually think of for PN diodes. The advantage of the lower forward voltage of Schottky devices may be lower than I initially thought.

It can be made quite compact if you don't mind using SMT parts. That is: no bigger than a Hartley oscillator built from the same size SMT's and axial inductors. My "little secret" (pun intended): BU4S584G2 single Schmitt-trigger inverter, so you'll need two of these. Accepts 18V Vdd, so you can get higher output voltages with the same number of multiplier stages. Or alternatively, use less stages for the same voltage. It comes in a nice small SOT23-5 package. Ideal for when you're having little board space. The circuit draws ~1.2 mA, Hartley oscillator ~3mA. These CMOS oscillators need a double-sided PCB with a ground plane underneath the oscillator circuit to prevent those nasty high dV/dt's from getting into your audio circuitry. You'll have to carefully layout the PCB to prevent capacitive, inductive, and common impedance coupling of these high switching frequencies into the audio path. Keep the circuit compact and keep audio tracks as far away from the circuit as possible.

There's just one disadvantage of using the BU4S584G2: it's single-sourced from ROHM. If they decide to pull out the plug and make it EOL, you're in trouble. But at the moment, it's still active and available through Mouser and Digikey.

Jan

 

[ccaudle]

single-sourced from ROHM

I'm glad you provided that link, I had not previously been able to find a single gate high voltage CMOS device. The vendors I was aware of with single gate devices like TI and OnSemi only have devices for up to 5V supplies.

I have been considering TI DRV8220 motor driver as way to get basically the same function with an H-bridge driver. Not a perfect fit, but runs at up to 18V supply voltage (abs. max of 20V), and you get two push-pull outputs in a 6 pin 1.6mmx1.6mm package (or 2mmx2mm with some additional features if you don't mind dealing with a no-lead plus pad on bottom package). No Schmitt input, so the small Rohm inverter might be better for an oscillator.

 

[ccaudle]

single-sourced from ROHM

Looks like Toshiba also makes that part:
Toshiba TC4S584 datasheet
Pretty cheap at Digikey.

 

[jp8]

Dang! You're right, good catch! Actually, that Toshiba part is the 18V version I was talking about. The ROHM is max 16V. I bought the ROHM . I mixed them up, apparently. Will order the Toshiba's with my next Digikey order.

Jan

 

[ccaudle]

I am trying to simulate this in TINA-TI, and have convergence problems with the diode and capacitor connections.
Will I need to put a high value resistor in parallel with each capacitor in order to get the simulation to converge? I put a 1G at the end of the chain so there would be a DC path, but I thought the series chain of diodes would simulate OK since they start out forward biased.

 

[ccaudle]

Snapshot of the schematic in TINA-TI. This is the first time I have tried to simulate logic gates in TINA, so there may be some details I am missing to get this to converge properly. Those resistors to ground at each cap to diode node are 100G, I was just checking if not having a DC path on each node was tripping up the simulator (that used to be a problem with SPICE simulators, might still be required, but I am not positive).