Solid State Vibrator.

[Station X]

This thread is rather long as it describes the trials and tribulations experienced when designing and building a Solid State Vibrator.

If you just want details of the circuit and Printed Circuit Board layout go straight to post #56.

https://www.vintage-radio.net/forum/...5&postcount=56

************************************************** *********************************************

I am currently restoring a WW2 military radio and if all goes well I hope to post a success story in the near future. The set's vibrator is in a sorry state. It has obviously been opened up several times for "attention", The contacts are blackened and the springsets are bent. Although I am quite capable of cleaning contacts and adjusting springsets, I decided that I wanted a more reliable solution, so opted for a solid state vibrator.

Commercially made items are available from Antique Automobile Radios in the USA or via their UK agents:-

http://www.radiosforoldcars.com/vibratorsF.htm

Unfortunately they don't do a seven pin version which is what I needed. I considered making an adaptor, but in view of the price of these vibrators I decided to make my own.

I claim no credit for the circuit which is taken from the VMARS website. The only change I made was to make both collector to base resistors the same value. The pictures tell their own story. I used the original base, keeping the old innards for possible future use. The chassis/heatsink is made from aluminium cut from the power supply cabinet of an old Sperry hard disk. The whole assembly slides into the original can, so the outward appearance is unchanged from the original.

The PSU section of the radio is still being restored, but I isolated the step up transformer and parallel capacitors from the rest of the circuit, plugged in the homebrew vibrator and switched on. There was a satisfying purring noise, but I'm not sure where it comes from. A voltmeter check showed about 14V AC at the input and about 900V AC across the secondary of the transformer. That looks good.

Once I've rebuilt the PSU I'll load it up with some wirewound resistors to check whether the heatsink is good enough with and without the can. If not a rebuild will be called for.

[OErjan_S]

yes, that circuit is a good one, and it is very well made by you, almost invisible.
I have made a few similar myselfI, have found another circui that also works, fairly similar though, here you have that.
oops, forgot to put the diode type in the schematic, they need to handle quite bot of both current and voltage, so 1N4005 or similar is preferred, i put 18V as the max, it is acually possible to get spikes of above 24V if the frequency is wrong (spikes of opposing voltage adding)

[Nickthedentist]

Very impressive work, Graham.

I hope it works as well as it deserves to.

Nick.

[Ed_Dinning]

Hi Gents, provided the contacts can be cleaned up and the springs are OK, it is possible to use the contacts to drive a MOSFET. This then does the switching of the transformer.
As Mosfets have lower dissapation than bipolars the can often be hidden within the can of the vibrator.

Ed

[Station X]

I'm hoping to carry out some more detailed testing today. If the present device is unsatisfactory I'll try the 555 timer and MOSFET approach. I'm a great believer in the KISS (Keep It Simple Stupid) principle.

[evingar]

To be honest, the MOSFET and 555 approach would be the one I would use. It's likely to waste less power, even if it is slightly more complex than the bipolar approach and you don't have the potential reliability issues of old mechanical components.

[Ed_Dinning]

Hi Gents, if you need a delay and are using the 555, use a 556.
One half is a bistable, the other is your delay monostable. All in one packet as they say.

Ed

[Paul Stenning]

Originally Posted by Station X

The only change I made was to make both collector to base resistors the same value.

That was probably intentional to provide some imbalance in an otherwise symmetrical circuit to make sure it starts up reliably. With the tolerance of all the components used though there would inherently be some imbalance so it shouldn't be needed to introduce some deliberately.

[Station X]

Today I connected the guts of the mechanical vibrator to a 13.8V supply to determine its frequency of oscillation. As I expected it proved impossible to do this with a frequency counter due to contact bounce.

I then used a scope which showed a rather asymmetrical waveform no doubt due to the contacts being maladjusted. However the frequency was determined to be 83Hz.

I then turned my attention to the solid state version. The power supply has a tapped secondary on the step up transformer for high and low voltages, which are selected by a HIGH/LOW power switch, which earths the appropriate tap. Connected across the secondary are two 22nF capacitors in series giving 11nF. The series connection is needed because of the high voltages which can be developed. The original specified caps were 0.02uF, but had to be replaced because one was leaky and the other of the wrong value (0.005uF).

With the rectifier valves removed the frequency was 138Hz in the LOW position and 131Hz in the HIGH position. The change is no doubt due to the increased inductance in the high position.

The circuit would not start reliably, but could be made to start by operating the HIGH/LOW switch. I will try changing one of the resistors to see if starting can be improved.

AC voltages across the secondary were 633 in LOW and over 1000 in HIGH. Voltages applied to the rectifiers will be half this.

I then inserted the valve rectifiers and measured the smoothed DC voltages at the output of the PSU. They were reasonable for off load voltages, but I forgot to record them.

The next step will be for me to calculate the expected current demand from the HT supply and load it up progressively to draw that current whilst monitoring the temperature of the 2N3055s. That will have to wait until another day though.

[Station X]

Did some more testing this afternoon. I calculated the maximum load on the HT as 60mA and three 20K wirewound resistors in parallel across the output of the PSU resulted in exactly that current being drawn with the HT steady at 400V. I left it running for about 10 minutes. The transistor cans and heatsink/chassis got hot, but not so hot that it wasn't possible to hold a finger on them continuously.

Alas starting is very unreliable. In fact it won't start at all with a load applied. Changing one of the 200 ohm resistors to 180 ohm as per the original circuit did not improve this. The switching frequency varied with load too.

So I have to report an UNsuccess story

I'm going to have another go using MOSFETs and a 4011 CMOS Quad NAND Gate. The circuit is from a battery eliminator designed by Tony Maher.

[kalee20]

I would definitely go for a 'driven' circuit (as opposed to self-oscillating), using MOSFETs as Ed Dinning says. The frequency needs to be fairly accurate as vibrators supplies have a capacitor across the transformer, which resonates with the transformer's inductance, to minimise switching losses. These are the capacitors you mention.

Also, I would add a divide-by-two bistable so that the MOSFETs are both driven with identical pulses. Mismatch of the FETs' 'on' time can bias the transformer towards saturation, leading to heating and loss of efficiency.

Potentially, you can end up with a very useful, plug-in, vibrator replacement.

[Station X]

Today I built the circuit shown in post 10. The CMOS Oscillator/Buffer/Invertor is absolutely bomb proof. It starts first time every time and has a frequency of 100Hz. Unfortunately the mark/space ratio is 60/40 rather than 50/50 which isn't desirable. I shall have to try building a driver using an NE556. No problem using one half as a timer with a 50/50 mark/space ratio, but I'll have to work out how the use the other half of the device as an invertor.

The MOSFETs get exceedingly hot. So hot in fact that I daren't leave the device switched on for more than about 20 seconds. They should be well within their current carrying capability and have big heat sinks. I'll investigate that tomorrow. I do hope the step up transformer primary isn't faulty.

[jimmc101]

As kalee20 suggested earlier, the only certain way to get an accurate 50/50 M/S ratio that doesn't drift is to use a divide by 2.
I would suggest using the 4047 as this has the oscillator and divide by 2 in one package. (Maplin and CPC stock it).
While checking availability I came across this which may help

http://electronics-diy.com/electroni...tic.php?id=661

My only comment is that some decoupling across pins 7 & 14 of the 4047 would not go amiss (say 100n)

Jim

[mhennessy]

Originally Posted by Station X

I shall have to try building a driver using an NE556. No problem using one half as a timer with a 50/50 mark/space ratio, but I'll have to work out how the use the other half of the device as an invertor.

Actually, you might have a problem getting 50% from a 556 - the standard astable circuit doesn't allow this. Bodges are possible, but these increase the passive component count such that there's no advantage to using the IC.

Have you considered driving the MOSFETs with a simple 2-transistor astable? It would only need 8 components, yet be as reliable as the CMOS circuit you tried, and could be built in less space that the CMOS IC...

The MOSFETs get exceedingly hot. So hot in fact that I daren't leave the device switched on for more than about 20 seconds. They should be well within their current carrying capability and have big heat sinks. I'll investigate that tomorrow. I do hope the step up transformer primary isn't faulty.

There are two possibilities: either they're not saturating, or they're switching too slowly. Investigation with a scope should prove which; I'd suspect the former. Given good devices with low RDS(ON), I'd expect them to stay cold with minimal heatsinking... Last time I built anything with a MOSFET, it switched two 12V 20W halogen downlighters with no heatsink...

Another possibility is that you need to provide "dead time" - that is, a short period when both switches are off. If this is the case, you might want to investigate using a classic SMPSU controller IC like the TL494 or SG3525 - while potentially very comprehensive, these ICs don't necessarily need too much fettling in this application...

[mhennessy]

Originally Posted by mhennessy

There are two possibilities: either they're not saturating, or they're switching too slowly. Investigation with a scope should prove which; I'd suspect the former.

Forgot to say - the switches not saturating is probably caused by the transformer core saturating, due to the 40:60 duty cycle...

Jim makes a good suggestion with the 4047 - just make sure that the CMOS IC is well protected from the HV side (protection diodes and Zeners might be worth using). Because I'm paranoid, I lean towards the transistor version because I find them more reliable/predicable. The duty cycle will be close enough to 50:50 - the mechanical version wouldn't have been perfect in this regard BTW

Mark

[Station X]

Originally Posted by mhennessy

There are two possibilities: either they're not saturating, or they're switching too slowly. Investigation with a scope should prove which; I'd suspect the former.

How would I check using a scope? I am seeing clean square waves on both gate and drain. I haven't displayed them both at the same time though, but could do as the scope is dual beam.

Jim. The CD4047 sounds like a good idea.

[mhennessy]

Originally Posted by Station X

How would I check using a scope? I am seeing clean square waves on both gate and drain. I haven't displayed them both at the same time though, but could do as the scope is dual beam.

How close to 0V does the drain get when conducting?

In principle, any deviation from 0V should only be due to conductive losses in the RDS(ON) of the MOSFET, which give I2R losses. But you might not have enough gate drive to fully "saturate" the MOSFET - check the datasheets or measure it directly with a bench PSU...

I'm not sure which devices you're using, but in the past I've used the BUZ11 which has an RDS(ON) of 0.04 ohms, or 40mV per amp if you prefer. Funnily enough, I used one of them to achieve what you're doing, only being a university project they wanted to make it really complicated, hence it had to use a SMPSU, and it had to run from a 2V lead-acid cell. Despite the high primary current (around 5 amps), and the HF switching, the MOSFET didn't need a heatsink. The start-up circuitry was neat, and by complete coincidence also involved a 2-transistor astable - nothing else worked at 2V! The main control circuit used 4000-series CMOS though - also to guarantee a particular duty cycle. Goodness knows what the HF switching did to the radio receiver

I'm sure you'll gave it working soon

[Station X]

The MOSFETs are NTD3055L104T4G surface mount devices.

Unfortunately Maplin don't stock the CD4047, so further development of the driver section has been put on hold until I can get one by mail order.

[mhennessy]

Originally Posted by Station X

The MOSFETs are NTD3055L104T4G surface mount devices.

Yes, these look pretty cool (pun intended!) - 12A at around 0.1 ohms. Looking at Figure 1, with a drain current of 4 amps, the VDS should be less than a quarter of a volt, so that's a power dissipation of around 1W. Out of interest, have you measured how much current your circuit consumes?

And how big is the gate drive waveform? As it's coming straight from CMOS I'm assuming it's basically the same as the input voltage to the whole circuit. As long as it's >5V you should be ok - 10V should be easily attainable.

If you're stuck for something to do while waiting for the 4047, why not try the 2-transistor astable? I only mention it out of idle curiosity

[jimmc101]

Maplin do stock the HCF/HEF4047 but do seem to bury it quite deeply.
The stock number is QX20W but it may be web only.

There are a couple of points to be wary of using the transistor astable:

For a supply over about 5v base emitter reverse breakdown can occur when the bases are driven negative.

The rise time at the collectors is very slow since the cross coupling capacitors have to charge via the collector resistor.(The other end is clamped by the BE junction of the opposite transistor).

See for instance

http://www.uoguelph.ca/~antoon/tutor...or7/xtor7.html

figs 12 &13

Jim