New head amp design

[Lawnboy]

still having trouble getting the head amp to work for my 24-line direct-view camera. so far ive built 3 versions of the original head amp (handbook design) and one of the new head amp with no success. the latest build of the handbook design was the only one that produced some sort of an image, although it was only an unrecognisable shape in binary color. the new head amp only dims the display when power is applied, even after waiting the required 10 seconds. the LED driver is Peter Smiths version, simmilar to the one in the handbook, and it has been tested.
here is the layout diagram that i made for the new head amp; maybe its something simple that i missed.

[Steve Anderson]

The only thing I can see that might be wrong is the 100uF capacitor (C1 in the diagram) appears to be incorrectly connected. Its positive end should be connected to the junction of R3 and R4, not the junction of R6 and R9.

This shouldn't affect the DC operating point, but could have a dramatic effect on the AC gain. Exactly what I'm not sure without building it. I'll have a go at simulating it.

The other thing is this sort of construction is not really suitable for amplifiers of this sort...it really does need a ground plane otherwise there's a big chance it will be unstable. I would also suggest putting a 100nF disc-ceramic capacitor in parallel with C1. I repeat what I said earlier quoting from the article itself...Take heed of the sentence, "Owing to the high gain (around 85db) good RF techniques should be used in the construction of the amplifier." The gain figure is a bit optimistic, but still RF techniques should be applied.

The only other thing is your sketch doesn't show the pinning of the transistors, I can only hope you've got it right.

Steve A.

Having simulated it with and without C1 there's very little difference, less than 1db which I was surprised at. Green trace below is with C1, red is without.

More...The input impedance is around 40k with a source of about the same the gain (as expected) drops by half and the 'bump' at the LF end vanishes. What the source impedance of the 'dome-cell' is I have no idea.

[Klaas Robers]

Steve,

is your simulation including the variable resistor of 22k that loads the dome sensor? And have you included the parasitic capacitance of the dome sensor? I don't know how to measure it, but I guess it is a few nanofarads.

You could simulate the dome sensor by a current source and a parallel Si-diode. The current is very small and drives the diode open to conducting.

I would have designed the circuit such that there is a reverse voltage over the diode. This increases the efficiency of the photo effect and it gives the diode a smaller capacitance.

[Steve Anderson]

I would have designed the circuit such that there is a reverse voltage over the diode. This increases the efficiency of the photo effect and it gives the diode a smaller capacitance.

Agreed. But to be quite honest I still don't really know what this 'dome sensor' is. Is it a photo-diode, a photo-transistor, or as Jeremy put it, "They are miniature silicon solar-type cells, like BPW81/A, SFH 2030 etc." The SFH2030 is a photo-diode, the BPW81 has a Schmitt detector circuit built in and neither is what I would call a solar-type cell....so I'm confused. What doesn't help is the lack of a datasheet or even a manufactures type number.

If I do make it to the convention this year (there's an 80% chance of that) I'll ask if I could purchase a couple of these devices and do some tests on them when I get back home.

Without an example of one of these devices it's hard to discern which mode it's currently being operated in. If the 22k pot wasn't there and the amplifier had a much higher input impedance it would be operating in photo-voltaic mode. This is not the best mode for this application as the open circuit voltage roughly only doubles for a hundred-fold increase of irradiation.

But with the pot and the amplifier it's 'sort-of' operating in a photo-current mode with the load impedance converting that into a voltage...almost a transimpedance mode.

As you suggest, and assuming this is a photo-diode and not a photo-transistor, reverse bias would reduce the junction capacitance and convert the device into a photo-detector whose output current is a linear function of illumination, but making allowance for 'dark current', i.e. leakage.

So now it's a current source, that is a very high impedance source. To maximize the output voltage the load impedance needs to be as high as possible and to preserve frequency response stray capacitance needs to be minimized.

Without wishing to be critical using a bipolar device with a source impedance of this magnitude doesn't yield a very good signal-to-noise ratio. With low impedance sources bipolars are best, but here a FET wins hands-down when dealing with meg-ohm sources. Here you're dealing with noise current, not noise voltage.

Also of note is that the BC109C was good in its heyday (the 70s), it now is easily surpassed by the likes of the super-beta MPSA18 or the KTC3200 both with noise figures under 1.0db as opposed to 4.0db. But as mentioned before only really usable with sources of 100k or less.

Of the FETs I've used the quietest one to date is the Toshiba 2SK30, it is specifically designed for high impedance audio applications, capacitor microphones for example. Another contender is the 2N5457, although not marketed as a low-noise device it does a pretty good job. MOSFETs even with their amazingly high input impedance have too much capacitance for this task and are generally much noisier.

Steve A.

[Klaas Robers]

Steve, I had one, but can't find it at this moment.

It is a silicon photo diode with a rather large surface area. I guess it is designed as an IR receiver for optical remote control of e.g. TV sets. The dome provides it with some directional properties. If you look staight into it the whole surface seems to be black. I have no idea about the opening angle. Please contact Vic Brown and he could send you one for little money by post.

In this circuit it is used in the photovoltaic mode. This has the practical advantage that you might use the loading resistor to control the sensitivity as a kind of iris control. May be that it even gives some form of gamma correction. However this is the way to get the maximal parasitic capacitance.

The circuit is built with semiconductor the club has plenty of. That is the reason that not transistors of e.g. the BC 458 series are used. A BC550C is equal to the BC109C, but lacks the metal case, so it has less stray capacitance. On the other hand I observed that the 2N3904 has a different pin layout in respect to the BC548 and BC109 transistors. For newcomers this is a special type of pitfall.

I just found the BPW34 and it is likely that this is the same photo diode, however not mounted under a dome and not available in het club shop. If you can use the BPW34 in your simulations you are not far away.

[Steve Anderson]

I just found the BPW34 and it is likely that this is the same photo diode, however not mounted under a dome and not available in het club shop. If you can use the BPW34 in your simulations you are not far away.

I have some BPW34s here, in fact I used one in the laser link item I posted here some time back, that item was also in the newsletter. I have used it in other non-NBTV applications too. Others have used it for long-range optical through-the-air communication.

It's available from most suppliers in the UK for around a quid. Make sure you get the plain BPW34 not the BPW34FA which has the black ambient light filter. There is also an enhanced blue sensitivity version, the BPW34B.

The sensitive area is 2.65 x 2.65mm and under back-bias arrangements fast with a switching speed of 20ns. Datasheet for each version attached.

I might have a Spice model for the device but how would you stimulate (spelling correct) in a simulator? Certainly mine (Simetrix) has no light simulation. It was more the amplifier I was interested in.

If the 'dome sensor' is simply a re-packaged BPW34 that would indeed be a useful photo-detector.

Steve A.

[Klaas Robers]

Steve, can you stimulate a simulated model with a current? If you send a current into the diode, in a polarity that the diode goes into the conducting state, that is the same as stimulating with light. Think of input currents of 10 nA and less.

You could also use a voltage and a series resistor of 1M. Then after the resistor you see the diode to ground and the 22k variable resistor to ground. But I think that due to the high voltage gain (80-85 dB) the voltage excursions at the input are so small that you don't see any current running in the diode. So the non linearity of the diode is not seen.

I also see that the capacity at Vd = 0 V is only 72 pF. This is low! So you can also replace the diode by a capacitor of 72 pF.

Good Luck

[Steve Anderson]

I also see that the capacity at Vd = 0 V is only 72 pF. This is low! So you can also replace the diode by a capacitor of 72 pF. Good Luck

True, but it's fairly typical of photo-diodes of this surface area, as you'd expect the capacitance figure is roughly proportional to the sensitive area.

Other detectors, e.g. Solar Cells have a hundredfold or more greater capacitance which is why the pre-amps for these have a rising frequency response. Using reverse bias I have easily obtained a flat frequency response from this device to beyond 10MHz. The NJL6163 should go to more than 40MHz! Not that you need that in NBTV!

You'll note from the 'Capacitance' graph further down the datasheet at a reverse bias of 10V which is easily derived from a 12V supply, the capacitance drops to around 15pF, again this is fairly typical of Silicon photo-diodes.

Reverse bias capacitance for a few randomly selected photo-diodes of around the same surface area at 10V...

BPW34, 15pF
BPW41 (IR only), 15pF
NJL6163 (@ 5V), 9.5pF


As for simulation, I doubt that it's possible to simulate something like this in photo voltaic mode without writing a proper Spice model file (something I've never bothered to lean). However, using reverse bias it simply becomes a current source and all simulators have those. I have used this method before and it's quite accurate.

Steve A.

[Klaas Robers]

It is not difficult to make a first order approximation of a photo diode in photovoltaic mode. It is a current source and a Si diode in parallel. As the current per mm² is rather low the open voltage gets never higher than 0.6 volts. This is the same with regular Si diodes when you send a low current through them, e.g 0.01 x max current.

But in this specific case, with the 22k in parallel the voltage will be always much lower than the 0.6 volt, so the diode will hardly conduct any current.

A reverse voltage will indeed lower the capacitance, but it is very difficult to keep voltage changes from the power supply out of the very sensitive input. In the photovoltaic mode this is no danger. I think this is the reason that this mode is choosen.

With this low capacitance of 75 pF and the resistor of 22k in parallel, the HF cut off is at about 200 kHz, so it will not be of any influence for NBTV. However you can prove it in the simulation.

[Lawnboy]

Steve, I finally got around to fixing the 100uf capacitor connection and retried the circuit, but as expected it didnt make a difference. i also marked the emitter pin on the diagram with little arrows. how does one make a ground plane? just a piece of copper clad pcb underneath the circuit? if i cant get this to work i might try the solar cell amp. according to past newsletters, people have had success with them in cameras in the past so it must be worth a shot.

[Steve Anderson]

how does one make a ground plane? just a piece of copper clad pcb underneath the circuit?

In essence, yes. A piece of single-sided un-etched PCB should do it. Place the non-copper side in contact with the underside of the existing circuit and connect the plain copper side to the 0V/ground of the circuit. Keep the lead for this as short as possible.

If you have access to a scope it would help to look at the output and see if the thing is unstable. However, sometimes hanging a scope probe on a circuit can make it unstable by virtue of its capacitance. Many (including me) have been caught out by this.

Steve A.

[Steve Anderson]

I also marked the emitter pin on the diagram with little arrows.

There are at least two flavours of the BC212, the standard BC212 and the BC212L, the pinning for these otherwise identical devices are different (who knows why?).

Also the only datasheet for the 2N3704 I have shows the collector as the centre pin, not as you've drawn.

You need to get the correct pinning for the devices you actually have. This is a perennial pain in the $%@!

Steve A.

I have since downloaded a couple more datasheets for the 2N3704 and that suffers from the same problem, different pinning even though the case (TO92) is the same...even from the same manufacturer!

I suggest that if your multimeter has a NPN hfe function you find the right lead combination that yields the highest gain. The difference will be quite obvious, there'll be just one combination that results in a gain of at least 50, probably well over 100. The rest will be at or close to zero.

A rather grotty example of the 2N370X NPN and PNP series datasheet is attached, note the different pinning! Yet the same case!

[Klaas Robers]

Indees Steve,

it is a bloody shame. Even the 3704 with the base in the middle had the pin of the collector and the emitter reversed in respect to the BC546-BC550 and BC556-BC560 transistors and so many other transistors that are placed in these cases. Now I have seen:
CBE
EBC
ECB
BEC
and how many other orders of the pins are possible? I always wondered why this was needed.

Klaas

[Steve Anderson]

...and how many other orders of the pins are possible? I always wondered why this was needed. Klaas

Agreed. In the good old days of metal can trannies it was always EBC from either the metal protruding tab (emitter) or a coloured spot (usually the collector). BC107-9 style. Even with the Germanium AC126 and the like it was never any different. (Yes, I do remember them!)

Once the plastic package came along it seems to have been a free-for-all.

Some sanity seems to have ruled in the chip world. Can you imagine a 74LS00 from Texas has one pinning, from Fairchild a different one and from Philips yet another?

Then you come to electrical specifications. Our good old friend the 4046 in its HC incarnation (74HC4046) has different specifications regarding the VCO section. (Not used in the usual NBTV motor-control application).

From Texas or Philips it's rated to a maximum frequency of well over 20MHz. But from Fairchild less than half this. It meets the Fairchild specification, but that's different to the Philips or Texas one!

Same generic chip...but different! Another example of specmanship....the standard 4000 series version of the 4046 (CD14046 or HEF4046) has a frequency span in the VCO section of over 50:1. Wonderful. I had an application that needed a VCO that could operate over a 16:1 range. Easy for the CD/HEF4046.

But like all 4000 series chips it's way too slow. So changing to a HC version one would expect the same....but no. Conveniently omitted from the datasheets for the HC version is a specification for its VCO range which in practice is only about 5:1 thereby making the device useless for the application I was hoping to use it in.

Steve A.

[Lawnboy]

i scanned the pinning for the transistors i used. they are in a TO-92 package so it is the center diagram. the BC109C came from the club as a TO-18 and i am assuming that the emitter pin is the one with the tag. i will give the circuit a try with a ground plane when i get a chance to work on it again.
also, i tested a solar cell that i had considered using, and it has a capacitance of about 40nf. im guessing this is too high for nbtv?

(added) actually i just realized that the only circuit that i got working with these transistors used a NPN. i have not gotten one working with a PNP yet. that is mostly because for a while i thought the diagram was a top view (notice how the package clearly states that it is a "bottom view" )