Fluke 8000A-01 Power Supply Modification

As much as I can get side-tracked rambling about Linux, my real passion remains electronics, or, if you want to call it by its fancy name, Electrical Engineering. After all, this is my field of study. My personal ‘lab’ is, apart from my oscilloscope – my pride and joy, it was a gift -, built on an extreme budget.

A bench multimeter has long been something that I’ve wanted for it. A bench multimeter is just much more practical than a budget-handheld one: It stays in place and you can always look to the same place to see your reading. However, usually these devices go for pretty high prices, as they keep their value very well over time and the second-hand market for them in The Netherlands is very small.

So I was extremely surprised when my bid of €15 on a Fluke 8000A was accepted. It had a small issue, for which I will explain my solution in this post, but at such a price, even an old multimeter like the Fluke 8000A can be a bargain – as long as all digital circuitry in there is in order.

Fluke 8000A Variations

Before you buy a Fluke 8000A, or any multimeter really, I would recommend going over the manual. This gives you an idea of the performance of the device. The Fluke 8000A, being a 2000-count device with a three-digit accuracy most of the time, is not the device to buy if you want to measure exact references or if milivolt precision is important to you. Luckily, it also does not carry the price tag such features come with, so if it’s not what you need it can be a nice multimeter for the price.

Unlike some of its contemporaries, it lacks continuity checking or diode measurements, so I would strongly advise against choosing this device as your only multimeter. You can get pretty good cheap multimeters, like the Aneng AN8008, from China – or if it carries them maybe your local retailer. It really is a steal, and I would recommend it if you’re not doing any work with dangerous voltages (although it has worked for me measuring 230V line voltages, do so at your own risk!).

There are many variations of the Fluke 8000A available. For starters, there is the plain 8000A, with a big transformer and filtering capacitors for stabilizing the supply voltage. There is also the 8000A-01, which requires batteries to be operated from net supply, an interesting feature I will get to later, as this is the model I got. Then there’s also a version with a low-ohms option – the 8000A only goes down to a 200 Ohm range in its default configuration, another reason to get a second multimeter just for that – which carries the third and last variation of the main PCB.

If you’re a bit confused at this point: There’s still more to choose. My version came with option 05: An integration option on the mA ranges – AC & DC – that is completely analog! An interesting and impressive feat of engineering which requires at least three separate references by my count. There is the earlier mentioned low-Ohms option, a data output unit option – to get your measurements to digital devices – and a high-current range option. Lots to choose indeed. There appear also to be some overlap in the model numbers used for each option, but that might be due to regional differences.

If you can, it could be worth checking what features a particular 8000A has before you buy it, as with all multimeters. The integration of mA might be worth a premium to you, for example, or if you could really benefit from the low-Ohms option, look for it!

8000A-01 Battery Issues

The model I received held a surprise for me when I opened it up. It is the 8000A-01 with support for battery power! The option for current integration was clear from the product pictures, but I had no idea there was even a battery powered option before I bought it.

So what’s the deal with this battery powered multimeter? It actually requires the batteries in order for its power supply to function. It really is one of the most interesting old power supply designs I’ve ever seen – though that’s not that impressive if you consider how few I have seen.

When I got it, it had a problem measuring resistances near the top end of its ranges. When measuring a 1.8kΩ resistance, for example, it would only go up to about 1.3kΩ. So, I opened it up and started trying to find what the problem was. I was met with huge battery holders.

If looking at old circuit designs like this doesn’t get you excited, you’re on the wrong blog.

Apart from it not working entirely as it should, it really was in pristine condition. I feel ashamed to admit that I did ruin some of the natural beauty this gorgeous piece of 80’s engineering holds, even if it was only the power supply. Modern designs can be pretty nice to look at too, but nothing beats a PCB like this. It’s double-sided too, and all traces are tinned copper on substrate – no ground planes, no fills, nothing modern like that here.

In the picture the battery holders are shown on the right. These are intended to hold 1.2V NiCd batteries. NiCd batteries are pretty hard to get nowadays, especially in the size that this puppy requires them. Four of them could set you back anywhere from €20 to €40, depending on where you buy them. As this multimeter only cost me €15 and I did not know yet that these batteries would turn out to be essential in the design of the power supply, I got to work finding the problem without the batteries.

The great thing about old gear like this, is that the manual includes full schematics, component specifications and even the PCB layout in this case. Measuring some basic things is therefore really easy. So, when I got to work I measured the DC voltages available from the power supply to start with: the 5V line was at 3.3V, whereas the symmetric 15V supply was even worse, at 9.9V. See the numerical relation there?

So, what to do? Well, I decided to measure the supply rails with my oscilloscope. As usual I forgot to make screen-grabs of what it looked like, but suffice to say that while the DC voltage was about 3.3V, there was an AC voltage of approximately 1.5V peak super-imposed on it. Typical for an unstabilized power supply. Looking at the schematic of the power supply, this actually made total sense.

Multi-Purpose Batteries

After we’re all done swooning over the gorgeousness of these hand-drawn schematics, let’s focus on what’s going on here. The light bulb visible in the picture of the internals of the multimeter is actually part of the charging circuit of the batteries! What’s going on?

Really, the power supply here should be split into two separate sections. There’s the part before the switch, which supplies a stabilized 5V, and the part after the switch, which generates a symmetric 15V supply from it. Let’s start with the 5V section.

The 230V mains is converted down by a symmetric transformer, which supplies two 5V AC windings – pins 5 and 7 – centered around a ground connection – pin 6. These 5V AC supplies which are in counter-phase are rectified with a single diode each; It is not entirely clear to me why they opted for a symmetric 5V AC supply rather than a single 5V AC line with a full-bridge rectifier, but it may have been to shave a few more cents off component costs. This unregulated, rectified 5V AC is fed to a 22Ω resistor and the light bulb. The batteries are connected to the other side of the light bulb.

The batteries actually play a crucial role in stabilizing the voltage. For one, they act as giant capacitors, smoothing out the rectified 5V – hence why I saw 1.5V AC on the 5V line without them. However, while they are charging their voltage is actually below 5V. The interesting voltage-current characteristic of a light bulb will cause them to charge at a controlled current while also forcing the voltage of the 5V line to be at whatever voltage the four NiCd batteries have across them when placed in series, acting as a sort of zener diode. When disconnected from the mains, the batteries would supply a very stable supply voltage of around 5V, albeit at some margin.

It should be clear now why my 5V line was at 3.3V with a 1.5V AC signal. My first attempt at fixing the problem was adding lots of capacitance to stabilize the voltage: About 10mF’s worth. This is a terribly stupid and bad idea! It does not respect the zener-like characteristic that the batteries play in this design. Even when placed at the point where the batteries are supposed to be, this causes the DC voltage to be much higher than 5V, approximately 7V. When I realized my mistake, I was afraid that I’d ruined the digital circuitry with these few seconds of too high a voltage. Luckily, this beautifully designed circuit turned out to be more resilient than that.

After verifying that the multimeter still worked by injecting the right voltages with my lab power supply, I concluded that if I wanted to have a stable 5V supply, I’d have to completely disregard the original design of the 5V circuit. So I put in a 5V switching power supply at the point before the switch, connected it to the mains input and called it a day. Or did I?

Oscillator Tricks

As the symmetric 15V is generated from the 5V, let’s look at how that works. These days special chips are available for generating such supply voltages with a few inductors and some charge-pump capacitors. However, back in the day, this was not the case. Apart from the ADC and some display circuitry, this device is wholly analog, without fancy PWM signals and switching power supplies.

Yes, this is the same image as before. It’s just too nice to look at not to put it in twice.

However, this power supply actually does act as a switching power supply, in a sense. The MPS6560 NPN transistors Q22 and Q23 together with the transformer and T2 and capacitors C23 and C25 (C24 is not present in this design) form a double Hartley-oscillator. R22 and R24 are present to bias the base of Q23. The oscillator operates at approximately 10kHz and generates a time-varying current through the coils of T2, which triples the voltage – the source of the 3.3V/9.9V relation from earlier – after which this AC voltage is rectified by a full-bridge rectifier and stabilized with some capacitors. The design runs at currents up to about 30mA at both supply voltages.

Personally, I think this is a great trick. This is not something you’re taught in a university class, or at least not one that I’ve followed. The low switching frequency limits the efficiency, but it’s amazing to me that something like this can be achieved without any microcontrollers or power MOSFETs.

I must note though, that by getting 7V at my 5V line, while it did not damage the ADC or other digital parts – which would be costly and difficult to replace – the maximum ratings of these MPS6560 transistors were exceeded and both of them were damaged. As these transistors are no longer available, I had to replace them by 2N2222A’s. However, the design of this power supply is so ingenious that it used MPS6560’s because they were audio transistors, which sacrificed gain for nicer transient characteristics. Both the 2N3904 and 2N2222 cause ringing in the circuit when switching, generating an audible whine at 10kHz. I’m still looking into a way to at least reduce this effect.

Conclusion

So, after the whole ordeal, I now have a Fluke 8000A that operates within spec, and only from mains power. However, in the process, I mutilated the power supply, which still hurts me quite a bit to think about. The hand-drawn PCB is quite delicate to solder on, and it’s easy to rip off entire traces if you’re not careful enough.

The multimeter after the whole modification ordeal, measuring DC current within its 1% accuracy spec.

However, personally I learned a lot from this project. I can wholly recommend projects like this. Even if hardware is old, if schematics are available it’s easier to figure out what’s going on and you might actually learn more than if you just grab a newer device.

Don’t go snapping up all Fluke 8000A’s though, because I for one will be on the lookout for another one, that I can hopefully treat with the respect it deserves.

Leave a Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.