Hafler 9505/9303 Power Amplifiers Repair and Measurements

Over the past few months a few different Hafler amplifiers passed through my bench. First it was a combo of a 915 pre-amplifier with a 9303 power-amplifier to accompany it, and some time later I got a 9505 power-amplifier. If you’ll have a look at the schematic you will see that while the pre-amp isn’t too exciting, the power amplifier has a few interesting points in its schematic. So I’ve decided these amplifiers are interesting enough to warrant a short post about them with some pictures and measurement results.

The 915/9303 combo was given to me for repair by a friend. The power amplifier was working well and only needed a check-up to make sure its still operating as expected (such as quiescent current for proper operating point), while the pre-amp was faulty. One of the channels of the pre-amplifier wasn’t working at all.

The 915 pre-amplifier is a very simple pre-amplifier if you have a look at the schematic. As a consequence it was also very easy to find the problem, RLY1 had a bad contact. Apparently it wasn’t the first time this part was replaced so the owner decided he prefers to short it instead which is what I did. While at it I’ve also replaced the op-amp driving the headphone jack so that it is somewhat more capable.

Fig. 1. Hafler 915 Pre-Amplifier – Bad Relay

I then moved to the 9303 power amplifier, which is the more interesting of the two. Apparently there are multiple variants of this amplifier, so finding a schematic that would match the exact variant you have can be a bit problematic. Looking at one of these schematics in Fig. 2, you can see that the output stage has a non-standard arrangement.

Fig. 2. 9303/9505 Schematic

The output stage isn’t using a Source-Follower structure, its instead using a Common-Source stage with 3/4 devices (9300/9500 series) to act as a gain stage. Not only that, the bulk capacitors are reused as the output coupling capacitors in this case. This structure is obviously non-trivial to stabilize across different loads presented by different loud-speakers, and as you can see there are multiple compensations components, as well as a local feedback loop around the output stage. I must admit I quite like this structure, you could argue about its pros and cons, for instance, it doesn’t need a high-voltage supply for the first stages to drive output stage properly. However, in my opinion, its an interesting structure, and I always enjoy amplifiers that are different in structure than the typical “text-book” approach most amplifiers follow.

Fig. 3. 9303 Inside Look

Getting back to the 9303 amplifier I was working on, “unfortunately”, there was nothing wrong with it. Bias was set right, and it worked great. So I gave it a little clean and closed it all up to go back to its owner. Then, a couple of weeks later I get a phone call about it, saying there’s something wrong with it. The problem description (along with the sounds the speaker was making when it happened), led me to believe the output was shorted which might have damaged the output stage, or hopefully only the fuses which are in the path of the output signal. Luckily, after getting the amplifier back and having a look inside, the fuses were indeed the only problem. I’ve measured it again to make sure there is nothing wrong with it. Fig. 4 shows the measurement results.

Fig. 4. 9303 Driving 145W Into 8ohm – Both Channels Loaded

The distortion according to the manufacturer should meet 0.07% @ 150W into 8ohm load. In this case I’m seeing much lower distortion (0.007%) @ 145W, although I measured at 1KHz only and it will degrade with frequency. This is more or less the sweet spot, and the number jump up sharply after this point. Considering the age of the amplifier and the fact the bulk capacitors are probably not at their peak, this is completely acceptable and doesn’t call for any work to be done, and it could go back to its owner.

Roughly around the same time the 9303 was on its way back to me, I found a damaged 9505 for sale in one for the local online stereo boards. I had no info about the state it was in, but the price was reasonable, so I decided to buy it and see if it can be fixed. If not, I could always reuse the chassis and transformer for another project.

First things first, the 9303 isn’t a small amplifier. Its quite compact for its power, but its a medium sized amplifier with a hefty weight to it. What about the 9505? well, it dwarfs the 9303 (as you’d expect consider its a 250WPC vs 150WPC amplifier):

Fig. 5. 9303/9505 Size Comparison

Inside look shows its a similar structure, but everything is super-sized. The transformer is a beast, the capacitors are huge (4 caps of 20,000uF/100VDC rating), there is an added XLR/TRS combo connector in addition to the RCA jack. Next thing you notice, is that its in a very bad shape. The paint appears to be swollen across the heat sink and a few other places. The top and bottom covers quite a bit of rust on them. But what’s going on inside? Well, its not good. A lot of corrosion across the PCB. Many damaged traces (rusted completely or close to it) appear, so component leads are so full of rust that they break off as soon as you touch them. Looks like I will have some work to do. Fig. 7 shows some examples, you can find many others that I haven’t circled in the same picture, and this is just a portion of the board.

Fig. 6. 9505 Inside Look
Fig. 7. A Couple of Examples of Rust on the PCB and Component Leads

I went over the board, removed any component that didn’t look right, as well as components that were in the way of reaching rusted traces. Then took a small micro-motor tool (similar to Dremel but much smaller and lighter) and started grinding down the bad traces. If you ever need to do this, make sure you take pictures of the board before doing so, as you might have to grind down component identifiers on the PCB. Once this is done, I add a stripe of copper tape, with glue on the back side, and cover it all with solder to keep the copper safe from corrosion. Once done, I use a green UV-activated solder mask liquid to cover it. This doesn’t just look nicer than exposed metal, its also safer as it prevents an accidental short if you place a component over the repaired trace.

Fig. 8. Fixing Traces Mid-Work

Fig. 8 shows an example of a few traces that were repaired before the green solder mask was applied, and before I punched the holes to for the components in the copper tape over the through hole drill points. Over all there were probably 20-30 traces that called for a repair.

Next I gave the PCB a good rinse, to with some OPA from both sides. Once that was done I moved to replacing the components I’ve removed. I’ve removed all 4 bulk capacitors from the board and tested them at load voltage to make sure they are all still in good shape. I was happy to see they were all in good shape and didn’t call for a replacement. Finally I put it all back together, hooked it up and started testing. All DC points looked good, so I moved on to testing with some signal.

Fig. 9. 9505 First Testing After Repair

I was happy to see that there were no additional surprises, the amplifier was working properly. I gave it some time to come up to steady temperature, readjusted the bias which was about 20% lower than expected on one of the channels, and adjusted the CMRR knob of both channels. I was able to reach the 75dB spec for CMRR as measured on my setup for both channels by fiddling with the knobs until it reached the optimal point. Next I repeated the measurements at full load.

Fig. 10. 9505 Driving 242W Into 8ohm – Both Channels Loaded, SE RCA Input

The maximum power I reached while meeting the THD value is 242W (with both channels loaded, its >250W with single channel loaded), slightly more with 0.1% maximum THD rating. This is at 1KHz only, which I’m using as a representative frequency, it can obviously degrade with frequency, I will probably measure again but over the entire frequency range once I’m done with the amplifier. At slightly lower power of 230WPC, the distortion is less than half that at 0.025%. Its worth noting that using a balanced input should drive this distortion down somewhat by reducing even order components of the spectrum. One more important note to add here is the voltage rating of this amplifier. It is a 240V unit according to the label on the back, while the mains voltage here is 230V. Crunching the numbers, this would turn a 250WPC amplifier into a 230WPC unit. Therefore this result is more than satisfactory.

Fig. 11. 9505 Driving 230W Into 8ohm – Both Channels Loaded, SE RCA Input

You can see the 9505 suffers from ~10dB higher 50Hz induced noise. This is expected considering its much larger transformer, and the fact I was measuring it with the case open and the signal wires running not far from it using the SE RCA input. I expect this will drop somewhat when I move to a balanced input and/or measure with the case closed to reduce coupling from the transformer to the input signal cable.

The 9505 is now operating properly and the performance is close enough to spec (just a few Watt’s shy of the 250W rating at the promised THD figure) for me to be happy with it. After all, its not a new unit, the bulk capacitors might have degraded slightly over the years. There is still work to be done with the amplifier. The TRS/XLR combo jack must be replaced. It is corroded too, and has a problematic contact on the XLR side. I’ve ordered replacements for it and will take care of it soon. Additionally, I need to address its appearance. I will probably sand off the top panel and apply a fresh layer of paint to make look as if it was new. The heat-sink as screws holding the case from the outside will be treated to a similar fate, although I might replace some of the screws to save some time. Once complete, I’m sure this amplifier will serve well for many years to come, and the output power of this beast might prove useful some day if I get a pair of power hungry speaker 🙂

Edit October 17, 2020:
The new XLR connectors got here, and were put into the unit. I’ve used a Neutrik NCJ6FI-H combo jack which seems to be the modern equivalent of the original part, and fits properly into the unit. While at it I gave the amplifier some TLC and fixed the paint job on the heat-sinks, top panel, and front handles. For the front handles and top cover I went the extra mile and sanded them down completely to make sure the new layer of black paint will look as good as new. This was especially important for the front panel as it did have rust in multiple points. Still have to wait for the new rubber pads for the bottom panel, but other than that its basically finished.

Once done with that, I put it back together, connected the XLR input and measured distortion once more. I see the balanced/single-ended input has no noticeable affect on THD or its components. This is reasonable considering the amplifier isn’t truly balanced throughout. While at it, THD was measured up to 20KHz to see if it was up to spec. I have measure its at ~235W output, and I see this is a very nice 0.023% at 1KHz (2nd harmonic dominant), well within spec. It goes up with frequency, 0.086% at 10KHz, and up to 0.11% at 20KHz (at these frequencies at max power the 3rd harmonic is actually more dominant). Is the THD of any real value at these high frequencies where harmonics are far outside the frequency range we can hear? Maybe, maybe not, but this figure is good enough to know the amplifier is operating properly, and a little overhaul will probably pull that back inside the 0.1% specified in the 9505 spec. These number are all with Z-weight, which means they “raw” numbers.

You can see the spectrum at 1KHz attached below. As you can see, the 50Hz pickup is significantly reduced now that the signal cable isn’t routed next to the power cable as it was before, and the case is closed. The 50Hz at output is actually similar to the 9303 I’ve measured before when both are driving the same output power,. When driving the 9505 to full power the 50Hz tone is even slightly lower compared to the signal amplitude which is now increased by a couple of dB. All of this results in an improved THD+N figure due to reduced noise/interference.

Fig. 12 – 1KHz Spectrum when Finished

This bring this session with the Hafler’s to an end (at least for now :)). I still have no idea what to do with this power amplifier, as I have too many amplifiers already, and 250WPC is way more than I need for any of the speakers I have, but I’m sure I will think of a good use to it. Perhaps I will even build or fix-up a pre-amplifier to go with it.

Leave a Reply

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