The Graaf GM20 is one of those amplifiers that, for better or worse, never leaves you indifferent. We are talking about a tube OTL (Output TransformerLess) design based on the 6C33, conceived to drive loudspeakers without an output transformer and with a very particular technical approach: when it is properly sorted out it delivers a big, fast and believable sound, but it demands correct maintenance and, above all, carefully selected components and tubes. In this article I document a complete overhaul of an original 1996 unit, which arrived in working condition but with electrolytic capacitors long out of specification. I also add a practical guide to adjusting bias and offset safely, together with some notes on the most common faults I have seen recurring on this model, including issues with the balanced inputs and the long standing case of the internal “inductors” that can cause an apparently unsolvable hum.
The goal is not to “modernize” the GM20 or to alter its character, but to bring it back to correct operating conditions while respecting the original design philosophy. On equipment like this, the real difference is made by maintenance done with good judgement: replacing what is genuinely subject to wear, such as electrolytics and unstable tubes, checking adjustments and wiring, and leaving alone high quality components that have no technical reason to be replaced just because they are old.
This GM20 was brought to me fully functional and in original condition, with no components replaced since it was built in 1996, and with the request for a general overhaul. As soon as I started the inspection I noticed a very common situation on machines that have effectively been “frozen in time”: practically all the electrolytics had lost value, both those on the PCB and the four large Kendeil capacitors. This is typical behavior, even when the amplifier “seems to work”: capacitance drops, ESR rises, and the circuit starts operating outside its intended conditions, with higher ripple, increased stress on components and adjustments that become more sensitive.
On an OTL like the GM20 this aspect is even more critical, because the power supply and the operating points of the 6C33 tubes directly affect output offset, residual noise and overall stability. In practice, before even talking about “temperamental” tubes, it always makes sense to be sure that the power supply is doing its job properly.
I therefore had to remove the entire PCB for replacement. This is not a “half hour job”, because on these amplifiers you have to work carefully, mark connections, avoid stressing wiring and connectors and, above all, put everything back together cleanly. A GM20 does not forgive rushed work, and many problems later blamed on “mysterious oscillations” are in fact the result of poorly executed previous interventions.
I also took the opportunity to thoroughly clean it from the dust and grime accumulated over 25 years. This cleaning is not just cosmetic: dust and dirt retain moisture, promote surface leakage currents where they should not exist, and over time can create unreliable contacts, especially in the presence of heat. In the photo you can see the four capacitors that had suffered the most, a clear sign of how much the unit had worked for years under real thermal stress.
At this point I want to warn GM20 owners about certain charlatans who want to “refresh” the entire amplifier, replacing not only dried out electrolytics but also all film capacitors and resistors. Do not let them do it, take the amplifier away before they get their hands on it.
First of all, the type of resistors used is of very high quality and, moreover, Mariani often selected them. A resistor, if it is not visibly burnt and not of carbon composition type, and there are no carbon composition resistors in the GM20, does not “degrade” over time simply because it is old. If the value is correct and there are no signs of stress, there is no technical reason to replace it just because it is old.
The same applies to the ICEL polypropylene capacitors: they are high quality components, individually selected, and unless they are clearly faulty, for example shorted, there is no reason to replace them. Above all, do not be persuaded by those who say “they are industrial capacitors, fit Mundorf, Jensen or whatever else”. The ICEL units selected by Mariani are functionally indistinguishable from the best polypropylene “big cans” on the hi fi market. Replacing them at random only introduces unnecessary variables and often makes the amplifier worse.
Photo of the PCB of a GM20 that has undergone the pointless replacement of all the original ICEL capacitors, individually selected by Mariani and perfectly functional, with Audyncap capacitors “whatever was in the drawer”…
Once the capacitor replacement was finished and the board reinstalled, I proceeded with testing and with bias and offset adjustment. Here I also clarify a frequent question about this amplifier: no, it cannot sensibly be modified to work with automatic bias without completely overturning the project and introducing additional complexity and failure points. The GM20 was designed with manual adjustments and, if properly maintained, it works correctly that way.
A problem often reported by GM20 owners is the need to frequently readjust bias and offset because the amplifier starts to hum. Part of the issue is related to the fact that, at the time it was designed, 6C33 tubes of far higher quality were available on the market compared to what can be found today. Mariani selected them extremely carefully: he rejected those that could not withstand 1 kV of internal insulation between electrodes and those that showed intermittent shorts when “hit” with a rubber hammer, something I personally witnessed.
In addition to this initial selection, dozens of tubes were run in on a large test frame where they stayed powered on for a week, and only then were they paired. It is said that GM20 units fitted with tubes selected by him could run for up to 20 years without needing tube replacement or trimmer adjustments, and this is plausible. With such strict selection, initial stability and early life drift are much more controllable.
With today’s available 6C33 tubes, unfortunately, it is no longer possible to rely on that level of stability. Even pairs sold as “matched” sometimes stay matched for less than a week, after which it becomes difficult to adjust offset because the tubes vary significantly during the initial run in, which effectively takes place inside the amplifier. For this reason I published the 6C33 tube tester project, which can help GM20 owners reselect their own 6C33 tubes properly before installing or reinstalling them in the amplifier. On average, adjustments and tube re pairing will be needed about every three to four weeks.
It is therefore absolutely wrong to throw away pairs of 6C33 that no longer match. As mentioned, 6C33 tubes change their parameters markedly during run in and over their service life, but this does not mean they are faulty or worn out. After a certain settling period they tend to stabilize more. The practical approach is to keep the tubes that no longer match, accumulate a sufficient number of them and then, calmly, look for new matches among used tubes to form new pairs to reinstall in the GM20.
Yes, it is annoying, but those who sell 6C33 today often do not run them in, or do not run them in long enough, before selecting them, so the match between two new tubes is destined to disappear quickly. So far, the experience of those who have built my 6C33 tester is that new pairs formed from initially rejected tubes tend to last longer on average than newly bought “pre matched” pairs.
How to adjust BIAS and OFFSET on the GM20
Disclaimer: do not open your amplifier unless you know what you are doing. Tube amplifiers operate with lethal voltages that can remain inside the amplifier even after it has been disconnected from the mains. If you do not know how to work safely on your amplifier, do not try. Ask a technician for help. If you die or your amplifier is damaged, I am not responsible. Click here for more details.
The GM20 is a beautiful sounding amplifier, but it is very important to adjust bias and offset from time to time, and always when you change tubes. It may seem complicated the first few times, but with a bit of practice it becomes a simple and repeatable procedure. When carrying out the adjustments you will be working inside the amplifier, so you will need: an Allen key to open the bottom cover, a small flat screwdriver for the trimmers and a reliable multimeter to read voltages.
Before starting, make sure the amplifier is on a stable surface, that cables cannot get caught and that you have enough space and light. Remember that in an OTL the output is directly linked to the active circuitry, so avoid any improvised or rushed operations. Turn the amplifier upside down and remove the cover.
Get a pair of power resistors rated at around ten watts, with a value between 22 and 33 ohm, and connect them to the speaker terminals, because it is not safe to power up an OTL without a load. The absence of a load can lead to abnormal conditions and, in general, it is one of those “things not to do” if you want to avoid unnecessary problems.
Switch on the amplifier and let it warm up. Inside you will find the test points where you can read the bias voltage. Other guides say it is very awkward to keep the meter probes steady while trying to adjust the trimmer. I add that it is also risky, because the probes can slip.
But there is good news: those who wrote other guides for GM20 adjustment often did not notice that the two test point pins are connected to a resistor located just about one centimeter away. This resistor is large and intentionally raised from the PCB, so ignore the test points, get two meter leads terminated with alligator clips and connect them directly to the resistor. This way you have your hands free, you reduce the risk of slipping and the measurement becomes much more stable.
With the meter set to the 200 mV range, adjust the “BIAS” trimmer until you read 30 mV. Then connect the meter to the speaker terminals, across the resistor you connected as a dummy load, and adjust the “OFFSET” trimmer until you read a value close to zero. Do the same for both channels.
Leave the amplifier on for half an hour or 40 minutes, because 6C33 tubes really take a long time to reach full operating temperature, and check again from time to time, making small corrections calmly. The practical rule is simple: no quick adjustments just after power on, because you will be chasing continuously changing values and it will seem impossible to make them settle.
Hum or low level problems on the balanced inputs
A fault I have often observed on the GM20 concerns the balanced inputs, with issues such as low level, low level on only one channel, hum, one channel dead and so on. More often on units that have already undergone repairs or tampering. The video below is self explanatory:
Explanation: unfortunately, a design flaw of the GM20 lies in the fact that the PCB barely fits inside its cabinet. In some units, due to tolerances, it does not fit at all, even if only by a small margin. To insert and remove it, force is almost always required, and this force is transmitted to the XLR connectors, because they are precisely what prevents the board from sliding in smoothly.
For this reason, units that undergo repairs or “tinkering” often end up suffering failures of the connectors themselves, which then have to be replaced. It is a classic case where the fault does not arise spontaneously, but is encouraged by repeated disassembly and inevitably aggressive mechanical operations in an area that is already critical by design.
Persistent hum problem on the GM20
Sometimes some GM20 units start to suffer from a persistent hum on one channel, or on both, that does not seem to be resolved by changing tubes or readjusting bias and offset. In this case the fault is attributable to the two “inductors” you can see in the photo.
I put “inductors” in quotes because they are not inductors. I know how they are made internally, but without having the schematic of the amplifier the exact operating principle of these two parts remains unclear. What has become evident, however, is that sometimes one of the internal windings goes open circuit, and when that happens hum appears at the speaker output on the faulty channel or, in some cases, on both channels.
Here too, repairers often indulge in imaginative diagnoses and creative “cures”. The point is that if the symptoms are consistent and everything else has already been checked, these units should be considered among the prime suspects, not among the last.
User experience has also shown that usually, when the first one fails, the second one fails shortly after. Several years ago I reproduced a clone of these units starting from an original one. See this page for more information.
If you are chasing a hum that “just will not go away” and you have already ruled out tubes, adjustments, power supplies and contacts, this is one of the very few paths that is really worth checking, because the symptom is subtle and tends to waste a lot of time for those who have not already seen the problem in person.
