J.C. Verdier LE220 and DE220 Amplifiers: Technical Analysis, Restoration, and Optimization

When examining various tube amplifiers produced mainly between the 1980s and 1990s, I have often encountered unconventional design choices and, in some cases, compromises that can introduce measurable limitations. In the specific case of the J.C. Verdier, I had the opportunity to analyze the original design and to identify some aspects that are open to optimization. I would like to highlight, in a respectful and professional tone, a few technical considerations regarding the bias system, because elements of this kind can affect correct operation and long-term repeatability of performance.

More specifically, I noticed that instead of adopting a separate cathode resistor for each tube, for each pair of tubes, or a fixed-bias system with dedicated adjustments, the design connects the cathodes of all four output tubes together. This configuration allows the total cathode current to be used to power the heaters of the two driver tubes. However, this design choice can introduce some critical issues: if one tube conducts more than the others, the voltage on the common cathode tends to follow this variation, influencing the bias of the entire group. This can result in current imbalance among the tubes, with a potential increase in distortion and the possibility of a residual DC component in the output transformer, especially when the tubes are not perfectly matched or age differently. Below is an approximate schematic of the circuit.

The trimmer present on the board allows adjustment of the grid voltage, but it does not allow independent and stable setting of the operating point for each tube or for each pair, nor does it allow long-term optimization and maintenance of current balance among the tubes. In practice, its purpose is to increase the total current in order to obtain adequate heater supply for the ECC81 and ECC83. In other words, the output tubes are also forced to perform a “service” function within the heater supply chain, a condition that can penalize optimization of their operating point and the stability of current balance among tubes. For these reasons, based on my experience, I considered it appropriate to intervene on the bias system to improve reliability, current balance, and overall performance of the amplifier.

In the two photos below, the two versions with and without controls

It is important to note that, in an industrial context, some choices may be driven by cost, assembly, or standardization constraints, and not necessarily by the pursuit of maximum electrical optimization or maximum long-term performance repeatability.

In the specific case of the amplifier under analysis, there are solutions that are unusual compared to what is today considered more “robust” in terms of bias stability. This does not automatically mean that the design is “wrong” in absolute terms, but it does mean that it can present room for improvement, especially when the goal is to reduce the effects of uneven tube aging and minimize conditions that can lead to current imbalance between the push-pull branches.

Fortunately, the cathodes of the output tubes and the heaters of the ECC8x tubes are connected via wires rather than PCB traces. This made it possible to intervene without irreversible modifications to the original layout, avoiding cutting traces and preserving the integrity of the PCB, with the exception of a few mounting holes for the “patches”.

In the photos below, the appearance of the PCB in the version with controls and the one without controls

Before starting the modification, I obviously repaired the circuit, which had a large number of electrolytic capacitors that had reached the end of their useful life. I began desoldering them one by one to test them on the bridge, and I had this surprise…

At least half of them had pins corroded from the inside; in practice, the capacitor had disconnected itself from the circuit. I would like to point out, however, that those that “survived” showed very good electrical characteristics compared to the standard replacement I had available at the time. Take a look:

To clarify the issue briefly: the originals, with at least 40 years on their shoulders, were dual 25+25uF units used with the sections in parallel. The total nominal capacitance was 50uF, while the measured value was about 52uF with an ESR of 0.26 ohm. By comparison, the new ones (red colored), produced a few months ago, with a nominal capacitance of 47uF, measured about 42uF with an ESR of 1.1 ohm. This comparison should not be interpreted as a general rule that “old is better than new”, but rather as a practical reminder: replacements should not be chosen only for capacitance and nominal voltage, but also evaluated for parameters such as ESR, tolerance, and ripple behavior, especially in the most critical parts of the circuit. Measurements were performed with an LCR bridge, under conditions consistent with what is visible in the photos. From a sonic point of view, such a marked difference in ESR can be audible, because it affects filtering and power supply dynamics.

Of course, there are modern high-quality electrolytics that would have been a valid replacement choice. However, in addition to the electrical aspect, I also had to consider a mechanical constraint: I needed a replacement with at least the same diameter as the originals, to ensure correct and clean mounting without compromising aesthetics. For this reason, in the three most critical points of the circuit, I added Audyn Cap polypropylene capacitors in parallel with the electrolytics, in order to improve high-frequency impedance and compensate for the limitations of the chosen replacement.

This episode is also an invitation to restorers to avoid unverified “blanket replacements”: correct practice consists in measuring components and replacing what is actually out of specification or at risk, then choosing replacements suitable for the application. In some cases, replacing still-valid components with replacements of non-equivalent quality can worsen the final result, even in sonic terms.

The correct practice in situations like this is to use a quality LCR bridge to verify the health of the component to be replaced. It is essential to carefully evaluate whether replacement is actually necessary and, if new components are chosen, to verify their characteristics to ensure they are up to the task. A second intervention I carried out was the replacement of the old red Wima polyester (not polypropylene) capacitors rated at 250 volt with axial polypropylene capacitors, which I had in abundance in my stock.

In the case of film capacitors placed along the signal path, the dissipation factor (D) is a useful parameter because it describes dielectric losses and therefore how much energy is dissipated as heat instead of being transferred ideally. In practical terms, a lower D tends to correspond to a “cleaner” behavior at high frequencies and lower loss introduction. For example, the D of the red Wima is 0.0032 @ 1 kHz, while that of the ERO polypropylene capacitor is 0.0001 (values comparable to those found in some “audio grade” models). Considering that the originals were 470nF and I had 220nF capacitors available, I opted to install two in parallel. Although I also had 470nF capacitors, they were rated at 1000 volt and were too bulky for the available space. This difference in D factor can translate into less high-frequency degradation and increased clarity, especially in the upper range.

What is the dissipation factor?

The dissipation factor (D) is a parameter that describes the dielectric losses of a capacitor. In practical terms, it indicates how much energy is dissipated as heat when the capacitor operates with AC.

In an audio circuit, a higher D can contribute to less ideal behavior at higher frequencies, introducing losses and, in some cases, reduced clarity in high-frequency response. For this reason, in points where the capacitor is located along the signal path, the choice of dielectric and construction quality can become relevant.

In general, many good-quality polypropylene capacitors exhibit lower D values compared to other families. This does not mean that a component “is worth” only that parameter, but it is one of the useful indicators to consider together with tolerance, stability, and physical size.

Once the recap was completed, I focused on modifying the bias system, building and sandwich-mounting two pieces of 1000-hole prototyping board on which I implemented a Blumlein self-bias, self-balancing circuit that I have already discussed in this article. In this way, I achieved, in a simple manner and without the need for periodic adjustments, a system that tends to keep the bias currents of the two tubes per channel balanced, with more uniform wear, lower distortion, and reduced risk of DC component in the output transformer.

The yellow electrolytics you see belong to NOS batches that I purchased years ago, electrolytic capacitors that in various practical applications have proven to be very valid even compared to many modern productions. I also changed the value of the control grid stopper resistors of the output tubes from the original 470k to 220k, because 470k is a high value for output tubes such as EL34, 6L6GC, and similar types. Under some conditions, it can increase the risk of operating point drift as tubes age and leakage currents appear, with potentially destructive effects as can be seen in some Unison Research units. It is true that many vintage datasheets report even higher values, but in my practical experience 220k offers greater stability and reliability up to end of tube life, with a negligible impact on drive capability in this type of circuit.

The last three modifications I carried out were adjustment of the NFB network to increase the damping factor from about 1.5 to 2 in the original circuit up to a factor of 6.1 (there was sufficient gain to do so without making the power amplifier difficult to drive), and the exclusion of the rear inputs served by a pair of sliding-contact switches, which were heavily oxidized and difficult to source as replacements. I also bypassed the Control/Direct switch, routing the signal from the four RCA jacks of the direct input straight to the two volume potentiometers, which were replaced with two new units. I would have liked to install an ALPS stereo potentiometer, but physically it was difficult to fit, so I opted to replace the two mono potentiometers with two identical new ones. I also powered the ECC81 and ECC83 directly with AC from the circuit that supplies the output tube heaters. The circuit now delivers about 14 to 15 watt RMS, with the tubes that were present at delivery time, not brand new but not worn out either, which however are 5881 and not EL34 or 6CA7 as specified for this amplifier. Let us take a look at the measurements:

The bandwidth at 1 watt is 20Hz -0dB / 43kHz -1dB

The bandwidth at 10 watt is again 20Hz minus a fraction of a dB and just over 40kHz at -1dB

I would like to emphasize that my analysis is based on direct observation, measurements, and before-and-after comparison. When I criticize a technical aspect, I do so because I see its practical effects on operation and performance repeatability. At the same time, it must be acknowledged that the output transformers of this J.C. Verdier are well made, with very good characteristics and, in terms of electrical behavior, comparable to high-level transformers. Furthermore, the graphs I publish, as you can verify by opening one full screen, have a vertical scale of “1dB per division”.

I continue with THD at 1 watt and at 10 watt, with distortion of 0.17 and 0.28%

Square waves at 100Hz / 1k / 10k

Listening tests immediately showed a very clean and detailed behavior. The amplifier was delivered with four 5881 tubes, one Sylvania ECC81, and one Sovtek ECC83, all in good working condition. Replacing the Sovtek ECC83 with a Philips Miniwatt resulted in a further improvement in high-frequency brilliance and overall detail. Compared to the original version, these modifications brought more air and control. A power supply section that is lightly filtered and not decoupled between channels is still present, but despite this, the final result in terms of performance and sound quality is superior to what can be heard from many amplifiers on the market with even more generous declared specifications.

Error in Tube Selection: When a Mistake Causes Serious Damage…

In another unit of the same model, I happened to see what can occur when tubes are replaced “by feel” without verifying real compatibility. In that case, the customer installed tubes belonging to a similar family, but not equivalent in terms of operating limits and dissipation compared to the correct variant specified by the circuit. This choice caused significant damage, evidenced by carbonized components, burned PCB areas, and melted sockets. The photos below document the incident and serve as a reminder of how important it is to select and replace tubes correctly, always verifying real compatibility, pinout, and operating conditions.

For a detailed understanding of the differences between the various tubes in the 6L6 family and the risks associated with unconsidered changes, you can refer to this article, click here. The subsequent images document the amplifier repair process. I carefully removed the degraded areas with a milling tool. Subsequently, I applied UV resin to consolidate the fiberglass and proceeded with replacing the melted sockets. This intervention requires precision and care, and demonstrates how important competent work is when dealing with damage caused by incorrect component selection.

In conclusion, I would like to point out that the amplifier chassis specifies the use of 6CA7 tubes, considered equivalent to EL34. Although the circuit can also operate with 6L6GC (and many people choose this option, sometimes for availability reasons), I considered it appropriate to suggest that the customer purchase a new matched quad of EL34 tubes, in order to remain within the intended specification and optimize the operating point. I recommended Tung-Sol EL34 tubes, one of my currently preferred production brands.

Note: The brown capacitors, thin and tall, aesthetically unappealing, were installed by a previous repairer. Although electrically functional and within specification for the intended use, I chose to leave them in place. In the previously described repair, instead, I paid particular attention to the visual aspect of the intervention as well, looking for red capacitors with the same diameter as the originals, in order to achieve a repair that was not only technically correct but also neat and visually consistent.

J.C. Verdier DE220 Control (old article from 2019)

This amplifier can also be optimized, should I happen to have another one in the lab again.

Curiosity: Several people ask what Double Ended means. Some speculate that this amplifier is a parallel single-ended or that it contains some “innovative” circuitry. In reality, in this case Double Ended is a commercial label: the amplifier is a classic push-pull in class AB with a cathodyne phase splitter, a widely known and used topology.

This unit was delivered to me with one channel muted, initially with suspicion of a faulty output transformer. Following inspection, I instead found a worn output tube and a failed resistor on the circuit, as well as several electrolytic capacitors that had reached end of life and were replaced. The output tube bias section adopts a design philosophy entirely analogous to that already described above, with the same limitations in terms of current balance stability. In this case as well, the global feedback is rather limited, with a low damping factor that can make the amplifier more critical when driving demanding loads.

The amplifier delivers 20 watt RMS with a damping factor of about 1.2. The output transformers, unexpectedly, show good bandwidth and would make this unit an excellent base for optimization work (power supply, output tube biasing, controls and connectors, and stronger feedback, where compatible with stability). Let us take a look at the measurements (Note: due to distraction, I set the graphs in linear mode instead of decibels. If I get another one of these units again, I will re-acquire the graphs correctly).

Bandwidth on resistive load (1 watt)