IF7 Amplifier

Description of My IF7 Amplifier

This amplifier is an attempt to mimic the First Watt F7, designed by Nelson Pass. Unlike some other First Watt amps, Pass has not published the official schematic of the amplifier in order to prevent cheap clones from appearing on the market. The amp was introduced in 2016 and apparently is still selling well (as of 2023) at a US list price of $3,000. Pass has hinted that the amp may be nearing end-of-life, so he may soon make the design available to DIY hobbyists. The user guide is downloadable from firstwatt.com. It has a description of the design and specifications.

I called my implementation the “IF7” with the “I” standing for “imitation”. I prefer this to using the word "clone".

The amp is an extremely simple, low-power, Class A design. It will produce about 25 Watts into an 8 Ohm load. At 1 Watt, It has less than 0.05% THD and can be tweaked to make second harmonic distortion dominant, with some 3rd harmonic and much lower levels of higher-order harmonics. Like all Class A amps, it runs warm. My implementation dissipates about 150 Watts of AC power at idle. My intention is to use it in a bi-amp setup, driving the top-end speakers directly. This will necessitate an analog crossover before the amp, covered in this project.

It has only 4 active devices, 2 complementary JFETs at the input and 2 complementary MOSFETs for output devices. There are 8 fixed resistors and 3 (4 in my implementation) variable resistors. There are no capacitors in the signal path as the amp is DC coupled.

The input JFETs are the well-known (and prized by DIY builders) Toshiba types that have not been in production for many years, but NOS parts can still be obtained on eBay. These devices are very linear and exhibit low noise. The output devices Pass used are “lateral” MOSFETs and are also out of production, and very scarce. Like many rare semiconductors, he made large batch purchases years ago and burns down his inventory building production amps built around specific parts. Fortunately, similar devices are now available from Exicon, which are what I used. The lateral MOSFETs exhibit tube-like distortion qualities and are extremely thermally stable. The circuit topology used does not need any thermal compensation once bias levels are set.

A unique feature of this amp is that it incorporates positive current feedback. This is actually an old technique, dating back to tube amp days, but has been rarely used until Pass resurrected it with the F7. He used it because the basic topology of the output stage has an inherently high output impedance, typical of common-source push-pull MOSFET stages. This results in a low damping factor and consequently poor control over the speaker system’s behavior. To solve this, he used a current-sensing resistor in series with the speaker load. The resulting voltage across the resistor is therefore directly proportional to the current through the speaker. This voltage is sampled through a resistor and fed back to the amp’s input. Since it is in phase with the input signal, this results in positive feedback.

The feedback can be adjusted so that the open-circuit output voltage does not decrease when a load is connected. Or more practically, only decreases by a small amount. When there is no decrease, the amp has an essentially infinite damping factor. Since Pass had to build a production amp that would work with all speaker systems, he fixed the feedback at an average value of about 150 to cover most speaker variants. In my implementation, I made the feedback variable so it can be “tuned” to specific loads. See the adjustment section for details.

I would say overall, this amp has a very neutral sound characteristic. It is not fatiguing to listen to and has a sweet and detailed upper end. If you can live with low power, this is a highly recommended amplifier for a DIY build.

IF7 Schematic

When the F7 was released by First Watt, it triggered a wave of speculation by DIYers in the Pass forum. People were extrapolating from the hints that Nelson was dropping there and in the owner's manual. An online review published by 6moons, complete with interior pictures, added fuel to the fire. Several were doing LTspice simulations of various topologies and one of them, from DIY Audio member lhquam got pretty close in this post. His simulation gave the same specs as the real thing, so several DIYers began constructing amps based on this topology. I came along much later, in 2020, and decided to try my luck at emulating the circuit.

This is lhquam's LTspice simulated circuit which he posted several years ago. I based my implementation on this topology, with some minor changes. He references the output "lateral" Alfet MOSFETs used by Pass, but those are no longer available. I used the Exicon lateral devices, the ECX10N20 & ECX10P20 which have similar characteristics.

There is general agreement among the DIY community that we will not publish detailed schematics and actual parts values of working implementations until Pass publishes his.

Stay tuned for the precise schematic of my implementation. I do not have direct knowledge of the actual schematic of the real First Watt F7, but I believe mine is pretty darn close.

Counter-intuitively (and unlike many complementary input stage designs) I selected NOS Toshibas on the input to be deliberately mismatched, with the 2SJ74s about 3ma lower Idss than the 2SK170s (see schematic below). The reason for mismatching is to favor the P-channel MOSFET on the upper (positive) rail with additional signal swing, due to P-channel devices having a higher Vgs threshold than N-channel devices. Since the bias pots affect both bias and signal levels to the gates, this is necessary. Otherwise, the signal coming out of the upper MOSFET would not be symmetrical with the lower one, resulting in higher distortion. This is one example of the type of "secret sauce" that Pass uses when tweaking his amps. This gives me the lowest distortion I could get using Exicon LATs in the output. I experimented with various adjustments and ended up favoring 3rd harmonic about 6 db down from the second, which is negative phase. Overall distortion specs are very close to published figures for the real thing.

I have a couple of enhancements planned. One is to use a pair of LT4320 "perfect" MOSFET rectifiers in place of the conventional power supply bridges to get another 1-2 Volts on the rails. I think the Exicons will be happy with the slightly higher Vds, putting them in a slightly more linear portion of the transconductance curves. Also, I may swap in a pair of more closely matched double-die Exicons and see if the higher transconductance is beneficial (when I originally ordered the Exicons, their distributor, Profusion, was not offering "binned" devices. Now they do, so theoretically you could request N- and P-channel devices from bins which would reduce the Vgs mismatch from the typical values. This would be better than buying lots of devices (expensive!) and trying to sort for the lowest Vgs mismatch.

This was my initial breadboard which I used to verify component values. I used old-school Tektronix silver-ceramic soldering strips (just for fun) and a small heatsink, which limited how much idle current I could test with, but for short periods, it was OK without getting too hot. The base was just a piece of blank PC board material and the strips and potentiometers were glued with superglue.

Note the simplicity of the circuit. Only 4 active devices! There are 4 potentiometers for adjusting bias, distortion, and positive feedback.

The breadboard worked on the very first attempt, powered by a bench supply running +/- 24 volts.

Here is a 3D rendering of the custom PCB I designed - one channel shown. The power MOSFETs are "underslung" on the bottom of the board and bent back so they can be mounted flat against the heatsinks. Faston PCB tabs are used for all the high-power connections. There is also a test point in the upper right used to measure the voltage drop across a current sense resistor so that bias current can be easily adjusted. The large caps provide decoupling for the supply rails.

IF7 Power Supply

The power supply is also very simple and typical of many Class A designs. Inrush current is limited by a pair of thermistors that have a high resistance when cold. This decreases a few seconds after power is applied. Voltage is stepped down to about 20 Volts AC by a dual winding transformer, which feeds two high-current rectifier bridges. The bridges provide about 24 Volts unfiltered (bumpy) DC to the filter capacitors. The resistors separating the pairs of capacitors provide some additional filtering of the DC. This is referred to as a “CRC” filter arrangement. The filters reduce the output ripple to about 6mV RMS at full load, thanks to the large amount of capacitance.

Note the “star” ground. The star ground is isolated or “lifted” from the chassis and AC line ground. This is very important to reduce hum caused by ground loops. The star ground is the reference for both channels’ power supply rails as well as their inputs and outputs. A third thermistor connects the star ground to the chassis and AC line ground for safety purposes. Should a fault occur, such as a transformer short, the thermistor will conduct heavily and blow the line fuse.

The amp should only be connected to a 3-prong grounding AC outlet for safety.

I used a 3-U "Dissipante" chassis ordered from the DIY Audio Store to build the amp. These are very high-quality chassis made by Modushop in Italy.

Here are the power supply components laid out on the baseplate. Note the single toroidal transformer. This is a 300 VA, dual 20 Volt secondary unit. The filter boards have 4 x 33,000 uFD capacitors for filtering for each power rail. The transformer is flanked by 2 bridge rectifiers. The power supply filter boards are from the DIY Audio Store; these are the Universal PSU V3 boards. The filter boards each hold 4 x 33,000 uF 35 Volt electrolytics. Two bridge rectifiers are connected between the secondary windings of the transformer and the filter boards. One board supplies +24 Volts and the other -24 Volts. The ripple current into the filters is about 4 Amps RMS when the rails are loaded to about 3 Amps DC. There is very little ripple on the DC rails (several millivolts) due to the hefty capacitors. The bridge rectifiers are over-rated at 35 Amps, but get warm since they dissipate about 5 Watts each. The baseplate keeps them cooled. I may replace them with LT4320 "perfect rectifiers" at some point - this will give me another ~2 Volts on the rails.

Here is a closeup of the AC surge suppressor thermistors and the safety cap which is connected across Line and Neutral.

Here is a closeup on the star ground and the ground lift thermistor connecting it to the chassis.

Interior of the completed amplifier. The amplifier boards are mounted on the heatsinks on the left and right. The MOSFETs are bolted to the heatsinks and insulated with Keratherm insulators and fender washers. All the wiring (except the input leads) is 14 AWG silicone insulated. The back panel has a power switch, fuse holder, isolated RCA input jacks, and Speakon connectors from Neutrik.

IF7 Adjustment (click to expand)

The amp’s adjustment can be tricky, and should only be attempted with decent digital multimeters (DMMs). The distortion adjustment is optional, since it requires a way to measure harmonic components. This can actually be done with a USB A/D converter and some computer software, but won’t be covered here.

Initial Setup - Power Off

Set the bias pots (P1 and P2 in the schematic above) so that the wipers are all the way at the supply rails - this ensures the MOSFETs are in a fully non-conducting state initially.
Set the JFET (P3 in the schematic above) pot to its approximate center.
Set the positive feedback pot to maximum resistance.
Short the input.
Connect a dummy load to the output - typically 8 Ohms.
Connect a DMM across the .1 Ohm resistor to measure bias current: 100mV = 1 Amp.
Connect a second DMM between the output and ground to monitor output offset.

Bias Adjustment

Apply power.
Check the voltage drop across the bias pots. It should indicate the JFETs are operating within their Ids specs - typically between 5 to 10 mA.
Gradually adjust the bias pots until the MOSFETs begin to conduct, alternating between the two. Keep the offset at a minimal level.
Continue to adjust the two pots until the desired bias current is reached - typically 1.2 to 1.5 Amps and the output offset is zero. Allow the heatsinks to reach thermal equilibrium and then re-check bias and offset.

Distortion Adjustment

Connect distortion analyzer and 1 KHz signal source.
Adjust input level for about 1 Watt output: if using 8 Ohm dummy load, this will be 2.8 Volts RMS.
Observe the output FFT, looking at the relationship of the second harmonic to the third.
Adjust the JFET pot so that the second harmonic is about 10 dB above the third.
This will affect the output offset, so compensate with small adjustments to one or both of the 500 Ohm pots.
THD at 1 Watt should be .05% or less.

Positive Feedback Adjustment

Apply a test tone to the input, typically 100-400 Hz.
Adjust input level to produce a modest amount of output power, 5-10 Watts.
Measure the RMS output level with a dummy load connected then disconnected.
Continue to connect then disconnect the dummy load while gradually reducing the resistance of the positive feedback pot, monitoring the RMS output.
When there is no difference in the output level between load and no load, the amp is effectively exhibiting an infinite damping factor. This is impractical and may result in instability, so back off slightly to produce a more realistic DF of about 150 (or whatever you choose).
I use this formula: DF = Vnoload/(Vnoload-Vload)
The above experiment can be repeated with actual speaker loads and interconnects, effectively “tuning” the amp to a specific load.

Page updated

Google Sites

Report abuse