f5 turbo v1
Nelson Pass 2/5/12
Intro
Five years ago I launched the F5 power amplifier project which culminated in
a production run of 100 amplifiers and a DIY article. It has proven to be a
popular design, and as of this writing approaches 2 million views on
DIYAudio.com.
The F5 is a nice simple design, sounds good and measures well. There are
other such offerings in the world, but this one seems to have struck a chord in
the mind of the DIY enthusiast as well as
paying customers. Of course it has
limited output power at 25 watts, so it is not surprising that there have been
numerous requests, complaints and even some independent effort for
M
ORE POWER. You want more? I give you more.
But First, the Original
If you are not already versed in the design and construction of an F5, I refer
you to the original article published in AudioXpress magazine and also to be
found at
www.firstwatt.com (download the F5 manual from the Products
page).
In addition to the details of the design, the article contains lots of tutorial
information about how gain devices work. If you have any intentions of
playing with an F5 Turbo, you will want to read this material, otherwise when
you email me a question about the stuff covered there you will receive an
RTMF
sort of reply. I should add that the nearly 12,000 postings in the F5
thread also cover some of this material, but it's pretty tedious reading them
all.
Assuming that you are already mostly familiar, here is the official DIY F5
schematic. It differs mostly from the original by the inclusion of P3, which
aids in tweaking the symmetry of the circuit for the lowest possible distortion.
A number of modifications to this have been proposed and even executed by
DIYers, including the removal of the current limiting protection circuit,
paralleling more and different output devices, balanced operation, cascoding
and of course bigger power supplies.
The documentation on the performance of these modifications is
unfortunately sparse, and I know that many DIYers won't tackle a project
without some “official assurance” that it will work. I played with most of these
things during the original development work, and I applaud the people who
have done variations independently – please don't complain that I have
copied someone else's achievement.
The F
5 TURBO V1
Increasing the power supply voltage is the obvious way to get more power out
of an F5. You can simply raise the supply rails to +/-32 Volts and get 50
watts into 8 ohms right away without other modification. 24 V AC secondaries
on the power transformer will do it. Don't forget to use higher voltage power
supply capacitors. Probably you should also upgrade R9 through R12 to 5
watt resistors. Depending on your heat sinking, you will probably want to
adjust the bias so that the power transistors don't run too hot.
As a rule of thumb, the output devices should not be operated at more than
about half their maximum rating, and generally the case temperature needs to
be under 100 deg C. For most amplifiers this means a heat sink temperature
of about 50 to 55 deg C., which is the temperature that you can put your hand
on for about 10 seconds.
The devices are biased at about 1.3 amps in the original circuit for a
dissipation of about 30 watts per device. We are not in a good position to ask
the original heat sinks to dissipate more, so if you increase the supply to 32
volts, you may need to operate at a lower bias point – I used about 1 amp
(0.47 volts across R7 and R8). This will still get you a decent distortion spec
to 50 watts. If you have difficulty keeping the bias stable, you can increase
the values of R7 and R8 to 0.68 ohms without significant penalty.
You can remove the current limiting (Q5, Q6, R15 to R18 of the original) as
long as you are cautious about shorting the output. I don't recommend
dropping the thermistors – you are going to need them for thermal stability.
Want to drive 4 ohms with this? You can do it, but I recommend that you
consider the upgrade to the F5 Turbo V1:
Here you see the F5 without the limiters and with an additional set of output
devices. With adequate heat sinking, you can bias them to 1 amp each, for a
total of 2 amps, which will operate the circuit Class A well above 100 watts
peak into 8 ohms. You probably take the bias to 3 amps with enough sink.
The more technically astute DIYer may note that the input Jfets are now
being exposed to greater voltage and dissipation, and this might be a
concern. With respect to voltage, the operating point of these devices is
around 30 volts, 5 volts over their rating. In actual testing these devices
break down around 43 volts, and I depend on Toshiba's famed conservatism
to carry the day. If you don't want to trust Toshiba as much as I do, then you
can cascode these devices, which is covered here later.
The average dissipation of the Jfets with a 32V supply will be about 28 volts
times the operating current. A 2SK170 or 2SJ74 with a 10 mA Idss will
operate at about 8 mA in this circuit, which gives a dissipation of about 220
mW. A quick calculation shows that its maximum junction temperature at 220
mW is reached with an ambient temperature of about 70 deg C. A wise DIYer
will either select a Jfet with a lesser Idss (say 8 mA) and/or see to it that the
Jfets gets some cool air or a little heat sink. Or you can cascode them.
You may also notice that the feedback resistors R7 through R10 have been
increased, increasing the amplifier's gain to about 22 dB and decreasing the
amount of feedback by about 7 dB. If we are going to put out more power it is
appropriate to have some more gain, and it gives us more margin for
feedback stability – the amplifier still is flat to about 800 Khz. Of course we
have noted that the 100 ohm resistors will start to cook at the higher power
levels, so we would be making some changes to them anyway.
Along the same lines you will also notice is that I have doubled up the Source
resistors to a parallel pair of 1 ohm values in order to increase the dissipation
from 3 watts to 6 watts.
I built a few V1 circuits into a standard First Watt stereo F5 chassis, and you
can see it here:
And here is the inside:
No, they are not for sale. These were biased low at 1 amp per channel and
gave the following distortion curve into 8 ohms:
Less feedback and lower bias means a little more distortion, and this does
about 0.005% at 1 watt, where the original did about 0.002%.
You can just see the slight hump in the curve at 0.05% and 20 watts where it
transitions out of Class A. This is a common artifact – you can go through
Stereophile's measurements of amplifiers and see examples where a high
bias Class AB amplifier does a similar transition.
Into 4 ohms the hump becomes more pronounced and the amplifier is seen to
clip at 100 watts output. If you have a more heat sinking available (a mono
version of the F5 chassis would do) you can do better yet with higher bias, in
which case the hump goes away.
At 1 amp bias, these sinks only run warm, but if you are after the highest
performance, you want to run as much bias current as you can.
Here is a photo of V1 doing a clean 1 Khz sine wave into 0.1 ohms showing
peaks of +/-23 amps. This amplifier will poop out slightly above 1 ohm.
I mentioned that it pretty much retains the bandwidth of the original in spite of
a doubled up output stage. Here is a 100 Khz square wave at 1 V rms.
Subjectively this is a little different than the original. The lesser amount of
feedback and larger output stage makes it seem a little more authoritative
while at the same time a little more relaxed. I say that because the listener
(me) experiences more relaxation.
One thing that many will appreciate is the addition of P3, which due to
inattention on my part was not public until recently. It allows the adjustment
of the distortion character between complete nulling of the second harmonic
distortion to an arbitrary ratio of second to third harmonic. It definitely alters
the sound, and reaffirms that things going on below 0.1% are audible. In any
case, you can adjust it to taste.
Here is what the waveform looks like when the second harmonic is allowed to
be about 0.03%: