At a first glance there's nothing difficult in building a unity-gain preamp: the well known and fairly simple source follower should do the job. However in my case it turned out to be quite an adventure. Why "HotFET" - please refer to this article.
|HotFET-Pre (c) - prototyped in a "networking" case|
The design is pretty simplistic. There were many calculations, tests and measurements on the way. But I could not recall every bump on that road thus will try to merely describe the key conclusions here.
Let's start with classics: the source follower as it is, decorated with the input capacitor blocking DC and an attenuator playing the volume control here.
|Simple unity-gain preamp: source follower|
There were issues with this creation:
- This design will exhibit quite high and thermally unstable DC offset on its output. Therefore it cries for yet another capacitor being added directly to the signal path.
- Distortions added by this follower are not negligible.
- Output impedance is also high, unless we deploy a power J-FET with high idle current. Shall you know the supplier selling such power devices at reasonable prices - please let me know ASAP!
Replacing the resistor in the J-FET's source brings us close to what Nelson Pass has made famous: "The First Watt B1 Buffer Preamp". Provided the transistors were closely matched - there will be negligible DC offset at the output and, what is even more interesting for us: much lower level of distortions. This schematic differs from "The First Watt B1" in a way it gets fed by the dual-power supply. I would leave this subtle difference at the designer's discretion. HotFET Pre can be used in different configurations.
|Source follower with a symmetrical current source loading it|
The reasoning behind biasing of HotFET Pre has been discussed in lengths already. Thus, let's add the dreaded MOSFETs in cascode. For the sake of simplicity the biasing voltage sourcesfor MOSFETs were depicted as single symbols. Perhaps I should have drawn batteries there instead...
The complete schematic of preamp (one channel):
|HotFET Pre (c) - schematic diagram|
- R1 - 50 KOhm
- C1 - 3 uF
- C2, C3 - 0.1 uF
- C4, C5 - 220 uF 10 V
- R2, R3 - 91 KOhm
- R4 - 1 MOhm
- R5...R8 - 604 Ohm
- R9, R10 - 22 Ohm 0.1%
- VD1...VD6 - green LEDs 1.7 V 20 mA
- VT1, VT3 - IRF610 (matched by Vgs(th) @ 30 mA)
- VT2, VT4 - J310 (closely matched by Idss)
I wanted to get low-pass -3db as low as I could. Well, 1/(R1*C1) ~= 7Hz - is not so bad already. There is an opinion that we should strive to get as low as 0.5Hz. May be next time...
Most parts were selected pretty arbitrary. Wherever parts precision or matching matters - it's specified explicitly.
There is one important addition if you compare this final schematic with the skeleton above: here we added resistances in J-FET sources (R9 & R10). These resistors increase the depth of local "degenerative" feedback. Despite my probably very harsh accent even in written English, this time I said exactly what I meant: they increase the depth of the negative feedback that was always there due to intrinsic resistance and inductance of real transistors. Nevertheless we do want this extra feedback because it improves the overall thermal stability and allows us selecting the biasing point à volonté. As a drawback this increased feedback plays to increase the output impedance a bit - let's live with that. Last but not least we have to decouple the followers from the load just in case there were too esoteric interconnects plugged into the HotFET Pre' output jacks and it still has to deal with heavy capacitive loads without oscillating.The old good green LEDs are apparently my preferred voltage reference devices for such low-current designs. They might not provide high precision nor extreme temperature stability, but that's really Ok in this particular case.Initially I was tempted to use modern 3-volts blue or super-bright green LEDs. Unfortunately these often behaved in a strange way: the voltage across a diode was suddenly dropping as if there was something braking and short-circuiting the LED, while at nominal currents all LEDs did shine in a stable fashion. Should anyone knew an explanation of such an effect - I would be grateful to get educated!Gate resistors are necessary in order to avoid parasitic oscillations. The devices we use are capable of very high frequency oscillations, so high that many scopes will simply not show anything. But the measurements and the whole functionality of the circuit may be screwed, should one omit these gate resistors.Virtually any Power Supply can be used, provided it is capable to deliver bi-polar voltages around 15..20 Volts with loads up to 30mA per channel. The amplifier has very good Power Supply [noise] Rejection Ratio. Also it's a very good behaving load to the power supply as it actually exhibits the current sink/source type of load.
The usage of huge banks of filtering capacitors of tens of thousands of mkF seems to be in fashion these days. Here we go out of fashion and use moderate capacitances. This allows much lower inrush currents and longer (therefore - lower) charging current peaks - the latter literally means less noise energy spread in the box. The relatively high ripple we filter then with a simple RC filter.
The proposed PS schematic was tuned for HotFET Pre and should not be copied blindly to other loads. Still, I do hope that main ideas may prove valuable for other reader's projects.
|HotFET Pre (c) - the Power Supply|
Requirements for the mains transformer are very relaxed here. I used some small one with two 12 Volts secondary. The transformer must be capable to deliver 120 mA according to its manufacturer spec. In my test implementation even the smallest transformer did not fit into the "networking" box I've chosen for this preamp. Thus I used RS765 (GX-12) type connector for low voltage AC power and put the transformer in a separate box.
In fact any rectifier diodes could be used here, but Schottky types (10BQ060 from IR in my case) yield lower noise due to their virtually immediate turn-off.
The input tri-wired choke is also optional. I used some 1/2" ferrite ring (scavenged from a PC power supply) with 15 turns of 0.21mm wire, made holding 3 wires together.
Possible variations. Utilizing Mr. White's cathode follower topology would be an option here. But that variation, while providing mightier load capabilities, would require a dead-stable power supply effectively nullifying high PSRR advantage we get with the simple layout.
Bench results: HotFET Pre Q&A; .
How does that sound? Sorry, but even if you managed to withstand my writing till here - please do not expect me putting up some glossy-magazine words about sound brilliance and such.
Honestly, that sounds as good as a simple but yet well matched preamp can do. It's balanced, gives what I would call a proper arrangement to the scene. After all I caught myself recognizing new subtle things that I never noticed before on the discs that I enjoyed listening in the headphones many times... But then I'd need to tell what I am comparing it to etc. You are welcome to pass by and listen yourself. Or build it and enjoy listening to your version 🙂
In place of a conclusion: I am convinced that a correctly designed and carefully built preamplifier is the must for any high-definition sound reproduction installation. Some may argue that in case of a very high input impedance power amplifier (for ex. a tube amp) a so-called passive preamp will do. I would still disagree: there's a very uneasy load introduced by interconnects. Thus if we want enjoying the music instead of going into the listening to how interconnects from the brand XYZ sound - the signal source must provide low output impedance.
Where to buy? Should you be interested in purchasing a DIY kit for building HotFET Pre, or a pre-built version of it - please drop me a note. I am considering to put up some kits for sale. Your request will definitely help speeding-up the preparations 😉
Update 2012.02.13: Coming soon!
HotFET Pre+ the elite version of this schematic using depletion mode MOSFETs.
- Twice lower the components count.
- Equal or better quality.
- No compromise, the best silicon-based buffer preamplifier for audio one can build or buy for money! (Till proven otherwise 😉
--- Stay tuned 😉
This content is copy-righted. No copies can be made without the written permission from the author. Referencing/linking to this blog or page are welcome. === (с) MyElectrons.com ===