Tube preamp with current mirror load and no attenuator in series with signal path.

The picture above is the version described in figure4. The transformers are in a separate small box, and a 2meter D9 computer cable connects the two together. Again simply since that’s what I had laying around at the time. I’d prefer a slighly shorter and larger guage type of cable.

Here’s a simple preamp with resistor loading. I’ve had this as my PC-system preamp for two years now and feel it sounds very well for it’s simplicity. The entire preamp consists of one 6922 tube, one volume pot, three resistors and one capacitor per channel. Not including the powersupply, which of course is important soundwise and absolutely part of the circuit. In this case I use tube rectifiers, but soft recovery diodes work well too, LC filtered. I’ve started favouring placing the volume pot at circuit output. Keep interconnects reasonably short and listen for ya self. A farely typical circuit.

Figure 1:

Tubes are great, but I refuse to enter a blind alley and swear by them in a way that limits my options. If done right, silicon devices can sound great too, and combining them opens the way for innovative designs that use the best of both worlds. I came up with the idea (fig2) while designing a phono section of using current mirrors and CCS to enable me to use lower voltages while keeping the plate current up even with high load resistors. I prefer to run tubes at a farely high plate current, but with high value load resistors, the supply voltage has to be much too high for my tastes.

The mirror holds the plate voltage almost constant and isolates the tube from the load. The latter allows more options for the load resistor, and I found no difference in distortions wether I use a 10kohm or 10ohm load resistor (R4). Since there’s very little voltage swing at the plate the circuit is very fast, only slight drop, less than 0.5dB, at 200kHz wich is as high as my function generator goes. PS: The circuit is now NON-INVERTING. The sound of the fig2 circuit blew me away, but don’t just sit there and think of me as an arrogant SOB, try it yourself!

Since the idea for the mirror load was to enable lower voltages, I replaced the transformer to find out if it still performed. With the 40volt supply as shown in fig2 the idle went down to about 4mA. It’s also quieter than the fig1 circuit, probably because R4 is referenced to ground and not the supply line. I’ve read many places that many tubes run very happy at low voltages, and the 6922 supposedly sounds best at 40-60volts. The noise of the tube supposedly also drops. Anyhow: the preamp sounds great with a 40volt supply. The gain increased to 7 (from –5.5 in fig1) even though the plate voltage and plate current is less, which is because of no negative feedback effect of the inverted voltage swing at the plates that you get with the first circuit. More ideal transconductance in other words

Figure 2:

Since the current mirror circuit can be loaded all the way to zero ohms without increase in distortion I came up with the idea of making the load variable, and thus eliminate the volume pot in the signal path. Actually it’s still in the signal path, but not in series with the signal, and if the mirror mod blew me away, after the next mod I was blasted all the way to Pluto…not sure I’m back on the same planet I left yet.

Figure 3:

At first I simply put the pot parallel with R4 in figure2. This is okey if the volume pot were a stepped attenuator, but being a carbon type, the current passing thru the viper made a scratching noise when I moved it that almost blew my woofers out of their baskets and my wax out of my ears. Since I like my 2dollar carbon pot and Q-tips are a little more comfortable I had to do something about that.

So in came a powersupply that sports a negative rail. A constant current source that matches the current of the mirror ensures that the voltage over R4(and pot) is zero. Instead of a simple resistor at Q1’s collector to set the current, I use a JFET constant current source, which vastly improves PSNR. It’s borderline using a JFET at this voltage, which is why D1 is there, but I like the simple CCS it makes. If it weren’t for the slight voltage drift C1 could be removed and it would be a very nice DC-500kHz circuit. (The output offset drift is a few hundred mV at max volume, so if poweramp has input cap, it may be removed, no lethal voltages will be at output.)

Another bonus I didn’t think about with having the attenuator like this is that the problem with varying capacitance at the viper depending on it’s position no longer exists. That means it’s bandwith is constant no matter where the volume is set. Try passing squarewaves thru a typical attenuator setup, stepped or carbon or whatever, and the more the viper moves from the top, the more rounded the square waves become.

The sound is awesome. To my surprise it seems the bass benefitted the most, although the entire range improved. It’s as if I got another octave downwards out of my PC-system. With the speed, or high bandwidth of the new preamp, I was expecting more details in the upper frequencies, not at the lower as well. Of course I’ve tried it in my main listening system, but this is my PC preamp. I’m in the process of modding my main preamp, but time is not abundant.

PS.BEWARE.WARNING.ACHTUNG. Since the voltage rails don’t always come up at the same moment, the circuit will have a rather huge turn-on hump. Therefore either turn it on before amp, or hold over your ears when turning it on. At least in my case I needed to change my pants the first time I plugged it in…I simply handle it with care now, but the next time I build one it’ll include a delayed mute relay at the output.

Never mind the above scare! The reason for the mentioned turn-on thud is that I used the same tube rectifier for the positive voltage that I had for the preamp in fig1, and simply threw in a solidstate rectifier for the negative voltage. The tube comes up slowly, while the SS rectifiers get on line like Marines fresh out of bootcamp. That gave me a rather unsymmetrical voltage rails upon turn-on.

After a few weeks…

I really liked the results of the figure3 circuit, and decided to make a less hasty put together preamp and even a PCB layout for it. I also tweaked it a little with a current sink at the tube’s cathode. I do not need the high gain that figure 2 and 3 produced. So I used some 1kohm Caddocks I had laying around for load. Now the gain is a more usable 3x. This is still a bit high for me, and I think 500ohms would be perfect, but I only have cheap metalfilms at that value at the moment. The load resistor is the most important soundwise, and the Caddocks are very good. (R1, R3, and R9 in figure4 are important if you want to spend money on exotics). I removed the J-FET for biasing the current sinks, and simply improved the powersupply filtering (larger capacitance). It really doesn’t matter, just bias the current sinks anyway you like so it runs at the same current as the tube. Q2 is the current sink connected to the cathode, and is biased so that the DC-voltage over R3 is zero. This increases the idle current a little and the dynamic range as well as drive at the output. Just another way to use transistors to help tubes run happy at low voltages. I also increased the emitter resistor values a bit to help match all the currents.

Figure 4:

The tube I used is still the 6922 or 7308, but the schematic CAD software doesn’t have those in it’s library so the 6SN7 is shown in the figure. There’s no reason not to use any other tube with this type of configuration, but not all tubes are as happy at low voltages as the ones I used here. The transistors are BC640s and 639s. I’ve used IRF9510 MOSFETs for the mirror with equally good results.

Note: (10march,2002) Regular volume pots do not have zero value at minimum settings. The alps I use has a resistance of 5-10ohms at minimum, therefore the preamp does not go completely quiet. It’s still only as loud as a whisper. If total muting is required, either use a stepped attenuator, or have a mute relay at the output. Another thing to be aware of is that no tubes are identical, so when tube rolling, the plate current and thus offset will vary. When an offset exists, there will be a scratching noise when moving the viper on the volume pot. No big deal, but just be aware of it.

I also recommend using a slightly larger resistance for R2 than for R11 and R12, so as to let there be a slight positive voltage over R3.

The powersupply.

I had a 50VA transformer with 2x28volt secondaries laying around, and that gave me +/-39volts with solidstate rectifiers, or about +/-35volts with tube rectifiers. The voltage isn’t critical, anywhere from +/-25volts to over a hundred is fine, so don’t worry. I selected +/-40 so I can use more common transistors like the BC640 and BC639. Just be aware that you must adjust the resistance (P1 and R13, figure4) that sets the sink current according to supply voltage. I use a hybrid tube/SS rectifier as shown in figure5. The diodes enable tube rectifiers to be used in a fullwave config without a centertap transformer. For such low voltages I permitted tortureously low limiting resistors and even with 5600µF capacitors nothing has blown up yet. I’ve checked the voltagedrops over the 10ohm resistors, and there isn’t much of a charging current to worry about. The datasheets for the EZ80 rectifier tube demands minimum limiting resistors of 100ohms or so, but I think that applies for much higher voltages, since it’s not nice to apply high voltages to cold tubes. I think with the low voltages here the tubes are okey without them. The nice touch with tube rectifiers is that they do not conduct untill the filament has heated the cathodes properly, so the voltage increases slowly from turn-on. No nasty turn-on thumps. Rather fitting with a hybrid powersupply for a hybrid preamp…

Figure 5:

Figure6 shows the typical powersupply with solidstate rectifiers. Use soft recovery diodes for best sound. The filter caps may of course be much larger in both size and quantity, whatever suits you. Allthough the preamp only draws around 50-60mA I recommend using a good sized transformer around 100VA or more. Remember, the powersupply is in the signal path! As always, use star grounding, with the origin at the powersupply.

Figure 6:

The PCB:

For some reason I am not able to paste a copy of the circuit layout here. I’ve tried to get a good black and white scan, but this didn’t turn out so good. Here at least you gat an idea of one layout solution. I had to estimate the hole positions of the tube’s pins. They match well the tube’s pins, but not a pcb-socket. Hopefully I’ll get a proper pcb trace here soon..

This shows the outlines of each component. Not very easy to see whats what, but perhaps it’ll be helpful in getting your own layout.

Here’s a photo of the PCB and tube socket. The slim cabinet I had doesn’t allow for the PCB to be just below the tube socket, os it’s done like this. The square white and black things are the caddock resistors, mp930 types, I use for R9.

Various.

I really recommend the mirror load and attenuator setup as in fig3. It’s simple enuff that if it doesn’t appeal to you, no real effort was wasted.

For them’s that care, here’s a shot of 20kHz squarewaves with the volume set at my typical listening level, they used to look like sawblades.

Here’s a picture of the underside:

Circuits drawn with Circuitmaker, free download from: www.circuitmaker.com