Tube DAC. DAC with tube output. Multibit version of the circuit

For DAC or CD player. At first glance, the circuit may seem somewhat complicated, and some of the functions contained in it (such as a phase shifter) are of no use to a simple music lover.

Today we offer you a version of such an amplifier with just one tube (in each channel), without unnecessary marketing tricks, but, as before, with good characteristics and high sound quality.

When testing for various Focus groups for different types of DACs revealed one common result - the sound quality of a CD player is higher, the better the signal at its output is filtered from HF components. That is, the output filter should have a fairly steep frequency response decline at the edge of the audible range.

If in the 90s analog filters were most often used, recently digital filters have become increasingly popular. It is not surprising, because with the relative simplicity of implementation, they show much higher performance compared to analog filters. Meanwhile, test results showed that listeners prefer CD players with analog filters, since digital players, although they have higher characteristics, use clock signals in their operation, which leads to an increase in the level of HF noise.

Well, if the filter is not only analog, but also made using tubes, then in addition to cleaning it from HF noise, it makes the sound of CDs more pleasant, “warmer”, and eliminates digital harshness of the sound.

The filter circuit is shown in the figure:

Click to enlarge

The amplifier has an absolutely flat frequency response throughout the entire audio range. The roll-off begins at 20 kHz (-0.5 dB) and is -24 dB/oct. This allows you to very well filter out all the high-frequency noise of digital audio reproduction and make the CD sound similar to vinyl.

In addition, the circuit has a low output impedance, which significantly reduces the requirements for interconnect cables.

The power supply diagram is shown in the figure:

Click to enlarge

Here we see the traditional heating of lamps with direct current to reduce the background level of the network.

The high-voltage stabilizer is made according to an electronic choke circuit with current protection.

To improve the separation of stereo channels, each channel is assembled on a separate printed circuit board. The power supply is designed to operate with two channels.

This design can be designed as a separate unit or, if space allows, built into an existing CD player.

To achieve high sound results, the capacitors in the circuit must be of the highest quality. Capacitors C4 and C7, not indicated in the diagram, are places on the printed circuit board in case it is not possible to find capacitors C3 and C6 of the required capacity and you have to connect several in parallel, or to change the cutoff frequency of the filter.

Listening showed a significant improvement in the sound quality of a CD player with such a filter. This is a fairly cheap and easy way to upgrade your device to a higher price category.

Printed circuit board drawings and element layouts can be downloaded

Printed circuit boards in SLayout format (rar archive, 47 kbytes).

The article was prepared based on materials from the magazine "Elector"(Germany)

Happy creativity!

Editor-in-Chief of RadioGazeta.

28 comments on “A simple tube filter for a DAC or CD player”

Kirill:
February 27, 2017
Does it make sense, in your opinion, to “raise” the heat potential?
Chief Editor:
February 27, 2017
Here the filament is powered by constant voltage. So it's redundant.
Please note that the lamp filament circuits are connected in series! Keep this in mind when repeating the design.
If you don’t want to assemble a stabilizer for incandescent circuits, then yes - “raising the incandescence won’t hurt.
Kirill:
February 28, 2017
Permanent - this is understandable. However, the potential difference still remains, since the filament circuit stabilizer sits on the ground. I agree that, in theory, there shouldn’t be a background. However, isn't this harmful to the lamp itself? Unclear...
Chief Editor:
March 1, 2017
This is called “I heard a ringing, but I don’t know where it is” :)
The filament potential is raised when it is powered by alternating current, only so that there is no background. Since the background incandescent voltage (50Hz) successfully penetrates through the heater-cathode section (which is essentially a diode). By raising the potential, we block this diode and block the path for the background.
If the filament is powered by a constant voltage (and even a stabilized one), then there should be no background there, so there is no need to block the diode. We save two resistors. This does not affect the lamp itself.
It is important for the lamp that the MAXIMUM ALLOWABLE heater-cathode voltage is not exceeded. Indicated in reference books. Typically occurs in cathode followers and high-power output stages.
Kirill:
March 1, 2017
Why, I just know where this ringing comes from - from the datasheets. For example, for a 6F5P lamp the maximum permissible cathode-to-heat voltage is only 100 volts. Structurally, this lamp is close to the ECL86/PCL86, so I believe this is also true for it. In the presented scheme, this condition does not seem to be met.
Chief Editor:
March 2, 2017
To me, “doesn’t seem to be followed” and “not being followed” are very different things.
What voltage do you think is present in this circuit at the cathodes of the lamp?
yuriyruss:
March 9, 2017
PCL86 and 6f5p are completely different lamps according to the datasheets. It’s impossible to put them in the forehead. It is necessary to recalculate the entire voltage offset circuit. Later, when I check this filter on 6f5p, I’ll post here the resistor values and voltage on the lamp.
Chief Editor:
March 10, 2017
Chief Editor:
March 10, 2017
By the way, we didn’t write in the article that 6F5P is an analogue of PCL86.
This is confirmed on a lot of other sites.
They definitely differ in the heating circuits.
Sergey Khraban:
July 18, 2017
Hello! Please tell me what type of zener diodes D1-D3 and D4 are in the power supply?
Chief Editor:
July 19, 2017
D4 - BZX55C18 (or analogue), KS218Zh, KS508G, 1N4746A
D1-D3 - NTE5157A, 1N3045 and similar.
Sergey Khraban:
July 19, 2017
Thanks a lot! All the best!
kagantsov:
October 5, 2017
There are errors in the 12V PP power supply: 2200 the electrolyte must be turned over, otherwise it has + on GND (it makes a noise). + and the AC needs to be re-wired, it turns out that the AC goes to the microcircuit and + goes to the variable input. It turns out to be nonsense on the print, it goes 100%. Correct or warn that there are errors. I'll also look at other PPs later. I want to assemble this device. If I collect it and everything is ok, I’ll share my PP. Thank you.
kagantsov:
October 5, 2017
The situation is the same in a 330V PP power supply with a diode bridge.
Chief Editor:
October 5, 2017
Printed circuit boards in pdf format from the original source.
Boards in SLayout format from Mars Corporation.
Neither one nor the other has been verified by the editors.
Thanks for the info!
In any case, care and caution when repeating any design will not hurt.
Mars_Atlant:
October 5, 2017
Good evening.
Thank you for the notes, I corrected the silk-screen printing and sent it for archive updates.
Chief Editor:
October 5, 2017
The archive with boards in SLayout format has been updated!
kagantsov:
October 5, 2017
So everything fits together, but the diode bridges have AC in the middle. Sorry for being stubborn, but if you do it, it’s good.
Chief Editor:
October 6, 2017
Perseverance is welcome! We are both in favor of working completed structures.
By the way, I would like to hear your opinion, impression, etc. later. about the scheme...
kagantsov:
October 6, 2017
OK. But it won't be soon. Time is short and projects are moving slowly. I finished the acoustics 2 weeks ago, it took 2.5 years to complete. Well, maybe it will be more fun with a filter.)
Mars_Atlant:
October 6, 2017
Good morning.
In part, you are all right about the diode bridge, but this type of diode bridge also exists with a different pinout, the order of the legs. You can check it out for yourself online.
I “made”, or rather sketched, the PP according to the PP photos presented on this resource.
Everything corresponds to the original materials, so as not to cause confusion if questions arise about this design.
You can also post your versions of the software on the forum. Perhaps this will make it easier for someone to assemble this structure.
Good sound everyone.
camper:
October 11, 2017
I used lamps 6n2 and 6p43, the sound pleased me, although I don’t know what the 86 lamp sounds like, I didn’t find it, (maybe the 6n1 sound seemed a little harsher) I lowered the supply voltage to 250V
Chief Editor:
October 11, 2017
PCL86 is very similar to our 6F3P (and this one is like dirt) and to ecl82.
You just need to be more careful with incandescence - for the listed lamps it is 6.3V!
kagantsov:
October 13, 2017
Good afternoon. I just have 2 PCL86 lamps, and the difference between the 6F3P is only in the filament? The power supply remains the same - 330V?
Chief Editor:
October 13, 2017
Does religion not allow you to look in the reference book?
It clearly states: the maximum voltage at the anode for the triode is 250V, for the pentode 275V.
We draw conclusions based on the information received.
Grey:
August 14, 2018
Hello! I want to use the circuit of this high-voltage stabilizer to power the Morgan Jones ULF at 220 V. There are a lot of similar schemes on websites; I’ve almost figured out the method for calculating the elements. But they lack R2. There is a similar circuit in “Modern Hybrid Amplifier” dated 08/02/2014, but there the R2 rating is completely different. Please tell me the purpose of R2 and how to calculate it for a 220 volt circuit.
Chief Editor:
August 14, 2018
R2 here is a type of small decoupler (filter).
The denomination is not very important.
For significant current consumption, it is better to remove it completely so as not to reduce efficiency.
And so you can leave it at 100 Ohm.
Grey:
August 15, 2018
Thanks a lot! All the best!

Add a comment

Spammers, don't waste your time - all comments are moderated!!!
All comments are moderated!

You must leave a comment.

The main thing in our business is to take the right start! I don’t have to worry about building a product line from cheap consumer goods to the very high-end. Therefore, I can afford to immediately choose the digital-to-analog converter chip I like and build a design around it. So, the “mystical DAC” was taken as the basis " as it is called on the Internet. I will not make a big secret out of a small microcircuit, but let’s still keep the intrigue for a start.

Build a good DAC I’ve been planning for my beloved since the last century, but somehow I didn’t get around to it and more priority tasks took over. And here, to my delight, a customer appeared, on the one hand, able to appreciate good sound, on the other hand, willing to put up with a certain level of “homemade” in the finished device. Naturally, I will make every effort to ensure that my clients are satisfied with their choice. What my “pre-production” products lose in comparison with serial devices of popular brands is:

some of the editing is done with cobwebs on mole rats, and not on print, which has a positive effect on the sound quality, but, alas, will not be available in production samples;
I don’t skimp on little things like a surge protector or shunt capacitors, which, by the way, has been caught by recognized authorities more than once;
My “brand” is not yet very widely known in narrow circles :)

Let's start, pay attention...

Where to begin? That's right, it's best to start with a ready-made device, even a simple one, but containing key components. In China for US $ 50 I purchased a generally good kit for self-assembly of a DAC. As I already said, the Chinese economic genius is not distinguished by any special technical talents, so everything in that set was at a minimum, exactly according to the datasheets. Except that the creators of the set built, as it seemed to them, very high quality: they stuck " KRENOK" with garlands. But the kits came with very appropriate R-core transformers.

At this stage, the task was not to somehow specifically control the digital receiver or DAC, so the hard-wired minimalist S/PDIF->I2S->DAC chain suited me quite well.

I didn't consciously try to find a DAC with a USB input. The reason is simple: the computer generates a lot of noise and there is no desire to let all this garbage into the audio device. Of course, there are methods, but I still haven’t come across a single DAC with proper decoupling of the USB input (devices for 1K green and higher, as well as products from Russian “left-handed” audio don’t count).

I consider it necessary to note that despite all my quibbles about the circuit design, etc., the quality of the printed circuit board is simply excellent!

Taking control of the situation into our own hands

In the documentation for the DAC, in one place it is written that the analog power leg must be bypassed with an electrolyte of 10 μF and ceramics of 0.1 μF. In the diagram, leg 18 is bypassed exactly like this.

A little further in the same document it is said that it is advisable to bypass the input on pin 17 with an electrolyte of 10 μF and ceramics of 0.1 μF. The developer acted in full compliance, dutiful comrade, just great!

Another place in the documentation says that 17 leg Can run it straight to analog power. This is what we see in the diagram :)

What’s funny is that not only in the circuit, but also on the printed circuit board, everything is laid out like this: with two electrolytes and two 0.1 µF capacitors, with a short one right between the 17th and 18th legs of the chip (the path to the capacitors from the 17th leg goes under the chip body) :

Everything came just this dirty from the factory. How I washed it is a different story :)

For those who are especially curious: the pitch of the legs of the microcircuit body is 0.65mm.

I once came across a gorgeous picture from my friend Vadich-Borisych on VKontakte: " resistance is futile". Here, it inspired me, it is as useless here as the duplicated shunt capacitors in the diagram above, I redrawn the “circuit” especially for you:

I needed to control what was happening on the 17th leg. I had to cut him alive. It’s good that they haven’t put a jumper under the chip yet - the prospect of unsoldering one leg of the SSOP case is somehow not encouraging.

Mediocrity goes overboard

What digital-to-analog converter is complete without operational amplifiers?

That's right, only high quality DAC. So I simply did not solder the modest filter on the NE5532. Maybe it was worth it to have something to listen to for comparison and make sure how unconvincingly deep loop-backed op-amps play... But I already have a CD player from a venerable manufacturer, which very diligently plays the very mediocre sound of op-amps, although hidden behind the sonorous name HDAM and soldered into small screens. And there are plenty of other similar “samples”.

Study, study, and... think!

Perhaps, without exception, on all DACs from manufacturers from the “heavenly Empire” I see the same locomotives from “KRENOK” (the photo on the right is not mine, caught on the Internet). By fanning out serial voltage stabilizers, the developers are obviously trying to achieve better power supply isolation and reduce the penetration of interference from the digital part to the analog part. Unfortunately, the masses lack what I call “current thinking” in circuit design. In fact, everything is simple and... a little sad.

Look at some LM317 from the output side. You will probably find a 10 µF electrolyte and a few other small containers. Now let's estimate the time constant in this circuit: just look at the datasheet and make sure that the output resistance of the "crank" is very small, which is what the developers of the integrated stabilizer sought. To be honest, I’m too lazy to count now, but interference with frequencies from, say, 100 kHz and below the roll “sees” right at its output, that is, the control electrode and, as it was designed, transmits these pulsations “upstream on command,” diligently trying to maintain the voltage on its way out.

Current fluctuations reach the output of a higher voltage stabilizer. Following the same logic, fairly high-frequency current changes still flow almost unhindered throughout the entire chain of stabilizers. And they whistle and make noise to everyone around.

I see the only rational grain in the use of two linear stabilizers in a row is that small precision stabilizers usually do not tolerate high input voltages, and kits for self-assembly of DACs often fall into the hands of soldering riggers, who often do not even bother to look into documents for the used components. And the kits should still work...

The spread of sufficiently high-frequency interference can be easily prevented by adding... ordinary resistors to the circuit. Simple RC filters by entrance linear stabilizers will provide excellent decoupling of RF pulsations in both directions, sharply reducing the “distance” in the circuit where surge currents reach (including the “ground” wire!)

So the power supply has undergone major changes on the board. Alas, it was not without a couple of cut tracks and hanging installation.

Sometimes a small resistor is much more effective than a large capacitor:

We respect the heritage of our ancestors

Instead of a stupid bridge, we install super-fast diodes in the rectifier, which significantly reduces the current “shocks” when the diodes are turned off. This technique is quite popular and quite meaningful, so we will use it too:

By the way, it is precisely the lack of understanding of how to decouple linear stabilizers at HF that leads meticulous developers to start installing a separate transformer for each block of the circuit. Another very popular, but also costly solution to the problem of series stabilizers: the use of current source-parallel stabilizer combinations. In this case, everything is in order with the decoupling, but the power has to be dissipated with a considerable margin.

Let's not demand too much from the "whale"

A separate article is needed to describe a series of experiments with various stabilizers. Here I’ll just note that, to the credit of the developers from the Middle Kingdom, the LDO stabilizer they chose, lm1117, is perhaps the best option among the mass-produced and relatively affordable integrated stabilizers. All sorts of 78XU, LM317 and others like them simply rest due to the incongruously high output impedance (measured at 100 KHz). Alas, the precision LP2951s also went into the same basket. The TL431 behaves a little better in a shunt stabilizer circuit, but it has its own story: TL431 can be very different, depending on who made them. 1117 wins by a landslide. Alas, it also turns out to be the noisiest stabilizer. It rumbles and squeaks, both with and without load.

I had to assemble the stabilizer myself, using discrete components. From just two modest transistors, following the HotFET ideology, we managed to “squeeze out” everything that in an integrated design requires dozens of transistors and still falls short. Of course, to ensure the work of the “sweet couple”, several more active components were required... but that’s again a completely different story.

An interesting result of macro photography: I didn’t notice with the naked eye that the board was not completely washed off from the flux.

Polymers rule the roost

The latest modification aimed at achieving the most accurate sound transmission was the “smoothing” of the power supply.

In critical places, the usual (albeit good ChemiCon) aluminum electrolytes from the kit were replaced with solid-state aluminum Sanyo OS-CON. Since I collected two identical sets in parallel, it was possible to arrange “A/B” testing. The difference is barely audible, but it is there! Without a signal with conventional electrolytes, at (very) high gain, there was a certain “noise space” in the headphones. Polymer electrolytes take us into the absolute.

Sanyo OS-CON - purple barrels without a notch on the lid.

If you don't want to think with your head, work with your hands

On almost all boards and DAC kits using the CS8416 digital receiver, the Chinese put a toggle switch so that the user can choose between an optical and copper S/PDIF input (the photo on the right is a typical example caught on the Internet). So: there is no need for a switch there, the receiver chip can easily listen to two inputs without any outside help, be it a crude toggle switch or a smart microcontroller.

I’m sharing with you a trick I spotted on a demo board from Cristal Semiconductor themselves. It is enough to connect, for example, copper S/PDIF to RXN, and the output of the optical TOSLINK receiver to RXP0.

I hope there is no need to explain how this works? 😉

Even in the reference design, the companies screwed up and forgot the shunt capacitor in the TORX power supply :)

Economy or illiteracy?

It can be very useful to read the documentation of manufacturers, especially those that make the very microcircuits that audiophiles then swear by. I’m revealing the most secret secret: reference design board, evaluation board and similar “probes” from manufacturers usually contain examples literate the use of those same microcircuits. Moreover, it is not at all necessary to buy all these boards, and the price tags for such “samples” can be very different: 50, 400, and can exceed a thousand greenbacks. But, my dear developers, the documentation for all these boards is publicly available! Okay, good to teach.

So, what the Chinese did not read, or what they saved on: modest shunt ceramic capacitors of 1000 pF in parallel to 10 μF and 0.1 μF. It would seem - why, because with such capacitors we bypass frequencies from tens of megahertz and higher. The audio range is considered to be up to 20 kHz, well, up to hundreds of kHz. But no one has canceled the digital part in the digital-to-analog converter. So it is precisely the interference at tens of megahertz that freely walks through inexpensive self-built DACs, causing all PLLs to tremble in fear and thereby creating ideal conditions for the occurrence of terrifying JITTER.

Another popular way to save on matches

The vast majority of manufacturers of both digital audio sources and digital-to-analog converters save 30...50 cents on each device. We, the users, pay for this. Read details.

What's high-end without lamps?

I am amused by the hordes of tube-DAC and tube-headphone-amplifier "s in the price range from one and a half hundred to hundreds of dollars that have flooded the market recently. People seem to like how a light bulb hisses and distorts at 15...24 volts anode. However, analysis of all the problems of such DACs and pseudo-tube amplifiers for headphones is a topic for a separate article, but not just one.

(photo on the right is an example, I don’t have such a lamp-tac)

Rich topic. I just skimmed the surface here and didn’t touch on the analog part at all. And how interesting it can be to properly lay out the “ground” or organize simple and yet convenient control of the device. And what are attenuators worth - after all, they can be selected with different resistances, built according to different topologies, and connected in different parts of the path. Coordinating sources with load is a very, very interesting question, you know!... But for today it’s time for me to wrap up.

BOM, or Bill of Materials

Of course, the matter is not limited to fifty dollars. The ceramic capacitors from the kit were replaced with film. Schottky diodes, high-quality electrolytes, and a lot more had to be added, not to mention the housing. And, of course, my HotFET amplifier: only 2 (two) amplification stages from the DAC output to the headphones or amplifier output. Neither more nor less, but in the amplifier itself I counted 32 transistors in the stereo version. Yes, all transistors are JFETs and depletion MOSFETs. No way I can’t fit into the green fifty kopecks even in terms of components 🙂 Note that this is without any audiophile esotericism. Well, yes, I also have my own opinion on this matter. After all, there are people who believe that by installing the “right” components, any circuit can be made to sound. If you, dear reader, are from their ranks, teach me, I will listen, argue, listen and tell everyone about my experiences right on this site.

So where is the promised freebie???

Friends, this article is just thoughts, notes in the margins, it was written hot on the heels of remaking a Chinese DAC. I myself would never get involved in such an adventure again: although it turned out well, it was too expensive in terms of time and effort. And I don’t recommend it to anyone. When I dealt with that set, the poison simply oozed out, which was reflected in the article :) I apologize for the slightly arrogant style of presentation, and if I did not live up to your expectations and did not offer the distribution of almost free high-end DACs to the population 😉

If you were interested, please let me know. There is still a lot of material in the bins, but the strength, motivation to publish and formalize all this comes mainly from reviews and comments from my readers.

The Aune T1 is a tube USB DAC with a built-in solid-state headphone amplifier that has sold over 50,000 units. Worldwide.

Main characteristics

1. External linear high-quality power supply, which implements additional filtering. This solution helps eliminate noise from the power source.

2. The DAC is implemented on an asynchronous USB controller SA9027 and a PCM1793 chip.

3. Aune T1 Mk2 USB DAC is an external sound card, DAC and high quality headphone amplifier in one package. Aune T1 can also be used with powered speakers as part of your home hi-fi system.

4. Aune T1 works under Windows 7, 8, Vista, XP, Mac OS operating systems. Can be connected to iPad. No additional driver installation required.

5. The headphone amplifier module is made separately and can be replaced later when the corresponding upgrade is released. The lamp must warm up completely before playback begins. When you turn on the device, the light bulb heats up for 30 seconds, after which a white indicator lights up under it, and only then the device begins to function. The Aune T1 Mk2 USB DAC also features switching amplification modes.

6. New modular design. Multiple boards within the device are powered independently, eliminating crosstalk. The DAC also features a safety shutdown that prevents damage to your headphones or speakers when the device is turned off.

7. Aune T1 Mk2 USB DAC is made of high-quality audio components: ALPS potentiometer (Japan), WIMA capacitors (Germany), ENLA electrolytic professional audio capacitor and so on.

8. The amplifier will pump headphones with a resistance of 30-600 Ohms. The amplification circuit is OP+BUF.

9. Aune T1 Tube USB DAC has one line input and one line output.

Video (promo, English)

Specifications

Tube: 6922EH Electro-Harmonix (Made in Russia)

Frequency response: 20 Hz - 20 kHz

SNR: >=120 dB

Output power: 1000 mW/32 Ohm, 400 mW/120 Ohm, 150 mW/300 Ohm (maximum 20 V)

Output Impedance: 100 ohms, 10 ohms (headphones)

USB interface:

Data up to 24 bit / 96 kHz

Operating systems: Windows XP/Vista/7/8, Mac OS

Power: AC 220/110 V

Size: 155*97*40 mm (L*W*H)

Contents: power supply, USB cable, adapter 6.35 - 3.5 mm

To go further in the design of amplifiers, I ran into the problem of a quality source. I really needed a good DAC. I was not fully satisfied with the quality of those that I had at home and that I had listened to before. If this is a classic DAC with operational amplifiers at the output, then this usually leads to problems in reproducing the upper mids and highs. The middle becomes slightly grating, harsh, as if there is sand or metal in the voice, especially at high volumes. With tube DACs, not everything is all right either - often there is no good bass or a flat, inexpressive sound, and besides, for some reason, developers really like to install a cathode follower at the output, which, although it reduces the output impedance, but in my humble opinion the sound to put it mildly, it does not decorate. In general, I came to the conclusion that I had to do it myself.

Why did I choose Ad1955? Its output is designed for an I – U converter with a current of 3 – 5 mA of positive polarity. And here is a wide range of options for connecting to a high anode voltage in such a way that the output current of the DAC chip passes through the lamp.

Yes, of course, I wanted a DAC with a tube output. And given my weakness for cascades with a common grid and transformers, the output was planned on my favorite 6E6P lamp with a transformer output. The choice of this lamp is also due to its low internal resistance in the triode, as well as its high transconductance (30 mA per volt), and in the case of a cascade with a common grid, this gives a lower input resistance - and this is very good for I - U DAC converters, for which the input resistance should tend to zero. It is logical to make the input I - U of the converter on a germanium transistor connected according to a circuit with a common base. This is where the scheme was born. According to my rough estimates, the input impedance of my hybrid cascode is somewhere on the order of 1 Ohm. How did you calculate? We take the formula for calculating the input resistance of a cascade with a common grid Rin = (Ra + Ri)/(u +1). The lamp load is 3.3 KOhm, the 6E6P itself in the triode has about 1500 Ohms. Add and divide by 30 - this is the gain of the lamp. It turns out 160 Ohms. This is the input impedance of a lamp connected according to a circuit with a common grid. Now for the transistor, the lamp is a load Ra. I don’t know the internal resistance of a germanium transistor, but we take roughly 50 Ohms, then if its Kus is about 250, then (160 + 50) / 250 = 0.84 Ohms.

If someone finds 6E6P too emphasizing the middle, then it can be replaced with 6ZH9P, 6ZH11P or 6ZH49P. Only in this case, you should pay attention to the fact that the collector of the transistor is connected to terminals 1 or 3 of the lamp socket (and not to terminal 6) - then you can simply plug in the one that seems more melodious to you.

I present the first version of the scheme, although I am sure that it will have to be finalized, because there is no limit to perfection...

In order not to do the digital part myself, I took a DAC board for AD1955 on e-bay and removed the operational amplifiers from it, also unsoldered the 2K resistors from the power supply positive according to the datasheet from the AD1955 outputs, and left 100 pf (capacitors C1 and C2 in the diagram) those that were on the board. I'll give more details a little later.

I tried a transistor stabilizer as a power supply, but it still turned out to be the best-sounding tube doubler on the 6N1P, which was later replaced by the ECC99. The reason for using this rare lamp is simple - to package my DAC I used a case from a Chinese Lite DAC, which died for a long time, thank God I didn’t throw out the case. Both network transformers, the network button and the input/output connectors came in handy. Here is the power supply diagram:

As you can see, the 6E6P filament is powered by direct current, but unstabilized.

Now a little about listening. The source was a Denon 1500 CD player and compared it with my DAC, the signal was supplied via an optical digital cable. The amplifier is my cascode for 6E5P - 2A3. Speakers - wideband in OYA from 3AC505. The first impression was very bad, I was very upset and was about to take my creation to the closet in the company with other unsuccessful projects. I found my DAC to be overly harsh on female vocals and trumpets. But then - lo and behold! - it turned out that it was me who mixed up the inputs on the switch in front of the amplifier - something that I was disappointed in - it was just a Denon DAC, but my DAC gives an excellent presentation of the material! And the timbral balance, stage width, and emotional richness will be higher than with Denon. In general, he sings cleanly, detailed, transparently, and what especially distinguishes him from my signature Denon is the very soft delivery of vocals and the upper mids and highs in general - no ringing, no excessive harshness at almost any volume, in general - much more natural. Here it is appropriate to talk about the “coloring” of sound. As in colorimetry, when talking about color, it is important to answer the question - what is accepted as the standard for white? If we take transistor sound as this standard, then yes, the lamps provide “coloring”. But in my understanding, tube sound is the standard of white. And the operational amplifiers at the output (by the way, always used with deep OOS) give a slightly metallic color and a slightly unnatural upper register, which IMHO is not typical for live performance. Overall, I was very, very pleased with my creation.

Here are its characteristics

– output voltage at 0 dB – 2 Volts;

– noise level – less than -80 dB, there is simply no way to measure less;

– total harmonic distortion at the maximum level – less than 0.15% – again, I can’t measure it more precisely yet.

– inputs – optical and SPDIF;

– outputs – unbalanced 2 Volts and balanced 10 Volts;

– output resistance – at the unbalanced output – less than 100 Ohms, balanced output – about 2 KOhms;

– the circuit does not contain OOS circuits.

Here is what the device looks like packaged in the case and a photo of the entire set of listening equipment.

The output transformers were wound to order at the Audio Instrument company, for which we bow to Sergei Glazunov. And also - read on the forum http://www.diyaudio.ru/forum/index.php?topic=4180.0. My first attempts (not entirely successful) to make a DAC using only tubes are in another thread on the same forum http://www.diyaudio.ru/forum/index.php?topic=1267.570.

Updated June 6, 2015. I had to adjust the diagram a little. Firstly, at peak volumes, excitation (resonances) was observed and therefore it was necessary to add capacitors C3 and C5 to the lamp grids, as well as C1 and C6 to the anodes. Also, due to voltage drift at the output of AD1955, it was necessary to stabilize the bases of the transistors using a 3.0 volt zener diode D1. Well, nevertheless, I replaced 6E6P with 6Zh49P - out of all those listed earlier, it seemed to me the most balanced in timbre.

****************************************************************************************************

— a multibit digital-to-analog converter made on four industrial 18-bit AD5871 DAC chips.

- a tube headphone amplifier with an impressive power of 8 W and the ability to replace the tubes with “solid-state” amplifier modules, which are purchased separately.

The devices are designed using a fully balanced amplification topology.

Appearance

All Schiit devices are made in the same style, and models from the upper price category are no exception. No sapphire crystals or diamonds in the handles, no unnecessary overpayments for gilding the case and screens. However, now the cases are full-size and look harmonious in any audiophile equipment rack.

The controls are still minimalist: a single button on the DAC that selects the desired input.

On the headphone amplifier, in addition to the volume control, there are switches for gain and selection of balanced or unbalanced input.

The rear panels of the devices are also laconic.

The Gungnir DAC has USB, optical and two coaxial inputs, one of which is BNC. It should be noted that BNC is a connector specifically designed to transmit high frequencies (as opposed to low frequency RCA). BNC is also optimal for high-quality digital signal transmission.

There are two pairs of unbalanced RCA outputs and balanced XLR outputs that operate simultaneously.

The Mjolnir 2 amplifier has balanced and unbalanced inputs on the rear panel, as well as outputs for connecting other equipment, such as a power amplifier for speaker systems.

The power switches on both devices are also located on the rear panel. And if in the case of a DAC that consumes a relatively small power of 20 watts, you can close your eyes to this and leave it constantly on, then in the case of a headphone amplifier that consumes 45 watts in idle mode and has a limited lamp life, this is quite inconvenient. At least in a rack you can’t easily turn off the power. This is exactly the case when you have to pay for the beauty and design of the front panels with convenience.

Passport specifications

Gungnir Multibit

Digital-to-analog conversion chip: Analog Devices AD5781BRUZ ×4 (two per channel, balanced circuit)
Digital filter: proprietary closed-loop type with bit precision function, implemented on Analog Devices SHARC DSP processor
Analog Path: Fully discrete JFET buffer stages for balanced output and JFET summing stages for single-ended output, direct coupled
Operating frequency range: 20 Hz - 20 kHz, ±0.1 dB; 1 Hz - 200 kHz, −1 dB
Maximum output amplitude: 4.0 V RMS (balanced output), 2.0 V RMS (unbalanced output)
Total Harmonic Distortion (THD):<0,005% (20 Гц — 20 кГц, при полной выходной мощности)
Intermodulation Distortion (IMD):<0,004% (измерены по стандарту CCIR)
Signal to Noise Ratio (S/N): >115 dB (relative to 2 V RMS)
Inputs: coaxial S/PDIF (RCA and BNC), optical S/PDIF (Toslink), USB
Supported formats: up to 24 bit/192
Outputs: One pair of balanced XLR connectors and two pairs of unbalanced RCA connectors
Output impedance: 75 ohms
Clock recovery: Bit-accurate at all native sample rates via Adapticlock analysis and VCXO/VCO regeneration
Power supply: two transformers (one for digital, one for analog) with 8 control stages, including separate power rails for critical sections of the digital and analog paths
Upgradability: separate USB input and DAC/analog circuit boards, replaceable
Power consumption: 20 W
Dimensions: 406×223×60 mm
Weight: 4 kg

Mjolnir 2

Operating frequency range: 20 Hz - 20 kHz (−0.1 dB), 2 Hz - 400 kHz (−3 dB)
Maximum power at load impedance:
- 32 Ohm: 8.0 W RMS/ch
- 50 Ohm: 5.0 W RMS/ch
- 300 Ohm: 850 mW RMS/channel
- 600 ohms: 425 mW RMS/channel
Total Harmonic Distortion: Less than 0.005% (20Hz - 20kHz, 1V RMS)
Intermodulation distortion: less than 0.006% (CCIF test, 1V RMS)
Signal-to-noise ratio: over 104 dB (unweighted, relative to 1 V RMS, in low gain mode)
Interference: Less than −75 dB (20 Hz − 20 kHz)
Output impedance: 1.0 ohm (high gain), 0.3 ohm (low gain)
Gain: ×8 (18 dB) or ×1 (0 dB), front panel switch
Topology: Tube voltage amplifier or LISST solid state voltage amplifier, Crossfet push-pull parallel output stage, Non-inverting single voltage gain stage
Power source: a special transformer for the output stage of Cyclotron 4, powered by filter capacitors with a capacity of more than 65,000 μF, plus a separate transformer with a voltage of 200 V and storage capacitors with a capacity of more than 4000 μF - for the high-voltage discretely controlled input stage
Inputs: a pair of balanced XLR jacks and unbalanced RCA jacks, switchable using a toggle switch on the front panel
Outputs: 4-pin balanced XLR, 6.3mm minijack, pair of 3-pin XLR pre-outs, one pair of unbalanced RCA
Power consumption: 45 W
Dimensions: 406×223×60 mm
Weight: 5.4 kg
Approximate price: 76,500 rubles (only with 6BZ7 lamps) at the time of preparation of the review

Internal structure and measurements

The internal structure of the Gungnir Multibit DAC will cause a positive reaction in any engineer. A hardened audiophile may think that for such a price, gold parts and film capacitors were underused. But wait, Schiit engineers have prepared another surprise for you!

On the motherboard, according to the accepted Schiit concept, there are separate modules for a multi-bit DAC and a USB input. Note that this does not reduce the cost of the device, but allows for higher quality wiring and operation of individual components compared to if everything were mounted on one board.

Particularly touching are the personal thanks to the engineers written on the boards; this once again confirms that the device was designed by people for people, and not by obscure OEM manufacturers solely for making money. The DAC has many interesting solutions aimed at improving sound quality.

The USB receiver is made on the familiar CM6631A controller, but has galvanic isolation, and done correctly: the master oscillators are located on the “clean”, galvanically isolated side. Yes, it is more expensive, yes, it is more difficult to implement, but this is the only way to get a good result. And then it's done. So you can safely connect the DAC via USB to your computer and not worry about interference and ground loops. Note that in our case, Windows 10 independently found and installed the required driver. It was not possible to install USB drivers from the official website.

The S/PDIF receiver is made on the old familiar, but no worse, CS8416 chip.

Also on the motherboard, in addition to transformers, rectifiers and primary stabilizers, there is a rather interesting phase-locked loop unit, with its own oscillators at frequencies of 22.579 and 24.576 MHz. This proprietary technology is called Adapticlock and serves to further suppress digital signal jitter.

On the multibit DAC board, in addition to the AD5871 itself, there is an Analog Devices ADSP-21478 digital processor, which is used for digital signal filtering. After it and before the AD5871, according to all the canons of building high-quality DACs, there is a reclock made on separate digital D-trigger chips.

The post-DAC filter-amplifier is a separate, audiophile topic. Surprise! It is implemented on JFET field-effect transistors using non-OOC circuitry. Yes, there are microcircuits there, but the signal does not pass through them, they are only needed to maintain zero at the DC output. This is a gift to those who believe that negative feedback in the audio path is evil. Yes, this affected the measurements, but not the sound.

Objective measurements were carried out when working from USB under Windows 10.

In this case, the measurements can be characterized simply: the manufacturer did not sneeze at them, the zero-emission concept and sound were put at the forefront. And the technical specifications with many zeros after the decimal point, posted on the official website, serve rather to avoid unnecessary excitement among people who listen not to sound, but to graphs. We do both.

The tube spectrum of distortion produces field exhaust, and, apparently, this was done intentionally.

To verify this, the measuring card was connected directly to the output of the DAC, before the JFET filter-amplifier.

In this case, we see very low distortion of the AD5781 DACs themselves with a typically multi-bit signal spectrum. Just for fun, I listened to this version. Let's just say: without a harmonizing filter the sound is not very good. Despite their low distortion, DACs subjectively sound very sharp.

We also ran the J-test test file, which allows us to show flaws in the construction of the digital part, recalculation or increased jitter of the digital signal. The result is perfect: no side interference beyond the main comb. This confirms the very high quality of design of the digital, “pre-audiophile” part of the device.

A few words about the multibit converters used. The AD5781 is one of the best multibit converters produced today, but they are also very expensive, about $40 apiece. There are also AD5791, 20-bit precision, they already cost $100 apiece and are used in the top Schiit Yggdrasil DAC.

Despite the costs, we support the manufacturer in using parts that are produced here and now, and not obscure old warehouse leftovers or even Chinese fakes. This guarantees the quality and repeatability of product characteristics.

The post-DAC filter and output are made using JFET transistors in a SOT-23-5 package marked XL, which could not be identified. Wima foil capacitors and Nichicon KW electrolytic capacitors are also used.

The filter circuitry is fully balanced, so the unbalanced output only receives half of the output signal, essentially only two DACs are working instead of four. This affects measurements and subjective sound quality, so we do not recommend using an unbalanced connection to the amplifier, although it is possible.

The Schiit Mjolnir 2 headphone amplifier is designed according to a similar balanced ideology. But here the emphasis is already placed on the output power and stability of the power source to drive the most complex load.

Powerful 30-watt transformers, a capacitor bank with a total capacity of 65,000 uF, IRF610 output transistors capable of dissipating 54 watts of power, connected using the proprietary Crossfet topology - all this allows the amplifier to operate even with an 8-ohm load connected to the balanced output.

The manufacturer uses transistors of the same structure in both amplification arms, which ensures their excellent identity and lower distortion.

The heart of the amplifier, which provides the main signal amplification, is an electron tube with JFET transistors at the input. The present microcircuits are used only to maintain zero at the DC output; the signal does not pass through them. The amplifier's circuitry is unique and does not resemble standard solutions.

The unbalanced output is implemented quite differently; a separate bipolar transistor amplifier is provided for it, although the main amplification is also provided by the lamp. The maximum power of the unbalanced output is limited to 2 watts. Balanced and unbalanced outputs operate simultaneously, but at the same time independently of each other.

The soft glow of a vacuum tube warms the soul of any music lover. Tubes produce a special sound in some way, emphasize voices, timbres of instruments, muffle and soften the sound, veil garbage in the recording... But what if you want to cheer yourself up and listen, for example, to several Death Metal albums? Schiit has taken care of such desires by offering a solid state tube replacement circuit - LISST. Essentially, this is a two-stage amplifier mounted in a housing. And there will definitely be no veil with him - checked!

Technically, the Mjolnir 2 amplifier gives a very good impression, and in terms of circuit design it is not inferior to the Gungnir DAC, but what about the measurements?

For testing, we used a professional balanced Lynx L22 card, and in most cases the measurement results were limited precisely by its quality, and not by the amplifier.

Regardless of the use of a 6BZ7 vacuum tube or LISST solid state circuit, the balanced amplifier performs excellently into a 300 ohm load. When the load is reduced to 32 Ohms, only the second harmonic increases, which will not affect the sound in any way.

An unbalanced amplifier is more demanding on the load, and with a power of more than 100 mW into a 32 ohm load, distortion grows catastrophically. At a 300-ohm load, nothing like this happens. Therefore, for maximum quality, we still recommend using a balanced output.

The output impedance of the balanced output is about 0.8 Ohm, it perfectly damps any headphones, prevents uncontrolled resonance phenomena, which ultimately provides natural and dynamic sound.

Listening

We won’t bore you with listing the musical material used for listening; it all depends on personal preferences. Note that we were not limited to any specific styles of music; we listened to the Gungnir Multibit DAC on different systems, and used different headphones with the Mjolnir 2 amplifier, from Oppo PM-2 to Audio-Technica M50x. A person who has “grown up” to this class of audio technology knows perfectly well what he wants to get in sound and what he will use to listen to his favorite music.

The Gungnir Multibit DAC can be described as a source that is ready to play almost any music with high quality. It cannot be called too harsh or, conversely, too delicate. But a multibit heart is undoubtedly more predisposed to dynamic and fast music such as rock and metal, and pop music performed by Gungnir will sound much more interesting. Of particular note is the way the voice is reproduced: it seems much more alive. And in general, the mid-frequency range is reproduced perfectly. Instruments are reproduced separately, there is no mess of sounds at all. In many old familiar compositions, many new sounds were heard that were previously masked in the general rhythm. The bass is elastic and punchy, but at the same time there is not a hint of buzz. In some systems there may not be enough bass, but where there is bass, it will be excellent. High frequencies have never been the strong point of multibit DACs; in this case, the manufacturer tried to make them as neat and comfortable as possible. Yes, there is some roughness, but it is by no means dirt! Quite the contrary, in high-quality headphones you can hear much more detail in the high-frequency range than when using modern “sleek”, boring DACs like the AK4490. The limiting factor here is the quality of recording and mastering of the phonogram itself. Correct sound reproduction of the high-frequency range is also facilitated by a very low level of jitter; when connected via the USB input, the DAC plays perfectly! This is how we recommend listening, but if you have a high-quality audiophile digital source, you won't be disappointed with the coaxial connection.

To summarize, Gungnir Multibit can be characterized as the most neutral and even, perhaps, slightly detached source. He does not dump the musical picture directly on you (and deal with it as you wish), but presents it, delicately and accurately revealing the intention of the composer or sound engineer, without hiding or embellishing the details. The listener is more in the role of an observer, contemplating the riot of sounds from the side. You won't find yourself in the center of the orchestra or on stage next to the musicians. But you will hear everything. This is how they listen to music. Dynamic, fast, open. This is why old multibitniks love it, and this same trait has been preserved in the new Schiit Gungnir Multibit.

We admit that, having received this DAC for testing and carrying out preliminary measurements, we were somewhat disappointed with the low measurement results, and there was no wow effect during the initial listening. However, the longer this DAC was in our possession, the more we liked it, revealing more and more secrets of old familiar compositions and without causing any irritation even after very long listening. The sound is very comfortable, but at the same time clear - a rare combination. That is why it is recommended that when purchasing equipment, you conduct a fairly long test, at least for several days, and only then draw conclusions. Perhaps the option that “hooked” you in the store with its dynamic and bright sound will blow your mind in just three days at home. With Gungnir Multibit everything turned out the other way around.

The sound of the Mjolnir 2 headphone amplifier also deserves the most flattering epithets; it is not for nothing that it is the most expensive in the manufacturer’s line.

The amp's sound is completely neutral and clear, and the bass control is amazing. Listening with headphones through this amplifier allowed us to discover new subtleties of bass parts: for example, it turned out that in some compositions the sound engineer deliberately introduced distortion into the bass, which was never noticed when listening on speakers. In general, the bass has ceased to simply set the rhythm; it has become an interesting object of observation. The amplifier controls any headphones so masterfully that it seems that you can connect external speakers to it and listen to music in peace.

The use of 6BZ7 vacuum tubes or LISST solid state modules is purely a matter of taste. In both cases the sound is very good. The tubes tint and soften the mids, but give some veil in the highs. Solid-state modules provide the cleanest high frequencies, a wide stage and a neutral sound; they are more preferable for modern electronic and rock music.

To understand the subjective difference, we tested the amplifier's balanced and unbalanced outputs using Oppo PM-2 headphones and replacement cables with corresponding connectors. The outputs have only minimal differences in coloration and sound quality, but balanced outputs generally produce higher-end sound, resulting in better bass control and cleaner sound at the same volume. If you do not go beyond a reasonable volume, then both outputs have excellent sound, the difference is purely tasteful. If you want to give in, remember that the balanced output in high gain mode has a maximum amplitude of 20 volts! Probably, I still needed to connect the speakers.

Among the features of the amplifier, it should be noted that due to good bass control, it does not “inflate the low frequencies,” so if the headphones have a rolloff in the low-frequency region, then the bass will end up being not enough.

conclusions

And made for each other! You understand this when you turn them on together and just start listening to music. You can spend more than one hour contemplating music, only time forces you to stop listening. Isn't this the main criterion for the quality of technology? We think yes. Schiit engineers think the same way, creating products that are worthy of admiration both in technical terms and in terms of the transmitted sensations.