Tube amp with usb - diyAudio
Go Back   Home > Forums > Amplifiers > Headphone Systems

Headphone Systems Everything to do with Headphones

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 9th June 2015, 02:57 PM   #1
diyAudio Member
Join Date: Mar 2011
Default Tube amp with usb


I have begun planning a headphone amp as a summer project. I want to use a tube amp since it's interesting hardware and I also want to use a DAC. This project is not really aimed at perfect audio but more to learn before finishing my masters in EE.

I have been looking at this circuit JEDAC - yet another take on a DIY high-end USB Audio DAC and though that design but with a different amp stage would be fun. I'm looking for a suitable tube amp design to connect this with. Any suggestions for what design I should use?

I have access to a proper lab and have worked with high voltage before but my audio knowledge is rather low sadly...
  Reply With Quote
Old 10th June 2015, 01:35 PM   #2
diyAudio Member
Join Date: Nov 2004
Location: Germany
That DAC design is... how should I say it... strictly hobbyist level.
A fancy DAC chip is wasted on a single-side PCB design (I'm sure the guys over the in the Digital Line-Level section will have more to say on that), one with a nice big ground loop running around the circumference no less (those are the perils of mindless ground fills). You'd have to fix the layout at the very least, which on the upside would provide an opportunity to learn using Kicad. Finally, no kind of performance verification, at all. "High-end" - yeah right. I wouldn't be awfully surprised if performance turned out to be about on the level of the onboard audio that you already have. (Your average Chinese "high-end" DAC board tends to clock in at about 95-100 dB of dynamic range.)

How much time, money and effort are you willing to invest? I would not advise setting your standards too low needlessly - external constraints will mean you'll have to do that anyway. Shooting for the moon and ending up in upper stratosphere is better than deciding that the rooftop is quite high enough from the get-go.

You should also consider your priorities. DAC wise you could take a shortcut by just getting a pre-assembled ODAC board (which is the only form they come in as you would not be able to procure the USB receiver chip on the free market). Performance of these is well-documented and very good.

Building some kind of tube-based headphone amp still is relatively easy... but one that is actually some definition of "good" is quite a different story. On the one hand, the relatively large mechanical dimensions of typical tubes mean that they are suitable for relatively high supply voltages, with parasitic capacitances that are quite voltage-independent, giving good linearity... but on the other hand, their current capabilities tend to be quite limited, and voltage noise tends to be rather on the high-side... maybe the 10 nV/sqrt(Hz) ballpark with a following wind in a good low-noise design, easily a multiple of that otherwise. Hence they're inviting designs with high internal levels at high impedance. (As you may be able to guess, low capacitive coupling and potentially shielding from stray electric fields are advised.)

Which is all good and well, but there's still that pesky outer world to interact with. Input wise, a preamp stage in front of the volume pot (cheating with semiconductors allowed) would get things sorted if you insisted not to use an input transformer of 1:2...1:3.16 turns ratio - though the latter could keep you out of trouble when it comes to ground loops, which would otherwise form almost inevitably (a lot of tube gear tends to be IEC Class I kit that ties its audio ground to safety earth somewhere, as does your average PC).
But then there's the output side. If you still have some insensitive 600 ohm cans (K240M/DFs, K340s or somesuch), a decently beefy cathode follower would drive these to high levels as-is. A power-hungry ortho at 40..50 ohms.... well, not so much. And something sensitive and low-impedance may show uncomfortable levels of hiss and hum. Which means you'll probably want an output transformer with multiple taps. Now those come with problems all of their own. A high-impedance primary (for good LF response) tends to mean that bandwidth is noticeably restricted, making more than a bit of negative feedback hard to apply (like 10-15 dB for speaker amps, which ranges somewhere between nothing and counterproductive - see Putzeys on negative feedback), which is unfortunate because they tend to be a major source of distortion themselves due to B-H nonlinearity. Things should be a little less bad for headphone dimensions, including size and weight, but basically these things still apply. Besides, good transformers are never cheap, assuming you can find a suitable off-the-shelf part at all and don't have to go through the fun part of winding one yourself.

Microphonics are another problem that requires some attention.

The good thing about designing with tubes in this day and age is that you can "cheat" using semiconductor parts to form current sources, gyrators and all that good stuff that keeps loop gain, linearity and PSRR high. So if you need a transconductance amplifier of very low voltage noise for high voltages for some reason, you can always cascode a low-noise BJT and a tube. (Though it's not like there aren't any high-voltage bipolars around, even if they are getting harder to find with the demise of CRTs.) The bad part is that capacitors and resistors suitable for high voltages are getting harder to find. (Less so buffer electrolytics - thank switch-mode power supplies which usually need 400...450V= parts primary-side - but rather other stuff. Standard 1/4 W leaded metal film resistors can sustain about 250 V= if memory serves.)
  Reply With Quote
Old 10th June 2015, 02:25 PM   #3
diyAudio Member
Join Date: Mar 2011
Thanks for a very informative answer!

I have been looking around for a while now and have narrowed down my project a bit.

For the DAC, I will build PupDAC Overview,
but with a new layout. I'll do everything in Altium, and then etch my own prototype board, if that works I'll order some better pcbs of the new design. Hopefully it will be a bit better than the standard laptop sound card.

For the amp I have not really decided on a design, I have a rather low budget on parts that can't be sampled, so an output trafo is sadly not possible. I have no moral problems with using semiconductors, this is not intended to be a hardcore build.

So far Morgan Jones design The Morgan Jones Mini Tube Headphone Amplifier | HeadWize seems nice but I don't know how it will differ to say Hybrid Headphone Amplifier | EEWeb Community with 6DJ8 tubes. I would prefer to not use 230V ac and have a 12V dc instead, even though I have worked with up to MV earlier...

I have a headset in mind when planing this, I want an amp that can be used from 32 to 400ohm or something like that.
  Reply With Quote
Old 12th June 2015, 09:46 PM   #4
diyAudio Member
Join Date: Nov 2004
Location: Germany
Originally Posted by Evil AnAnAs View Post
For the DAC, I will build PupDAC Overview,
but with a new layout.
Have you worked with mixed-signal devices before? Otherwise you would probably be well-advised to read up on grounding in such environments.
Originally Posted by Evil AnAnAs View Post
Hopefully it will be a bit better than the standard laptop sound card.
Distortion wise, quite possibly. Output level wise, definitely. Dynamic range wise, not so much, given that the PCM27xx receivers are quite long in the tooth by now and limited to 16 bits and 48 kHz tops (according to the first USB audio spec from back in the day). So that's like 98 dB of theoretical dynamic range, a bit less when dithered (which is advisable). Standard-grade onboard audio will usually benefit slightly from 24-bit operation, so that's about the same ballpark.

Note that while modern onboard audio can be a lot better than its reputation, some drivers have their quirks, and there are some other pitfalls as well. Notebook speakers are often set up with the "Full range speakers" checkbox unticked in Windows, which then sadly applies to both speakers and the headphone output. Some chips or their drivers will also goof with certain bit depth / sample rate settings. Older Realtek driver versions from 2011/12-ish sounded bad (and posted some really lousy RMAA results) in 44.1 kHz, 24/44 in particular. The IDT 92HD90 in the Dell I'm typing this on (please excuse any typos, the keyboard is decent but not a Cherry G80) sounded like it had a built-in smiley face EQ in 16/44, but things went to normal in 24/44. That should tell you how important performance verification is.

These days I can spot gross irregularities by ear (basically, if things sound way off from my Rockboxed Clip+ even when using the FiiO E11 to provide near-zero output impedance, something must be amiss), but generally speaking a good measurement setup is the more flexible option. If you know how to defeat ground loops, a good soundcard will go a long way.

Originally Posted by Evil AnAnAs View Post
So far Morgan Jones design The Morgan Jones Mini Tube Headphone Amplifier | HeadWize seems nice but I don't know how it will differ to say Hybrid Headphone Amplifier | EEWeb Community with 6DJ8 tubes.
6DJ8... ah, the good ol' ECC88 on this side of the pond.

TBH I would pick neither, but there are lots of things to be learned from both.

MJ is a classic tube design. No / very limited feedback, not much inherent power supply rejection (PSRR), hence requiring the kind of effort on the power supply side that used to be common back in the day. In optimized form it puts out 83 mW per channel into 300 ohms and 10 mW into 32 ohms, which should tell you that it is current-limited - no surprise though with the 6DJ8 totem-pole output stage run a bit short of 10 mA (which is about the maximum for that kind of "small-signal" tube; normally one would use something more powerful in an OTL design, like EL84 or EL34). And that while consuming 975 mA @ 6.3 V (so ~6 W) for the heaters alone. At about 22 (27 dB) at no feedback, voltage gain is rather on the high side for a headphone amp.
The article contains several helpful practical hints, like twisting the power supply wiring or providing extra supply filtering for critical common-cathode circuits. Incidentally, the John Broskie mentioned is the TubeCAD guy. He's got some really good articles online (even if they can go over most people's heads, yours truly often being no exception), including one on "totem pole" (SEPP) output stages as used in the MJ where he discusses their PSRR problems, an unfortunate downside to their push-pull abilities. Those would arguably work better with a negative power supply, but then the previous stages wouldn't be happy with that. Anyway, maybe that's part of why some people prefer SE outputs - when you have a circuit that has good PSRR otherwise and the output stage becomes the bottleneck.

The ExtremeCircuits design is a 3-stage open-loop buffer (so no voltage gain, rather a bit lower than unity), all single-ended, with CCS loaded output stage. Due to it being all followers, PSRR should be less pathetic than in the MJ (essentially, voltage gain in the common emitter case becomes PSRR in the EF case). I would prefer some more than 27 mA of standing current (maybe 3-4 times that, which would require a beefier T2 in a better case like TO-126). Also it looks like heater voltage isn't especially well-defined. As shown, we'd be looking at about 9.6 V, which is neither fish nor flesh and with the series connection as described would result in underheating. Tubes don't take kindly to either over- or underheating in the long run. No mention of plate current is made, probably because it's some really pathetic value at the low supply voltage; the 6DJ8 with its higher transconductance should perform somewhat better there.

Both of these circuits have their issues. The MJ has more gain than you typically need when run open-loop, yet open-loop gain still is too low to be much use with negative feedback ("go big or go home", while reportedly coined in reference to Intel's eDRAM sizing for Iris Pro graphics equipped processors, applies here as well - 40 dB of spare loop gain would be desirable). And it needs a a fancy classic tube amp power supply.
The ExtremeCircuits design, while less picky about power, provides no voltage gain at all and would require some redesign to fix the heater issue.

A design that strikes me as simple and clever is Sijosae's Multi-Hybrid (some Bravo Audio amps are based on this). Still open-loop with too much gain or rather noise, but enables relatively low supply voltage (24 VDC) and decent PSRR by using a CCS in place of the plate resistor (PSRR = Ri / Re(Zccs)), and SE output stage current is cleverly recycled in the heater (which works due to the 6922 being a series / parallel heater design at 2x 6.3 V / 150 mA, so you can obtain either 150 mA at 12.6 V or 300 mA total at 6.3 V). The latter means that some output swing is lost, of course, but it still is pretty decent. You couldn't pull this of with a 12V supply.
Originally Posted by Evil AnAnAs View Post
I would prefer to not use 230V ac and have a 12V dc instead, even though I have worked with up to MV earlier...
And next you'll tell me you want to use that spare router SMPS that's stlll floating about, or what?

(At the very least, use a conventional transformer supply. Switch-mode supplies potentially are nothing but trouble for audio. Pesky common-mode noise that's hard to get rid of without a safety earth connection, and with the latter you're easily completing a ground loop if the supply designer wasn't cautious enough to include a resistor. Pure DC/DC converters tend to be much less of a problem IME.)

Look, if engineers had always "preferred not to", we wouldn't have gotten anywhere in the 20th century. When dealing with Real Men's Hardware[tm] (which includes vacuum tubes), you may just have to man up. Have you ever tried picking a suitable operating point for a simple tube circuit (like a common cathode amplifier) using the curves provided in the part's datasheet? Or a JFET, at least? If not, I suggest you do that now. The 6DJ8/ECC88 wouldn't be a bad pick, it's about as good as tubes are going to get. Shoot for Va ~= mid-supply, 3-5 mA. I suggest you keep a few Vpp of output voltage swing in mind, too.
Originally Posted by Evil AnAnAs View Post
I have a headset in mind when planing this, I want an amp that can be used from 32 to 400ohm or something like that.
Then you should pick your headphone jack carefully or otherwise ensure that any kind of 4-pin plug makes proper contact, i.e. regardless of whether it's an Apple or Android pinout. Not sure how many good 4-pin sockets are out there among parts distributors. One of those with the last ring and sleeve connected should get things sorted.

For lowish-mid-highish impedance cans of decent sensitivity (not counting Hifiman HE6s here) you can usually get away with ~2...3 Vrms max. (I'm a quiet listener and 0.8 Vrms on a Clip+ will do with 102 dB/V HD580s on any but the most dynamic classical recordings, and even then it's just a matter of gain on the quiet parts. I've used that player with rather insensitive 600 ohm cans for portable listening with the compressor on, and that still worked out, but of course the compressor was a must.) Output current wise, 30-50 mA is quite OK. Output impedance wise, 5 ohms should still be quite easily acceptable and should help you keep stability in check. Output noise floor, I'd say 30 µV (20 kHz BW) abs max, but preferably 10 µV or under. Hum, not sure, maybe 10 dB more.

The kind of amplifier concept I'd be after would be more along the lines of my 3-transistor preamp. Two gain stages for plenty of open-loop gain, hence plenty of negative feedback possible, dominant pole compensation, healthy device currents.
Granted, there is no such thing as a complementary "p" tube, and setting DC levels in an all-"n" setup gets a bit messy, but it's doable. This 2-transistor phonopre shows a typical arrangement. Note the two feedback paths, one being for DC only.
The input stage would be a hybrid cascode with a low-noise small-signal transistor and a tube (maybe even bootstrapped as optionally suggested for the 3-transistor amp), with a tubed second stage. I might also bump the simple follower at the output to a CFP, possibly CCS loaded.

Something which is hard to get rid of in these single-supply designs is power-on pop noise. At the very least, the output has to move to about mid-supply when powering up, which through adequately-sized output coupling caps may give quite some excursion at the output. Split supply generally is much better in this regard, but requires some rethinking / redesign since, for example, input transistor feedback must now be referenced to ground and a new current path to the negative rail has to be established (resistor or preferably CCS for best PSRR). (Which, incidentally, is where the original Leach preamp design goofed, giving essentially no -PSRR as a result.) Ditto for the output stage.

So much for today's novel...

Last edited by sgrossklass; 12th June 2015 at 10:15 PM.
  Reply With Quote
Old 13th June 2015, 02:25 PM   #5
00940 is offline 00940  Belgium
diyAudio Member
Join Date: Sep 2003
Location: Paris
First point: where in the world are you ? Parts availability and prices depend a lot on this... What is your budget btw ?

Second point: for which headphones are you designing your amp ? It is much easier, in particularly with tubes, to build an amp according to the headphones you want to use than a jack of all trades.

Third point: I wouldn't bother redesigning a pcb for the pupdac, especially not one you could etch yourself. Dual layer pcb from China are so cheap that it is not worth sacrificing all the good things you get from proper pcb manufacturing.

Fourth point: indeed pcm270* are the only usb receivers you can get as parts. And indeed, they're limited to 16bits. I don't think that's a problem for audio reproduction (it could be for audio recording and production). If you insist on building something yourself rather than getting an already built dac (or simply getting a pupdac kit), maybe have a look at this (go directly to post 5).
  Reply With Quote


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
Music Angel Yayi USB Tube amp - 6N3 to 6P1 Push Pull john65b Swap Meet 2 7th July 2015 10:41 PM
tube-amplifier-shaped-usb-speaker trigger Tubes / Valves 11 23rd January 2010 08:55 AM
USB tube DAC£¬High-quality£¬Perfect design Jacky Ling Vendor's Bazaar 12 8th August 2009 08:44 AM
Built the DDDAC MKII USB audio card, modified it to feed SRPP tube amp line stage wa2ise Tubes / Valves 0 8th June 2008 03:21 AM
USB powered NOS USB DAC with built-in headphone amp dsavitsk Digital Source 15 10th August 2006 06:11 AM

New To Site? Need Help?

All times are GMT. The time now is 09:13 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2