A bespoke audiophile tube DAC: measurements and discussions. And politics season.

As per the title and image, this post will be a nebulous one because we're not looking at an easily-accessible mass produced product. 🤔

The reason is because we're looking at what appears to be a one-of-a-kind device that's not advertised, the company making it doesn't have any web presence with only a handful of posts I can find over the decades referencing the company/person who makes what appears to be typically tube-based products.

I came across this unit from an audiophile friend who commissioned this designer/maker for a good sounding DAC with tubes. This was what he got; the first iteration at least.

As you can see, it's a 2-box unit with identical metal chassis except for some cut-out differences for the volume control in the DAC section, a 6.35mm headphone jack (not pretty!), and digital input selector (coax, optical, USB). We can see the tubes sticking up from the DAC box. The box below is the (presumably linear) power supply with its own power button and a "dummy" knob that I guess is just part of the generic case.

I was told that the DAC chip inside is the ESS ES9038Q2M which we've explored previously like here and here.

The boxes are metal and of basically identical size: ~7.5" x 13" x 3". The DAC box is quite light, about 1.95kg, and the power supply weighs a bit more at 2.5kg.

Let's have a closer look at the tubes:

Since it's not mine, I did not open it up to have a look at the internal workmanship. On the top, we can see the 3 vacuum tubes - a single 12X4 rectifier and 7119 dual-triodes presumably for the analog output stage with tube shields that could help block hum and interference.

We can see the rear digital inputs - S/PDIF Coax, TosLink, and USB-B. I wondered initially if it's based on a generic CM6631A board. RCA output level is labeled as "1.8V" and impedance labeled as "86kΩ" which is extremely high for output impedance of line-level devices like a DAC or CD player! No buffer stage? Normal CD-player RCA out should be something like 100-500Ω!

The dual 7119 tubes are GE-branded and the 12X4 is branded Sylvania.

Nothing much to see looking behind the power supply box. DIN cable connector and N. American IEC A/C plug.

With 25" power cable.

Cosmetically, this looks very DIY. I guess beauty is only skin-deep if the device performs very well and makes sweet music.

[As a group, of course many audiophiles do care about appearances and we've touched upon that.]

Let's get to the measurements...

I. Oscilloscope & Filter:

Oh my, right from the start, we can see that there's a problem! Into a 1MΩ digital oscilloscope (reasonable high-impedance match), the RCA visibly clips at 0dBFS all the way down until -7dBFS at around 1.9Vrms out. No kidding, on the rear it says the output level should be 1.8Vrms!

This might not be an issue if not for the fact that the front volume control, even when turned lower, does not fix this clipping above -7dBFS! This means that we can't just feed bit-perfect data into this without the top 6dB of data always clipping. Even with high dynamic range, low-average level, classical albums for example, every time we have loud transients, this is a problem, and easily audible.

Because of the clipping, the impulse response looks... clipped. We can see that the filter used is a sharp roll-off linear filter. Absolute polarity is maintained.

Let's have a peek at the frequency response and phase:

Reasonable flat response with what looks like subtle upper frequency rippling. Nothing worrisome on the phase graph, generally floating around 0°.

II. Resolution

I fed the DAC's RCA out into my E1DA Cosmos Scaler which has 100kΩ input impedance. Unfortunately this is just a little higher than the 86kΩ output impedance on the DAC as labelled! While 100kΩ is a reasonable value for most consumer preamps, the two similar impedances will form a voltage divider so the ADC will see roughly -5.5dB lower level which will add noise, hum and distortion. I kept the cable between the DAC and measurement system as short as I could (6") to minimize issues. So, while the distortions and noise would be higher than what you would find if we plugged this DAC into a 1MΩ pre-amp (which is uncommon!), the results will still show the gross limitations in the measurements.

I let the DAC, power supply, and tubes warm up for ~1 hour before running these measurements. Vacuum tube shields on.

Because of the high-level clipping, I could not properly run the RightMark test. So let's just do some FFTs and stepped sines using REW.

First, here are some 1kHz sine FFT:

As shown above with the visible clipping on the oscilloscope, I had to drop the input level to -7dBFS to escape clipping. At -7dBFS, the THD+N is -56dB into the 100kΩ load which is only about 10-bits of digital resolution.

It made no difference if I used the Coax, TosLink, or USB. Both the TosLink and Coax are capable of 24/192 input but the USB is a really old-skool "USB to IIS Out" interface which is locked at 16-bits and 48kHz! This suggests to me then that the USB controller must be one of the old ones like the PCM2706 maybe, not even the CM6631A I speculated about above!

Let's have a peek at the distortion amounts across output levels:

The DAC is incapable of less than 0.1% THD+N.

[Even the CheapDAC'22 gave us about -70dB (0.03%) at 1kHz - just sayin'.]

Here's the linearity plot down to -100dBFS:

That's pretty ugly clipping at the upper end and the signal deviates beyond 0.5dB by -80dB which is high compared to most modern DACs. Notice that this is about the same number as the FiiO K11 R2R DAC measured recently but without that top end clipping; so basically this tube design has made the modern sigma-delta ES9038Q2M chip perform at about the same linearity equivalent to an R-2R converter using 0.1% discrete resistors at low levels. 🤔

For completeness, let's get a reading on the non-clipped Triple-Tone TD+N:

Triple-Tone TD+N = -53dB. Notice some L-R channel imbalance of ~0.2dB.

III. Jitter

Given the issues above, I know it's silly to be looking at jitter. But here's the 16-bit J-Test which was identical for TosLink and Coax:

And here's the USB input which is fixed at 48kHz so I used the 24-bit J-Test even though it is only capable of 16-bits:

All about the same with -80dB sidebands close to the primary signal.

For those who enjoy looking at tubes. 😉

IV. Subjective and Summary

I trust that we've seen enough when it comes to measurements.

No need to beat around the bush, audiophiles. At this stage, this is a very poor DAC. With such strong imperfections, we should be able to hear what the measurements are showing us. Indeed, with all that clipping near peaks, distortion is easily heard and I did not have fun during my subjective listening sessions even before putting this on the testbench.

Here are a couple of AMPT recordings - listen for yourself:

Bespoke Tube DAC AMPT (bit-perfect playback)

Bespoke Tube DAC AMPT (non-clipping -7dBFS peak)

24/96 capture through RME ADI-2 Pro ADC

With the bit-perfect playback, you should be able to hear the distortion with loud transients in the classical sample, and it gets really bad with dynamically compressed content like the pop tracks (last two Benjamin Clementine and Dua Lipa samples). Notice that when I applied -7dB digital volume reduction, the clipping has been avoided and it sounds much better even though noise floor and distortion are higher. Despite the resolution being only ~10-bits, I think some people who don't mind or like the extra euphonophilic "warmth" of high non-linear distortion could still enjoy this presentation even though it's not really "high-fidelity" reproduction.

Obviously, if this were my DAC, I'd be highly unimpressed especially knowing the kind of resolution the ESS ES9038Q2M should be able to produce (eg. even the inexpensive Topping D10s), among other issues I would not expect with modern DACs.

While I have no expectation of great high-resolution performance through tubes, I've advised my friend to go back to the maker to fix the clipping issue above -7dBFS at the very least.

Another thing is the very high output impedance. If this specification is accurate, the DAC demands that the pre-amp or amp it's connected to should have something like 1MΩ input impedance ("Hi-Z inputs")! Very few (pre)amps would be able to satisfy this. Presumably a buffer at the output would be the most reliable fix and something the DAC maker should consider implementing depending on what preamp/amp it's going to be paired with I suppose.

Also, that USB 16/48 limit is primitive in 2026 especially when the S/PDIF inputs can accept up to 24/192; even an inexpensive CM6631A I²S board can handle 24/192 USB 2.0 (we talked about the CM6631 way back in 2013!).

As for the headphone output, it's not a proper headphone amp with low impedance. Probably just the line out in the form of a headphone jack bypassing the front volume knob.

Estimated output impedance @1kHz ~7Ω. Obviously high for
proper headphone out!

In my opinion, regardless of what others may testify about the magical sound quality of this or that maker of exotic hi-fi gear, it's essential to run measurements to ensure that the device achieves the level of quality you're expecting. When it comes to audio gear, there is obviously no magic. In the past, I've measured another device that was tube-modded for another friend at significant expense and I also found it lacking.

I believe the owner of this box will be able to bring this DAC back for changes to make sure the result is satisfactory. Maybe I'll get a chance to have another listen/look at the next iteration.

[Finally, a reminder that quality control and safety standards are clearly issues to consider when using products that have not been officially tested or come from very small companies. A practical consideration is to make sure your home insurance will cover for negligence and DIY mistakes (see discussions like this) if investigation for a fire reveals that it's because of "exotic" products.]

--------------------

Politics season...

While I generally steer away from saying too much about politics, it really has been quite interesting times with a barrage of elections locally and abroad.

Here in British Columbia, we just had our municipal elections which saw major regime change in Vancouver with a new left-leaning city hall. Earlier this week, the province of Alberta clearly decided to stay in Canada (silly to think otherwise!).

Today, as I post this article, BC will be voting for the provincial leadership in a snap election with the conservative right party possibly making gains. I'll be getting some popcorn ready for this evening's coverage.

And of course, the mother-of-2026 elections, November 3rd is the US midterm just around the corner.

Permit me to recycle a 'cartoon' image I posted in April 2025, shortly after the tariffs were brought in:

Image made before the East Wing of the White House got demolished in October 2025. Sadly, it appears nobody rational has been able to reach that man with the executive orders pen on the golden throne.

Recently, I've been pondering on a word I rarely ever use; evil. No, not the bloody horror-evil of slasher movies as we get closer to Halloween, or even the spiritual possession evil of demonic spirits. Evil actions/inactions in today's world I think are more the serious, culpable wrongdoings that inflict terrible harm, sometimes done with malice, and other times, just callous indifference to the suffering of others.

[A few examples for posterity of the horror, ridiculousness, and incompetence: here, here, here, here, here, here, here, here, here, here, here, here, here, here, here, here, here. Makes debates about Trump selling Bibles quaint in comparison, just a couple of years ago.]

In all those ways, I think the current American Republican administration under Trump, his loyal lackies, enabling aides, and ultimately those who voted for him will at some point need to answer for the forces they have unleashed in what otherwise should be a leading, respectable, freedom-loving, democratic nation. We are certainly not living in any golden age of America; rather, it's a reality-TV-inspired, corrupt, manipulative, truthless, idiocracy.

Of course that's not to say others are without blame on the Democrats' side - as bad as this current government is, we obviously must never idealize the other side! However, the magnitude of "bad" here is phenomenally transparent.

While the local elections are of great importance, November 3rd will be very interesting for us around the world as we watch the results streaming in. Come on Americans, I'm sure there's a much brighter path to go than following an obviously unintelligent, obliviously narcissistic, perfidious, poorly spoken, possibly delusional, lascivious, dictator-wanna-be octogenarian and the callous incompetence he has collected around him. Apparently he's on the ballot - "just one more time" he says.

[Speaking of an unintelligent man, I seriously doubt Trump would even get 30/30 on the MoCA as he claims if administered properly. I'd wager 24 points tops using one of the variants he hasn't seen already, on a good day. Stuff like this among the countless other nonsense should forever be stamped in the historical records to remind us of one of the lowest points in the modern history of elected officials in western democracies. 

BTW, it's actually quite normal for 80-year-old men to score less than 30, but with Trump, it always has to be "the best", "perfect", and "nobody's ever seen this before"; so dumb! 🙄🤥]

I don't think there's any value left in this Trump reality-TV series - entertainment or otherwise. Hopefully we've reached the beginning of the end of this season; alas, with two more years to play out.

--------------------

Let's end with something relaxing and light; here's Dave's True Story's "Dog's Life" from their album Nature (2005, DR10):

Audiophiles, I hope you're all doing well, and enjoying the music this autumn!