• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Summing differential output to single ended with tubes?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
So, like a total tube nut, I'm working up a output stage for a dual mono PCM1794 DAC, using tubes. The way that these chips output signal is through a current source (grossly oversimplified, but close enough for basic details) and then they develop the signal voltage by varying that current, which I will be picking up off of a resistor feeding each grid of a long tailed pair, sitting on a CCS. So good so far, right? I plan on a buffered output that is basically the Heretical linestage by Sy, with a DC protection, servo, and startup mute/dethumper circuit. All tubes will be Russian 6N16B.

My dilemma, is I'm not certain exactly how to go about summing two differential outputs into a single ended output in the most elegant fashion.

Here's a typical circuit, just I'm using a LTP instead of opamps-



And the recommended summing circuit for dual mono use, single chip per channel, picture another LTP instead of the XLR type of output-




So, each pair of outputs per chip will each feed into a LTP, and then both of these will feed into either another long tailed pair, which may provide too much gain, or into a buffer circuit of some sort. Since I own zero balanced gear, I don't need a balanced output, so I need to sum the output into a single output somehow.

I thought of using a transformer, but good ones are too expensive for what they are. I have a pair of Edcor WSM on hand but they are earmarked for another project, and might not be up to par for such a circuit anyways.

Is this as simple as just taking the signal off of a single side of a the final LTP circuit? Will taking signal off a single side hurt the balance?


I was looking for some sort of balanced input buffer stage, but other than John broskies unbalancer circuit, I didn't see much... I also don't like the idea of having the output floating at half B+ through a capacitor, and would rather have the output at ground potential via the Heretical output being at ground...


If I'll be fine pulling the input to the buffer from a single side then I'll just go with the Heretical. I suppose I could wire up two buffers per channel and toss on a pair of XLR jacks, just to be one of the cool kids. It would give me options to build fully differential nonsense later should the ideas pop up.
 

PRR

Member
Joined 2003
Paid Member
....varying that current, which I will be picking up off of a resistor.....

The old-old DACs expected a ZERO impedance load. If the voltage swung, linearity got bad.

I know recent ones have relaxed that; but it still needs to be a low-low voltage. Low enough relative to grid-hiss to limit dynamic range?

The chip implementation is usually a summing node, NFB to input. You can try this on a tube. But there are FAST transitions here, and it is difficult to rig a tube to follow clock pulses and stay linear.

Given two inverting summing stages, balun is easy. Feed the output of one to the input of the other.
 
From my own testing in the past, and a few others writeups, the linearity is not affected appreciably at all up to around a volt or so, and even then it's negligible. I'm shooting for around 300mV across the I/V resistor or so, maybe up to 500mV depending on how things go. The PCM1794 is pretty friendly to passive I/V conversion. The 6N16B tubes I'm planning on using are very nice performers, and have pretty good noise performance. I plan on DC heaters (maybe CCS fed all in a string?) And it will be in a metal case with attention to lead dress and compactness.

Last time I did this it worked out fantastic with single chip conversion, using the 6CG7 as a LTP, and I summed the signal with OPA2132 opamps, but I wanted to do it all tube this time, and dual mono.
 
Last edited:
Unbalancer%20Schematic.png

Like this?
 

PRR

Member
Joined 2003
Paid Member
This page shows passive 270r or 130r output resistors, and the builder is very happy.

4mA peak in 270r is 1V peak which would indeed be "large" compared to grid hiss. (Perhaps 120dB S/N, less than this DAC promises but more than you can use in a practical playback system.)
 
Hi,
I'll be following this thread as I like pcm1794a and 6n16b. Actually, what I use now is pcm1792a, the software control version, on a asus stx ii sound card. One chip in stereo mode though. I have taken the output directly from the chip via a 2 ohm I/V resistor to feed the tubes grid. First schematic shows the current revision of the tube output stage. Plenty of gain allowed to apply nfb but I think it does breathing oscillation now. I can't measure it. It didn't start like this and have some more ideas for it - second schematic. A CCS on top of the cathode followers - J. Broskie's trick - hopefully will make the buffer stage something like a PP OPT, summing the differential signal while retaining balanced output. Perhaps nfb won't be necessary then. Third schematic shows the CCS I'm planning to use, found here http://waltjung.org/PDFs/Sources_101_P2.pdf
 

Attachments

  • Preamp schematic.jpg
    Preamp schematic.jpg
    299.2 KB · Views: 394
  • IV.JPG
    IV.JPG
    294.5 KB · Views: 373
  • CCS.JPG
    CCS.JPG
    121.9 KB · Views: 162
Neat. I was thinking of throwing in some feedback, but the more I look, the more I think I want local loops around each stage. As of right now I'm thinking of just having individual cathode resistors under each cathode on the LTP (on top of the CCS) to reduce gain and give a little more linearity, and then I'll just run it straight to the Heretical buffers. Assuming I use something like 33R I/V resistors, the PCM1794 will give 7.8mA peak with the 10k ref resistor... I should get 0.2574V at each grid, which assuming a gain of 25 max for the 6N16B, will give me 12.5x per triode, which gives 3.2175v peak per phase, 6.435v peak-to-peak/4.5Vrms... I can reduce or raise the value of the I/V resistors as needed to change output level. Assuming the cathode degeneration and real world losses, even if we get that cut in half I'll still be happy. If I add an additional LTP before the buffer I would think that global NFB would be necessary to keep the gain from being excessive...

As it stands right now, I'm looking at (per chip) two differential pairs, and a pair of cathode followers per channel, for three bottles per side. This will give balanced output, and I can always add an RCA jack for each output for single ended use, but I have a spare set of XLR jacks to give it balanced output capability. I have a volume control LCDuino project going too, and I think that I'll put the Edcors in there with a motor pot for both volume control and Balanced to SE signal conversion options.
 
Last edited:
I'm pencilling in 47R as the max I'd like to use for now, back when had access to nice audio test equipment I found that once you go much higher you start to get a smidge more (But still pretty negligible) THD, but it was still exceedingly low, and not much worse than an opamp implementation would run on my eval board at the time, but all my testing was single chip stereo, so it should be even better going dual mono.
 

PRR

Member
Joined 2003
Paid Member
Speaking of grid hiss, how would I be able to ballpark a realistic noise floor for this setup?

Pencil 2 microVolts per tube over the audio band. (RMS summing: two tubes makes 2.8uV.)

It varies A LOT with the tube. You may find that in a crate of 100, a couple touch 1.5uV, most run 1.6-2uV, and a dozen run 3uV to 100uV.

It varies with current, except high current reduces hiss in the KHz range but typically increases 1/f hiss (random rumble) below (even above) 1KHz.

There is data saying that a "good" tube at well selected operating point can approach 0.5uV over the audio band.

The rule in recording is "don't bet on it!". Even if a preamp shows 0.5uV on the bench, tubes drift, sometimes quite a bit. Tube mike preamps aimed for low hiss almost invariably take a 1:5 to 1:10 transformer from 100r mike to grid.
 
I thought of trying out some microphone transformers, since this seems to be a similar situation where they may work well... Looks like Edcor has some affordable options that may be suitable for much less than all the Gucci types that you see the Golden Ears pushing.

Using such a low mu tube it looks like the noise won't be too much to worry about, especially if I err on the higher side for my I/V resistance to get a bit higher voltage to swamp it out.



Well, started up on the boards. Doing each pair of LTP/CCS circuits on a single board, and a pair of cathode follower/CCS each on single boards as well, for two boards per channel in addition to the DAC and PSU boards.

attachment.php


Gotta order some more caps and put in headers for off-board connections (PSU, signal) but the buffer modules are pretty much done.
 

Attachments

  • IMG_20180213_204623.jpg
    IMG_20180213_204623.jpg
    807.2 KB · Views: 341
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.