Sauermann Amplifier

[gsbrva]

If you call this math, then I don't see where it's wrong.
If this isn't math at all, it's also no wrong math.

Thanks for the explaination again.  I think my confusion came from expecting an advantage from a more precise interpolation (out of band of course).

For the example I asked about (two adjacent samples differing by one LSB), 4 bits at 16x upsampling does not give the same shape you could get with more bits(before analog filtering).  I was thinking if you used more than 4 bits that the amplitude of any sampling artifacts should go down and the effective digital filter frequency should go up.  4 bits seems a rather arbitrary choice.  Why not 3 or 5 or 8?  You seem to already have proved that x16 sounds better than x8, so why not go for even more smoothing?

It all goes back to my inclination to preserve as many bits as possible for "something".  I'm not sure what the use would or should be and you make a pretty convincing argument that there is no reason to avoid use of bits for digital volume control.

I can see that your design choices for XX and the NOS1 dac are mutually supporting.  I respect that thinking.  It seems to me that your plan is based around attenuating the "loud" rather than boosting the "soft" recordings.

Would it be an equally valid choice to load the PCM1704 for linear full scale, implement lossless digital "gain" and use more bits for upsampling?  That would preserve the minimum parts count you have achieved.  I realize the gain question is tricky.  I guess it all depends on the usefulness of extra bits.

Dunno, just thinkin.
Greg   

[PeterSt]

The digital gain is already in there (auto - and limited to what it can take).
Sorry ...




PS: But max what I ever encountered was 4.5dBFS (possible boost).

[manisandher]

I never used XX's built-in vol control for a long, long time. But (quite a while ago now) I did a comparison of the following three different attenuation techniques:
- XX's digital vol
- Audio Synthesis Pro Passion passive (Teflon insulated high purity silver conductors and precision bulk-foil Vishay resistors) with a very short IC to power amp
- Pass Labs X1 preamp (certainly not the last word in preamps but considered good value for the £5K I paid for it ~12 years ago)

I still have the Pro Passion and the X1 and could re-conduct the comparison if necessary. I can't find the thread where I posted my findings, but I do remember my ranking. To my utter dismay (as a strong analogue attenuation advocate up to that point) it was:
1st) XX digital vol control
2nd) Audio Synthesis Pro Passion
3rd) Pass Labs X1

IIRC, the Pro Passion was very close to XX's vol control, but lost some of the dynamics. The X1, even with it's pretty advanced vol control was waaaay behind the other two.

For a discussion I'm having with Gerd right now I thought I'd open the ProPassion passive attenuators up to see what's inside. I bought these over ten years ago and paid ~£700 back then for a fully balanced setup. But from the picture (only one channel shown) it looks like I was short-changed - they don't look 'fully balanced' to me - but anyone, please correct me if I'm wrong. Maybe this explains why I prefer the sound of XX's internal digital volume control?

Mani.

[PeterSt]

Hi Mani,

Before I start thinking about what I actually see - how can it NOT be balanced while you will be having some balanced interlink connectors in it, them ending up in a balanced amplifier ?

The only thing I can think of is that in that case the sound is 6dB less without you knowing it.

But further ? You will recall how you measured DC Offset from the NOS1 (the version without the metering). Well, find yourself a 50 Hz test signal (but 200 or so will still work, 1000 I'm not sure) and do the same at the end of your balanced interlinks. One change : set the Voltmeter to AC;
With your posted picture I indeed imagine one of the pins not showing a voltage ... (and notice that what you see should be reversed for plus and minus on the both pins)

Peter (hoping he did not overlook something or made a mistake otherwise)

[PeterSt]

Start laughing - I am just trying to be creative ...

http://www.audiosynthesis.co.uk/propassion.htm

From in there, and it seems to be a key phrase :

The original and best selling PASSION controller now has a one input dual-mono partner!

If that means that it physically can have one XLR input (like from one channel) but it outputs in "dual mono" up to doing that by means of one XLR connector not changing the phase underway (all far sought !!), you'd have a balanced control.
Use two of these boxes for balanced stereo.

But I am over-creative here.
Peter

[manisandher]

Haha, I think they added the 'dual-mono' line when I ordered mine. I was originally using them in a multi-amp system and I needed to match the gain betweeen the sub-amps and the rest of the system. But I needed one attenuator on each side, so they made a pair for me (see attached).

The more interesting line for me is "It can be supplied with top quality WBT RCA sockets for unbalanced use or with Neutrik XLR plugs, sockets and additional attenuators for fully balanced operation."

So where are my additional attenuators for fully balanced operation? Surely there should be two rows of resisters per channel, one for each phase, no?

Mani.

[PeterSt]

I was looking at that line too, but I envisioned that since you have XLR connectors that extra set of resistors was virtually assembled in your case. Virtually, because apparently that is not the case.

But now assumed that one of your pins will be dead, there's not much balanced operation in order. So, no noise rejection plus an unnecessary low output (6dB less).

Or ?

[Nick]

Mani hi,

I have a ProPassion silver of the same era, mine is a single ended version (RCA inputs and outputs). The shalco switch internally has two layers in mine (left and right but otherwise looks identical to your picture. I cannot see how the version you have could be balanced. Room for improvement there if you are looking for balanced end to end connectivity.

Best,

Nick.

[CoenP]

Hi,

This looks like a fully balanced control to me. Two 3k5 resistors from input to output for each phase and a "pinch" resistor on the (very nice) switch.

Off course neither the input nor the output impedance is constant. It varies with each volume setting. But that being constant was not claimed i presume.

Regards, Coen

[manisandher]

Well, I always was terrible at electronics. So could someone explain how a single 'pinch' resistor could be used for both phases?

Mani.

[CoenP]

Well,

I'll give it a shot.

The pinch resistor kind of short circuits at the output.

 When volume is completely down, plus and minus are tied together at the output. Input "sees" 3k5 to ground on each phase, impedance at the output is zero ohm.
When completely up, the driving device (DAC) "sees" the 3k5 in series with the input impedance of the amplifier (600 ohm per phase?). This will allready act as a voltage devider.
In between the "pinch" resistor is parallell to the 2x amplifier impedance. The smaller the resistor value the more both phases are short circuited, the more the signal is attennuated. The same thing but in different wordings is saying that the more attennuation, the more the plus and minus are summed at the output pins. Complete summing leads to zero signal at the pins!

Attenuation range is probably limited and the pinch resistors will be small in value (which is advantagious since these are dominant for the output impedance of the attennuator).

I can't think of a simpler balanced attennuation device.

Regards, Coen

P.s in the single ended community this circuit is known as a "shunt" attennuator with a fixed value resistor in series with the signal. (my personal preferred way of building an attennuator dispite some obvious theoretical drawbacks).

[manisandher]

Hi Coen, thanks for the explanation. Really appreciated, and I think I get it. So apart from a varying output impedance, this seems like a pretty neat little 'trick'.

I would have used these passives in the upper 1/3 of their range (from -18dB to -6dB) and maybe here the output impedance of the passives and the input impedance of the poweramp played a part in the difference in sound I was hearing. I can't even remember which poweramp I originally used but whichever it was, maybe its input impedance was lower than ideal for the passives. But I do know that all interconnects were 1m or less in length.

In any event, the digital volume control in XX sounded the best... for whatever reason.

Mani.

[PeterSt]

[this post has been on the screen for the whole day because I hardly dared to post it ... ]

In any event, the digital volume control in XX sounded the best... for whatever reason.

Answer : Because the signal now goes through that "fine" 3K5 resistor(s) ...

All I did before was concentrating on all the other resistors, thinking "would those be the best for audio ?" - and next I never saw that this other XLR would be the output. Or actually, that this other again would be the input.

So, the resistors on the pot won't matter a thing (they can be lousy ones), but that 3K5 does. Now all is relative ...

Suppose you are looking at a pre-amplifier, and this would be part of it. Envision a preamp built in discrete fashion. There will be these parts (resistors) here and there and everywhere. This is one more. They all could be of the best type, and this is just one more of it. Now the theoretical (or mathematical) problem : In the NOS1 is exactly one resistor like this (per phase and channel) and it should be the best I could find (for thermical noise, deviation). I assume it is better than we see here, but let's say it is of equal quality.
... Now you added as much "distortion" as there already was. So, twice as much now.
Keep in mind : all is relative.

Additionally that 3K5 resistor can't be the best for passing on the higher frequencies.

I am not sure anymore whether I tried exactly this setup in-DAC but I imagine not because I don't recall using one (mono) pot only for one channel. But then I don't think this is a shunting attenuator (as you seem to suggest Coen) because I see no ground involved. So, just (slowly) shortcutting the both phases is what's happening here (but correct me where you can !).
Heck, I just don't know anymore in what type of (volume) setup I applied this kind of thing; maybe it was behind a transformer gain and that sure didn't work because of the transformers. Point is : I recall this as the most easy "attenuation" which I didn't see myself at first, but it was not this situation. IOW ... maybe I should try this now ...
I just don't see how it can harm the signal and all I need to do is change that one resistor into a somewhat higher value so the one resistor there just remains that ...

[CoenP]

Hi,

On a more technical level, the currents from the inpot voltage are flowing to both phases to the outputs because of the pinch resistor.

So current on the plus flows also through the pinch to the minus output (and input throught the 3k5) and current on the minus via the pinch to the plus output (and input). These currents are summed at the output, hence the blending of positive and negative currents (iaw attennuation of the phases).

Note: to clarify the above, this is when either the plus or minus has the positive voltage. The increased current through the series resistor will cause the attenuative voltage drop at the amplifier input impedance. On the negative leg this positive current will cancel some of the voltage developed across the amplifier input impedance.

No groud is involved. It is not nessessary and is a lot harder to make symmetric (maintaining the balanced nature of the signal).

Since signal current flows through the pinch resistor, i would presume that quality would matter here. Such is also my experience with the se shunt (to ground) attennuator. Since the current will be less than through the series resitor(s) one might argue that is is less important (but not unimportant).

Imho you are trading of dac liniarity for resistor noise, parasitics, hf cutoff and thermal issues. The constant low output impedance of the dac also will do no harm. Furthermore resitors do distort measurably, as you can ee one of the linear audio magazins.

Regards, Coen

[PeterSt]

Since signal current flows through the pinch resistor, i would presume that quality would matter here. Such is also my experience with the se shunt (to ground) attennuator. Since the current will be less than through the series resitor(s) one might argue that is is less important (but not unimportant).

Thanks Coen.
This is what I have been looking into yesterday;
So, indeed the current will flow through the pinch resistor, but IMO that can't do much. Where it sure can matter is at the other side : the output which is feeding this. This is the DAC in our case, the NOS1 in particular. So, if you envision something like a gain stage in there, now current flows through there. Can still be harmless, if only the device(s) in question can have it. And, with a proper balance of the series resistor and until where attanuation is allowed, it looks like it can be totally harmless. Key here is this attenuation, which I personally already never thought to be from full to zero. So, in the NOS1's particular case I'd say that something like 20-24dB of attenuation is just perfect; it will allow to dial in the optimum THD figures coming from the digital attenuation as discussed heavily earlier on in this topic. And otherwise, of course, something like 24dB will be a very convient range of operation.

But it has to be tested (measured). The shunting to ground at least never worked, I think partly (?) because of influencing the whole grounding scheme with it. This, while the whole lot is a pure differential setup. I think that by attenuating one of the phases only the slightiest off, the whole balance (zero point) is drawn to somewhere, and this with four of these points that can go off (2 phases per channel). All this stuff is super tricky because all *depends* on its virtual ground, while there's also DC corrective stuff which could be called "mechanical". Not that I saw it, but it can easily oscillate I think, and I wonder how that's seen on the analyser when the frequency is sufficiently high. Think about the data itself too (the voltage of it) and what happens when one phase is exponentially higher than the other (I'm not even sure whether this is linear or not) when the level of the music changes. So, already that can bring things out of balance with nasty results. Just thinking out loud.
And for that matter, if you'd look at the output level of the both channels of the NOS1, you'd see that both will not be 100% equal. It may differ 0.02dB or so. This not only can't be achieved in equal fashion, but also not consistently per DAC. Thus, "nothing in the signal path" is quite true, but from the beginning to the end there's a lot of circuitry per phase (starting at the transformers). So, we can't say there's nothing on the PSU for components, and that too is differentially setup (with 0.1% resistors, but still). On that matter we'd be better off if the differential signal would have been created with some OpAmps, but *that* wouldn't comply much to "nothing in the signal path" ...

But this is why I now like this means of balanced attenuation. It just can't do something else to the plus opposed to the minus, right ? But keep that current under control. It just should work, if it's only dedicated to the outputting source. And this, of course, is under my control here.
So I guess I will have *another* round of attempting an analogue volume control. But later.

Thanks again,
Peter