Metrum Acoustics Octave: A NOS digital filter-less DAC

[K3cT]

I haven't seen this DAC being talked about here so I will just leave this here. Designs-wise, the DAC interesting because apparently it uses 4 "industrial-grade" non-oversampling DACs (2 per channel) whatever they are and it doesn't have a digital filter.

Price-wise, it seems reasonable enough. The most expensive model with black anodized case and external PSU will set you back €710 or a little less under $1000. Shame it's still single-ended and has no built-in USB.

Here is the home-site and a white paper posted there detailing this NOS digital filter-less concept. I wonder what do our resident DAC experts like Filburt and Colin have to say about this as the concept is quite intriguing.

[NekoAudio]

"Industrial grade" might mean they are designed for more extreme temperatures. That's sometimes what the term means when applied to some other electronic components. And so it is safe to use them in a car or other environments where a "regular" component would fail.

[RudeWolf]

The Legend from 6mmons review:

"For the last 20 years we have been manufacturing electrostatic transducers and related electronics. We also design transducers and systems for non-destructive ultrasonic investigations. Based on this background we subsequently attempted to design a DAC which would not be based on the regular AKM, Burr Brown, TI, Crystal and Wolfson chips which are ubiquitous in consumer audio. Instead we wanted an ultra high-speed part such as you'd find in industrial applications. After many years of experimentation we finally identified an extremely fast chip that's useable for 16 or 24-bit audio but handles sampling rates up to 15 Megahertz. Of course some glue logic was required to match this chip to the standard audio formats. Due to the very low glitch energy of our resistor ladder network, we need no digital filter or oversampling. Our machines are very deliberately without frills. Paralleling four industrial DACs per channel with integral voltage outputs improves low-level information and noise floor for the digital data which enter via the receiver chip. Because our DAC chips are relatively expensive and usually only seen in fast data acquisition systems for industrial use, we kept functionality and cosmetics simple. We avoided multiple i/o ports and a hefty chassis to maintain instead an attractive price. One obvious advantage of our high-speed chips is that they exhibit excellent impulse response without the pre/post ringing so common in today's oversampling converters. To avoid mutual interference between digital and analog data we use a six-layer printed circuit board. After further developments we eventually asked certain reviewers for performance feedback since they had broad experience with standard audio converters across many different price points. They reported back that our €350 Quad DAC competes up to €2000. We have now managed to duplicate this unfair advantage with our Octave version. It sells for just shy of €700 but performs up to €5000. We ended up with a total of three production models, the Duo (one chip per channel), the Quad (2 chips per channel) and the Octave (4 chips per channel). It is hard for me to describe the performance of these machines in English but we already have a rave Dutch review. This prompted many sales in the Netherlands. Due to exposure in the international audio forums global sales inquiries have followed. Because of our unusual price-to-performance ratio and the uncommon core parts one does not usually come across in home audio, would you be interested in a review?"

[Beefy]

I think industrial grade actually refers to the fact that these are housed in a Hammond case...

[Currawong]

There's an interesting thread on this on stereo.net.au. One guy likes it using software up-sampling. I'm quite tempted to buy one out of sheer curiosity as I have enough digital gear to experiment with it. Their site shows a perfect square wave at 1kHz, but I really would love to see if it could re-produce a sine wave perfectly at 10k, or 20k. I'd be surprised if it could.

[ecclesand]

Needs moar caps!

[Dreadhead]

There's an interesting thread on this on stereo.net.au. One guy likes it using software up-sampling. I'm quite tempted to buy one out of sheer curiosity as I have enough digital gear to experiment with it. Their site shows a perfect square wave at 1kHz, but I really would love to see if it could re-produce a sine wave perfectly at 10k, or 20k. I'd be surprised if it could.

If it can do a perfect square wave at 1k it most likely will have no problem with sine waves at 20+k because if it can get rid of the Gibbs phenomena it has a very wide bandwidth.

[arnaud]

So, does it mean this device is resampling heavily to the point that some brickwall filter is no longer necessary? What is the point of feeding it with software upsampled signal then?

Edit: ok, had to read twice to see it specifically mentions about non-oversampling process . There's got to be some resampling somewhere else I don't see the point of using very high speed converter? Also, is it really expected that a resistor ladder network can be so fine that you can simply ignore reconstruction filter?

Edited November 8, 2011 by arnaud

[Craig Sawyers]

I bought one of these around 10 days ago, based on Martin Colloms' recommendation in HiFi Critic, in which he billed it as a real giant-killer. Initial thoughts were exceptionally good, and as it warmed up progressively better. But I can't resist getting things right that are wrong. The first problem is common to most DAC's and CD transports - RCA connectors. I did a whole host of measurements using Time Domain Reflectometry, and basically putting a fast pulse into a 75 ohm coax terminated with an RCA, into an RCA socket, with a surface mount 75 ohm resistor tacked onto it - and it is a disaster. Massive reflections - and reflections add jitter.

The only way to do things *properly* is to maintain a clinical 75 ohm environment for the whole digital signal chain, and that means 75 ohm BNC connectors throughout. And most digital cables, if not terminated in RCA's are terminated in 50-ohm BNC's! In fact the only audio high-end 75 ohm BNC's are Oyaid. I just use regular clamp-on greenpar and RG302 teflon cored coax.

After doing that (and measuring the TDR response into the Octave) the DAC really started to sing.

Then I spotted the hokey little pulse transformer (visible in the picture above). This is a $2 part, a standard Murata ferrite cored pulse transformer. I replaced it with a far better (electrically) Lundahl LL1572, which uses an amorphous ribbon core. In the UK this is £35 - so not cheap.

After swapping transformers (invalidating the warranty of course) I can honestly say I have never heard a DAC sound that good regardless of cost. I've just ordered an Audio Note toroidal pulse transformer wound on a mumetal ribbon core, so it will be interesting if that improves it further, or otherwise.

I can't stop listening to it!

Craig

PS The DAC's are intersting. They are 16-pin devices (with the type number taken off, of course), so they are definitely not Burr Brown or any of the other audio DAC's, which have far more pins. So I think they must be DAC's intended for instrumentation or data acqisition applications. And they are supposedly R2R ladder DAC's too. I've done an initial trawl of manufacturer's data, but have not been able to find anything that looks like those.

[Pars]

PCM1702s and IIRC 1704s are 16 pin (DIP), but those appear to be SMD...

As for pulse transformers, did you think about the Newava S22083 or 22160, which seem to be the ones that Jocko recommends?

http://search.digikey.com/us/en/products/S22160/470-1004-ND/555462

Not silly $ either

I had noticed what you did regarding most pfeu pfeu BNC connectors being rhodium (insert material of your choice here) plated 50 ohm connectors...

[luvdunhill]

Did the board look hand soldered?

I would recommend the Scientific Conversion transformers. I even have one on hand i could sell you They measure insanely good and I'm using them on my ADC and they don't disappoint.

[Craig Sawyers]

PCM1702s and IIRC 1704s are 16 pin (DIP), but those appear to be SMD...

As for pulse transformers, did you think about the Newava S22083 or 22160, which seem to be the ones that Jocko recommends?

http://search.digike...-1004-ND/555462

Not silly $ either

I had noticed what you did regarding most pfeu pfeu BNC connectors being rhodium (insert material of your choice here) plated 50 ohm connectors...

Comparing the Newava and Lundahl pulse transformers, the interwinding capacitance is 0.5uH and 1.3uH respectivly, and the interwinding capacitance is the same at 15pF. A figure of merit is root(leakage inductance x interwinding capacitance), and this is proportional to the rise time. So the Newava should have a rise time 62% of the Lundahl - ie it is faster. But the primary inductance of the Newava is only 0.225mH, whereas the Lundahl is 40mH (180 times greater). The winding inductance is related to the droop in the top of the pulse, and smaller inductance gives more droop. Calculation for the Newava shows that the top of the pulse will droop to around 80% for a typical pulse length. The Lundahl has an unmeasurably small droop. Droop gives rise to jitter (because of the variable pulse length encoding of serial digital data) - so it is good to keep it as low as possible.

The low winding inductance of the Newava is all to do with use of a ferrite core, which has a relatively low permeability. The more expensive constructions use either spin melt amorphous ribbon, or mumetal, both of which have 200 to 500 times the permeability of most ferrites.

And yes - the 50 ohm BNC connectors is just plain stoopid on a 75 ohm system. If that was a RF power amplifier feed to a transmitting antenna, there would be so much backward wave that the amp would blow. And RCA! There are supposed 75 ohm RCA's (WBT do a cable RCA plug and maching chassis socket), but I remain skeptical until I measure the darned things. Might be able to actually do that during the next two weeks.

Now 75 ohm N-type! that would be the way to go with a clean sheet of paper!

Craig

[Craig Sawyers]

Did the board look hand soldered? I would recommend the Scientific Conversion transformers. I even have one on hand i could sell you They measure insanely good and I'm using them on my ADC and they don't disappoint.

Yes - the board is hand soldered. Nicely hand soldered.

I'll bite for the SC transformer (I've just been downloading and speed reading his AES papers; he knows what he's talking about for sure!) - I'll PM you.

[Pars]

Did the board look hand soldered?

I would recommend the Scientific Conversion transformers. I even have one on hand i could sell you They measure insanely good and I'm using them on my ADC and they don't disappoint.

Your measurements or theirs? I know you don't like Jocko much (and I'm not sure what I think), but SC was one that he always said lied thru their teeth on their datasheets...

http://www.diyhifi.org/forums/viewtopic.php?f=2&t=284&p=5709&hilit=Scientific+Conversions#p5709

[elnero]

I bought one of these around 10 days ago, based on Martin Colloms' recommendation in HiFi Critic, in which he billed it as a real giant-killer. Initial thoughts were exceptionally good, and as it warmed up progressively better.

I'm curious, what's the rest of your setup?

[Craig Sawyers]

I'm curious, what's the rest of your setup?

A hot-rodded and rather ancient CD player, converted to a transport. Multiple mains tansformers, a regulator glued to the top of every chip, clock board (my design) with OPA623 S/PDIF driver (1ns rise time), and focus and tracking servos run from a dedicated supply. Metrum Octave, Audio Research LS3 pre, Audio Research D125 power, Quad ESL57's

[elnero]

A hot-rodded and rather ancient CD player, converted to a transport. Multiple mains tansformers, a regulator glued to the top of every chip, clock board (my design) with OPA623 S/PDIF driver (1ns rise time), and focus and tracking servos run from a dedicated supply. Metrum Octave, Audio Research LS3 pre, Audio Research D125 power, Quad ESL57's

Very nice!!!

[Pars]

A hot-rodded and rather ancient CD player, converted to a transport. Multiple mains tansformers, a regulator glued to the top of every chip, clock board (my design) with OPA623 S/PDIF driver (1ns rise time), and focus and tracking servos run from a dedicated supply. Metrum Octave, Audio Research LS3 pre, Audio Research D125 power, Quad ESL57's

Nice! We need pics of the CD player

Edited February 4, 2012 by Pars

[Filburt]

Well, 16 pin (looks like TSSOP-16?), >1MSPS, buffered voltage output, and "industrial." That tends to narrow it down. I'm not sure what to make of "handles sampling rates up to 15MHz" or "useable for 16 or 24-bit audio." As far as I'm aware, there aren't any 15MHz 24 bit DACs; the fastest I've seen are the ARDA AT1401 (1.536MHz) and PCM1704 (768KHz). 15MHz is also an odd spec for a high speed dac. There may be one out there but I haven't seen one that's >= 16 bits and 16 pin. If anyone has seen a 15MHz 24 bit DAC of any sort, I'd be curious to know about it. Anyway with those caveats in mind, if you want just some guess at it, maybe DAC8581 but it's hard to tell from the picture. I don't remember the pinouts on many dacs of that sort and don't feel like digging through datasheets over it.

- I just noticed that one of the reviews Metrum links to has measurements of the DAC in it.

Edited February 5, 2012 by Filburt

[Craig Sawyers]

Nice! We need pics of the CD player

Coming; it is not a pretty sight though!

I forgot the headphone systems. Mostly KG stuff. Triode E/S feeding original Lambdas. Dynalo with outboard PSU feeding either AKG K701 or Beyer DT990. SRM-T2 clone feeding 007's from a Tent CD in the bedroom. And a Blue Hawaii in the other bedroom waiting to be wired up and looking for a pair of phones to be fed by it.

[bhjazz]

Craig, thanks for posting this. Your information about 75 ohm connectors - makes me think twice about both the consumer gera I already own, and my future DIY builds!

[Craig Sawyers]

I haven't seen this DAC being talked about here so I will just leave this here. Designs-wise, the DAC interesting because apparently it uses 4 "industrial-grade" non-oversampling DACs (2 per channel) whatever they are and it doesn't have a digital filter.

Price-wise, it seems reasonable enough. The most expensive model with black anodized case and external PSU will set you back €710 or a little less under $1000. Shame it's still single-ended and has no built-in USB.

Here is the home-site and a white paper posted there detailing this NOS digital filter-less concept. I wonder what do our resident DAC experts like Filburt and Colin have to say about this as the concept is quite intriguing.

Just looked at this again - there are 8 DAC's (hence Octave), ie 4 per channel. I think it uses the concept I first saw on Cambrige Audio CD players shortly after CD's were introduced - you time delay the digital signal to each DAC, so they do linear interpolation between the data points. Which gives you an extra two bits. Colloms measured a resolution of 18 bits, which would be the 16 bits of the individual DAC's, plus 2. It may be that they use something more sophisiticated than linear interpolation, but I have absolutely no evidence either way.

Notice the lack of anything resembling a PLL or crystal. I have no idea how they recover the clock - it is something in the row of four IC's near to the pulse transformer, but other than that I just don't know.

[pigmode]

The first problem is common to most DAC's and CD transports - RCA connectors. I did a whole host of measurements using Time Domain Reflectometry, and basically putting a fast pulse into a 75 ohm coax terminated with an RCA, into an RCA socket, with a surface mount 75 ohm resistor tacked onto it - and it is a disaster. Massive reflections - and reflections add jitter.

The only way to do things *properly* is to maintain a clinical 75 ohm environment for the whole digital signal chain, and that means 75 ohm BNC connectors throughout. And most digital cables, if not terminated in RCA's are terminated in 50-ohm BNC's! In fact the only audio high-end 75 ohm BNC's are Oyaid. I just use regular clamp-on greenpar and RG302 teflon cored coax.

After doing that (and measuring the TDR response into the Octave) the DAC really started to sing.

Just to be clear, are you replacing the Octave's built-in spdif RCA jack, or planning to use a verified 75 ohm bnc terminated cable with a BNC to RCA adaptor?

[Craig Sawyers]

Just to be clear, are you replacing the Octave's built-in spdif RCA jack, or planning to use a verified 75 ohm bnc terminated cable with a BNC to RCA adaptor?

Replaced the RCA jack with a 75 ohm BNC. Just to get it working I initially used a verified 75 ohm cable (RG302 terminated in 75 ohm BNC's) via a BNC to RCA adaptor. But during my swapping the Octave to insulated 75 ohm BNC I tested the cable-adaptor-RCA with 75 ohm termination on my Tektronix 1502 TDR - and it was truly horrendous. It was so bad that it was impossible to see anything approaching the 75 ohms termination resistor. The only way to go is to rigidly stick to everthing matching the characteristic impedance of the cable.

[Pars]

As Craig shows clearly, you really need to think of S/PDIF as a system. Changing out only the cable is probably a crap shoot. The transmit and receive circuitry/jacks and the cable make up the system. It is not a given that all 3 are done correctly...