Metrum Octave Dac - What are the Chips used

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This NOS Dac which is causing something of a stir is getting soem very glowing reviews including in Hi Fi Critic. In all the threads and reviews I have read up to now there has never been a mention of the manufacturer or chip number that is actually used. It is often mentioned that it is not an audio but an industrial one used for other things. However it should still be easy to get the chip number and who made it ?

I mentioned this Dac to my audio repair engineer and he sent me an e-mail that they were using TDA1543 chips which I would find pretty hard to accept as these are nothing like what is being described and are just cheap replacements for TDA1541 chips.

Does anyone have one of these Dacs and have you looked at what chip is installed . Would be interested if anyone can shed any light on this.
 
no, it is not TDA1543, ive seen inside pics in a review and they look nothing like them. its described as a very high speed industrial Dac chip and it is voltage out. its a NOS dac, but i think that is where the similarity to the TDA ends, so I would wonder a bit about your engineer, maybe he hoped you wouldnt do something like this.... sorry i dont know the part #, they have been tight-lipped about it. you would need to come up with some sort of uC or glue logic to use it for audio though
 
Last edited:
Here's the back story as reported on 6moons:

...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.

If you read the datasheet you'll notice there are a few things here which don't quite ring true - the claimed sample rate achievable is '15MHz'. Contrast that with the maximum data input rate of 50MHz - 16bits need to be fed in per sample, giving 3.125MHz flat out. The settling time (to 15bits) is quoted as 1uS giving a practical limit of 1MHz. I could mention further that its most definitely a 16bit DAC only...

<edit> There's a subsequent laugh further down the same article - Because our DAC chips are relatively expensive. TI's price is $1.85 each, 1k.
 
Last edited:
they are running several in parallel though arent they?

Yep - 4 per channel.

could they not interleave the samples on half for higher speed and parallel for higher bit depth?

They could use a DSP to oversample but it looks like they don't - the impulse (and square wave) response is devoid of any filtering effects. Paralleling doesn't increase the bit depth, oversampling might to a small degree.
 
Yep - 4 per channel.

thought i remembered that, didnt look at the article you linked, but i read it when it was released and i remember thinking using these chips was simply a way of having a unique selling point; rather than presenting any real performance advantage. I generally prefer to see an actual external i/u convertor designed for audio...ie. I prefer current out dacs

They could use a DSP to oversample but it looks like they don't - the impulse (and square wave) response is devoid of any filtering effects.
ahh, was just a thought experiment, i think its as much hogwash as you

Paralleling doesn't increase the bit depth, oversampling might to a small degree.
I meant only by way of lowering noise vs signal and thus presenting an apparent bit depth increase, however small. you know hifi, small becomes large
 
thought i remembered that, didnt look at the article you linked, but i read it when it was released and i remember thinking using these chips was simply a way of having a unique selling point; rather than presenting any real performance advantage.

I think there is a performance advantage from paralleling the chips - and Cees, the designer mentions it in the article on 6moons. Actually there are two advantages - better linearity because bit-weight errors in the DAC resistor elements get averaged out in the summation of outputs and also better drive capability.

I generally prefer to see an actual external i/u convertor designed for audio...ie. I prefer current out dacs

You and me both, in general :) Though there's an exception to every rule - on erin's recommendation I recently got hold of some PCM53 and there are two types, a V-out and an I-out. I'll see whether I can hear a difference.

I meant only by way of lowering noise vs signal and thus presenting an apparent bit depth increase, however small. you know hifi, small becomes large

Ah, I see where you're coming from now :) Its never going to bridge that huge divide between the claim (24bits) and the reality :D
 
I think there is a performance advantage from paralleling the chips - and Cees, the designer mentions it in the article on 6moons. Actually there are two advantages - better linearity because bit-weight errors in the DAC resistor elements get averaged out in the summation of outputs and also better drive capability.

agree completely, i was talking about the industrial voltage out dac chips as the unique selling point; paralleling is hardly unique :D its how the ESS get some of their performance, but IMO past a certain point the added complexity becomes ridiculous and not worth the effort; for me thats only 2-4 chips. it only reduces correlated noise though, a point not lost on Thorsten


You and me both, in general :) Though there's an exception to every rule - on erin's recommendation I recently got hold of some PCM53 and there are two types, a V-out and an I-out. I'll see whether I can hear a difference.
yeah there are a few exceptions, some of the AKM are pretty nice and i'm pretty interested in trying out the new AK4999. building buffers is never going to be as much fun as tweaking IV stages though and i cant see it being my new favorite, but all the same i have enjoyed their dacs in the past as Vout goes its about as good as it gets IMO; I like how they often have options for external OSF and reference, which is not totally uncommon, but adds to the fun.


Ah, I see where you're coming from now :) Its never going to bridge that huge divide between the claim (24bits) and the reality :D
of course not hehe, but it gives them and the marketing department a fragment to latch onto. it probably just has 24bit OS input by way of the receiver
 
Last edited:
agree completely, i was talking about the industrial voltage out dac chgips as the unique selling point; paralleling is hardly unique :D

Ah yeah I agree that using a non-audio spec'd DAC is a USP. It inspired me to start looking at other non-audio DACs so I really appreciate that aspect. Its a significant USP but it got blown rather out of proportion in that 6moons article.

yeah there are a few exceptions, some of the AKM are pretty nice and i'm pretty interested in trying out the new AK4999. building buffers is never going to be as much fun as tweaking IV stages though and i cant see it being my new favorite, but all the same i have enjoyed their dacs in the past as Vout goes its about as good as it gets IMO; I like how they often have options for external OSF and reference, which is not totally uncommon, but adds to the fun.

Aren't all the AKM parts low-bit though? So far I've not heard a low bit part which does dynamics as well as multibit - they all sound (to varying degrees) compressed to me. Just recently I was listening to a Cirrus 6 channel part (the precise part no. I've forgotten) in a surround sound decoder I'm playing with and in isolation it sounded ok, just a bit polite. Then in switching back to my TDA1545 prototype I found 'ah, no competition' :D

<edit> Ah I see you've edited your post since I replied. If added complexity scares the bejesus out of you, what will you say to my 32 * TDA1545 DAC which I'm planning to increase up to 41 * TDA1545? :D :D (They are cheap parts though)
 
Last edited:
yeah its interesting, but thats about where it ends for me, i dont really have the skills or patience to put together a custom uC or glue logic to make it work with audio.

actually no, there are a few multibit AKMs in fact thats one of the things that interests me, as traditionally ive been more of a DS dac man. they have quite a few 32bit and at least one of them is a multibit dac. AK4390 is a 32bit multibit dac and AK4396 is multibit but 24bit and i'm interested in trying it and the brand new AK4399, which i assume is delta sigma, but should be fun all the same and accepts DSD natively. I'm looking to experiment with the brand new Xmos multichannel i2s which will do DSD over USB. but i'm running titan for multichannel USB at the moment

i'm no stranger to complexity, the system i'm building is driving 2 (soon 3) separate synchronously clocked balanced dacs (sabres at the moment 2 x ackodac AKD12P) each with an IV stage tuned for the range its covering (Jfet on highs, opc's D1 varient on lows), so one drives L/R woofers and the other the L/R tweeters. digital XO on my mac with room correction/convolver and 4 x balanced poweramps 2 x LPUHP (highs) and 2 x 'the wire' lateral fet amps (lows); all tied together with a central MCU on beaglebone (raspberry pi is taking too long and may be folly) and switched/controlled and monitored with a network of xbee wifi modules. so this will be controlled with touchscreen or any network capable device like ipad etc over ethernet/wifi

but paralleling say...41 dacs is just silly and overcomplicated hehe. (why 41 specifically?)

i'm starting with 2 ways to get the system sorted, then I already have another 2 channels of 'the wire', 2 x Aleph JX and 2 x power jfet circlotrons mostly done waiting for another dac which will probably be multichannel rather than building another 2 dacs for subs and rears

traditionally ive been a headphone guy, but i started planning this system about 2 yrs ago and its getting pretty close. its been going together piece by piece, some of it is finished and just waiting for a proper chassis so i can draw a line under it and move on (2 of the dacs and 2 amps) but the rest is in various states of completion
 
Last edited:
well the software is not 100% certain yet and will be one of the last pieces to the puzzle as software developments will continue while i finish building up the system and require no other hardware or planning, I have 4 options i'm looking at

1. buy another mac mini or heaven forbid a PC (gasp) and run windows on it with Allocator as a dedicated headless crossover/convolver machine and slave it to my main mac so i dont have to deal with running windows on my main computer even for audio. Allocator for mac looks like it is never going to happen, its been saying mac version coming soon on the site for the last 2yrs

2. Combine puremusic and its crossover with something like IK media ARC for room but arc is limited to 96khz, so i'll see how much difference the convolver makes and maybe buy something more expensive. ARC is a good entry into it as it includes a mic, so not bad value. eventually the idea is to build myself a very nice ADC (the ESS, maybe even just the eval board) with nice analogue front end like a borbely 213, Samuel Groner Sassariente or Monte' generoso, or Jung Mic Pre which i can also use for recording as a side benefit.

3. perhaps buy the KRK ERGO for room correction and use puremusic for XO

4. forget the convolver and just run puremusic.
 
frugal-phile™
Joined 2001
Paid Member
I have pure vinyl running now on top of iTunes, makes significant difference. The XOs are kind of simplistic, but i did note a 1st order bessel filter -- not something "easy" to implement analog. By the time i get to it, i wouldn't be surprised to see that Robinson has evolved it further.

I already have a 4 channel firewire ADC/DAC and a (free) dual core G5 tower* so, except for the convolver (althou i really like the idea), we are along somewhat similar paths.

*until an Intel comes my way i have to keep it simple.

I do have an advantage most don't in that my room has been designed for hifi, and the FASTs that will be used are highly optimized acoustically.

dave
 
actually no, there are a few multibit AKMs in fact thats one of the things that interests me, as traditionally ive been more of a DS dac man. they have quite a few 32bit and at least one of them is a multibit dac. AK4390 is a 32bit multibit dac

You got my hopes up for a nanosecond there, so I downloaded the datasheet but the second line says:

Ultra Low Latency 32-Bit ΔΣ DAC

so no need to read that document any further :)

but paralleling say...41 dacs is just silly and overcomplicated hehe. (why 41 specifically?)

Ah, glad you picked up on that - I think I could start a whole new thread on that topic, but here's the short version. I had 32 DACs for linearity and output impedance reasons (low value I/V resistor makes passive filtering that much more practical) but found I couldn't compensate for the NOS roll-off with any practical passive filter that also sounded transparent. So I figured (as yet untried) that by adding on a handful of 'helper DACs' I could do the sinx/x correction by means of a transversal filter. That's a kind of FIR filter but implemented in hardware rather than software. Once I've prototyped it (its relying on a few so far untested assumptions) then I'll be sure to share it :)

Complexity has its place - that in my book is to hide complexity and end up with something nice and simple.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.