[Russtafarian]

Well, kind of…  Here’s how I put together a six channel 100 wpc Tripath digital amp using the chassis and power supply from an old Carver amplifier.  

I’ve been hearing about the great performance of Tripath-based amps from Bel Canto, Audio Research, Carver Pro, etc. for the past few years.  And all the recent Powerwave banter on AC really piqued my interest.  I was tempted to drop $100 to try one but I knew that the 15 watt p.w. would shrivel when asked to drive the current hungry Dynaudio woofers in my Gallo Nucleus Reference speakers.  I need more power!

So I started checking out the various evaluation boards that are offered at www.tripath.com .  Since I would have to come up with a chassis and power supply, I had to evaluate module performance (I wanted 6 channels @ no less than 100wpc), cost (no more than $500) and power supply requirements (the simpler the better).  I looked at some of the +12 volt boards which would have been the easiest to implement, but they didn’t have enough output power.

I decided on the RB-TK2150 Six Channel Reference board for $479.  The TK2150 doesn’t deliver quite as much power as the TA3020 used in the Bel Canto amps, but it comes close and the power supply requirements are easier to implement.  Plus, I could get the TK2150 in a six channel configuration, as opposed to three stereo TA3020 boards which would have cost nearly $1000.

Next I needed a chassis and power supply.  This is where the Carver comes in.  Boxed up in the garage I had a mid-‘90s Carver AV-405 multi-channel power amp that I’d used it in various configurations of my HT rig.  Lots of grunt for movie playback, but I never liked the way it sounded for music.  But it had a nice chassis and it sure was HEAVY, so I figured it must have a beefy power supply.  So I popped the lid to check it out.  I saw that once I removed the SS amp modules there would be plenty of room to mount the Tripath board in the chassis and I could use the existing binding post for speaker connections.  I did decide to replace the cheapo RCA jacks with some nicer gold-plated ones.  Plus, the big metal box would help block the RFI the Tripath chips would generate.  

As for the Carver PS, I found a big toroidal transformer and two separately grounded, rectified and filtered power supplies, one delivering +/- 60VDC and the other +/- 45VDC.  This is great because the Tripath board requires two separately grounded power supplies: one +/-39VDC to +/-60VDC supply for the switching mosfets and another +5VDC regulated supply for the digital circuit.  The +/- 60VDC supply with over 13000 uF’s of storage would work great for the switching mosfets.  Now I had to figure out how to get +5 regulated volts out of the 45V supply.

Because of my limited knowledge of power supplies, this was the most challenging part of the project.  Thanks to the help of Eli Duttman and Vinnie R. over at the AA Tweaker’s board and Kevin at Tripath I figured it out.  I needed a regulation circuit consisting of a 10 watt voltage drop resistor connected to the input of a 7805 5V regulator.  The regulator input is bypassed to ground with 47 muF and 10nF caps and the output bypassed to ground with a 4.7 muF cap.  The +5V output of the voltage regulator would then feed the Tripath board.

The trick was getting the right resistor value that would drop the voltage from +45 to the +8 volts that the regulator needed.  This involved knowing the amount of current draw from the tripath circuit.  I looked at the data sheet of the TK2150 and it said 50 mA.  No problem.  45V – 8V = 37V which is the amount of voltage drop I need.  37 / .05 = 740.  So a 740 ohm resistor will drop 45 volts to 8 volts with a current draw of .05 amperes.  But when I hooked it up, the amp didn’t work.  I was measuring +5V off the regulator but only +3V at PS input on the Tripath board.  After some discussion with Kevin at Tripath, it finally dawned on me that because the six channel board uses three TK2150 chips, the current draw was 150 mA, three times what I originally estimated.  Back to the calculations:  37 / .15 = 246.67.  So I plugged in a 220 ohm resistor and it works!

The rest of the project involved mounting the Tripath board to the chassis and connecting the board’s audio inputs/outputs to the RCA jack and binding posts.  I chose to desolder and remove the pre-mounted RCA jacks and speaker connectors from the Tripath board so I could solder my own wire directly to the board.  I used 30 gauge PTFE insulated silver wire for the RCA connections and wire hacked from an old run of Kimber 8TC for the binding post speaker connections.   I then mounted two of the heat sink plates from the original Carver SS modules back into the chassis on either of the Tripath board to hopefully provide some RFI isolation between to power supply and the Tripath board.

Next I adjusted the DC offset for each channel by turning the amp on and measuring the DC voltage between the + and – speaker terminals.  The blue trim pots on the board allowed me to adjust each channel for minimum DC voltage.  Before the amp left my workbench, I verified that each channel was working by hooking up a portable CD player and a cheap speaker.  The amp is now in my rig and working just fine.

How does it sound?  Great so far.  I’ve only had a few hours to listen to it and it will take a while for me to get a handle on its character.  I want to give it some time to break in, then I’ll compare it to my Quicksilver monos and my Museatex CAS-10.  My buddy has both the PS Audio HCA-2 and the Spectron Musician, so I’ll eventually haul my Tripath Frankenstein over to his place to do some comparisons.  

My objective up to this point has been to get the amp working, but there’s lots of future tweaking to be done: add a power cord inlet, rework the power supply with better diodes and caps, upgrade the input/output caps on the Tripath board, and maybe even input transformers like the Rowland digital amps use.  The possibilities are endless!

[StevenACNJ]

Russtafarian

Nice job

Do you know if Bel Canto uses the TA3020 Tripath evaluation board as is or are they modifying it. I know I just read something at 6Moons about an output cap replacement but other than that do they run it stock?

Thanks

[Russtafarian]

Judging by the pix in the 6moons article, Bel Canto has designed their own circuit board for the TA3020.  As to whether they are using the same parts as the eval board I don't know.

I am going to do the cap swap described in the article.  My board has two .15 uF caps on the output of the inductor.  The parts on order to swap these out for the caps described in the article.

[StevenACNJ]

I am going to do the cap swap described in the article.  My board has two .15 uF caps on the output of the inductor.  The parts on order to swap these out for the caps described in the article.

Since you only need two, why not put really high quality caps in that location like Mundorf Supreme silver & oil caps. E-speakers.com has them in a .15uf value.

[Occam]

Since you only need two, why not put really high quality caps in that location like Mundorf Supreme silver & oil caps. E-speakers.com has them in a .15uf value.

Those Mundorf caps may well be the 'bestest' audio coupling cap in the whole world, but their use in that specific position might be problematic.....

Per the Tripath pdf, those 2 0.15uf 100v stacked film Panasonic ECQ-V1154JM caps, 1 for the 12db filter and 1 for the Zoebel, are quite small, having 7.5mm lead spacing, 'set in concrete' on the pc board.

The Mundorf is a rather large sucker, by virture of its construction and materials. It would reqire rather long, contorted leads to be inserted in those positions, which is gonna add inductance, which is exactly what you don't want. And those leads, wending their way off the circuit boards ground plane, are going to be superb antennas, reintroducitng that roiling crap back into the output.

This does not address those technical specs that are pertinant in a board uncostrained use.  Some of the very best (on a technical high current pulse filtering application) caps are film/foil construction, but I've not been able to suss out those Mundorf's pertinant technical parameters.

While these cap's purpose is switching artefact filtering, arguably, these caps are in the audio signal path, in a shunt, subtractive mode.  The 6Moons review does give credence to the fact that a technically 'adequate' cap can be improved by the substitution of  a different, .\technically adequate cap, on a subjective metric.


Which leaves the possibility of using the hypothetical EVo caps. My WAG was, eventually,  the Vishay MKP1841 cap. But these are only available, in their lowest adequate voltage rating, in a 15mm leadspacing, which is going to incurr some of those penalties of added inductance and mini antennas. Same goes for other purpose built caps, the Rifa PHE450, etc....

Ideal merits of a given component aside, the major constraint in tweaking this specific board, is the board itself.

[WerTicus]

so we need a new board!

these tripath chips are all very interesting am i the only person who is looking at this one?

http://www.tripath.com/pb0105A.htm

[Occam]

so we need a new board!

And why would you say that? Life is full of compromises. But my Momma always told me " Ya dances with the one who brung ya"

Russ,

Congrats on your project! By no means was my intent to dissuade you (if that is your intent) from 'tweeking' this tripath board. It is simply that board layout constrains your choices given the concerns about EMI. The engineering term for the environment within such an amplifier is a 'roiling cauldron of crap'.

Given the constraints of the tripath RB-TK2150-2
(with your rails, I assume you're using the -2 version, which is fortuitous. The manufacturing engineer who specified 100V XR7 dielectric caps for bypassing the 33uf half bridge rail bypass elecrolytic, on the -1 version, should be horsewhipped.)

You might consider the following 3 modifications -

1. Replace the CX00, the audio input coupling caps (where X=1 thru 6) 2.2 uf Panasonic NHG elecrolytics with BlackGate 4.7uf 50V type N caps @ $3.20 ea.. They'll fit exactly. You could save a bit by considering the 'lesser' grade BGs (or anything else), but inevitably, someone is going to say you absolutely must try them... and a type C is only going to save you $2/channel. The tripath's achitechture puts a dc bias on the input cap which generally improves a caps performance.

2. Replace the 6 well specified Panasonic 33uf EB elctrolytics, CX13, with BlackGate VK 47uf 160v caps.  This is what LCAudio did for that role in their digital amp upgrade for their SE version. Make sure that the increase in diameter from 10mm to 12.5mm can be accomodated on the board. A Rubycon BXA might provide substantial savings over that $10ea cost, but unless you're buying 1000+, good luck sourcing them.

http://www.referenceaudiomods.com/Merchant2/merchant.mvc?Screen=CTGY&Store_Code=RAM&Category_Code=BG_N

3. The Tripath and CarverPro boards use iron powder, formula #2 toroids in their ouput filters. BelCanto and LC Audio have chosen to use gapped ferrites in those roles. Inductor manufactures, Pulse, Cambion and others have entered this bugeoning market. They both offer assembled toroids which should fit, and might work quite effectively. But unless you've the wherewithall to consult with one of their applications engineers (or access the equivalent), I wouldn't attempt this. Among their standard, purpose built offerings are 20uH toroids, which are quite close to your boards specified 18uH inductors. Any changes to the output filters inductor or capacitor values will change both its frequency and Q characteristics. See the TK2150 PDF for the formulas.

[Russtafarian]

Hey Occam, I appreciate the input.  This project has pushed me into new territory in terms of my understanding of electronics and digital.  I'm just a guy on strict budget wanting to get into a digital amp.  Just getting the thing working is a major victory for me.

I will look into the cap replacements you suggested.  I am considering playing with the high pass cutoff frequency of the amp, moving it from around 10 Hz to around 50-60 Hz to better suit my speakers.  Given the design, is this do-able or am I asking for trouble?  This would probably change the cap values too.  I'm also thinking about getting some of those ERS sheets to put in and around the amp.

[Occam]

Hey Occam, .....
I will look into the cap replacements you suggested.  I am considering playing with the high pass cutoff frequency of the amp, moving it from around 10 Hz to around 50-60 Hz to better suit my speakers.  Given the design, is this do-able or am  ...

Most doable good Sir. Those input cap calculations are bare nekk'd on page 19 of 2150 chip PDF that I suggested you download at the tripath site. Consider youself mildly chastised. With the 30.1k input impedance(Rx00), you'll find the pole frequency of the input high pass to be 53hz if you use a 0.10uf cap in the Cx00 positions. That low value opens up the possibilities of fillm caps with a 5mm lead spacing for that position. But I'd still suggest a BlackGate NX 0.10uf 50v @$2.20ea.

I'm an agnostic on those ERS sheets, lacking personal experience. But prior to spending that (not inconsequential) green on them, I'd spring for the BlackGate VK caps I mentioned previously.

You're actually in a great position for  a modder. With the six channels on your board, you can evaluate changes to a stereo pair, and evaluate them against a still extant baseline stereo pair. No need to rely upon, sometimes suspect, aural memory. This also allows incremental expenditures.

As the chipset for channels 1&2 are configured to provide the requisite VN10 for the other 4 channels, I'd also suggest that you do your initial mods on the 3&4 and 5&6 channels, because if you (inevitably) blow a chipset, if its 1&2's, you'll loose the rest of the channels. Only after your willing to move that baseline configuration upwards, should you muck with Channels 1&2.
FWIW

[Occam]

I commented earlier on the constraints imposed by the board's 7.5mm leadspacing for the 12 0.15uf capacitors, Cx08 & Cx09/channel (x=1 thru 6), the output flilter and the Zoebel caps, respectively. Tripath has used a stacked polyester ECQ-V capacitors in those important (IMO) positions. Digikey has just started carrying the EPCOS MKP stacked polypropolene box caps, same value and leadspacing. These, far better specified for the specific tasks 160v capacitors,  should (IMO) yeild subjective improvements, also.

Epcos part#  b32620a5154j
www.Digikey.com

search on the Part# which will briing up pricing and availability. If you click the Digikey Part# next to the 'Technical/Catalog Information', you can pull its  PDFs.