[Monolith]

Any talk of the parts on the output board? This should be easy to bypass, and should provide favorable results.

I have not gotten any further than our previous discussions.  I do not know how much inductor quality plays a part on the output.  In my brief thoughts on removing the output board, I have hesitated due to soldering the inductors directly on the binding posts, mainly for vibration concerns.  Dangling inductors just don't seem like a good idea, but it may not be a real problem.  I might have to manufacture some type of "mini" output board to support the inductor weight and still permit point-to-point wiring.  Just thinking out loud.

[Occam]

Which form factor provides the best heat transfer, 220 or 247?  I may end up ordering more than I need because of small part quantity prices/shipping/handling.

The TO247 would probably offer higher heat transfer efficiency through its lagrer metal tab surface, over that of the smaller TO220s. But their larger physical size might make implementation of a pin-for-pin module problematic given the pin spacing and geographic constraints of the ZR amp..... Given that you've some vertical space available above the present heatsink, but a constrained board area, you might find TO220s the most practical.

[_scotty_]

Monolith, check the Tripath site for their recomendations on inductor core materials and read the application notes on their modules.  It is unlikely that Carver deviated from their recomendations on this part. The inductor has to pass instantaneous currents of large magnitude without significantly adding to the output resistance of the amp.
An aircore might have lower distortion but it's series resistance can't be any higher than the ferrite cored inductor and it cannot have significant self-capacitance, it also has to fit on the board.The large coils proximity to the output MOSFETS is to minimize the possibility of ringing and the total amount of RF inside the chassis and the risk of self-contamination of the circuits. Moving the coils closer to the binding posts would increase the size of the antenna radiating RF energy and the amp may ring.The diodes whether they are TO220 or TO247 will have to be heatsinked in order to realize their amperage rating
without premature thermally induced failure. These packages only have their amperage rating in conjuction with the approprate heatsink. Good luck with this beast.  Scotty

[Monolith]

Monolith, check the Tripath site for their recomendations on inductor core materials and read the application notes on their modules.  It is unlikely that Carver deviated from their recomendations on this part. The inductor has to pass instantaneous currents of large magnitude without significantly adding to the output resistance of the amp.
An aircore might have lower distortion but it's series resistance can't be any higher than the ferrite cored inductor and it cannot have significant self-capacitance,  ...

Just to clarify which inductors (L404-407, 1uH, on the schematic) I was referring to:

Click  Here --->Output Board

The top four grey inductors is what I would need to relocate to remove the output board.  The four caps would also be case grounded from the binding posts.

The main larger inductors (red round things) (L402 & L403, 11uH) are on the main board and is what I think you are referring to:

Click  Here --->Main Board

P.S.  If you have any application notes to the diode heat sinking requirements from Fairchild, I would greatly appreciate them.

[_scotty_]

Monolith, the heatsinking required is a fuction of the forward voltage drop times the peak current the device can handle. In this case 30Ax2.1volts
equals 63watts per device worst case scenario. You can probably reduce this to half or one quarter because the devices won't be seeing a constant 30amps throughput under normal operating conditions. The exact amount
of heatsinking required is hard to spec, the amount on the stock rectifier bridge is fairly sustantial. The output board with its additional filtering might be eliminated with good sonic results, my DIY amp only has the basic two pole filter at 80kHz and sounds fine.

[Monolith]

I found a site with some heat sinks that are tall with back fins that might work.  One of the downsides about going up is that the disconnect for the transformer is near the bridge rectifier.  Not impossible, but it would make disconnecting the transformer more difficult.  Another alternative, though I have not looked at the specs, is the Hexfred bridge that M. Percy carries.  It might be an easier drop in replacement.  As far as its sonic virtues, I have never used it before, so I cannot attest to whether it would improve the sound.

I swapped out the LM833 (U202A/B) last night with an OPA2604 (unity gain stable, yes, there are better op amps, but not necessarily "drop in and forget").  Just wanted to test the waters  to see how much impact the op amps had on the sound.  There are 5 of these in each signal path.  Initial impressions are very favorable.  This is not an endorsement for swapping out the op amps, but rather that much improvement is available for completely bypassing these devices as suggested by Occam.  Despite not being able to access the parts under the wind tunnel, I will attempt to break the circuit and jumper on the exposed area of the board around the op amps and report back.  I have high hopes for a vast improvement.

DIY on Op Amp replacement Note - These are surface mount op amps and not in tremendously easy to access places.  Hot tweezers highly recommended.  If one wants to keep the high/low pass and limiter functions then swapping out the 5 op amps is highly recommended.  Unfortunately, unless you have some skill in soldering/desoldering surface mounted devices, this could easily end in disaster.  Be careful and weigh the "gain" versus the "risk" or have a highly qualified technician do the swapping for you.

Manufacturing note - It would seem to be a very desirable option to have the LM833 replaced by a higher quality op amp.  However, the price increase impacts the pricing of the unit (LM833 are approx. $0.30 versus $3.50 for OPA2604).  Possibly a redesign with the circuitry for these functions being removable via a daughter type card to allow easy and complete bypass for those not desiring those features.  Just a thought.  Or even just a by-pass switch via the front panel.

[Monolith]

Op Amp By-Pass Mod a.k.a. "Occam Mod"

If capacitors C201 (10uF NP Ch1), C202 (0.47uF film Ch1), C203 (10uF NP Ch2), and C204 (0.47uF film Ch2) are removed, this would permit convenient soldering points for soldering in by-pass wires for the incoming signals.

The signals (Ch1 & Ch2) need to be jumpered to a point after resistors R246 & R247 (respectively), which are immediately after U206, pin 1 and U207, pin 8 (LM833's) (respectively).  R246 (Ch1) & R247 (Ch2) are most likely surface mounts.  It might be necessary to remove these and temporarily tack the signal wire to the outgoing pad.  If someone who has access to the board could check and see if there is a better solder point after the resistors today, I would appreciate it.  Photos would be even better.  And actually trying it out and reporting back would be fantastic! Otherwise, I will give it a look tonight.

[Occam]

The previously discussed IXYS diode brigde's pinout and size is problematic, and my 'thin air choice' of individual Fairchild Stealth(tm) rectifiers, requiring assembly of a module and a fabricated heatsink, has its own attendant problems.

Its far easier to aquire a common pin and form replacement, such as BRUS3[4]W, made by the same maufacturer, EDI, that supplies the conventional bridge which we seek to replace.
http://www.edidiodes.com/PDF/Bridges/BRUS3-4.PDF

This would allow a simple swap, presto changeo, an "ultra-fast 50ns'' diode bridge, that fits, and uses the same heatsink. As I've not used them, dunno if it would be an improvement.......

[Monolith]

The previously discussed IXYS diode brigde's pinout and size is problematic, and my 'thin air choice' of individual Fairchild Stealth(tm) rectifiers, requiring assembly of a module and a fabricated heatsink, has its own attendant problems.

Its far easier to aquire a common pin and form replacement, such as BRUS3[4]W, made by the same maufacturer, EDI, that supplies the conventional bridge which we seek to replace.
http://www.edidiodes.com/PDF/Bridges/BRUS3-4.PDF

This would allow a simple swap, prest ...

Easy is always a worthwhile first attempt.  I will add your link to the first post.  If anyone is going to actually swap out the bridge rectifier, remember that the "W" in the part number is critical.  It designates "wire leads" as opposed to lugs.  This generally holds true no matter what vendor or part number.  Despite the linked PDF for the OEM rectifier only showing lugs, the actual OEM rectifier has wire leads ("M" versus "W" suffix).


Occam By-Pass Mod Update[/u]

Rough photos of the signal traces are now in the gallery.  When I get time I will post them using links to the first post.  Basically, R246 & R247 are located just to the left of the tunnel at the front of the case.  These are surface mounts.  These could be removed and the far right pads used to jumper the signal into the signal trace.  Other photos show that the two signal traces run around the nylon support of the wind tunnel and across the front of the main board.  They then go below the board using "vias" ("vias" are electrical links that link traces from one layer of a circuit board  to another, generally a hole drilled from one layer to another and filled with conductive material).  The "vias" on the bottom board appear to be copper trace "jumpers" and are only about an inch long.  These jumpers then go back to the top of the main board using another set of "vias".  After that, it's anybody's guess at this point.  I cannot see under the wind tunnel.  I did not have enough time to actually attempt this mod.  The downsides are going to be how to do a long term jumper.  Tacking on a signal wire will not be very robust.  I would at this point say that for a long term solution, the wind tunnel needs to be removed, the 0.47uF cap replaced with higher quality and a discrete resistor used in place of the surface mount.  This would bring more point to point wiring, but at the risk of increased RF exposure.  I should state that this mod, at this point, is hypothetical.  It may or may not help the sound and it is possible it may make it even worse due to the RF or other unforeseen issues.

[Occam]

For those considering the direct bypass mod (bypassing the circuity on page 3 of the schematic PDF)

The good news is that its overall circuit (the net of those 5 circuit path opamps /channel) is not phase inverting. This means that bypassing this circuitry will not invert the phase presented to the speakers.

And the bad news is - that same circuit has an overall voltage gain of about 14, (I believe?) Which means that the direct bypass will result in a substantial volume drop versus what would be with that 'pro' circuitry in the signal path, whether actually used or not.

This might be mittigated by increasing the gain setting on the input module, or alternatively -
While I don't have the schematic of the Input module, working backwards from the Tripath module,  leads me to believe that even at its highest gain, the Input module has fractional gain, and bypassing the Input module's opamps (as in the 8thNerve mod), might compensate for much of the 'lost' gain.

If someone who has the input module bypass could report on its volume difference, it would be very helpful.

[PeteG]

I'm doing the 8thNerve mod myself next week and Nathan said nothing
about gain loss, I'm sure he would have if there was.

I do know about gain loss when I disconnected the fan.

[_scotty_]

The Tripath module when utilized in the most basic form is an inverting power amplifier and absolute phase is preserved by reversing the speaker
cables.  Bridged operation is accomplished by one of the channels being inverted in phase by an opamp. Total eradication of all opamps would potentialy eliminate bridged operation of the amp. As long as bridged operation isn't desired I would do without the opamps in the signal path. If you check the Tripath application notes a basic power amp can be considerably simpler than the ZR series is with potentially better sound. The module itself has over current protection built into it and an external clipping detector probably is not needed either. The gain of the module and thus the amp should be adjustable if you can find the network that sets the gain on the schematic. R403 andC420 ,and R405 andC425 appear to be the gain setting nework.What their values need to be to result in the industry standard 26db of gain in conjuction with the servo-circuitry on the  input is unknown to me.

[Occam]

ERROR ALERT...... ERROR ALERT

In my previous post I stated that the 'professional' features signal path was non inverting, THIS IS WRONG! Upon rereading the the schematic, page 3 of the PDF, I believe that for this circuitry, channel 1 is non-inverting, while channel 2 is inverting. From a technical perspective, this is quite clever, as in normal 2 channel information where much of the power spectrum is in phase, this will tend to minimize power supply  'pumping'. I do not have the schematic for the Input board, so I will not hazard a guess as to how, or where, one of the channels adds the requisite additional phase reversal. If one simply bypasses the page 3 signal chain, the simplest 'fix' (not necessarily the best) would be to simply reverse one of the speakers leads. I'd caution against using the bridge mode until what is going on is clearer.

I'd appreciate it if someone would examine page 3 of the PDF schematic (accessable from the first post of this thread), and either confirm or correct my findings.

Sincere apologies for the error.

I'm doing the 8thNerve mod myself next week and Nathan said nothing
about gain loss, I'm sure he would have if there was.

I do know about gain loss when I disconnected the fan.

As a result of my typical run on sentences,  I obviously wasn't clear. What I am trying to find out is whether bypassing the input card (ala 8thNerve's mod of adding RCAs and bypassing the input cards opamps),  results in a gain increase, in comparison to leaving the Input card's opamps in the signal path?

Could you please elaborate on the gain loss when disconnecting the fan?

[PeteG]

Occam, Sorry misunderstood about the gain.
As far as disconnecting the fan their is a decrease in gain, I'm guessing
when the heat go's up in the tunnel the amp steps the gain down.

[Occam]

PeteG,

When I reread my own posts, even I can't figure out what I was trying to say.

[Occam]

Further study of Page 3's schematic, what I've refered to as the 'superfluous' circuitry, reveals that it is not as superfluous as previously thought. Bypassing this circuitry will have 4 effects -

1. It forgoes the highpass, lowpass, level set, limiting, and the 'signal present' functions, as previously described. For use as a 'audiophile' amp this should not be a problem.

2. It eliminates one of the channels' phase inversion. It is presently unknown where that inversion is reinverted. When the Input Board's schematic is made available, the answer should be clear. The presence of a dual opamp on each channel of the Input Board is reassuring, as it might logically provide the requisite phase inversion needed to provide that missing one, as well as that required for bridging.

3. It forgoes the signal gain of this block, and will have to be compensated for by an increase in level of the source preamp, and/or a change in gain jumpers on the Input Board.

4. Not previously described, the terminal voltage follower opamp, U206A & U207B, is configured as lowpass filter to eliminate  ultrasonic noise. As to whether this is required, to bandwidth limit for the Tripath chip, or to filter from a specific noise source such the switching 12volt floating powersupply needed for the Tripath chip, I do not know.


While the above (2,) 3 & 4 might be addressed by modification/replacement of the Input Board if necessary, this is not trivial, and goes substantially beyond an easily implemented or reversed 'tweek'.

[_scotty_]

At present my DIY TRIPATH amp employs no input bandwidth limiting. The
input wiring is shielded to reduce RF contamination. The opamps U206A and U207B may help eliminate RF as well as ultrasonic signals from the modules' input. I could not tell you if this is necessary. I think if you can get an uncorrupted signal to the modules' input the amp may sound better.
The  power supply impedence still needs to be lowered at the amps operational switching frequency. A problem for another day.

[Occam]

At present my DIY TRIPATH amp employs no input bandwidth limiting. The
input wiring is shielded to reduce RF contamination. The opamps U206A and U207B may help eliminate RF as well as ultrasonic signals from the modules' input. I could not tell you if this is necessary. I think if you can get an uncorrupted signal to the modules' input the amp may sound better.
The  power supply impedence still needs to be lowered at the amps operational switching frequency. A problem for another day.

Thanks for the info from someone who actually has experience with the Tripath. I'm just working from the schematic.

Your point about ps decoupling (at the rails of the MosFet half bridges) at the switching frequencies is excellent! I know that LC Audio offers the BG FKs on their amp for that purpose. What are you using on your DIY Tripath amp, value, voltage...?

The choices for ultra low impedance @ switching frequncy and high voltage (100v) is rather scarce if one chooses industrial, rather than boutique components. The obvious choices would be the Nichicon UPM or UHE and the Rubycon ZLs, which to the hobbyist is made from unobtainiun, requiring an order of 1000+.
Alternitively, slightly lesser specified caps, 100v, 100uf+-, are availible from Mouser.com. Specifically, the Nichicon UPW on Page 440 of their current catalog, and the Xicon ESRL caps on page 447. I've used neither, so I can't comment on their subjective qualities. Don't know where to stuff them, but leads should be short, as inductance will come into play at the switching frequencies. And then there is the possibilty of bypassing the bypass, with polyprops or polycarbonates....the rails do have .1uf bypass caps which I assume are ceramics, NPOs or XR7s.

[_scotty_]

I used 100mfd at 100v BG, that's all that would fit. This was a direct parts substitution for the existing caps on the eval board.

[8thnerve]

Bypassing the input card does not result in a phase mismatch at the output in my modifications.  Mono/bridged mode is achieved by sending the signal to BOTH CH1 and CH2 inputs, and ground to one of the available input grounds.  Both setups have been tested and prove functional and incredibly positive.  Both setups have also been tested with a phase tester and maintain absolute polarity from input to output.

I hope this helps a bit.

Nathan