|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
71.194.185.32
About a week ago I published a first installment on modding class D/T amps. In case you want to find the first post – here’s the link.http://www.audioasylum.com/forums/amp/messages/111691.html
This week I’m publishing a second installment focusing on the amp board signal path. And just a word of caution – modding the output section of a class D/T amp can result in premature amp mortality – so be careful.
Input Section Mods
The input section of the Evo is more interesting than it might first appear – but first a little more about the Tripath chip.
The Tripath chip includes an op amp input buffer, the gain of which can be set by outboard resistors. The internal op amp input buffers are not needed for amplification – in fact on both the Evo and the Tripath demo board – they are set to provide fractional (less than 1) gain. However, the op amps on the Evo board ahead of the Tripath chip perform several important functions:
- Changing balanced amp inputs to match the single-ended Tripath chip inputs.
- Inverting the phase of the left channel. Tripath describes this as “antiphase” and it helps to control pumping of the power supply at low frequencies. The reverse phase is “corrected” at the speaker posts by reversing the lead connections within the amp.
- Blocking the +2.5 volt input offset of the single supply op amps within the Tripath chip.
- Correcting DC offset with a servo.The first op amps encountered by the signal are Burr Brown OPA2134’s, one for the positive and one for the negative phase of each channel. These op amps are set up as non-inverting voltage followers with no resistor in the feedback loop. Their purpose is to block the +2.5V DC offset present on the input of the single supply op amps in the Tripath chip. The other choice for blocking DC offset is a coupling capacitor – the route taken on the Tripath demo board.
Although the OPA2134’s block the “outbound” DC offset of the Tripath chip – they do not block “inbound” DC offset from upstream components. So another OPA2134 is used in a servo to correct the DC offset (the chip is the one surrounded by four blue box caps). It also corrects DC offset inherent in the output of the Tripath chip, eliminating the need for the pot-adjusted DC offset correction used on the Tripath demo board.
Next in line is an INA2137 – a dual differential line receiver. It’s part of the OPA2132/2134 family of op amps. It differs from a regular dual op amp in that it has eight laser trimmed on-chip resistors, eliminating the need for external resistors. This chip is set to increase the gain by a factor of two and the left channel is set to invert phase.
At the Tripath chip, the internal buffer op amps are set up to reduce the gain again by a factor of .43 using exactly the same resistor values as those on the Tripath demo board. So the total gain of the front end is 1 x 2 x .43 = .86. Don’t ask. But all in all, the input stage of Evo 2 is reasonably well implemented so I decided to upgrade it, rather than do a total replacement.
I did not upgrade the OPA2134’s to OPA627’s largely because I didn’t want to use adapter boards. However, if the soon to be available OPA2827 op amps prove to be as good as some say – I may upgrade in the future. Unfortunately, the INA2137 will have to stay since it is the best differential line receiver available.
The biggest improvement to be made – aside from upgrading the power supply – is replacing the servo caps (C21 and C30). I replaced the two .33 uF, 63V Vishay (BC ) MKT370’s with .33 uF, 450V Auricaps. I used Auricaps both because they are relatively compact – and because I’ve has good experiences with them in signal path applications. As I’ve found in previous mods, the improvements from a servo cap upgrade occur across the frequency spectrum. This is because real (as opposed to theoretical) servos inject a bit more than just DC back into signal path. Be patient, signal path caps take awhile to break in.
There is also one other cheap trick. The Evo 2 used a four-layer circuit board with the top and bottom layers acting as ground planes. The ground planes over the input section are sensibly separated from the ground planes over the output section – and are connected with ferrite bead. However, the input section ground plane is coupled to the earth ground through capacitor C45 – creating the opportunity for noise to flow from the output section to the input section on its way to earth. So just remove the capacitor. This isn’t a big mod, but you will notice a little blacker background and better sound-staging. And it’s free.
And there is at least one other thing you might think about doing. I’m using the Evo with a Bent Audio NOH transformer-based passive pre-amp that gives me the option of breaking the ground connection between my SACD player and the Evo. I can report that the sound is significantly better with the ground lifted – a more spacious presentation with richer harmonics. So maybe Jeff Roland is on to something – his ICEpower amps have Jensen input transformers ahead of the amp modules. If I didn’t have the transformer-based volume control, it would be worth adding input transformers to the amp in my system. And isolating two potentially noisy components from each other (digital front end and class D/T amp) may be worthwhile in any system.
Output Section Mods
The output section of a Tripath-based amp is incredibly simple – which probably accounts for the superb resolution of detail. There are just two MOSFETs per channel – and only one of these two transistors is active at any given moment. With conventional transistor amplification, there are multiple PNP and NPN output devices. PNP and NPN transistors have different characteristics and there are production variations among transistors of a given type – so you’re listening to a choir of different voices – and that’s not really a good thing in audio reproduction. Ahead of the two MOSFETs are two gate resistors. And after the MOSFETs is a simple LC filter that reconstructs the output pulses into an analog signal.
In addition to the essentials, there are a couple of other parts that make the amp a better live-in companion. The first is a Zobel (RC circuit) that keeps the amp stable should you turn it on with no connected load. The second is a relay that connects the output to the speakers after the potentially noisy power-up sequence is complete. This second feature is not present on the Tripath demo board or in many other Tripath-based amps.
Starting at the beginning of the output section, the metal oxide gate resistors (R3?, R4, R5, R6) are a prime target for upgrading. Metal oxide resistors are used because they are sturdy and cheap. The replacements also need to be sturdy with at least a 1 watt rating – so Caddock metal film resistors immediately come to mind. I used Caddock MV217 5.7R axial lead resistors that fit the board nicely and have a 1.5 watt rating. These are not stock resistors – so they have to be specially ordered from Caddock – and you get to pay $31 each for the small run – but at least you can get them.
This mod needs to break in a bit – at least 24 hours of continuous power up. The Caddocks give a richer more natural sound with improved micro and macro dynamics. Cymbals have a more realistic character and detail (already good) is better still.
Next up – let’s take a look at the inductors. The inductors used in the Evo are exactly the same inductors specified by Tripath – T-103 Micrometal cores with 27 turns of 16 gauge solid core wire for and inductance of 11 uH. The inductors are shrink-wrapped – probably to keep them together during production.
The fact that the inductor uses a powdered iron core is not necessarily a bad thing. Using an iron core means that much less wire is required which minimizes DCR. It also minimizes the size of the inductor which helps to minimize its ability to broadcast noise. And the filter frequency is set at well above the audio range – minimizing any issues with using an iron core.
But, there are a number of things wrong with the inductors. First off, they have been wrapped in heat shrink material (polyolefin). You can make a no cost improvement in the sound by simply removing the heat shrink material. (Don’t try to remove the heat shrink material with the inductors in place. They can be damaged if not removed first.)
Now that you can see the inductors, you may notice that the core is really too small for 27 turns of 16-gauge wire. Toroidal inductors should only have about an 80% fill so the input and outputs don’t capacitively couple to one another – and the fill is about 100%. And then there is that large-gauge solid core wire, which is never a great for high frequency reproduction.
So, I got some bigger T-130 cores from Eastern Components and some 16-gauge OFC stranded magnet wire from Solen and wound my own inductors. The bigger cores need 30 turns of wire rather than 27 to achieve the same inductance. And the bigger cores need to be turned 90 degrees and will hang off the edges of the board slightly. This has the added advantage of aiming their centers (where the noise is concentrated) away from the output transistors.
And here are a few tips on winding your own inductors: (1) Glue a 5’ length of wire to the inductor core at the center point with a little quick setting epoxy glue. (2) Wind the wire 15 turns in both directions from the starting point using leather gloves. Wind at least one extra turn in each direction to hold the wire in place for gluing. (3) Glue both ends of the wire at the 15 turn point with epoxy. (4) When the epoxy has cured, unwind the last turns, cut the tails short and tin the tails in a solder pot. (5) Solder the inductors in place, keeping the inductors on insulated portions of the board. (6) Glue the inductors in place with silicone. You could use epoxy– but not if you ever want to remove the inductors.
It’s definitely worth the trouble to wind your own inductors. You’ll have to be a little patient because they won’t sound their best until the amp has been powered up for about a week. But after the hard work and waiting, you’ll find that the recessed, something thin treble performance is greatly improved. What’s more, the bass has better pitch definition and the midrange is just more fluid and natural sounding.
The other component in the LC filters is a .22 uF capacitor (C13, C16). It’s a Vishay (BC) MMKP383, a double-metalized polypropylene film cap designed for switching power supply applications. Bel Canto changed to this cap after production of the Evo 2, Gen II started and it was available as an upgrade until the Tripath-based products were replaced by ICEpower-based products. I was a little reluctant to replace these caps since Bel Canto plugged them as the holy grail of capacitor options – and maybe they are at the 50-cent price point. However, .22 uF 200V PPMFX Mulitcaps sound infinitely better. The treble becomes fuller and more natural sounding – and the bass tightens up and the already detailed midrange becomes even more detailed and the entire presentation loses a bit of constipation. And you’ll notice that the amp runs cooler (as in cold) as well.
The Zobel is not in the direct signal path – but the quality of parts (poor) still has an effect on the sound. I replaced the inductive wire-wound resistors (R13, R27) with Mills wire-wound resistors and the Panasonic Mylar caps (C17, C18) with more .22 uF 200V PPMFX Multicaps. This is a minor mod, but the better parts improve high frequency performance. Of the two parts – the cap has the most effect.
As you’re choosing replacement caps for both the LC filter and the Zobel – keep in mind that the caps need to be robust – and the lower their impedance the better. That’s why I selected MultiCaps and used the metalized film variety (rather than foil) for their self-healing properties. Failure of these caps might not be pretty.
And you’ll be pleased to learn that there is one component – the feedback resistors (R9, R10) that don’t need to be replaced. Bel Canto used Caddock MK132’s – one of only two audiophile parts in the amp.
And of course no set of mods would be complete without a few power supply mods – which will be covered in Part III.
Here’s the parts list for this set of mods:
2 ea Auricap .33uF, 450V Capacitor $15.50 PartsConnexion
2 ea Mills MRA-12 15 ohm Resistors $7.00 PartsConnexion
4 ea Multicap PPMFX, .22 uF, 200V Capacitor $14.12 PartsConnexion
Solen 16ga Litz Wire $23.70 Solen
4 ea Caddock MV217 Resistor 5.7 ohm $126.32 Caddock
2 ea Micrometals T130-2 Core $2.26 Eastern Components
Total $188.90
Follow Ups:
I have enjoyed reading your rework of the D/T amplifier. I find it very disconcerting that you have paid premium dollars for the stock amp, and then you have spent nearly 25% more for parts to make it approach "right".What is so disconcerting is to read where such a high priced product has such mediocre parts (ribbon cables, etc.). Considering the time and effort you have put into the EVO; I would find it interesting to read what mods you would consider for the much less costly Tripath amps such as the Sonic impact, SI Super-T, Trends Audio TA-10, 41HZ amps, Charlize, etc.
Even if you don't take the time to modify one, I would like to read your preliminary analysis as to what mods should be considered after hearing one of the less esoteric D/T-amps.
DaveT
Dave,Actually, I didn’t mind spending some money on the Evo – and buying it on the used market helped ease the pain. Mid-market products like the Evo are built to a price point – which always involves compromises and opens the door to modding. I find that modding is much cheaper than buying “cost is no object” products – and I learn and have some fun along the way. Moreover, sometimes I can get to sonic places that can’t be reached by just purchasing equipment – no matter how much I have to spend.
However, your point about parts quality in the Evo (and other Class D/T amps) is well taken. Every Class D/T amp I know of has less than great parts quality – yet many sound quite good and get “A” ratings. With a little more development and better parts, I think they have the potential to be awesome.
The amps based on Tripath TA2024 and TK2050/51 chips are similar in many ways to the Evo and some of the same tweaks would apply. There is just less to tweak since more power supply components and the output devices are integrated into the chip. For amps using these chips, much has been written about tweaks to connectors, power supplies and input caps. But from what I’ve seen, output filter mods have been largely neglected – yet they are some of the most potentially effective mods. The mods I describe to output filter and Zobel caps, resistors and inductors would all apply to any of the integrated chip amps.
The Evo doesn’t use input coupling caps – but if I were replacing the input coupling caps on another Class D/T amp I would use Auricaps if I were feeling budget minded or space constrained – and Mundorf MCap Supreme Silvers otherwise. And for electrolytic power supply caps – I can’t say enough good things about Rubycon ZA and ZL caps. They are great for both analog and high speed switching applications making them perfect for Class D/T amps in which one supply feeds all.
I agree with you regarding the spending of money on parts. Like you, my allure to this hobby is to making modifications to equipment with the hope of improving the sound. My experience with T-Amps to date, has awed me. The T-amps seem to improve with a minimal part swap/enhancements.I have read of people swapping out inductors for air-coils, but I have yet to try it. There seems to be both pros and cons for this modification, and the result seems to yield diminishing returns for the effort employed. I am willing to try your zobel mod if you would try to describe the sonic impact.
Auricaps are my choice for coupling. I prefer the sound that comes from a high amperage linear power supply over that of a low amperage SMPS. I have been experimenting with the tank capacitance on the supply feeding these amps. My experimenting has been limited to SIs, HZ41 Amp6, and a stock Trends Audio 10.1.
I have read various threads on this and other forums about the D/T-amps, and I found you selection of the EVO for modification a unique undertaking. Even at that, I still find the "shortcuts" (quality and economic)taken in such a highly touted product of concern. I guess it all comes down to "buyer beware" .
Dave,When modifying the output section, swapping out the capacitors in the LC filter has the greatest effect – so I would start there.
Changing the inductors to air core inductors is not necessarily helpful since it will increase the radiated noise and DCR. However, rewinding the inductors with OFC or silver magnet wire of the same gauge would be helpful (if the inductors are toroidal). The wire in the inductors is as long as the wire in a pair of interconnects – and it’s just ordinary magnet wire. But it’s true – winding inductors is a bit labor intensive.
If the inductors have shrink wrap material around them – consider removing it if it looks like the inductors will remain mechanically stable. The shrink wrap material is like having the worst imaginable insulation on an interconnect. It’s probably best to remove the inductors before trying to cut off heat shrink material. This operation requires some care since the insulation of the magnet wire is easily damaged.
Changing out the Zobel capacitor and resistor is icing on the cake. It makes a difference – but not as much as working on the LC filter. High frequency performance is improved a bit – and the sound becomes a bit more relaxed and less electronic in nature.
Adding a very large cap to the power supply is a move in the right direction. None of the little T-amps have enough power supply capacitance. The manufacturers are probably just trying to keep the amps small and cheap. And if you haven’t upgraded the power supply cap on the amp board with something like a Rubycon ZL or Pansonic FM – do it. And use the largest capacity cap that will fit the board.
This post is made possible by the generous support of people like you and our sponsors: