Amplifier:Sure Electronics TK2050 2x100W - Modification

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This is an article in progress. I've committed a brain dump from memory to the Wiki for now and have started to trawl through the thread where all the discussion on this amp is going on. So far I've got to the start of page 88.
| border="0" cellspacing="5"
| Name || Sure Electronics 2x100W TK2050
| Type || Chip
| Application ||
| Class || D (Also referred to as Class.T)
| Approx Cost || Starting from $39.99 for board
| Designer || Sure Electronics
| Thread || TBC

[h=The Project]%1[/h]


TODO: more introduction copy here, maybe want to create a Manufacturer's namespace within the Wiki so info can be dumped there and cross-referenced - note can also put details of products of note for that manufacturer

Sure Electronics are a Chinese company selling out of Hong Kong. They produce amplifiers based on the Tripath chipset.

TODO: some images of the amp unmodified

[h=Versions of the amp]%3[/h]

Model AA-AB013 (PDF Manual & Schematic)
  • 1.0 - Passive heat sink (glued on with a silicone adhesive ... tough to get off, poor thermal transfer)
  • 1.1 - Addition of 5V fan to heat sink (which is still glued on)
  • 1.2 - Larger heat sink PC Northbridge heatsink (pin connection), 5V fan soldered to board's 5V regulator

Model AA-AB32181 ($44.99) (PDF Manual & Schematic)

Tripath TK2050 Datasheet

Please note that this amplifier is already bridged, you will not be able to bridge the two channels to make a mono 200W amplifier.

[h=More volts]%3[/h]
It’s been reported by a number of members that increasing the voltage delivered to the amp improves the sound. The TP2050 chip has a maximum voltage of 36V, so it is not recommended to exceed this rating unless you are ready to perform some serious component replacement.

A good switching power supply will deliver a clean enough supply. Sure also sell a number of SMPS from Meanwell and both the 24V 14.6A (can be dialed up to 29V) and the 36V 9.7A (dialled down to 30V) have been reported to work well. Though it's recommended to add a short between the COM and the Ground of the meanwell:

I thought I share this with you:
Mean Well S145-24 with some 10mF and a 0.39 Ohm resistor in series with floating COM, picture taken at 20mV/div and 5us/div compared to COM tied to GND, at 5mV/div and 5us/div.

And I can hear this with a B1 buffer in front of the amp.



I wonder what the scope pattern would look like with the power supply actually powering the amp at high listening volume and with the supply outputs left floating as in the poor example? Would the amp pull the negative supply leg to audio ground and stabilize it with out the need for the jumper? It would also be interesting to see if the music power demand can modulate the supply at all. The MeanWell supplies sound very stiff to me.

The effect is swamped by the switching noise of the amp feeding back (through almost 20mF, mind you) to the PSU. The first attached pictures shows the scope reading taken at the PSU output terminals with the amp powered but idle at 20mV/div and .5us/div. The second with a 85Hz signal at quite loud, 20mV/div 5ms/div. Both taken with the cap btn COM and Earth in place. If the cap is removed the result look slightly different, but the signal basically spans the same voltage.



So that's close to 40mv ripple? coming back from the 650kHz and almost as much from the 85Hz which would be the .012 seconds per cycle wave in the second shot. Doesn't look all that impressive but the MeanWell power supplies certainly sound great. You added an additional 20,000uF capacitance? I would have thought that would knock the 85Hz ripple down more than that.

I replaced the tank caps on board with 4 x 1800uF Pana FM and I have 3 x 3300uF Yageo low-ESR caps and a 0.39Ohm resistor btn board and PSU.

Wow! That is crazy that all that ripple, even a simple 85Hz tone, can walk right through all of that filtering especially with the resistor.

I would be curious to see the phase versus the speaker output on a B channel. Is it possible that the inductors in the power supply are pumping like a current source and actually boosting the voltage as demand from the signal modulates the positive rail? The patterns don't seem to jive with the sound improvement I get when changing my linear supply to the MeanWells. They are in a whole other league. Much faster and more dynamic sound.

Several forum members prefer to, and have successfully run the amp from a battery, however a word of caution:

Batteries tend to be low impedance at low frequencies - and DC. But at higher frequencies the impedance can climb pretty high.

A class-d amp is going to be throwing a lot of high frequency noise back at the power supply. Battereis don't always handle that well. But you can always decouple them and add filtering, just like any other power supply.

(Emphasis added)

[h=DC Offset]%3[/h]
It’s been noted that a few of these amp’s have shipped with a DC Offset that is outside of the +/-10mv that Sure states on the spec sheet. This may be because different input voltages produce different offsets and Sure may have tuned the amp using a different input voltage to you. Either way it is a good idea to check out and tune the DC offset yourself to ensure that your chosen voltage doesn’t throw your speaker cone into orbit.

Please note, to adjust your DC offset accurately you'd need an oscilloscope. The high frequency switching in a class.D amp will confuse most digital multi-meters giving you a constantly shifting range of a few millivolts. An analogue meter may be less sensitive and therefore easier to get that 0mV offset.

TODO - Check details & numbers
  1. Place a load (X ohm - I just used old cheapo 8 ohm speakers ... wonder if different load would change offset?) across both channels
  2. Switch your multi-meter to 200/2000mV sensitivity
  3. Power on the amp with no source
  4. Measure the voltage across the output terminals.
  5. Adjust the resistor (R? - TODO number and picture) till the meter reads as close to 0mv as possible.

Note, as with Sure’s other Tripath amplifiers it is possible to change the resistors (R5 & R26) to a higher turn variable resistor to gain more control over the adjustments, however unlike the 4x100W amp this isn't as mandatory as the resistor seems to allow for a reasonable range of offset adjustment, though as it is REALLY sensitive you might just want to do it for your sanity.

[h=Heat Issues]%3[/h]
TODO: Brief discussion of heat issues with Amp leading to additional revisions & active fan; however active fan leading to some interference in sound.

It’s been noted that this little amp runs hot, very hot. In fact some of the first owners had amps burning themselves out because of the heat. Sure have released a number of versions mostly aimed at addressing this issue. The current version (1.2) is supplied with an actively cooled heatsink (Sure use an MSI branded heatsink normally used for cooling a computers northbridge chip) which keeps the amp at a far more reasonable ~45C.

Whilst the change of heatsink appears to have helped with the heat dissapation it is just curing the issue though, not the root cause for the heat. Further the introduction of a fan powered by the boards +5V has introduced noise to the signal path.

Some forum members, such as Panomaniac, Elifish and v-Bro, have speculated around and attempted to find the cause of the heat, any success they have will be reported here.

Having worked with the TC2000 controller a bit, I know it's a nice chip. But it does seem to drive the output chips hot -or sometimes just warm - depending on the chip.

I suspect this has to do with the BBM (dead time) being too short. Pins 25 and 26 contol this, tho the datasheet does not say how. Usually both pins are pulled up to 5V. I would guess that pulling the pins low or open would change the BBM, just don't know how.

Will have to do some digging to find out!

... I think that big mistake is that the config pin(24 on TP2050) is connected to the ground instead of connecting it to the vdd pins(21,22 on TP2050). Now these two pins are connected on the 5V line via 10k resistor. I increased value to the 47k but nothing significant happened, I was hopping that I will reduce heat but without any luck.

Regards Ales

[h=Chained SMPS issue]%3[/h]
TODO: More detail on this and what has been investigated
Audible humm (at 1ft) when using one SMPS to power more than one board

Producing near audiophile quality sound at a budget

TODO: Copy table from Sure's site ... adjust data for table, or table for data?
border="0" cellpadding="2" cellspacing="0"
| class="tcat" | Name
| class="tcat" | Description
| Volatge Gain || value
| Bandwidth || Value
| Phase || Value
| Sensitivity || 8 Ohm: Value, 4 Ohm: Value
| S/N Ratio || Value
| Rated Consumption || 100W
| Idle Consumption || 100W
| Harmonic Distortion || %THD or plot
| Harmonic Spectrum || FFT plot/plots
| Square/Step response || Oscilloscope Trace
| Stability || Phase margin + scope traces into capacitive loads
| Output configurations || Can it be bridged? Parallel drive?
| Output Impedance || Value or plot/plots

  1. Take layout image and highlight component areas, e.g. 5V regulator, Power caps, input caps, feedback resistors, HBR, Common model filter, Zobel network (missing) and differential mode cap.
  2. Take schematic out from datasheet PDF

[h=Bill of Materials]%2[/h]


TODO: Need to pull in respective lists from Audio1st, Dr Vega and Sendler. Audio1st never pulled together a priority list as such, he just full out replaced all caps on board + a powerline mods etc. Dr Vega hinted at replacing some of the SMD components making up the feedback section of the amp ... need to ask a question about that.

There are a number of modifications which can be performed on this amp, both in the signal line and for the amp's power. These are listed below.

my take:

- Input caps
- tank caps (and thus depending on your PSU also a soft start)
- Input and feedback resistors
- output filter incl. toroids
- voltage regulator
- re-route power rails on PCB and remove diodes (amp likes every bit of voltage)
IV. Might not be worth doing
- re-wiring pins 24 on both TP2050s to 21,22 (Vdd) instead of gnd
- exchange C22&27 for 470pF as per the TC2000 data sheet

A very good list, ElFishi. I think I'd add the Zobel filter to group I or II, more for safety than sound quality.


I have some spare time while I wait for the new sure board to replace the one I scorched. So I completed the list of mods and things around the board I did. Maybe you find that interesting.

PSU Mean Well 24/145:
- cranked up to almost 30V
- 0.47uF MKS btn GND and COM

Sure board:
- changed PCB traces per audio1st suggestions (
- removed D1,2
- added 0.1uF parallel to C6 from the underside of the board
- exchanged C22&27 for 470pF as per the TC2000 data sheet
- re-wired pins 24 on both TP2050s to 21,22 (Vdd) instead of gnd
- removed tank caps from rails and replaced with 2 x 2 Panasonic FM 1800uF/35V on the rails
- replaced C3 with same type Pana FC
- added 0.1uF/100V MKP4 parallel to C3
- replaced C4 with standard-middle-of-the-road cap 100uF/10V
- replaced R2 with 2 green low-current LEDs, V_F=2.0V
- replaced R3 with 470Ohm
- added diodes 1n4001 btn Vadj/Vout and Vin/Vout
- replaced output filter

12uH self-wound toroids: T68-2 with 46 windings of 0.6mm enameled wire, sets of 2 plain windings followed by 1 clove hitch to double up the wire layer on the inner perimeter of the toroid
.47 uF/250V MKP4 as common mode capacitors
.22uF/630V MKP4 (had those in the box) as differential mode capacitor
.22uF/250V MKP4 as Zobel capacitor
10 Ohm/1W Metal Oxide as Zobel resistor

- replaced R16,30 with 22kOhm metal film
- interrupted traces btn input RCA and screw terminal
- removed R11,27 and replaced with 22kOhm metal film connected btn old solder pad and screw terminal
- removed R4 to open loop over SignalGND, Earth and B1
- added modified Pass B1 buffer ( Buffer Preamp.pdf, to input
- 1nF/1000V MKP4 from speaker terminals to chassis

B1 specs
- R1 2.4 Ohm
- C1,2 2 x 3 x 3800uF/35V Panasonic FC
- C3 1uF MKP-10
- Cx00: 1uF MKP-10
- Cx01: 2.2uF MKP-10, directly linked to input resistor of sure board
- Rx05,x05: dropped
- 50k Vishay conductive plastic, log taper
- added 1N5401 to input of positive rail, bypassed by a 1k resistor for a controlled discharge of the caps (otherwise the sure amp makes an ugly noise when going down)

Soft Start/ Mute On Off Circuit
- 24V/0.4W relay 8A 2CO contacts (tyco RT424024)
- 2 white LEDs (~6V drop) in line with coil
- 47 Ohm/ 5W resistor to fill the tank caps up to operating voltage for the relay
- 1N4148 parallel to relay coil
- 0.39 Ohm/5W resistor to filter the power supply and to sense the current
- TL081 as comparator (TL081 can work in high-side applications)
- 2 x 10k voltage divider to drive a
- BS170 that mutes the amp when power is switched off and the tank caps unload via the SMPS and the current sensing resistor
- 100uF btn MUTE and GND to keep the amp mute a little longer
- 3 x 3300uF/35V Yageo SY tank caps
- 1800uF/35V Panasonic FM next to output (close to amp power input)

piccie of the board w/ toroids and schematic of the "mute on off" circuit attached

[h=General Modifications]%2[/h]
A number of modifications to the amp generally which whilst not improving the sound will help with other things, such as heat.

[h=Change the heatsink]%3[/h]

Many users have reported good results from changing the heatsink to a larger passive sink. The Zalman ZM-NB47J is known to cover the area required well, though as Sure use a northbridge heatsink (a component on a modern day computer) you may experiment with any northbridge sink to find one which meets your space requirements. The Zalman sink though has the largest underside surface area which makes contact with the chips at 37x37mm and is known to cover the chips.

These Zalman ZM-NB47J heatsinks are cheaper, cover the whole chip and run at least as cool at 44C even though they look smaller. $7.

TODO: Heat sink fan (advice for removing - version 1.0 (1.1?) was glued)

Heat up the heat sink with a hair dryer until you are able to twist the heat sink off (be careful touching a hot heat sink, you'll want appropriate protection, or you'll have a nice imprint of the head sink on your palm. Also make sure you don't twist the chips off, so not too hot, eh?)

Remove the plastic fan. Use some hot air from a blow dryer or a heat gun carefully on the heatsink. The glue looks more like a silicone rubber than an epoxy so it does eventually respond to some heat. Pry up on the heatsink with a moderate amount of pressure, well before the point of breaking or digging into anything and start working some heat on and off of the heatsink to slowly start bringing up the temp until the glue finally softens and it comes off.

[h=Signal Path Modifications]%2[/h]

Below are a number of modifications which can be performed to improve the components on the signal path, this should lead to an improvement of your sound.

[h=Off board input capacitors and volume pot]%3[/h]
TODO pull audio1st pic from #242


[h=Removal of "bonus" diodes]%3[/h]
There appear to be two diodes like soldered to the bottom of the board underneath each of the phono input connectors which aren't listed on the schematic. Members have reported an improvement in the sound having removed these.


BJ 6CA are TVSs (Transient Voltage Suppressors), something like diodes that clip signals over 6V. BTW it seems that TC2000 can handle only 1,5Vrms... (look at the datasheet)

[h=On-board Input Capacitors]%3[/h]

  • Remove phono connectors
  • Remove capacitors C16, C17, C24 & C25
  • Take 2.2uF capacitor of choice and solder between +ive of Phono connector (center pad) and -ive of C16/C24 (circular pads) respectively.
  • Burn in capacitors and enjoy sound.

While we’re at it, Col’s point applies to the input filter, too. The input cap blocks DC from entering the amp and being sent to the speakers, where DC will burn up the voice coils real fast. Also, since Tripath amps put a 2.5vdc bias on the signal input, the input caps also keep that bias DC from feeding back into your pre-amp or CD player, where it might cause harm.

So we need input caps to block DC both directions. The input cap works with the input resistor to make a high pass filter. The input resistor sets the gain in combination with a couple of feedback components. Everyone seems to accept Tripath’s input resistor value, which is around 20-22kohms. With the value of the input resistor fixed, we can see what various values for the input cap give us for the input high pass filter. Tripath says it should be below 10 Hz so it doesn’t roll of the bottom end.

Tripath pretty consistently recommends 2.2uF for the input cap. Sure uses a 1uF in parallel with a .22uF for a total of 1.22uF. I think they did this to save space. The 1uF is a tiny surface mount and the .22uF is a higher quality polypro film. So lets check the filter values these produce:

2.2uF = about 3.5 Hz
1.22uF = about 6.5 Hz
.22uf = about 35 Hz

So we can see that Sure’s 1.22uF is less than 10 Hz, well “within spec” for Tripath. Even the better quality polypro cap is passing audio down to 35 Hz.

The value of the input cap isn’t critical. Anything above 1uF is going to get you below Tripath’s recommended 10Hz. I’ve personally used everything from .47uF to 10uF on Tripath amps.

Since the audio signal goes through the input caps, the quality of the cap is far more important than the value. Get really good polypro film or foil or paper-in-oil, or whatever you like, but get the best you can find.

I’m partial to “Vitamin Q type” paper-in-oil (PIO) caps and think they sound great on Tripaths (real Sprague Vitamin Qs weren’t made over .68uF, but similar ones are available in larger sizes). On my current Sure 2*100W, I’m trying Dayton polypro foils. They sound great, too.

To me, DIY is about experimenting, not building to formula. Understand what is needed and why, and then try different solutions to see how they sound. I’m as much interested in the knowledge as the amp.


[h=Modulator Feedback]%3[/h]

While browsing the Tripath docs, I came across the section on Modulator Feedback. One resistor, Rfbc, needs to be matched to the supply voltage "for best signal-to-noise ratio and lowest distortion."

On the Sure 2*100watt board these are resistors R13, R17, R25, and R31 (one for each output from the TC2000 chip).

The board comes with 15kOhm resistors, which are correct for a 32v supply. Applying Tripath's formula gives these values for other popular voltages:

24v = 11k
26v = 12k
28v = 13k
30v = 14k
32v = 15k
34v = 16k
36v = 17k

Since the audio signal passes through these resistors, you might want to replace them with some nice metal films, like the Holcos that PartsConnexion has on sale cheap right now.

While you're upgrading resistors in the signal path to metal films, you also might want to upgrade the input resistors (R11 and R32). Keep their values at 22kOhm.

If you run with any of your gain switches "On," upgrade R18, R19, R20, and R21, too, keeping their 22kOhm values.

I have not yet done any of these updates, but I've ordered the Holcos. I don't expect them to make a lot of audible difference, but who knows, right?


... Another thing I noted, however, with the sure board is that they ignored the recommendation to offset the values of the feedback capacitors (c19,21,22,27) for each channel. I exchanged C22&27 for 470pF as per the TC2000 data sheet. I can't hear a difference, but I feel better now...

The HBR caps are high frequency bypass caps to filter the DC power supply just before it enters the chip. They filter out (mainly) any RFI AC that the board traces may have picked up.

The larger caps that Sure uses just filter down to a lower AC frequency than the Tripath spec. Since all you want is DC, filtering lower frequency AC is good, but probably unnecessary since there's unlikely to be any.

This is an example of Sure giving us something better than the Tripath spec. Changing them back to the Tripath spec actually "degrades" the amp, although it is unlikely you will hear any difference.

Another way to look at Sure's larger HBR caps is to consider them as mini-tank caps that provide a little extra power reservoir very close to the chip. A common mod on the T-amp was to provide a lot more capacitance right there, usually around 680uF. So if you want to replace your HBR caps, don't replace them with Tripath's .1uF, replace them with 680uF monsters!

I plan to add extra capacitance very close to the HBR caps, but I'm going to do it on the back of the board where there's more room.


C19, 21, 22, 27. How critical are these to the sound? Should we upgrade?

IMHO all upgrades you want to do are worth doing, just for what you learn. As to improving the sound, I doubt if it will make an audible difference. They are part of the feedback loop and not directly in the path of that. And smt caps tend to be very fast. I doubt if you could improve them much, perhaps going to Oscons. You might want to upgrade resistors R15, R22, R28, and R33. They also bleed off the feedback loop.

But the real gains in the feedback come from replacing resistors R13, R17, R25, and R31. They are directly in the feedback signal path and should be matched to your power supply voltage (see my previous posts).

Anyone else want to weigh in on replacing surface mount caps. Can you improve them?


TODO: Output caps, inducors & zobel network
[h=Swapping out the Output Capacitors]%3[/h]

The cutoff frequency of an LC-low pass filter is
1/(2 \pi \sqrt(LC))

The TP2050 datasheet says
"The TP2050 works well with a 2nd order, 80kHz LC filter with LO = 10uH and CO =0.47uF".
That's what I am going to try (haven't started yet), even if it gives a cutoff of 73KHz.
The schematic in the same datasheet lists 15uH and .22uF resulting in 88kHz cutoff. Higher, but maybe not decisive.

The sure board otoh is stocked with 22uH and .68uF yielding a cutoff at 41kHz. Given that the attenuation of the LC filter is sloping and starts well before the cutoff the fear is that the filter transmission could affect audible highs.

Of course Col is right. Tripath’s data is not carved in stone. These are all compromises to balance various factors. Tripath tells you what you need and gives you a solution that will work. It’s not the only solution possible. For example, when you read the Tripath info about the output filters on the spec sheets for different Tripath chips, they talk about the same problem, but even Tripath comes up with difference solutions on different datasheets.

The Tripath chips, like all switched amplifiers generate a lot of electrical switching noise, most of it above 100kHz. While this is well above the audible range, it can modulate with other frequencies to make audible distortion. Plus, it just isn’t good to send the noise along to speakers sucking up watts and heating your voice coils. So, Tripath recommends a second order filter at around 80kHz.

We’re lucky with Tripath because their switching frequency, while variable, is usually above 100kHz, which is higher than some of the other class D amps. The higher frequency means we can use a simple second order filter that is cheap and doesn’t harm the sound much.

On different datasheets Tripath recommends LC combinations of 10uH/.47uF and 15uH/.22uF. Sure, apparently, used 22uH/.47uH. These give the following filter points:

10uH/.47uF = 73,412.76 Hz
15uH/.22uF = 87,611.99 Hz
22uH/.68uF = 41,148.57 Hz

I replaced the Sure 22uH coils with 10uH, but I have not yet replaced the caps. That means I’m currently running with a filter point of:

10uH/.68uF = 61,033.19 Hz

This is one of Tripath's recommended combinations. But when I replace the caps, I’ll use .33uF instead, which gives me a filter point of:

10uH/.33uF = 87,611.99 Hz

That’s the same value I’d get if I had used 15uH/.22uF.

All of these combinations work just fine. Sure may have chosen a lower frequency to filter out a little more noise, at the risk of rolling off the top end of the audio spectrum a little – a very little, less than 2dB at 20kHz for 4 ohm speakers and effectively none for 8 ohm speakers. And remember, we're talking about the tweeter here. Even if your tweeter is 4 ohms, chances are your crossover has a resister on the tweeter bringing its effective impedance near 8 ohms, maybe more.

Since I’m using a 4 ohm tweeter without a crossover (I have an active crossover), I’m pushing my filters up a little higher. But only for the psychology of it, since I can’t hear anything above 16kHz anyway.

So Col’s right: don’t get crazy over the output filters. What Sure did works just fine. I didn’t replace my coils to move my filter point, I did it to get better quality coils. Upgrading the coils gave me a chance to play with the filter frequency, but that was a fun side effect, not the reason to do it.


[h=Homemade Inductors]%3[/h]

It wasn't too hard. A little time consuming but not any more so than winding a magnetic core. I started with 60 turns on a 1/2 inch wood dowel. Slide the cylindrical coil off the dowel and shape it into a toroid by spreading the coils on one side to become the outside of the toroid. Small tie straps hold the ends together and then some spray automotive clear to give it some strength. I am going to try some 14guage litz wire with a special low profile cable tie (7483K35) inside the toroid to compress the inner diameter a little, and one more tie around the outside before the clear to make it solid enough to reduce magneto-striction and microphonics.
Use the calculator to get your value in the ball park. Compute the value for a certain number of turns at the outer length (circumference) and the same turns for the inner length and your value should be 1/2 way between those.
Coil Calculator - Single-layer and mutil-layer coil calculation in javascript
If you want more inductance, you can increase the dowel size to 3/4" and make it to a final value toroid of 10uH with 62 turns. A 1 inch dowel will get you all the way up towards 18uH with the same outer diameter that I have.

[h=Off the shelf Inductors]%3[/h]

Arjen Holder's inductors (10uH 3A) from eBay.

The TA2022 docs are talking about the saturation capacity, not the power handling capacity. If fact, they recommend 22 gauge wire, which has a continuous power capacity of less than 1 amp.

They clearly state that bobbin style coils can't meet their 10A peak saturation requirement. They recommend Micrometals type 2 toroid cores. That is the core that Arjen uses.

The stock Sure coils are bobbins. Replacing them with Arjen's Micrometals toroid coils brings the board up to meet Tripath's most strigent requirements.


According to Tripath, the core material is critical. Those don't say what they are. Here's the quote from Tripath's datasheet for the TA2022 (which has the same requirements as the TK2050):

"Output inductor selection is a critical design step. The core material and geometry of the output filter inductor affects the TA2022 distortion levels, efficiency, over-current protection, power dissipation and EMI output. The inductor should have low loss at 700kHz with 80Vpp. It should be reiterated that regardless of the systems maximum operating current, a 10A rating is required to ensure that peak current conditions will not cause the inductor to saturate. During a short circuit event the inductor current increases very quickly in a saturated core (see figure 6), compromising the current protection scheme. A 10A rating is sufficient to ensure that current increases through the inductor are linear, and provides a safety margin for the TA2022. There are two types of inductors available in the 10A range that offers some EMI containment: they are the toroidal type and the bobbin (shielded) type inductor.

In bobbin construction, a ferrite shield is placed around the core of a bobbin inductor to help contain radiated emissions. This shield can reduce the amount of energy the inductor can store in the core by reducing the air gap, which can lower the peak current capability of the inductor. Typically, a 7-10A shielded bobbin inductor will not have the peak current capability necessary to ensure that the core will not saturate during short circuit events; this is why they are not recommended for use with the TA2022. Also it should be noted that shielded bobbin construction is not as effective as toroidal construction for EMI containment.

Tripath recommends that the customer use a toroidal inductor with a Carbonyl-E core for all applications of the TA2022. This core has a high peak current capability due to its low-µ Carbonyl-E metal powder. A distributed air gap increases its’ energy storage capability, which allows for a small footprint and high current capability. Carbonyl-E toroidal iron powder cores have low loss and good linearity. The toroidal shape is ideal for EMI containment. Also, EMI can be further contained by sizing the toroid to accept a full layer of windings. This aids in shielding the electric field. Tripath recommends:

- Micrometals ( Type-2 (Carbonyl-E) toroidal iron powder cores. The specific core Tripath initially verified and used on the EB-TA2022 was a T94-2 (23.9mm outer diameter) wound to 11uH with 19AWG wire. Since then Tripath has determined that much smaller Carbonyl-E toroids will not saturate during high current events. Tripath has
also used T68-2 (17.5mm outer diameter) and the T60-2B/60(15.2mm outer diameter) cores wound to 11uH with 22AWG with good success. If a smaller core is required, core outer diameters as small as 15.2mm (T60-2) work well, but core temperature effects should be tested. The T60-2 core did not saturate during short circuit testing, but maximum core temperatures must be considered and multiple layer winding must be used to achieve 11uH. Multiple winding can increase winding capacitance, which may cause ringing and increased radiated emissions. Bank winding techniques can minimize this effect. It should be noted that at core temperatures above 130C the single build wire used by most inductor manufacturers should be replaced with a heavy build wire. Micrometals does not provide winding services, but many companies purchase directly from them and provide completely finished inductors. Pulse Engineering has assigned a part number for the T68-2 wound with 44 turns of 22 AWG single build wire. The part number is PA0291.

- Amidon Inc./American Cores type-06 (Carbonyl-E) toroidal iron powder cores. Tripath has used T690-06 (17.5mm outer diameter) cores wound to 11uH with good success. Amidon carries type-06 cores in the 23.9mm to 15.2mm outer diameter range. They have assigned a part number for the T690-06 wound with 44 turns of 22 AWG single build wire. This part is approved by Tripath and is 690064422."


I was away for a bit and technically still am.
I just wanted to chime in with a few remarks:
I did finish my self-wound toroid cores to make an output filter with 12uH inductors, common mode caps of 0.47uF, a differential mode cap of 0.22uF, and a Zobel of 0.22uF and 10 Ohms. It was the result of some simulating with pspice. I don't have the results at hand but can post them in a few days.
The transmission of this filter when simulated with a 4 Ohm load is flat up to 20kHz and then goes down.
My simulation results with the filter design posted by swkbkk looked less appealing and I didn't try to build the filter. The transmission has an overshoot starting around 10kHz, IIRC, with a maximum around 40kHz (not much - a dB or so) and less attenuation at the higher frequencies.

[h=Zobel network]%3[/h]

The schematic from Sure details a Zobel network (physically located ahead of the Inductors and under the heatsink). Sure however have elected not to include this network on the boards, which would consist of four parallel resistors (47 ohms each) and a capacitor (0.47uF). This has lead to the highs on the board not sounding great.

You have zobel caps on your board (C34, C48) and resistors? My sure board didn't have zobel components on board, thats why it was a bit rough on highs. After adding zobel filter the sound was much nicer, also change these grey caps(output capacitors) with something better.

Regards Ales

I think that all these boards from sure are without zobel network.

If anyone thinks that amp has too much highs, then he should put zobel filter on the output 10ohm resistor and 470nF cap in series(cap is on positive side).

Regards Aleš

Time to put your own Zobel network in then ...

[h=No Caps:eek:]%3[/h]

TODO: Pages 86-88 have discussion on running capless in very specific situations, i.e. driving tweeters. Revisit and pull out info

[h=Power line modifications]%2[/h]

TODO: Power caps & Soft Start

TODO: post soft start schematic from ElFishi’s post 513, currently redoing in Eagle

An ideal power supply would deliver unlimited clean current instaneously at the rated voltage. Of course no such power supply exists.

We usually run a power supply that is not large enough to supply all the current the amp needs to run constantly at full power. But that's okay because audio signal are variable and don't require constant full power.

My Sure draws about 4.4 amps at full power. My power supply delivers 2.94 amps. So the Sure board has six large caps to store some extra energy and deliver it when the chips need more amps than the power supply can deliver. We call these storage caps or tank caps.

The bigger the caps, the longer they can provide the extra power (we're talking miliseconds, here). If your power supply is underrated and you play your music loud, espcially if you listen to organ music with deep pedal tones, you can certainly use some extra stored power.

But the caps do something else besides store extra power. When the chip wants current, it wants it right now!. It takes time for the current to run down the power cable, go through the diodes and run around the traces on the board before it reaches the chip.

By putting the caps close to the chip (as Sure has done), the current can reach the chip faster. This gives you crisper attacks on percussive and other sharp sounds.

As you might expect, not all caps are created equal. And the properties you want in tank caps are not the same as what you want in signal caps. You want tank caps that are basically big and fast. Fast caps help with detail in the music and also filter power supply noise better.

Favored brands/models are Panasonic FM or FC, Elna Silmic, and Blackgates. Given that you need at least 50v caps, the Elnas are out (they only go to 35v). People love Blackgates, but I've never used them, they're too expensive for my taste.

That leaves the wonderful Panasonics. The FMs are newer, better, and cheaper, but come in limited sizes. The FCs are the old standbys. They're great, cheap, and come in any size you want. Digikey carries Panasonic and Elna.

So, adding a big storage cap at the power supply connection is easy and will help give your bass more foundation, depth, and power. But it won't speed thing up much, even if you use a fast cap, because it is on the wrong side of the diodes and a long way from the chip.

I'm going to try adding a couple of Panasonic FM 680uF 50v caps on the back side of the board, right under the TP2050 chips.

Don't forget that also adding some fast extra capacitance on the 5v lines will help both the TC2000 and TA2050 chips. I haven't looked to see where to do that, yet.

So, if you do this, how much difference will you hear? I expect it will be subtle, but real. Sure has already done real well, giving us more storage capacity than the Tripath evaluation board has. They also used multiple smaller caps which are faster than one big one, and they placed them well on the board, close to the chips and gave us an extra 6.8uF right at each of the chip pins.

But who cares if we make the sound better. We don't really do these upgrades to improve the sound, do we, we do them because we can!

One last caveat before you add 500,000F to your board. When you first turn on your board, the power supply sees the storage caps as basically a dead short. If the power supply has to fill huge amounts of storage, it will complain, maybe even die. If you want to add massive amounts of storage, you should implement a soft start circuit to protect your power supply.


[h=Re-routing power lines]%3[/h]
TODO: Flesh out the text.

The Sure amp provides an element of flexibility by having power blocks on both sides of the board. This not only gives you a choice about which side you plug your power in depending on your enclosure, but also allows you to daisy chain a number of amps together in series (though I don't know of anyone who's tried this or what impact it has on the sound).

If though you have placed your amp and don't intend of using the extra power block it's actually making your power lines longer than they need be. Audio1st posted a mod where he has re-routed these power lines in order to minimise this distance and replaced a lot of the capacitors in the power line to lower the ESR.

This is a one way mod as it involves cutting the traces, do this and you won't be going back ...

- Removed C1 and C2 and replaced with a single 100nF Wima film cap
- Moved C6 to underside of board. Replaced C4 with a 100nF Wima film cap.
- *Added a 10uF Pany to the adjust leg of the 5V reg, detail of the 10uF/50v cap in parallel with R2. Position to adj pin1, neg to C3 ground below:

- Removed D1 and D2 and re-routed the power supply.

The suggested re-routing

The original path of current on the board:

The re-routed path of current on the board:

*Audio1st later replaced C4 to a 100uF Panasonic FM, this has now been changed to a 10uF Tantalum. The reason behind this was an article he read:

That's surprising! One would expect that the Panny, with 10 times the capacitance and lower ESR would give a more stable, solid, and quicker power source.

Any ideas why the tantalum sounds better?


I am like you and read lots of info on the internet, it can be conflicting sometimes.
Anyway, I found the link below which if I made any sense of it said that using a low esr cap on the output of the reg was a step in the wrong direction. Panasonic FM quickly removed..
Now it sounds better with the Tantalum to me, but may have been even better with the original

Sendler and v-Bro later disagreed with this point.

I saw that. I disagree totally. I've seen many tests of caps for digital supply and a stack of electrolytic/ Oscon/ ceramic always measures the quietest and stiffest. Oscon and chip cap must be placed within millimeters of the device and have a good ground.


The 5V reg. I would certainly think benefits from a low ESR cap, the suggestion of higher resonance frequency being a negative influence to the stability I have serious doubts about too. Actually I think the matter is way exaggerated.


Read about it, try it out and take what you think sounds better.

[h=Experimental Mods]%2[/h]

A section for mods which have been theoretically proposed, but no-one (that we know of) has stepped up to the plate and given it a shot (yet).

[h=More power]%2[/h]

TODO: there has been a discussion a few times, started around page 83, but again later on in the thread (past the 120 mark IIRC, involving v-Bro) talking about swapping out the TK2050 power chip and replacing this with a T.I. chip (?). This would require some serious component replacement (possibly not possible on the board size) to allow for the extra voltage, but the suggestion was to be able to drive this board at around 60V

important remark by the mods..

one comment her: the "EVA" statement form the forum.
Congratulations, you have destroyed the amplifier.

It's still working because it's low power and it can handle the huge amount of parasitic inductances that you added (for the moment ), but electromagnetic radiation may have increased around ten times.

If you did similar modifications to any high power class D circuit (or something less forgiving like UcD), it would explode next time the volume is turned up for some time.

The most ridiculous part are the big class D output capacitors attached with wires to the PCB. This filter is intended to provide 12dB/oct attenuation up to 10Mhz, to prevent RF from reaching speaker wires (antennas!), but now it's giving up below 1Mhz due to the parasitic inductance that you added.

Replacing three SMD electrolytics, which are paralleled for reducing inductance and preventing resonances, by a big axial electrolytic, which is the highest inductance type, is also an award winning dumb idea. Inductance is typically 1nH per mm of terminal spacing. The big axial cap is around 20 times more inductive than the three small ones in parallel (which where there for a good reason).

The huge input capacitors are also funny because they act as antennas and pick up all the RF radiated by the rest of the circuit due to the modifications (plus any ambient RF willing to disturb the input op-amps).

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.