Touch hardware mods

[SBGK]

pictures requested by HM, 5 x sanyo, Oscon SEPC 470uF 6.3v (6x9) + micro clock

[HumanMedia]

Thanks SBKG, I can see exactly whats going on now.
After poring through Caad's notes and compiling them together, I can see theat there are quite a few more hardware tweaks worth applying.
I know that SBKG doesnt want any more TT3.0 settings or more hardware mods, but there is still much room for improvement..... :-)

Also very interesting that Caad measured the jitter of the stock clock to be very low, but marred by lack of regulation on the clock power and the implementation of the hex inverter. Maybe the big gains from the added clock board is the onboard voltage regulation and maybe not the clocks themselves?
I'm also wondering if Caad's comments about keeping the leads as short as possible apply to ALL bypass caps and not just the hex inverter at U14 or the SPDIF output at U34?
He states that each 1mm of the cap lead will introduce around 1nH inductance which will end up as jitter
So should all bypass cap leads by trimmed as short as possible?

Now can anyone inform me on where to buy these tiny SMD caps that are all through the Caad mods?
Do the Mousers of the world sell these in small quantities??

[UV101]

Yep, mouser sell caps..... ;-)

[UV101]

?

[HumanMedia]

The art here is not just blindly copying, but in understanding what each individual compnent does and then how best to treat it and this is path I'm treading.......

I hear you. I am a large bounding audio dog. Willing to learn new tricks. Yet (too?) eager to get to where the path leads.

Lead the way UV101

[andyjayh]

I hear you. I am a large bounding audio dog. Willing to learn new tricks. Yet (too?) eager to get to where the path leads.

Lead the way UV101

Haha, I can relate to that statement

[JohnSwenson]

For all you brave souls attempting to mod a Touch I thought I would attempt to point out some misconceptions about some of these mods and maybe provide some insight into where you might wish to go from here.

First and formost is about the bypass cap mods I see here. It seems like most people have a misconception about what these are for, and hence a fair amount of mis-application of caps. It seems like the prevailing wisdom is that these caps are there to filter out noise from the power supply. This is not true. The actual reason takes a little bit of explanation which I will attempt to cover. Its a little esoteric which is why its probably not talked about much.

First some background. A digital signal is frequently called a square wave, when you look at it on a good scope you see steep edges and sharp corners. If you look at such a signal in the frequency domain you find a series of harmonics of the main frequency going many orders of magnitude higher than the main frequency of the square wave. (most of you will recognize this as the infamous Fourier series) These high frequency harmonics have to be preserved from the transmitter chip to the receiver chip in order to preserve the steep edge of the square wave. For example in the Touch you have a 22MHz clock, in order for that signal to be received with low jitter at the receiver you need to maintain the harmonics to at least 200MHz, preferably higher.

What happens when chip A sends a square wave to chip B? Its not just voltage, charge moves from one chip to the other to charge (or discharge) the capacitances in the receiver and the wiring between them. Moving charge is called current, it goes in loops, its not a one way street. If current flows out of the transmitter to the receiver it HAS to flow back from the receiver to the transmitter somehow. This is usually refered to as the "return current". Unfortunately this return current is frequently ignored, but it turns out to be one of the most important aspects of high frequency digital design.

If return current is even thought about at all its usually assumed to "go through the power supply", but think about what this means. The PS has lots of wire, traces on the board etc, all this means a lot of inductance. Now go back to the 200MHz of harmonics on that clock signal, whats going to happen when it is forced to run through all that inductance to use the power supply as the return path? It gets majorly messed up! (sophisticated technical term) The higher the frequency the more it gets impacted. The result is that our beautiful sharp edges are not sharp any more when the higher harmonics get messed up by the inductance in the return path.

So what is the solution? The ground plane. It provides a very low inductance path for return currents to go between chips. But it just connects to the ground pin, how does the return current get to the power pin? The infamous bypass capacitor! The cap blocks the DC, but the high frequency signals from those sharp edges go right through the cap and into the power pin, completing the current loop. THAT is the prime purpose of the cap, to provide a low impedance path for the return current.

So what are the requirements of this cap? Why do different ones matter? The cap value has to be large enough that the impedance at the lowest frequency of the return current is low. What is low? Well that is one of the major debates in engineering. If you are trying to build very low jitter systems I would recommend that it should be less than 1 ohm. Others will say that 10 ohms is sufficient, but I personally go with 1 ohm. So why not just always use a large value cap and not worry about it? The problem is that caps also contain inductance as well as capacitance, in general the higher the capacitance, the higher the inductance. Remember we want to have low inductance to preserve those high frequency signals.

The result is that you make a trade off, you need a high enough capacitance for the lowest frequency, and the lowest inductance to not mess up the high frquencies. This is where choosing the right type of cap and the right value comes into play. Physically smaller caps have lower inductance, in particular SMD caps have quite low inductance, but its hard to get high capacitance out of a small SMD cap.

The data sheets for different caps will usually show graphs of impedance with frequency which take both effects into acount. These are a great resource for choosing a proper bypass cap. It takes a lot of the guess work out of this, figure out the min and max frequencies in question and find a cap which shows the impedance to be less than 1 ohm over that frequency range.

Take for example the clock circuit in the Touch, the minimum frequency is 22MHz and the max is at least 200 MHz. Look at some spec sheets and you find that a simple .1uf SMD cap easily meets the criterion. There is no need for any higher capacitance since there is no frequency less than 22MHz in the return current. Thus those big through hole caps shown in pictures in this thread are unnecessary. The extra capacitance they offer is not needed, and the extra inductance can actually be harmful. When you parallel caps you wind up with resonances caused by the combination of capacitance and inductance of the two devices. These resonances can actually cause the impedance at certain frequencies to be greater than either cap on its own. Its very difficult to predict these resonances and figure out if they are going to be harmful or not. The best bet is just to not parallel caps if you don't have to.

So what do you do if there is no one cap that can meet the requirements? This is where it really gets fun, now you have to make compromises. Take for example a DAC chip. These deal with fequencies from 20Hz to hundreds of MHz, there is NO WAY one cap can deal with this. The best bet is to break the frequency range into chunks and look at the best way to deal with each. For the low frequency range the inductance of the power supply is not really a problem, so relying on the PS for the low range works well, this is good because if you couldn't you would need a HUGE cap to deal with those frequencies! The very highest frequencies need a small SMD cap. The intermediate fredquencies work best with an electrolytic cap. This is where some interesting tradeoffs occur. The solid polymer caps such as Oscons have a lower impedance at higher frequencies than regular electrolytics. These caps come in several different packages from small through hole to small SMDs. Its possible to make a pretty good tradeoff with an SMD solid polymer that has enough capacitance to handle the low fequency part and not too much inducatnce to mess up the high frequencies. If you can't figure out a single cap that does it all you have to start paralleling caps and try and get around the resonance effects.

In my latest DACs I've actually started doing something that should get me drummed out of the audiophile comunity, it goes against all the sacred truths, I'm actually using Murata 22uf 1206 ceramic caps to bypass DAC chips, I know, The HORROR of it all. These have very low impedance at high freequencies and high enough cap to handle the lower frequencies which are too high for the PS and they do it all in one cap, no parallel resonances. The results sound very good, even though they are ceramic.

Now a little on that clock board I have seen in this thread. It looks like a very well designed board, I would NOT add all those extra through hole caps, the probability is high they will actually make it worse. I hope from the above it's obvious why this would be so. The other aspect of its use is how to connect it to the Touch board. From the above discussion it should be obvious that its important to achive a good low impedance ground connection between the Touch board and the clock board. A thin little wire for the signal and one for the ground is not going to be optimal. A short piece of thin coax would be best, preferably one with a braided shield. Something like RG-174 or better would work well.

I hope all this makes some sense and can lead you to a better understanding of how to make intelligent modifications to digital audio circuits rather than just dropping caps in various places with out understanding why.

John S.

[JIJ3]

Great post, John

I also use the Muratas to bypass all digital chips in my circuits. Also a small SMD ferrite inductor on the supply line to each chip. However, this is on a 4-layer board designed for these things

The problem with mods is, it is very hard to add these caps to an existing board. I have used small SMD mounting boards to mount chips I have added to the circuitry inside the Touch, and then used the Muratas to bypass these chips. Hard to do with chips on the Touch circuit board itself. I just leave them alone.

Jack

[HumanMedia]

Now a little on that clock board I have seen in this thread. It looks like a very well designed board, I would NOT add all those extra through hole caps, the probability is high they will actually make it worse. I hope from the above it's obvious why this would be so. The other aspect of its use is how to connect it to the Touch board. From the above discussion it should be obvious that its important to achive a good low impedance ground connection between the Touch board and the clock board. A thin little wire for the signal and one for the ground is not going to be optimal. A short piece of thin coax would be best, preferably one with a braided shield. Something like RG-174 or better would work well.

Huge thanks for your time and knowledge, it really does help a lot (especially for noobs like me).

As far as I know the existing oscillator modules (those silver components) are removed from the SBT board and the clock board fits into some of the holes that are left.
Also when you talk about the "through hole caps" does this mean the 3 caps that are added to the clock board?

[UV101]

?