The Uber-Controller and DynaFET Build

[j4cbo]

I've been lending a hand with board layouts for a few projects, notably for a build of Kevin Gilmore's DynaFET design, and doing so made me realize two things: (a) there aren't any DIY projects for amps with a truly high-end feature set, and ( I want one, but I can't really afford it. The DIY crowd has some amazing analog designers and has produced some really astounding designs, but the supporting and surrounding circuitry often isn't much more than a pot and maybe a selector switch, and sometimes a power LED or relay control circuit. There's so much more that could be done to make amps more convenient, self-monitoring, and just plain nifty. So I decided that this was an opportunity to take matters into my own hands.

To that end, I'm designing an input, attenuation, and control board which I'll be using in some amps for myself and Icarium. We're doing this as a a collaborative, long-term project, such that I'm contributing much of the design and construction effort in return for having parts for my own build subsidized; I could likely complete four or so of these by the end of the year, and so we're looking for other people who are interested in financing the build and getting a completed amp in return. The final cost will probably be somewhere in the $1500-$2500 range for a finished amp, depending on what parts, design, and case decisions we wind up making. Send one of us a PM (soon!) if you're interested. I also plan to release all the designs, code, and whatnot for the controllers as I develop them. This isn't a commercial, money-making endeavor; I'm in it to get myself the coolest amplifier I possibly can, whatever that involves.

Without further ado, here's the feature set:

- Four selectable inputs; each can be configured as either single-ended or balanced.

- Two Sowter input transformers will be switched in as needed, to convert single-ended inputs to balanced, or vice versa. If the source and headphones are of the same configuration, the transformer can be switched out of the signal path entirely; otherwise, it is used to convert the signal as necessary. If all the inputs and outputs will be balanced, the transformers can be left out entirely to save on build cost.

- A four-channel (true balanced stereo), 7-stage, 128-step relay attenuator (similar to the Twisted Pear Joshua Tree) for level control. The output of each channel of the attenuator is connected directly to the corresponding amp board.

- Four DynaFET boards. The controller board could also be used for other amplifier designs, though, such as the Beta22 or Dynahi, but this build will be based on the DynaFET.

- Both unbalanced and balanced output jacks; the full configuration would include 4-pin XLR (balanced), 2x 3-pin XLR (balanced), and one or more standard unbalanced TRS jacks.

- A monitoring board to sense current on each amplifier's power rails, rail voltages, output offset voltage (differential voltage in balanced mode), and supporting remote temperature sensors on the heatsinks and other parts of the amplifier's case. This can be used to immediately shut down the amplifier in case of a fault, and also to aid in manual adjustment of bias and offset pots.

- External power relays to switch the AC input power.

- A front-panel UI based on a dot-matrix VFD display module and an Immersion PR-1000 haptic rotary encoder. This will work something like: normally, the display shows volume and some temperature/current statistics. Press the button for input selection. Press again for an advanced menu, with full voltage/current/temperature readouts, input configuration, and so on. The selected input and volume will be stored even if the power is unplugged. If you've used a Slim Devices Transporter, the interface will be similar.

- An RS-232 port, to allow the whole thing to be controlled externally, by a Transporter or other such device.

Other things I'm considering designing in are:

- Level detection, to enable automatic volume matching for A/B testing of two different sources. This would make the amp essentially a source or headphone shootout-in-a-box.

- Input detection to automatically switch to an active input.

- A reduced configuration for single-ended amplifiers, leaving out the balanced inputs and conversion transformers, but still including the other features.

There are a few electronics design issues to be figured out. The main question at the moment is how to arrange the attenuator and transformer. The style of attenuator I'm using has a variable input impedance, but transformers work best driving a fixed input impedance, so that's very much less than ideal. We could also put the attenuator before the transformer, since the attenuator does have a fixed output impedance, but that means that the transformer will be dealing with much lower levels most of the time, or develop an additional shunt circuit to make the input impedance of the attenuator constant. I'll be working on that issue in the coming weeks.

I've started doing the schematics and board layouts for the input selector, and am playing around with sensing circuits for the power rails. No pretty pictures yet, but I figured I should post this now to gather interest and input. This thread will be the central discussion point for the project. Here goes nothing...

Here's a big ugly block diagram:

[Grahame]

Sweet!

Sir, I applaud your efforts!

I have been watching the various threads on this matter with interest.

From a personal perspective, I've been interested in a High End SS amp, having seen what various people have done with the B22, naamf amongst others.

Given the likely cost of such an endeavor , the incremental delta for adding decent control, monitoring and display over and above the power supply, amplification and casework seems worthwhile. (If I'm spending this much on amp, it should do double duty as a speaker amp, and thus have a remote control of some sort!, etc)

Given the block diagram, how much of it is tied to the use of DynaFet gain stages, e.g could you substitute B22 boards?

I think there is a lot of mileage in this approach, and look forward to your progress with interest.

(Disclosure: As a Transporter owner, I'm very interested in the direct external control - you could imagine the duet/controller controlling the amp functions -etc)

[j4cbo]

It really isn't tied to the DynaFET in concept at all. I will be designing in pads for sense resistors on the DynaFET boards, but the current sensing could be easily built off-board as well for a B22 or some such; that's about it as dependencies go.

And, yeah, I have a Transporter and Controller too; controlling the amp power and volume from SC is definitely in the cards.

Looks like all the spots in this build are full now.

[Beefy]

I look forward to your progress as well

While there is no way that I can purchase or be involved right now, I would be *very* keen to see options for a cut-down version of the controller. An input selector, bal/un-bal converter, relay attenuator and AC relay in one unit would be so many different kinds of awesome......

[Smeggy]

Yous guys is crazy!

[Pars]

Very ambitious! Were you planning on implementing this in a somewhat modular fashion? I might be interested in the volume controller and the I/O incl. muting, etc. for a preamp project, but would not need items such as temp. monitoring, current monitoring, etc.

[Icarium]

My understanding is that he is going to make the board very modular/scalable in that you can mostly just populate for whichever features you want.

[Beefy]

My understanding is that he is going to make the board very modular/scalable in that you can mostly just populate for whichever features you want.

Sweeeeeeeet!

[luvdunhill]

so, here's my opinion on the matter. If we're considering 10K:10K input transformer, a 10K rotary-style attenuator, such as the PCB mount DACT units would be something to consider as an option for this project. Second, would be the equivalent using relays. This is what I did for my preamp. Yes, lots of relays, I think I have around 70 DPDT relays. Also, with a firmware update, you could accommodate a logarithmic ladder version. This would be a poor choice with the input transformer, but a nice option if you're doing balanced only. Finally, with yet another firmware update (or better yet, toggling an option in the uC) a hybrid with 2 switches closed (at most) could be used, which might prove a possibility with the transformer. Note, all of these versions can be built to any value, 10K, 50K, 100K, etc. by choosing the resistor values appropriately.

Obviously, depending on the maximum number of relays allowed for, there would be different step size depending on the version selected. I think think would be okay. In fact, a daughter board could be supported that would add more relays. This approach would be extremely flexible, and in my opinion this would superior option to the TPA JT attenuator. For example, this setup could easily accommodate a transformer volume control (TVA) or autoformer as well as resistive attenuation.

Thoughts?

[luvdunhill]

Jacob: Do you have any idea what the impedance range on the hybrid with 2 switches closed (at most) would be, if we scaled the input impedance to start at 10K? I think this would be a worthwhile calculation.

[Icarium]

Unfortunately we've had a member of this group build pull out. If anybody was interested and missed out please contact me asap.

This is the current estimate of cost.

240 + 120 + 120 + 100 + 300 = 880 ... this covers parts cost for transistors/sigma22 kits/sowter input transformers/power transformer and chassis.

If we go with cinemags then shave about 40 bucks off that for 840. Toroid also might be 20 dollars cheaper.

Okay I got an estimate from j4cbo for what's left. 200 for all the parts to the controller board and the boards themselves. 10-20 bucks each for dynafet boards (4x) and 100-150 for the passive parts and then other parts like jacks/feet/knob/power cord jack. Let's estimate that at 100 bucks (Unless you want to go fancy).

200 + 40-80 + 100 + 100 bucks = 390 - 480.

This is a pretty good estimate of what's left (But not a guarantee).

1230 - 1360.

Add another 100 dollars for possibly more chassis work like drilling extra holes and etching.

1330 - 1460.

This is I believe a good estimate of how much it will take part wise to finish this project.

So if we have 4 people in then the cost of a complete build subsidizing j4cob's parts cost is

1330/4 - 1460/4 + 1330 - 1460 = 1662.50 - 1825 for a complete build with the feature j4cob listed above.

That is if we have 4 people to help subsidize. Currently we have 3.

Anyway if this sounds like something you are interested in... please let me know. You have about 1 week to decide.

The numbers above are no guarantee.. it may end up slightly more expensive or slightly cheaper, but should be pretty close. I've had it verified by two DIYers (Btw if the numbers seem off to anyone please let me know and I will tweak). Also if you want to drive costs down even lower you can sacrifice single ended to balanced conversion and what not.

P.S. For DIYers we have 8 board/transistor sets extra for any that might be interested. This is for Kevin Gilmore's Dynafets. We got extra parts for the sake of the matching effort. If you are interested in some of these boards/transistors please let me or luvdunhill know as well!

[Icarium]

Spot gone pending first payment.

4 boards/transistors pending with that spot.

4 boards/transistors sets gone to digger

Looks like everything's gone. Might be squeezing Smeggy in.

[luvdunhill]

Jacob / Others:

Has anyone considered using CMOS FET switches? I notice Analog Devices and Siliconix among others sell these. I know that both BAT and Ayre uses these in some of their equipment.

[j4cbo]

I've gone ahead and designed a constant-impedance logarithmic stepped attenuator. This relies on the fact that it is driving a fixed output impedance, as would happen within an amplifier or preamp, and in turn presents a fixed impedance to its source. I wrote a program to calculate out and verify all the stages, and simulate the actual impedance and attenuation for each stage.

Here's the source code and sample output. It's set up for a 7-stage (128-step) attenuator, 0.5dB per step, with source and load impedances of 15k ohms, but that could easily be changed. The program takes into account available resistor values and selects the nearest E96 series (1%) value.

Code, written in Standard ML: http://b.j4cbo.com/stuff/attenuator.sml

Output: http://b.j4cbo.com/stuff/attenuator.txt

[swt61]

Color me impressed!

Nice work. I'll be excited to watch your progress here.

[luvdunhill]

Jacob:

Nice! I would prefer something other than constant slope, as I have rather low output sources... Looks like I could change the calculation of stageAttenuations to be a piecewise function to do this?

[Icarium]

Looks like PFKMan is thinking about bowing out.. if anyone is willing to purchase his spot (Pay him for what he's paid for up to now) please let me know asap.

Deadline for this is sometime in the next couple weeks since we are ordering the prototype boards/toroids soon and when that happens the opportunity may be gone given the hassle of figuring out cost retroactively.

Total for the build including what you'd need to pay PFKMan is looking to be ~1600-1700 depending on options.

He opted out of the cinemag input transformers which is a savings of ~100 bucks (Transformers + subsidy for the builder).. but if you are interested and those were something you wanted then you can still acquire them on your own.

[luvdunhill]

Jacob:

Would you consider adding a way to mount a small PCB in place of the input transformer? I'm thinking similar to what Nelson Pass does for the UGS boards in his preamps. He uses two banks of DIP pins to connect the daugherboard and you'd route up the necessary I/O signals up via these pins. I have a solid state SE to Balanced design I'd like to try down the road a bit, and it would be nice to just snap it into the boards.

[j4cbo]

Marc: Should be doable; I'll just add some extra mounting holes and pins.

Here's our display:

Noritake GU140X32F-7003

[Smeggy]

Mmmm, tasty

I'm very jealous of your skills and know-how.

[swt61]

Mmmm, tasty

I'm very jealous of your skills and know-how.

I taught him everything he knows.

[Edwood]

Looking forward to seeing how the Uber Controller turns out.

Looks like a great all-in-one solution for other Balanced Amps as well. Would love to have something like this controller implemented in a DynaMid or Beta22.

-Ed

[Icarium]

Unfortunately to get the full benefit of the Uber Controller you need boards designed around it since the microcontrollers which deal with current sensing/temperature sensing are located in our case on our version of the Dynafet boards. This turned out to be a cheaper option then not having them on the boards.

Of course if you dont want the fancy display/auto shut off features related to current sensing/temperature sensing then you'll be fine putting one of these + the input board in any amp really.

[luvdunhill]

Jacob:

How hard would it be to implement the attenuator as a stand-alone device? Would you be willing to post the schematic, so I could do the board layout?

[swt61]

So guys, does it look likely that the uber controller will be available in time?