diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Solid State (http://www.diyaudio.com/forums/solid-state/)
-   -   New home brew amp module (http://www.diyaudio.com/forums/solid-state/213020-new-home-brew-amp-module.html)

CBS240 20th May 2012 10:08 PM

New home brew amp module
 
7 Attachment(s)
Hey SS gurus!

This is my new SS amp module, straight from my mind, aided with some rare aggressive motivation:p, and is almost complete except for the rest of the PS caps and to finish soldering all the parts in. This is the first proto-type for this module which is a surprise in that I didn’t screw up the first one. I usually at least screw up the first build somehow.:rolleyes: Guess this means I must be getting better at this.:D BTW, I found a company that offered 3 PCBs for $10 each: 2 layers, top/bottom solder mask, and top/bottom silk screen. And they can do 6 mil trace, and 10mil vias! Pretty damn good deal if you ask me.:up: My minimum trace is 8mil (just one small place) and minimum via is 12mil.

Along with two separate channels, this module features on the PCB, the power supply, both amplifier channels (Av~30dB), control logic with DC detection/protection and clipping detection. The small daughter board mounts onto the front of the enclosure and has a momentary switch and a 3 color LED. The LED is blue for stand-by, green for active, and flashes from green to red for ~0.3 Sec and then back to green if either output is driven into clipping. The logic has DC detection/protection that will activate if either DC output exceeds +/-0.6V, switching the module back into stand-by mode. The heat sink used here is just an old junker.....I'm a fan of using old junk parts.:D

Instead of using relays to disconnect the speakers from the output stage during power up and power down, it is the output stages that are switched on and off using solid state. This eliminates the relay contacts in series with the speakers. The high current is switched to the output stage circuits by two Mosfets. I chose to use Trench type vertical fets here because of the lower Rds on. These will have two states, either full on or full off. The switching action of off to on state is deliberately slewed to prevent turn on thump, but the turn off action is very fast. There is no sound what so ever when the circuit switches back to stand-by mode. To add another layer of protection, each switched rail is fused. Notice that the voltage regulators do not use Zener diodes for the voltage reference. I detest Zeners used as references in voltage regulator circuits.:whazzat: Zener noise is unnecessary.

Due to the nature of a public forum, I will discuss the power supply, and the logic control. The output stage circuits are a bit different than what is normally done, but I learned quite a bit from discussions here about the concepts so I will discuss the output stage as well, that is if anybody cares. I chose to use planer stripe type vertical Mosfets for the output stage, they are more suitable for analog applications. The rest is personal IP. Forgive the photos of drawings, I do not have these drawings entered into a computer yet.:shy:

Some photos… photo 1, photo 2

CBS240 20th May 2012 10:46 PM

1 Attachment(s)
The output transistors are mounted directly to the brass blocks with a screw and HS grease. That brass came from signage scrap, and so being a Gilding metal is 80-90% copper to zinc. It has good thermal properties. The block is energized, energizing the output transistor drain by the tab, and then the drain pin energizes the driving stage and also makes contact with the thermal feedback device. The brass block is insulated from the bracket with mica and grease creating a surface area 4X that of the TO-220 tab, significantly reducing the thermal resistance of the mica layer. Planer stripe mosfets do not have secondary breakdown like characteristics and are limited in Pd by peak die temperature. Also they happen to be cheap, P-ch ~$1.60, N-ch ~$0.69.:D Most people don't try building a Hi-Fi amp using these devices as source follwer bias in class AB because they are designed for switching and have a nasty exponential transfer, but I believe they are exactly the type of transistor to be driving a reactive load like a speaker. It's all in the gate driving stage.;) By eliminating the mica pad between the TO-220 tab and the heatsink, the thermal resistance is lowered significantly. The density of the brass block acts as a large thermal capacitor, coupled with little thermal resistance to the transistor die, allows for some abuse. This arrangement creates a thermal filter that reduces the peak temperature spikes. The average temperature is higher but the peak temperature is kept below the point of melt-down, thus I can crank much more power from these output devices. These are rugged devices indeed, each channel can drive at least 80Wrms at 3R! As far as Fairchild Mosfets go, the planer stripe types are denoted with FQ, FQP=TO-220, FQA=TO-247. Obviously the TO-247 components can dissipate more power, but if you notice the datasheets, FQPXXX is the exact same die as FQAXXX. If you can get a more ideal thermal connection to the FQP parts, they will dissipate as much or more power than a FQA part traditionally mounted with mica insulator. The concept with the brass thermal capacitor surprised me with the amount of power that can be sucked through a single pair of TO-220 mosfets.


Mannn, it sounds good. Quite possibly the best amp I've made thus far!!:cool:

BTW here is the logic circuit...

Going to take a break for a while and listen, but I will post final photos when I get the rest put together....still have to add the Zobels and input filters...

AKSA 20th May 2012 11:22 PM

CBS,

Very nice circuit, elegant layout.....

Could you share details of the company you found - $10 for double sided protos is outstanding!

Hugh

CBS240 21st May 2012 12:14 AM

RushPCB

Click on "3 PCB special". The quality of the PCB is very good. This being the first build of the first prototype, I had to change some SMD components around to fine tune the circuit. It is easy to damage a PCB when removing and replacing SMDs, at least those smaller than 0603.

CBS240 23rd May 2012 09:41 PM

6 Attachment(s)
I’m getting closer to being finished with the first build of the first proto-type. The only thing left now is soldering in the final power components and caps, then finding a suitable enclosure. The daughter boards that hold the screws where the speaker wires are attached are mounted in quite solidly. The daughter board closest to the output transistors also have the output Zobel filters and coils. I added the heatsinks for the rectifiers and replaced the temporary fuses with the ones I designed for, 4.5A 0603 SMD. Each channel output stage bias is ~100mA. The small green and yellow wires are just so I can monitor output bias and are just temporary.

I don't see why folks don't use more SMDs. So many components are turning to surface mount. In this circuit I use several varieties of SMD packages. Some of the less common packages for power amp circuits that I used are: SOT-23; SOT-323; SOT723; SOT-923:magnify:; SOT-89; SOT-563; 0603; 0402; and even 0201:magnify::magnify:. 0201 may be small, but the power dissipation is no more than a few mW for these, most are uW. They allow you to cram a lot of circuitry into a small space. Also there are some 0402 resistor arrays. These really save space.;) This is an advantage in an amplifier circuit. IMHO, power is not needed until you reach the final stages. There are few transistors larger than TO-92 that are even remotely suitable for a VAS. The components in this circuit are not undersized, there was meticulous calculation involved.:D Don’t think that because the outputs are TO-220 devices that they are not up to the task. I just cranked from one channel about 100Wrms sine wave into 2R resistive and it did not fail:tongue: I think it can handle a reactive 4R speaker with no problem.

I do believe this is the most difficult amp circuit I've conconcted so far. I'm thrilled it was not all for nothing.:D

Some more pics.

dimitri 26th May 2012 11:27 PM

Very interesting,
May I ask several questions?
1) how 18V is set - by voltage drop of EC stage on 221 resistor?
2) is this 18V used also to power input stage?
3) as I got the input stage is dual differential, correct?
Thanks,
Dimitri

CBS240 27th May 2012 01:31 AM

1) yes. The voltage drop across the 220R resistor is 6V, set by the current required by the driver and pre-driver/EC. The +/-18V is a bootstrap powered by the output stage, ultamately limited by the zener/diode clamps that prevents overcharging of the gates.

2)No. The input stage is powered from the high voltage rail and +/-40V regulator circuits. When the control circuit is in stand-by, the low voltage high current +/-24V is switched off (B+/B-).

3) the IPS is a complementary J-fet source follower. It is preceeded by a buffer that sums the input signal with the output of the DC servo.

dimitri 27th May 2012 02:05 AM

Thank you,

What was the idea behind low-voltage bootstrapped rails for driver/EC - low power dissipation and SMT parts?

And the IPS is loaded by shunt feedback network, correct?

CBS240 27th May 2012 04:44 AM

The bootstrap rails lowers the power on the SMT parts. IMO the driver stage should be bias well into class A even if Mosfets are used as outputs. They do require significant gate charge in order to achieve high current slew, even though planer stripe fets are touted as requiring less gate charge per conductance than Hexfets. Although Iím fond of these devices, not everything about planer stripe fets is better than cellular types, they seem to have a greater dependence of Vgs vs. Ic to Vds, an extra distortion but this is nothing the EC canít deal with.;) In addition to the bootstrap cascode, there is a traditional cascode. The reason for this is to protect the components used in the EC loop that were chosen to maximize BW. RF devices have very thin base layers and cannot withstand Vce much above 12V. But they can maintain current gain at very high frequencies. I noticed that when the outputs are bias close to class B or cutoff there is a small ~7Mhz oscillation originating from the EC amp, but it disappears when output stage bias approaches 30mA. While oscillation frequencies this high can be hard to track the source, I suspect having bias below this point essentially pushes the PM of the EC amp into instability due to the required increase in voltage gain, or my compensation is too light. (ref principles of HEC) However with bias greater than 30mA (bias is ~100mA), I have not noticed this oscillation. If the EC amp is compensating to this high range of frequencies I suspect the nasty distortions of these 'switching':rolleyes: devices is being nulled to a more than accepta

qusp 27th May 2012 12:35 PM

very cool indeed, loving the SMD constellation, in one of the photos there is a patch in a slightly out of focus shot that looks like a dirty smudge until you look closely and notice its simply a cloud of tiny SMD. are you hand soldering the o2o4 or have you a home reflow setup? I would be quite interested in building this amp, but I gather by your earlier post that this may not be possible as its to be a fully commercial venture? or just a commercial kit/ready made board?

I too dont quite get the aversion to SMD, to the point folklore would have us believe that its an evil cost cutting exercise, undertaken by faceless men in order to steal our enjoyment of music ;) nevermind the fact that its also higher performance for small signal (and even some large signal) new higher performance parts are designed all the time and almost none will be available in PTH.

i'd love to comment on the input stage and VAS but...... =) intrigued at your results with the TIM


All times are GMT. The time now is 08:21 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2