Anyone interested in a digital amplifier project?

Status
Not open for further replies.
Jump to Latest
#81
Discrete classd amplifier

Hello all,

This is my first post on diyaudio ever.

I have read most of this classd thread, and I must say I am very interested in designing such an amplifier. I am especially attracted by the high efficiency ratios of this technique.

However, I would try and take another approach: I would like to build a classd amplifier from discrete components, or only standard components such as opamps, or timers.

I am not primarily focused on sound quality: the amplifier has to produce "big" sound and not to much heat. I would then review the design of the amplifier and optimize the sound quality. Then only I would consider using specialized components 🙂

Do you have any orders of magnitude of what could be achieved this way?

And then, does anybody have any schematics about discrete classd amplifiers? The links I seem to find on diyaudio archives only refers to the above mentionned specialized components.

Thanks in advance and keep this thread alive!

Koni
 
#82
I built an analogue class-D amp many years ago. It didn't sound very good, but anyway, here's how it worked:

You start by setting up an opamp oscillator which gives you a triangle-wave output, with peak-peak amplitude as large as the largest expected peak-peak audio input. Generate that at, say 70 or 80kHz. You feed your audio input and this triangle wave into opoosite inputs of a comparator, and what comes out the other side is a PWM square wave which has it's duty cycle modulated by the amplitude of the input audio. Ta-da!

Use the resulting PWM signal to drive a switching output stage, and you've got all the basics. Low-pass filters are a good idea at the input, and virtually required at the output, of course.
 
#83
The most famous one is Motorola AN1042. This is a fully discrete design.

There is also an old AN by Hitachi but I don't think that you can get the modulator IC anymore (HA13003).

The Peavey DECA series amps from the mid eighties was discrete. It is a refinement of a patent by Brian Attwood (who is also
named as inventor on Peavey's patent amongst some Peavey guys). You can find this very detailed patent on the web.

The same basic topology is used by Crest nowadays (double feedback loop around the output filter !!!) but uses less discrete transistors than Peavey back then. I assume the schematics schould be available on the web also.

I once built a class-d amp for my thesis. It is halfway discrete (i.e. it is using comparator ICs and the MOSFET drivers are also IC types. If anyone is interested he can have the schematic.
There was not much info around about class-d amps when I developed it so it leaves much room for improvement.
I can also supply some ideas for tweaks.

It was intended as lab power source but I always kept audio usage in mind during development (I still have an application where I could use something like that). We even listened to music through it. Though THD was quite high with 0.44% (2nd order was dominant) music sounded nice and sweet, the only disturbing thing was some high-frequency hiss.

Regards

Charles
 
#84
You know what's really scary? Take a look at the picture on page 24 of <i>System Design Considerations for True Digital Audio Power Amplifiers</i>. The reference design PCB is sitting on that amazing anti-static surface: a straw mat! Tatami is ESD safe? Hrmm.

Anyway, I think I am convinced I need to build this thing. They have an entire stereo amplifier on 12 square inches of board! For the cost you can't lose.

I have a great big list of questions. Their reference design really goes berserk with the snubbers. The bulk power supply has a snubber, a bulk capacitor, and an RLC filter before it hits the chip. Do you think all this is necessary with a linear regulated supply instead of a switcher? I can see dumping the inductive traces and keeping the rest.

Speaking of the linear regulated supply, I'm thinking LT1083 with a pot controlling the output voltage, separate supply all the way back to the transformer for each channel. Thoughts?

Filter inductors are going to be a pain in the ***. I don't exactly see anything in Digi-Key that meets the requirements.

P.S. I forgot to mention that this seems ideal for direct coupling to DSD, although I don't yet have an SACD player.
 
#85
Hello,

I think I know where those snubbers came from. Take a look at patents that I listed in my second post on page 4 of this thread.
They should be there regardless of switching or linear power supply.
Regarding inductors I see no real problem there. You can ask for samples from Coilcraft. They are very generous with samples.
I have another problem. I finished my PCB with DIR1703, AD1896 and TAS5015, but i cant get AD1896 upsampler anywhe. Does anybody know a distributor that would carry those chips?

Regards,

Jaka Racman
 
#88
Well, I just assumed that they were available. Perhaps it is like with TI that they pre-announce products about 6 months or so before they become available in any way shape or form 🙂

Sorry about the noise if that is the case.

Another point: Do you know of a simple PCM to DSD converter? It would be really easy to build a switching amp with a DSD signal, but then again, I quite like the TAS5015 approach.

Petter
 
#90
Hi! I'm first time here

I have to excuse my bad english first...🙁
Since IX 2002 works here TAS5015+TAS5110 powered by two OPA549 (heh! It is not warm! ) from two separate path (2x100VA trafo ect). DIR1701 (512fs) + 8x loJitter(about 2ps)PLL+(/2/4 divider (depends 44-48-96-196kHz fs) solved HiFreq problem.
Major problem-that is low ESR power source.... really, batterry of 10x2200 mikroF loESR C (ie 2miliohm) per channel - isn't too low...🙁crazy...
Previously I use switch power, but ... I can't regulate, control it properly...Maybe I'll come back to this idea? OPA is so easy to use...🙂
PCB based on EVM project (85x90mm TAS5015+2x5110 and output,LM317for 22V, 2x5V, 5x3V3 regs, DIR and PLL on separate 20x50 mm PCB), coil 7,2/22mikroH/ from Panasonic on output, PCM1804+THS4131 for analog input (3+MM/MC RIAA preamp) 7x digital ins...
Soldering? Weller's microwave🙂 PowerPad on TAS 5110? Solder copper heat via PCB😛
OPA and TAS - cold... only trafos abnormal hot🙁?? pulses ?
OPA need R (0,1 Ohm) or 5mikroH inductor on output - else is instabilly in some ~~~~ !!!!??/ rrrrrr cases? voltages ???(ech.. my english...🙁)
So sound is ... beauty for me🙂 Detailed, precision and colour. It is best think I done (incl. PMD100+PCM1704K DAC 🙂
Works fine with loImpedance ProAc Response (! even 2Ohm !)
For 8ohm speaker need change coils to 22mikroH

Regards
JaroMi
ya.ga@wp.pl
 
#93
Capacitance

4x2200(22 miliohm)+4x1000(39 mili)+10x680(47mili)+6x100(??) mikroF+some ceramic ... = 20F /2 miliohm ESR per channel... HitanoEXR, ElnaCEW, Cerafine (100mi).
Stable? None. Without any additional elements OPA isn't stable. Hot🙂 Very hot! But I add 0,22 ohm (or/and) 1-10mikroH inductance on output and problem was solved (or I think only so...🙂 OPA was cold...
So pity I come here so late, but I can't access in Internet before....
About V-VII (??) 2002 I build first PWM on TAS5012+5100+DIR1701 based on EVM 's layout, then I get 5015 (free samples! THX Texas!)... but I thought, that's popular chipset in DYI world. :bigeyes:
I'm first? 🙂 Great!:clown:
I've problem too... Any power-on/off any electric gear in my house (or probably not only) - light, refregerator, lift, TV(.... all....) makes muting... It isn't MUTE signal from DIR, but ? loss of sync? phase errors? I don't know. This effect isn't only in TAS, but maybe in all my DIR DAC's, PWM's... Anyone have any idea? Massing I think is good, Filter before trafos, SPDIF in/out with transformers and decoupling.... AES/EBU the same... I cant solved this....🙁
help....plisss...
:bawling:
Thank you!
And don't afraid TASes... great and easy (?) chipset
 
#95
Rookie said:
I wonder, what happened with Brian's project? :scratch:
Click to expand...


Hi,
Sorry about the long time - no post.

My project has taken a couple of detours (typical for me)...

Several months ago Panasonic discontinued their SA-XR10 receiver and replaced it with the SA-XR45. (The amplifiers in these are based on TI's Equibit chip set.) I had an opportunity to pick up one of the old SA-XR10s on a real cheap clearance price. Since I already had the schematics for this (to use as another reference design), I bought it for experimentation.

A couple of comments about the unit as it came out of the box:

I was quite surprised at the sound quality. Especially when I fed it with a 96KHz coax S/PDIF signal from a 2 channel DVD-A source.

The amp seems very effortless, with excellent imaging and depth (subjective descriptions that I hate to use, but what else can I do here?) Frankly, my brain is having a hard time with the idea that a chrome plated Panasonic A/V receiver that cost a little over $200 can sound this good.

Thanks to the RIAA, you can't do a direct digital connection from a SACD or DVD-A multichannel source (at least until encripted Firewire is standardized). I have to hook up the analog outputs of the players to the receiver's analog inputs (and it only has one set of inputs, so I have to manually swap the cables when I want to switch from DVD-A to SACD). It still sounds quite good with the analog interface but...

It's easy to compare the analog interface (digital to analog then back to digital) against the S/PDIF interface for two channel DVD-As and regular CDs. Not surprisingly, the S/PDIF link is much better than the analog patch job. The imaging, in particular, seems to get a bit 'fuzzier' and flatter with the analog connection.

By buying this receiver, I was also finally able to start listening to the multichannel disks that I had been collecting for over a year. This has taken a substantial amount of time to relocate the system to a new room and get the multichannel placement worked out (If you think a stereo setup can dominate a room...).
I'm using a set of matched B&W DM602s3 for all channels with a pair of ASW650 subwoofers (I'm using the sub's built-in analog amps for the time being). These are a fairly generic set of audiophile speakers, but I felt that they were a good quick starting point for my purposes while I was working with new electronic designs. I had been quickly convinced back when I was first listening to multichannel DVD-As that having matched speakers was very important (as opposed to the typical A/V setup with two 'main' speakers, two smaller surrounds, and a 'voice matched' center speaker).

One of the compromises that Panasonic had to make with this unit was with the output reconstruction filter. Instead of matching the filter to an individual driver (located close to the amp), they had to set it up to accomodate a variety of different speakers (with long cable runs and passive crossovers). In this respect I think they did very well. I have a 20+ year old pair of DM7mkII speakers that I always considered very difficult to drive. Most amps would just sound flat and dull with these speakers. A very few amps (generally with MOSFET output stages and large low impedance power supplies) would get these speakers to come dynamically alive with wonderful imaging. The SA-XR10 had no problem at all driving these.

Even so, I suspect that customizing the output reconstruction filter to individual drivers is one place that I'll be able to get a good improvement.

The other main area I want to play with is trying out different power supplies. I had just gotten a Tek P5200 diff probe to start poking around the stock power supply with this last weekend (the supply is floating, so I couldn't use a regular probe). It's a little premature for me to make any comments yet. The type of power supply to use has been one of the biggest questions I've had with my own Equibit design. Now I can try out different supplies before I build my own amp.

Short term, I'm presently laying out some digital interface boards so that I can bypass the D/A - A/D patch job for multichannel SACDs and DVD-As. I was so impressed with the difference between the analog and S/PDIF interfaces for the regular stereo stuff that I wanted to have the same improvement for the hi-res multichannel stuff.

The approach I'm taking is based on TI's MuxIt chipset. It uses separate LVDS clock and data lines (I'll be sending them over standard CAT5 cabling). I'll be serializing the IIS lines in the DVD-A player and deserializing them inside the receiver just before the Equibit section. The MuxIt chips also feature a PLL so that the signals are buffered and reclocked just before the amp.
For the SACD, I'm using an NPC SM5816A DSD to PCM converter before the MuxIt transmittter.

(If you look at the bottom of Jorge's pictures of the inside of the SA-XR10, you'll see the 9" ribbon cable with a fold going from the amp's DSP section to the Equibit section. Panasonic is sending the IIS MCLK signal over this ribbon cable! This is hardly optimal for low jitter. There might be some room for improvement here, although maybe the PLL in the TAS5012 Equibit chips is good enough at reclocking to make this less of a problem.)

Anyway, it appears that Jaro has beat me to making one of these from scratch! It's good to hear. (I've been thinking a lot about battery power myself).

Regards,
Brian.:cubist:
 
#98
About power...

I try buck converter (based on TPS5120 works ~190kHz, 5-10mVp-p@5A ) - works great. All higher (over ~50kHz) frequencies was filtered by LC output filter. But - max Uin=27V, Uout-max=20V... Uout-min=1,5V 🙁 Overheat, overcurrent, overvoltage protection... very hard to controll...
Opa is so easy🙂 and so clean... I think in this power stage that is very critical point.
As Brian said-sound is extremally precision imaging and depth. Hue, present and live. And precision. Very precision. I did never heard before sounds, I hear on RyTM'5015 and Focal's tioxid🙂. Even Beyerdynamic's DT931...
But when power source was 2x2200mikroF (10 miliohm ESR @100kHz... but @384kHz????)only sound was ... ugly... resonant ...
I have put much more C in power patch, but then🙂🙂
I discovering my collection again 🙂🙂
I based on CD/DAT source . In plain stereo.
Now I waiting for SRC4192... I'll try "100MHz XRef" concept described previously by Rookie and Bryan (in TAS5015 Master mode).
Then divide to SClck, BitClck, WClck, reclocked...
And again sorry my bad english... I can write russian more fluent...🙂🙂
Regards
 
#99
Is it better to use (excuse my newbieness) the clock generated by the 1703 or use an externall PLL?
I havn't really got a clue when it comes to clocking schemes, does anyone have any whitepapers or general info on how its surpose to be done?
From what I gather The 1702 generates a clock for the audio and the 5012 syncs with that? If thats the case can you use a lower jitter source (the 1703 states 75ps)?
Basicly I want to make a full amp section (TAS5012 and TAS5100a) and have a Pic controlled DSP on another board along with a toslink input and a ADC for use with radio.
Thanks
fr0st
 
Status
Not open for further replies.
Top Bottom