Class A hybrid power dac project

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,
can it be simpler?
 

Attachments

  • TDA1543.jpg
    TDA1543.jpg
    25.7 KB · Views: 234
Hi Ontoaba,

No particular problem with CFP, apart from occasional slight instability when using high capacitance, low inductance speaker cables, my preferred. However any amount of feedback does stuff (good and bad), so I want to be free of it as much as possible for this project. And ditto those long tailed pairs etc especially in high gain ccts.


Brief interlude, the amp is playing in the office on my good speakers as I write, so good, so good.... anyway


Hi WuYit,
yes that circuit looks simpler, but it does not drive the speaker which is what I am doing here, therefore, over all, no, probably not simpler.


Just so no confusion, here are some stripped down circuits of what I am aiming at, everything is fairly interchangeable device wise, just pick your flavour, the tube fet one would probably work as the “most minimal” if gate capacitance etc don't get in the way, all tube one will sort of work, got to love that volume control, with variable output damping as well.... mmm a bit more thought required there maybe, but it would work and I bet it would sound pretty good...
 

Attachments

  • variations on a theme.PNG
    variations on a theme.PNG
    57.9 KB · Views: 224
sulk

Grrr, I was wondering if anyone else had discovered the “2x oversampling” by inverting WS and of course they have..., I don't quite subscribe to GuidoB's friends (see below) observation of 16dB down at 20k, more like 7dB in my case, (2x nos 3.5dB). So unfair... never mind fortunately only three wires to rearrange.

I must say I had noticed it felt maybe a little rolled off, but did not measure frequency response recently, and have been just enjoying having it working, not really knowing what to expect.


theory:
non os : -0.8 dB at 10kHz, -3.5 dB at 20kHz.
shift -2.5 dB at 10kHz, -16 dB at 20kHz.

Somebody that knew what i was doing wrote me this (i didn't back then):

> Hi Guido,
> Your DAC would implement the equivalent of a 2-point moving > average filter, i.e. with impulse response h=[1 1]. You can read > more about those here: SAIC: Products: X-ray passenger vehicle inspection VACIS XPL
> This filter will have a lot of high frequency roll-off, -2.5dB at
> 10kHz, -16dB at 20kHz, so I am not sure what you mean
> with "high freq rollof of a true non-os dac is gone". On the
> other hand this filter has a very good time-domain response
> with no ringing, and it will reduce aliasing distortion quite
> significantly. I would think, although I have never heard one,
> that it makes a non-os DAC sound much more pleasant.
A wise man once said “there is nothing new under the sun” the more I know, the more convinced he is right. One of the joys of having ones memory erased on entry to this adventure playground I suppose.
 

Attachments

  • 2x oversampling fr feb 18th.png
    2x oversampling fr feb 18th.png
    26.4 KB · Views: 200
  • no oversampling fr feb 18th.png
    no oversampling fr feb 18th.png
    34.4 KB · Views: 170
Hi, h.
Seem that your variation with MOSFET doesn't work, it has no back current to that volume pot, and only work as voltage follower. EF itself is feedback, it has little output correction through limited Vbe and its constant (or almost constant) current gain that draw current from input.

Anyway, YAC516 has 8x oversampling, but still 24bit 96kHz will do better. I have 32bit from AT&T but I don't know it is too much for audio:confused:.

May be the most significant effect is that volume pot (passive damper, I am waiting your link working, or do you have another archive?).
If you allowed small global feedback as active damper, may be there will be easier.
How about built another new with feedback that will compared to your only one. (just a thought)
 
Had a brief diversion to the digital line level forum to sort out the -3.5 dB or -0.6 dB at 20kHz conceptual problem I had, the long and short of it is that my amp has an indicated 47% "THD+N" at 20kHz, and I embrace the "distortion".... This is one of the reasons I did this project, to break a few rules. Please see below distortion and amplitude measurements taken at 1 watt output recorded from my pc (ARTA m-audio) and Sound Technology 1700B distortion analyser.

see below.

Hi Ontoaba, I must point out the above diagrams are for indication of general topology rather than working circuits, the obvious mistake is the tube's cathodes are not on the same rail as the -ve output rail, but need to be at least 40 volts lower. I don't understand why you think the mosfet circuit won't work. It is a very common output configuration. You are right none of them are particularly linear, but good enough for this project.
 

Attachments

  • 47 percent.PNG
    47 percent.PNG
    25.9 KB · Views: 153
Very big differences. There is something wrong with your SoundTech 1700, probably another failure not repaired yet. Your amplifier is not type that distorted by slowness or lags.

7% distortion and up are easily noticed by ears, <5% is OK, that you may embracing with it.

The mosfet problem isn't only that tubes operating voltages, but also the damper will not functional. All tubes one also need 5W pots.

Hi, what if using your project to also find the truth about feedback, you may find many things and noticing the differences between your mosfet and bjt circuit, and you may like my linearized EF to be used too. I have another more precise and complex version with mosfets, using single rail with switching assistant.
 
Hi Ontoaba. Sound tech works fine, its "problem" is a distortion measurement bandwidth of >100kHz according to data sheet, unlike the m-audio which stops at 22kHz... My problem is what to call the distortion, because it is not really THD, may be just TD... It is a pity my site is not up, I think it is quite broken and not sure how to fix it short of moving it to a new host. yes i agree about the all tube circuit, it is a bit silly...
 
It is strange... hmm.. let me think... ARTA work with soundcard, if the soundcard has 20KHz maximum frequency, that it will still measuring H2 at 10kHz and H2,3,4 at 5kHz and up to H20 at 1kHz. I remember with shorted tantalum that been removed from SoundTech, did you replace it with another or just remove it? may be that cap is crucial.
Low distortion figures isn't everything. I ever try to built an amp with very low distortion at resistive load, but my other amp with much higher distortion beat it.
Haa.. I remember something. How about test your soundtech with another amp.
I am also interested how ARTA using microphone configuration in L/R setup, it is worth to be tried.
 
Interesting project.

Way back 1980 I had design a high power DAC that was to drive the drive the speaker directly without an analog buffer.

This was before the advent of digital audio which didn't hit the market until 1984.

I never got it off of the ground because of non linearity problems due to changing speaker impedences.

So I ended up scrapping the idea as the technology was still too new for the day.
Also when cd's finaly started to come around I was not happy with the way they sounded.
Although they were (and still are) very impressive the sound was not to my liking.
Even back then I felt that what is now todays 24bit standard should have been implemented.
It wasn't until the 24 bit format came about that I started to like and apperciate digital audio.
Anyway, keep at it. jer
 
My original inspiration for this project was the AD768, which is a 16 bit dac which can kick out 20mA, but as I prototyped the output stage I realised I could prove principle more quickly with TDA1543, just using 10 in parallel. In practice, I found I got sufficient performance using just two. The C10 and 13 give me a bit of HF roll off so I don't totally clobber the output tranies, the ECC88 can easily cope! The convinient place is across the volume pot, but would give a variable time constant.
A true multi bit power dac would be an interesting, if one could make it work, but a bit ambitious for me...
 
a few months later

The amplifier is built and tested (and marked) and my life has slowed down sufficiently to post the full circuit.

I am very happy with the amplifier, and it has bought about a shift in my perspective so much that I am wondering how to “de transistor” my home stereos. I even toyed with the idea of acquiring a “vintage” valve receiver, those guys are a bit pricey now days! So I am thinking of building my own, or a “happy hybrid” which would be an analogue input version of the amplifier, possibly with modest feedback...

My original suspicion that a single device, be it valve, transistor or whatever, providing modest gain and with modest (or less distortion), should be fairly indistinguishable from each other in the same circuit, is gone, I now hear transistors every where (at least I think I do!).

Various caveats about the circuit, If you wish to use /copy the PSU put a resistor in series with the voltage setting pot RV1 to stop shorting the input power supply in to the controller chip when the pot is reduced.
 

Attachments

  • spdif inputb90.png
    spdif inputb90.png
    161.4 KB · Views: 132
  • classa final90.png
    classa final90.png
    162.3 KB · Views: 134
  • power90.png
    power90.png
    172 KB · Views: 127
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.