I know it had me scratching my head also, my TZ resistor is 4.7kohm, all I know is that my buffer had a 100kohm to ground on it's input and when I lifted it the 1khz signal 0bdf from my test disc, TZ output of the 1khz signal jumped by around 5 to 10% So I left it off now it's whatever the fets input are.
Cheers George
To Mick/others who have found how good this AD844 is, I just found another improvment you can do. As I discovered for myself as well as Pedja Rodic saying it as well, that the buffer of the AD844 is good but not great. I questioned this myself and asked why would the great Barrie Gilbert who designed this put in an average buffer stage. So now that the input i/v stage is stacked and improved things so much feeding into my discrete fet buffer, I decided to give the internal buffer another listen, and I was shocked!!! as it now sounded better than my discrete buffer. All I can say with previouse conclusions about this internal buffer were unfounded, as it was simply showing up the faults of not having stacked i/v stages more clearly. And now with stacking the I/V section has finaly live up to Barrie Gilberts guru status. Also this buffer (bi-polar) has a better than 2000Volt per micro second slew rate as well as being a very low 15ohm output impedance and having 80mA current drive and is short circuit protected, what more could you ask of a buffer.
PS I am only using the first AD844 buffer, the others are not stacked only the I/V stages.
Cheers George
PS I am only using the first AD844 buffer, the others are not stacked only the I/V stages.
Cheers George
That's fascinating.
Do you have a photo to show how you do this please?
I wonder why you are only using the first AD844 buffer? Seems like a waste not using the other buffers with all that current.
Last edited:
If you mean the stacking, just sit them ontop and solder pins 2 3 4 7 & 5 to the one underneath and use whatever free floating pin 6 for the buffer output which ever one gives the least dc offset.
To scared to try it, might blow something, one has already great output impedance and current, couldn't see anyone wanting more/better than this
Cheers George
AD844
Hi George that is great news on the buffer. I'd think that means the current mirrors are starved for current in a big way. So that maybe the only limiting factor for the 844. I wonder if the designer was forced to limit the current in the mirrors for temperature reasons? Power consumption? I hope to have a budget for some chips next paycheck. Eager to try stacking them.
Dave
Hi George that is great news on the buffer. I'd think that means the current mirrors are starved for current in a big way. So that maybe the only limiting factor for the 844. I wonder if the designer was forced to limit the current in the mirrors for temperature reasons? Power consumption? I hope to have a budget for some chips next paycheck. Eager to try stacking them.
Dave
I would dearly like if someone has nerve and spare AD844's which I don't, to stack pin 6 buffer as well, as I love the sound one buffer gives with stacked i/v stages.
But with a test cd at full 0dbf 1k sine wave I can see that single buffer when it's output is loaded down with a 100ohm load a slight xover distortion starts to appear to the naked eye on the scope of that 1kohm sine wave. (rail volts do not drop)
It may not mean anything as when will the pin 6 see a 100ohm load, never, it maybe the 80mA output current limiting of the AD844 comming into play.
It says in the data sheet that it can drive 50ohms load to + - 2.5v Maybe that is when feedback is applied?
Cheers George
But with a test cd at full 0dbf 1k sine wave I can see that single buffer when it's output is loaded down with a 100ohm load a slight xover distortion starts to appear to the naked eye on the scope of that 1kohm sine wave. (rail volts do not drop)
It may not mean anything as when will the pin 6 see a 100ohm load, never, it maybe the 80mA output current limiting of the AD844 comming into play.
It says in the data sheet that it can drive 50ohms load to + - 2.5v Maybe that is when feedback is applied?
Cheers George
Last edited:
Hi George
A bit late on contributing, but he does not approve using it in non-feedback mode a la Pedja. But that shouldn't stop ya!
Cheers, Joe
Actually not so Joe, I have been conversing via email with Barrie Gilbert and he is most fascinated at what we are doing with his AD844 design, he will try to find and if so email me the complete circuit schematic, not just the simplified version, but he did say it was a long time ago, he needs to look into his paper archives and not to hold my breath.
Cheers George
Cheers George
One very well-know audio designer of a very famous brand, uses the 844 with the output on pin 5 and adds his own buffer. Indeed you could use something like BUF634.
I did a head-phone amplifier using that exact combination. Keep in mind that if there is no feedback around the BUF634, then the DC will shift, but interestingly, even though you are not using the 844 internal buffer, it is still internally connected to Pin 5, so we can still use the in-built DC trim of the 844 to null the output of the added buffer. I know it sounds weird, yet it occurred to me it would work and then actually did it. The offset will now be transferred to Pin 6 but the added buffer will be nulled. And quite stable after everything is thermally settled.
Another thing, what value are you using on Pin 5? If that value is low, then it figures that the output Z at Pin 5 is that resistor divided by the load value minus the resistor value.
You say you use 1K5 and then connect a highish 50K load to that. Then the output Z is effectively a low 44 Ohm. Keep the load impedance high and the Pin 5 value resistor low, and you will have low Z (in effect there is a kind of current output effect going on here). The lower the value load, the output Z goes up. So keep the two values as far apart as possible to get the lowest output Z.
Can you see what I am getting at?
Why not try to listen to it without the buffer?
Use the Trim to null the DC at Pin 5 with the load in place- and stick with that load. Or DC offset will creep back into the picture if you change the load value.
But this is only a pure DIY thing, you wouldn't do it commercially without a buffer, no way. But DIY, hey, we can do this no problem - and set it up to exactly for our system only - for our own pleasure.
Cheers, Joe
Last edited:
He has changed his tune then, and good for him. Via Vacuum State and Allen, he stated a different position then. But that is now quite a while ago when I noticed Pedja was using the 844, so was LC Audio in Denmark. I talked to Allen about it and Allen contacted him and I can only go by the answer we got then.
But, hey, that is a good thing. I think that is great.
Here is something to chew on, see below:
Cheers, Joe
Attachments
Chewed on that already Joe, the discrete fet buffer I was using in past posts got a approval tick from John Curl and I think the internal buffer of the AD844 is better now that the I/V are stacked and not being current starved.
And Pedja has not heard stacked 844 i/v's and with the TDA1541 (4mA) it really needs it X 3, at least as Mick from Supratek found out.
The PCM1704/2 only needs 2 stacked as it is only 1.2mA, one is getting current starved clearly audible once double stack is heard. So one with the 1541 was in big trouble with one and Mick found that 3 was the lowest amount.
Cheers George
And Pedja has not heard stacked 844 i/v's and with the TDA1541 (4mA) it really needs it X 3, at least as Mick from Supratek found out.
The PCM1704/2 only needs 2 stacked as it is only 1.2mA, one is getting current starved clearly audible once double stack is heard. So one with the 1541 was in big trouble with one and Mick found that 3 was the lowest amount.
Cheers George
Hi George
You are right, haven't yet read the thread, but made a start. The idea of stacking 844's is an intriguing one. As you have probably surmised, most of the idling current is taken up by the front-end and not the buffer, so if you wish to absorb the DAC's current, that I understand is problematic. I, on the other hand, would not do it that way.
Most modern current DACs are differential, and relative high current, a la Burr-Brown, so making a differential circuit would entail two AD844 and possibly both Pin 5s resistively joined together. Indeed that works. But even that would mean excessive current, so...
But there is a better way to use the 844 differentially that needs only one 844 and also makes sure that only a fraction of the DAC's current needs to be absorbed by Pin 2. For a start, don't ground Pin 3. But it only works with DACs with no offset voltage, but offset current is OK. Modern Burr-Browns are OK.
It's getting late here, but I will show it shortly, let's see how tomorrow pan out.
But it's interesting that the 844 has been 'discovered' on diyaudio. It is an extraordinarily flexible device, in fact been using it for about ten years, in a number of ways in custom electronics, for things like microphone preamps, where two are used, Pin 2s tied together via a small value resistor and Pin 5s resistively joined together to set gain. But you need to select the 844s as they are not equally low noise - and selected ones are excellent.
BTW, the schematic shown above was a headphone amp and did need an external buffer, specially driving as low as 16 Ohm - but otherwise I agree, the internal buffer is not that bad. (It can drive 600 Ohm Sennheisers of yore.)
Will post a differential I/V shortly, promise. And trust me, it's a ripper, even if some may find some aspect to pick to pieces, some can't help themselves, when actually hearing it would dispel any notions that are excessively negative.
Cheers, Joe
Last edited:
Oh, just one final thing:
Yes, I realise the method used. Done 1541 DACs going way back in the 80s. But I now prefer to use differential DACs with no voltage offset and only (if they have it) current offset. You can then cancel it all out and then use the Trim facility of the 844. But you can't do that with single-ended DACs. Then you have to cancel it out or other technique.
The schematic/topology I will post will show this. It's only just another way to implement 844 as I/V - but now how to do it with differential DACs. I am sure some will find it interesting, it is not in any way to be seen as a criticism of what is your preference, just another option. The more choices we have... etc.
Cheers Joe
Yes, I realise the method used. Done 1541 DACs going way back in the 80s. But I now prefer to use differential DACs with no voltage offset and only (if they have it) current offset. You can then cancel it all out and then use the Trim facility of the 844. But you can't do that with single-ended DACs. Then you have to cancel it out or other technique.
The schematic/topology I will post will show this. It's only just another way to implement 844 as I/V - but now how to do it with differential DACs. I am sure some will find it interesting, it is not in any way to be seen as a criticism of what is your preference, just another option. The more choices we have... etc.
Cheers Joe
Last edited:
I myself am a Multibit fan and have never been impressed with Delta Sigma or Hyperstream. I find them smooth/detailed but boring, properly implemented multibit leaves them for dead with jump factor drive and dynamics and can be equally as smooth. And that's not just me that thinks this, a lot of heavy hitters feel the same way and that we're doing the same as (Beta V VHS) all over again for ease of manufacture of the cheaper SD/HS dac chips over the far more costly to produce R2R Multibit chips.
Mick Maloney (Supratek) as we speak is already doing his ES90018 Sabre dac in dif mode he will be letting us know here when it's done, and on SNA. If you read these pages he's already done 3 stack 844's into his TDA1541 and was most impressed compared to one.
Now let's get back on track and find the volunteer to stack the buffer also on the 844.
Cheers George
Mick Maloney (Supratek) as we speak is already doing his ES90018 Sabre dac in dif mode he will be letting us know here when it's done, and on SNA. If you read these pages he's already done 3 stack 844's into his TDA1541 and was most impressed compared to one.
Now let's get back on track and find the volunteer to stack the buffer also on the 844.
Cheers George
Last edited:
Maybe you could ping some of the heavy-hitters here:
http://www.diyaudio.com/forums/part...-feedback-amplifier-ad830-lt6552-tda856x.html
to see if they know how to contact him.
Cheers,
Jeff
http://www.diyaudio.com/forums/part...-feedback-amplifier-ad830-lt6552-tda856x.html
to see if they know how to contact him.
Cheers,
Jeff
You know the ones they only frequent the DIYHiFi.org • View forum - Digital Stuff forums.
Cheers George
Cheers George