Hi George
I read somewhere in this thread that when stacking 2-3 ad844 and use the buffer of one chip the sound is quite good. I am wondering wheter an external buffer is needed.
It seems that Pedja connect pin 6 to his jfet buffer. is it better to take the voltage out from pin 5 i.e top of the i/v resistor?
Why run a buffer into another buffer. Just take the signal from the TZ into an external >100kohm buffer.
And yes you can use the 844's own buffer just one or stacked, it sound very good, but just doesn't seem to have the drive that an external OPA627 has.
BTW can anyone help me with my last post on the Lightspeed Attenuator thread? post there if you can.
Cheers George
AD844 opinion...
The thread has grown over the years. Originally George started it to share his results with the often mentioned PCM1704 series DAC's. Others have shared about using the AD844 with other DAC's, commonly the TDA1541 and A series. At the moment I have a version of George's original AD844 I/V with a triple stack of AD844's working behind my PCM1704's. It is a good sound. I have the pin 6 outputs tied together via 100 Ohm series resistors. You only need to use an AD844 null circuit on the lowest buffer. Would a Buffer off TZ be better? Yes. I think based on having used that configuration with TDA1541's it would be more dynamic. So... This is just my opinion based on my experience. Stack 3 to 4 AD844's. Take TZ pin 5 output to a high impedance buffer. BUF03's serve me well. OPA627 DIP also work beautifully in this roll.
AD844 current mirrors...
AD844's do have a datasheet input of 50 Ohms as I recall. When we stack them in a sense you do decrease the input impedance. Since a DAC is looking to sink or source a current it actually wants to be terminated into a direct short. So a triple stack will look more like a 12.5 Ohm load. The current mirrors average their current across the TZ resistor. 1.5K Ohm is pretty standard on a TDA1541, A. So keep the original TZ resistor value. I also would not change the cap across the TZ resistor unless you can measure frequency response. If you can measure that then look at the response around 19 Khz to 20 Khz. Should just be starting to roll off out there. Adjust capacitor value for desired roll off.
Thanks, guys. I don't have any good 1n8 capacitor so I am using 1nf at the input of ad844 to ground. does the value of this capacitor matter either with one or multiple ad844?
I plan to use a discrete buffer as I can get one ready made off eBay at reasonable cost. it seems to be well thought of by several members on diyaudio.
I plan to use a discrete buffer as I can get one ready made off eBay at reasonable cost. it seems to be well thought of by several members on diyaudio.
Last edited:
1n8 capacitor
The value of that cap should remain as close to 1n8 as you can. Say 5% tolerance. You could parallel caps to get close. It will directly affect the roll off so if you can measure it that is fine. My preferred choice is Polystyrene however I have used Polypropylene in that position with no issues. The value does not change unless you change the TZ resistor. So in fact if your using a 2.0K Ohm resistor 2 nF is likely right on. So look at it this way. A single or a triple stack stays the same. Unless you change the TZ resistor. A buffer is a great idea. Member suggestions are great.
As for I/V, read POST#1317. You can either choose AD844's internal buffer (through pin6) or followed by external buffer (through pin5)
AD844 opinion...
That is correct. High impedance is key to taking the pin 5 as an output. So that means you would want to choose a JFET input type buffer (BUF03) or op amp OPA627 (DIP) preferred. The nature of Pedja's DDNF circuit also means there is an advantage to doing that there as well.
Pedja Rogic DDNF circuit boards
Yes. OPA627, I have experience with. Very good audio chip.
Hi guys,
Last night I quickly built the circuit, and this is my 2nd attempt. Good news is that it works perfectly. In my opinion, a better solution to my OPA861 I/V+buffer board. I am not saying OPA861 is inferior, but my implementation included ultra low noise +-5V LDOs which may be the cause of that. Anyway, I like the stacked AD844 very much. The external buffer I like are still ADA4627 or Muses02, both have their advantage. ADA4627 is more precise, whereas Muses02 produces more relaxed sound. It's hard to choose between them. I still need time to make the judgement...
PS. I am not including input 1.8nF cap as Pedja's circuit. I/V resistor is 1.8K with two 3nf foil caps in series (=1.5nF). This foil cap is very inexpensive but has performance almost equal to Soshin SE99 mica
Last night I quickly built the circuit, and this is my 2nd attempt. Good news is that it works perfectly. In my opinion, a better solution to my OPA861 I/V+buffer board. I am not saying OPA861 is inferior, but my implementation included ultra low noise +-5V LDOs which may be the cause of that. Anyway, I like the stacked AD844 very much. The external buffer I like are still ADA4627 or Muses02, both have their advantage. ADA4627 is more precise, whereas Muses02 produces more relaxed sound. It's hard to choose between them. I still need time to make the judgement...
PS. I am not including input 1.8nF cap as Pedja's circuit. I/V resistor is 1.8K with two 3nf foil caps in series (=1.5nF). This foil cap is very inexpensive but has performance almost equal to Soshin SE99 mica
Last edited:
Hi George,
Yes, I did. (I followed the same pattern as my stacked ad1860s - solder all pins together.) It turns out the sound is very good. Is there any downside if I keep it like this way? Do you think the stacked TZ pin will have more power to drive the IV resistor?
Thanks.
Stacking...
I would caution against that. You are kind of assuming that all the chips behave the same. In my experience they are close however tying the pin 6's together without 100 Ohm safety resistors could cause some unintended problems. You may have some bias issues internally since the null pins all tie together as well. TZ = more power? Well in a sense no. You are averaging the current mirrors. Which makes for more accuracy. AD844's have actually very low internal current likely due to die heating concerns. So.... It works. If you like the sound enjoy it. I am not telling you to undue any of the pins. I do it as George pointed out in the other post. Play it and see if any issues occur. Do the op amps get really Hot?
I would caution against that. You are kind of assuming that all the chips behave the same. In my experience they are close however tying the pin 6's together without 100 Ohm safety resistors could cause some unintended problems. You may have some bias issues internally since the null pins all tie together as well. TZ = more power? Well in a sense no. You are averaging the current mirrors. Which makes for more accuracy. AD844's have actually very low internal current likely due to die heating concerns. So.... It works. If you like the sound enjoy it. I am not telling you to undue any of the pins. I do it as George pointed out in the other post. Play it and see if any issues occur. Do the op amps get really Hot?
Thanks for the warning. I will try to isolate those unnecessary pins tonight. I am not using pin6. Do I need to un-tie it also? As George has mentioned, pin 2,3,4,6,7 can be tied together. BTW, I didn't look into buffer OP yet. I tried single AD844 without problem, and then staked them for listening.
Last edited:
Leave all pin 6's float, never tie them together, if using an external buffer.
Only if you want to use the 844's buffers stacked, don't tie them directly together, they each need a series 100ohm isolating resistor on each leg then the other side of the resistor can be together and the output taken from there.
Cheers George