I've just started playing with an Atmel microcontroller and a PGA2310. The PGA2310 is capable of driving loads down to 600Ohm. So a buffer on the output of PGA2310 is perhaps not really needed. But is it recommended to place a buffer there anyway? I have some 2SK170BL which I was planning to use to make some borbely buffers, is that a good solution? Perhaps I should use another buffer or maybe no buffer at all? Suggestions?
One more thing, I have two pretty nice powersupplies (discrete regulators) that I want to use. One powersupply per channel. The PGA2310 have two channels, I just want to use one. Should I ground the unused channel on the PGA2310, or just leave it?
/Freddie
One more thing, I have two pretty nice powersupplies (discrete regulators) that I want to use. One powersupply per channel. The PGA2310 have two channels, I just want to use one. Should I ground the unused channel on the PGA2310, or just leave it?
/Freddie
Good questions Freddie. You should probably let your ears decide about the buffer. Surely in most situations an external buffer will just reduce the transparency. Why, if i may ask are you using a single ic per channel? Do you reckon they will work better as dual mono? This is probably also worth auditioning. You should make sure the unused input is grounded and the volume byte for that channel set to zero.
Actually, this seems to be the better way to use those chips (one chip per channel). Audionet are said to do this, and their amps are getting some of the best reviews (for those who want to believe in reviews).
Most DACs are differential anyway, you just have to take the signal before the last stage which is ususally a single opamp configured to work both as difference amp and low pass. It may make sense to incorporate the filter function of this stage separately into the + and - channels.
For non-digital signals like radio, you can use an inverter to symmetrize the signal.
Using an output buffer might be a smart idea. I am a little wary of the op amp used inside the PGA. It might be fine if the load is 10 k, but below or with a long cable, I'd probably use a good op amp for a buffer.
Greetings,
Eric
Most DACs are differential anyway, you just have to take the signal before the last stage which is ususally a single opamp configured to work both as difference amp and low pass. It may make sense to incorporate the filter function of this stage separately into the + and - channels.
For non-digital signals like radio, you can use an inverter to symmetrize the signal.
Using an output buffer might be a smart idea. I am a little wary of the op amp used inside the PGA. It might be fine if the load is 10 k, but below or with a long cable, I'd probably use a good op amp for a buffer.
Greetings,
Eric
Output buffer
Yes, you're right my DAC is differential. It is easy to take out the balanced signal before it gets buffered, filtered and converted to a singel ended signal.
That could be nice, any suggestions for a simple but good inverter? An OP-amp configured for gain -1 perhaps?
I have some OPA627 that I could use. Otherwise a simple buffer with bipolar transistors, or maybe the borbely jfet buffer based on the white cathode follower might be good. But the jfet buffer is probably not very good driving low impedance loads.
Thanks for the suggestions
/Freddie
Yes, you're right my DAC is differential. It is easy to take out the balanced signal before it gets buffered, filtered and converted to a singel ended signal.
That could be nice, any suggestions for a simple but good inverter? An OP-amp configured for gain -1 perhaps?
I have some OPA627 that I could use. Otherwise a simple buffer with bipolar transistors, or maybe the borbely jfet buffer based on the white cathode follower might be good. But the jfet buffer is probably not very good driving low impedance loads.
Thanks for the suggestions
/Freddie
Yes, an opamp is the best and easiest way to make an inverter. As there is no common mode voltage, even mediocre op amps will do nicely.
I have been using an OPA-627 (with a four transistor unity gain buffer INSIDE the feedback loop) in my amp for years and years. There might be better and cheaper op amps by now. I believe the AD797 (of the "old" league) might be a better choice here because
a) being a bipolar cascoded device, it might have even less common mode distortion
b) you'd be driving it from near zero virtual impedance, so its excessive current noise does not hurt
All in all, DAC op amps have to work much harder because of all the HF junk, so selection is much more critical there than for a volume control buffer.
I have been using an OPA-627 (with a four transistor unity gain buffer INSIDE the feedback loop) in my amp for years and years. There might be better and cheaper op amps by now. I believe the AD797 (of the "old" league) might be a better choice here because
a) being a bipolar cascoded device, it might have even less common mode distortion
b) you'd be driving it from near zero virtual impedance, so its excessive current noise does not hurt
All in all, DAC op amps have to work much harder because of all the HF junk, so selection is much more critical there than for a volume control buffer.
The problem with the PGA is that you have no access to the feedback loop. So if you use any of these buffers which have no overall feedback loop, you have the worst of both worlds, i.e. a mediocre opamp with low open loop bandwidth and lots of feedback and a "no-feedback" buffer which has considerable residual distortion. This is going to satify neither school of thought...
Well.. One alternative is to forget about the PGA2310 and get an LM1972 instead. The LM1972 have no internal opamp, and should have a buffer at the output. Or run the PGA2310 without external buffer.
However, why can't I place an as you call "no-feedback" buffer after the PGA2310 if the buffer itself has low distorsion without feedback? I have seen some designs that have the outputstage (buffer) outside the feedback loop.
/Freddie
capslock said:
I do not think that the PGA is that bad a performer. i would estimate thd @ 20k to around .004% given from their datasheet and the bandwidth filter they use. and this is 3Vrms into 600R load.
If it should be better then there is PGA2311 (+/-5V suppply) in a-grade.
at 2Vrms (limited by supply voltage)
.0002% distortion at 1KHz
.001% distortion at 10kHz
and i would estimate .003% at 20KHz
is this "mediocre opamp" and mediocre "analog switch" performance?
Yeah, I wondered passingly about the PGA2311, how can it be better if it is only a +/- 5 V design?
Well the specs of the OPA2134 (that's essentially what is inside the PGA) are certainly good, but there are some golden ears that say that these are not enough. I would assume one needs to look at high order distortion products (mainly from crossover switching, probably hidden in the noise) or intermodulation distortion. There are certainly better op amps (AD797, OPA627), and you can always go discrete.
My experience with Op amps in my DAC is that there seem to be subtle differences between op amps with nominally adequate specs (such as the 2134 or the 2604) and some of the really good ones.
The no-feedback school is going to cry for another reason: what use is a unity gain, discrete, no global feedback buffer if the signal has already passed through one of those nasty op amps using lots of global feedback?
Well the specs of the OPA2134 (that's essentially what is inside the PGA) are certainly good, but there are some golden ears that say that these are not enough. I would assume one needs to look at high order distortion products (mainly from crossover switching, probably hidden in the noise) or intermodulation distortion. There are certainly better op amps (AD797, OPA627), and you can always go discrete.
My experience with Op amps in my DAC is that there seem to be subtle differences between op amps with nominally adequate specs (such as the 2134 or the 2604) and some of the really good ones.
The no-feedback school is going to cry for another reason: what use is a unity gain, discrete, no global feedback buffer if the signal has already passed through one of those nasty op amps using lots of global feedback?
capslock said:
its all about the analog switches. It looks like when they go down in supply voltage, the Ron and Delta Ron get up to ten times lower (+/-2.5V or +5V) than with +/-15V supply voltage ... Something to do with the process they are using. And geometry of the mosfets used in the swithes.
And +/- 2.5V supply is more than adequate if you use the switches right... In virtual ground mode....
capslock said:
I agree perfectly with you... run an IMD test with 18KHz+19KHz tones and check for harmonics ... Wonder if ... only the test equipment is good enough to reveal the differences of the really good opamp designs....
Maybe a settling test would be an good indicator too.....???
capslock said:
I agree with you.
capslock said:
Thats another story...
Only .. distortion introduced thermally could be minimized by adding a buffer....
But it would be very simple to make an active attenuator with high perfomance swithes and a discrete I/V converter with no feedback and a discrete buffer.... should be very easy to get sub -80dB THD.... When done right...
the sims shows it... just have to find some time to do real life tests....
just wanted to post some FFT but even in zip-file the size is 170K
so i am only going mention the results
at 1KHz :
350mVrms - 2nd : -105dB - 3rd : -150dB
700mVrms - 2nd : -100dB - 3rd : -135dB
2.000Vrms - 2nd : -90dB - 3rd : -120dB - 4th : -150dB
IMD 19k+20k 350mVrms each :
all other components is 100dB below 19k+20k.
so total IMD is like ~ -92dB..
This is a discrete I/V without feedback.... and without the buffer...
so if you can live with that you can use a discrete No global feedback I/V or add a buffer and make the VAS node high impedance add a 33pF cap to the node and add a feedback resistor... and there you go...
so i am only going mention the results
at 1KHz :
350mVrms - 2nd : -105dB - 3rd : -150dB
700mVrms - 2nd : -100dB - 3rd : -135dB
2.000Vrms - 2nd : -90dB - 3rd : -120dB - 4th : -150dB
IMD 19k+20k 350mVrms each :
all other components is 100dB below 19k+20k.
so total IMD is like ~ -92dB..
This is a discrete I/V without feedback.... and without the buffer...
so if you can live with that you can use a discrete No global feedback I/V or add a buffer and make the VAS node high impedance add a 33pF cap to the node and add a feedback resistor... and there you go...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.