I made that comment because there have been some heated debates on IC's vs discretes on the forum . . . and here we have a product reviewed by someone independent and it got a good rating. ( this pre-amp was also one of Stereophiles 'Product of the Year'). Also, there was another debate that raged on about line transformers as well.
I am drawing the conclusion that there is a lot of prejuduce against these integrated solutions. It might be on an AB test a Goldpoint Attenuator is a bit more transparent, but I doubt a well engineered PGA23xx would come out as c***p.
I am drawing the conclusion that there is a lot of prejuduce against these integrated solutions. It might be on an AB test a Goldpoint Attenuator is a bit more transparent, but I doubt a well engineered PGA23xx would come out as c***p.
Bonsai said:
B,
Call me cynical, but I think the prejudice is against:
1. Relatively expensive equipment using relatively low performance parts for the sake of convenience and cost-cutting.
2. Reviewers giving good reviews to this equipment because a) their job / income depends upon it, and b) they are swayed by the badge on the equipment and the company reputation, and hence want to be seen to be 'a guru capable of recognising the wonderful quality of these expensive products'.
Personally, I think these prejudices are justified.
All of the integrated solutions seem to be designed down to a price. I hope one of them turns out to be an excellent performer, but I have my doubts.
Interesting thread. : )
The WM8816 needs an external inverting op-amp... so this may play to your advantage as you can roll your own...
I heard bad things about the PGAs too, from someone who measured them. Their S/N seems to be all over the place especially for settings between 0dB and +31.5dB.
But I'm still happy with my PGA2320 based preamp (not for long though).
I heard bad things about the PGAs too, from someone who measured them. Their S/N seems to be all over the place especially for settings between 0dB and +31.5dB.
But I'm still happy with my PGA2320 based preamp (not for long though).
switching into (virtual) ground was a widespread practice with the 40xx series, because of their severe common mode issues.
Modern solid state pots don't do that, they are marketed to be 'true potentiometers'.
Anyway, if it is a ss-potentiometer or small preamp like the PGA series, I havn't seen a good implementation yet.
So I wouldn't dare to blame the part.
regards
Looking at the part, I think there are a few basic things I'd do differently if I go for this:-
1. Use the PGA2310 - +-15V supply rails for decent headroom.
2. In the application circuits and in the Elektor implementation, they connect the controller directly to the digital inputs. In my book, this is a no-no because of noise feed through. The digital control signals need to be isolated via some opto's, and it goes without saying that there should be no physical conncetion between the digital and analog sections - completely isolated power supplies (but see the point below about decoupling to chassis)
3. Clearly, the supply needs to be clean - e.g. +15V well regulated. I'd then decouple these further right at the chip with 22 Ohms in sereis with supply with 100uF to ground at the supply pins - this will give around 40 or 50db of power supply noise rejection at 20kHz
4. The inputs should be well terminated wrt to RFI - already discussed at quite some length on a few threads on the forum - so for example 50 Ohm in series with each input and 100pF to GND and then GND decouplled via 0.1uF or so to Chassis (digital side also decoupled like this to chassis - kill any AC Vdiff between analog and digital ccts)
5. Output isolated from capacitive loads via 50 Ohm resistor (nothing different here that would be the case in most SS amp circuits). note th e PGA is supposed to be able to drive quite heavy capacitive loads - the data sheet mentions 1000pF and 600Ohm.
1. Use the PGA2310 - +-15V supply rails for decent headroom.
2. In the application circuits and in the Elektor implementation, they connect the controller directly to the digital inputs. In my book, this is a no-no because of noise feed through. The digital control signals need to be isolated via some opto's, and it goes without saying that there should be no physical conncetion between the digital and analog sections - completely isolated power supplies (but see the point below about decoupling to chassis)
3. Clearly, the supply needs to be clean - e.g. +15V well regulated. I'd then decouple these further right at the chip with 22 Ohms in sereis with supply with 100uF to ground at the supply pins - this will give around 40 or 50db of power supply noise rejection at 20kHz
4. The inputs should be well terminated wrt to RFI - already discussed at quite some length on a few threads on the forum - so for example 50 Ohm in series with each input and 100pF to GND and then GND decouplled via 0.1uF or so to Chassis (digital side also decoupled like this to chassis - kill any AC Vdiff between analog and digital ccts)
5. Output isolated from capacitive loads via 50 Ohm resistor (nothing different here that would be the case in most SS amp circuits). note th e PGA is supposed to be able to drive quite heavy capacitive loads - the data sheet mentions 1000pF and 600Ohm.
Quote
"I heard bad things about the PGAs too, from someone who measured them. Their S/N seems to be all over the place especially for settings between 0dB and +31.5dB. "
Yes, I can accept this - if the supply was noisy, or there was digital feed through coming into the device, you would get a serieopus degradation in s/n especially if you then set the device up to amplify (i.e. between 0dB and +31db gain range.
I think if you use a chip like this, the same precautions that you take in laying out a discrete cct need to be applied - so isolation, separete supplies etc are the order of the day.
"I heard bad things about the PGAs too, from someone who measured them. Their S/N seems to be all over the place especially for settings between 0dB and +31.5dB. "
Yes, I can accept this - if the supply was noisy, or there was digital feed through coming into the device, you would get a serieopus degradation in s/n especially if you then set the device up to amplify (i.e. between 0dB and +31db gain range.
I think if you use a chip like this, the same precautions that you take in laying out a discrete cct need to be applied - so isolation, separete supplies etc are the order of the day.
Imagine a parallel universe where volume control had ALWAYS (!!) been accomplished with an a digital pot such as the LM1972 or CS3310. Then without prior president someone invents and introduces the three pin analog pot. Imagine the amazement at the utter simplicity of this virtually passive device.
sam9 said:
Yes, there is always a cultural factor in these issues, isn't it?
BTW, the CS3310, or the later 3308 and 3318 are much better than the dated 1972 in terms of linearity at higher frequencies and higher levels. We always have the tendency to throw all these chips on one heap and assume there all the same, more or less, but they aren't. Just like different types of output power transistors can make a big difference in power amp performance, so it is with programmable attanuators or volume controls. It pays to search for the best.
Jan Didden
The problem is not with the clock signals into the chip. If you stop all the digital control signals (i.e. put then into a fixed state), you still get a lot of hash on the control lines - this is the stuff that ends up going into the chip. If opto's are used - and you decouple correctly to GND on both sides of the opto barrier, you can get rid of this hash. Another option is to to put the controller into 'shutdown' mode after making any set-up changes to the PGA2310 - this reduces uController I/O line noise a lot, but does not kill it completely. You need the opto's to do that.
Re the undefined input resistance - the chip Rin is typically 10k Ohms. I expect any source components to be able to drive this type of load easily. Might be some tube output devices struggle a bit, but then I can always arrange a HiZ buffer on one of the input sockets for that purpose.
Re the undefined input resistance - the chip Rin is typically 10k Ohms. I expect any source components to be able to drive this type of load easily. Might be some tube output devices struggle a bit, but then I can always arrange a HiZ buffer on one of the input sockets for that purpose.
You could also put the serial outputs into a hi-Z mode during sleep and use small-ish pull-down resistors on them. That should squelch any noise that Bonsai might be seeing. Should've done this on my own PGA2320 controller...janneman said:
If you do a shutdown or sleep of the controller, the clock in the controller chip stops until wakeup when you want to change the level. There is absolutely no hash or whatever coming from anywhere. Opto's may give you a warm and fuzzy feeling but that's about all.
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.