Re: Datasheet schematic, page 1
Hi Carlos, My first thought was exactly the same as yours, but I wonder if it might be because the input impedance of the circuit is low. Looking at the formula on the datasheet 10uf is actually about right for the resistor values they chose.
Carlos, what voltage rating for those caps would you use?
carlosfm said:
Hi Carlos, My first thought was exactly the same as yours, but I wonder if it might be because the input impedance of the circuit is low. Looking at the formula on the datasheet 10uf is actually about right for the resistor values they chose.
Carlos, what voltage rating for those caps would you use?
Re: Re: Datasheet schematic, page 1
From a quick check at the datasheet I don't have so clear what's the input impedance of this chip on the non-inverting input.
For the inverting input it's ok, but I was not talking about those.
If it's polyprop you're after, I've never seen much lower than 250V. Mabe there are 100V.
MKT (polyester), with very rare exceptions, are nothing special = veiled.
There are some bipolar electrolythics that are quite good, I have some. From Roederstein.
Very rare to get decent sounding electros for DC coupling, MKP is a safer bet, usually.
Russ White said:
From a quick check at the datasheet I don't have so clear what's the input impedance of this chip on the non-inverting input.
Russ White said:
For the inverting input it's ok, but I was not talking about those.
Russ White said:
If it's polyprop you're after, I've never seen much lower than 250V. Mabe there are 100V.
MKT (polyester), with very rare exceptions, are nothing special = veiled.
There are some bipolar electrolythics that are quite good, I have some. From Roederstein.
Very rare to get decent sounding electros for DC coupling, MKP is a safer bet, usually.
Here are some options for low Cin (but none really great as far as I am concerned)
1 Rin 16K Cin 1uf gives 9.94hz corner. Ick though, lots of resistor noise.
2 Rin 8K Cin 2uf gives the same, but is only a little better.
3 Rin 4.7K and Cin 4.7uf gives us 7.20hz still high R.
4 Rin 3.3K and Cin 4.7uf gives 10.26hz R getting better.
On the other side if you you go higher than 10uf for Cin/Ci you could use lower resistor values in the circuit which would have a lot of benefit.
Rin of 680R and Cin of 22uf gives 10.63hz which is not bad at all, and if you bump Cin to 33uf you get 7.09hz.
So what I was thinking was using non-polar nichicon muse 33uf electrolytic caps bypassed by 1uf MKTs.
That should allow me to use nice low value resistors to set gain and impedance.
Food for thought.
1 Rin 16K Cin 1uf gives 9.94hz corner. Ick though, lots of resistor noise.
2 Rin 8K Cin 2uf gives the same, but is only a little better.
3 Rin 4.7K and Cin 4.7uf gives us 7.20hz still high R.
4 Rin 3.3K and Cin 4.7uf gives 10.26hz R getting better.
On the other side if you you go higher than 10uf for Cin/Ci you could use lower resistor values in the circuit which would have a lot of benefit.
Rin of 680R and Cin of 22uf gives 10.63hz which is not bad at all, and if you bump Cin to 33uf you get 7.09hz.
So what I was thinking was using non-polar nichicon muse 33uf electrolytic caps bypassed by 1uf MKTs.
That should allow me to use nice low value resistors to set gain and impedance.
Food for thought.
Ci will potensially see some voltage, so I'm thinking min100V here.
since it's part of the NFB, i aint touching it, since it will probably change operation parameters, and i dont have the knowledge to compensate for it.
Cin is wildly overrated, i'm gong for a standard 4.7 uf and 3k3 series resistance or somthing in the region. this is an HV opmap after all, and i simply dont see the need to use theese extreme componen values.
perhaps i'l toss the filter out all together. I'l see when i get thing together.
now, as for my atempt at pcb work, i'm still working on it, and probably will for some time, making as close to a whole amp as possible.
transistors will be mounted on a heatsink, connected by wires, not an a board.
some keywords here is double rectifiers, chassi mounted or on the pcb.
40.000uf total, 20kuf pr rail. footprint for 80V caps.
the lm4702 gets it's own heatsink.
the feedback circut will be based on metal's work, though i'l be making the changes i see fit.
I'm really trying to make this a high quality piece of work.
-Marius
since it's part of the NFB, i aint touching it, since it will probably change operation parameters, and i dont have the knowledge to compensate for it.
Cin is wildly overrated, i'm gong for a standard 4.7 uf and 3k3 series resistance or somthing in the region. this is an HV opmap after all, and i simply dont see the need to use theese extreme componen values.
perhaps i'l toss the filter out all together. I'l see when i get thing together.
now, as for my atempt at pcb work, i'm still working on it, and probably will for some time, making as close to a whole amp as possible.
transistors will be mounted on a heatsink, connected by wires, not an a board.
some keywords here is double rectifiers, chassi mounted or on the pcb.
40.000uf total, 20kuf pr rail. footprint for 80V caps.
the lm4702 gets it's own heatsink.
the feedback circut will be based on metal's work, though i'l be making the changes i see fit.
I'm really trying to make this a high quality piece of work.
-Marius
Page 10.demogorgon said:
10uf is precisely a function of having the 1.8K Rin and Ri, which I am sure you know. It easy enough to get the C value lower, but you pay the price of more resistor noise. Icky.
Bypassing higher value bipolars with lower value film caps is a good compromise which should result in lower total noise then using smaller caps and larger resistors.
But do whatever you are comfortable with. Really there is no wrong way.
Cheers!
Russ
Russ White said:
Page 10.
10uf is precisely a function of having the 1.8K Rin and Ri, which I am sure you know. It easy enough to get the C value lower, but you pay the price of more resistor noise. Icky.
Bypassing higher value bipolars with lower value film caps is a good compromise which should result in lower total noise then using smaller caps and larger resistors.
But do whatever you are comfortable with. Really there is no wrong way.
Cheers!
Russ
Ummm...any problem using a parallel combination of smaller caps for the input filter to get to the magic 10uF????
Personally I suspect they are using a non polar electrolytic on the demo board.
demogorgon said:
If it is important enought to make the datasheet, it is important enough to design for as optimal values as practical, but I am sure there are many combinations which may sound good enough to you. As for me I like to keep away from itroducing any noise I don't have to.
Cheers!
Russ
Reading on page 4 Rs max is 100K, if you used 3.3K for Rin you would need something very close to 100K for Rf and Rs to get the correct bias/gain. I personally would prefer to operate more within the envelope then on the edge.
With 680R to get the 32x gain you only need 21K Rf/Rs much more reasonable.
even with 1K (which the datasheet actually suggests) 30K Rf/Rs would work fine, just slightly less gain (31x).
100K for a feedback resistor is very high.
:EDIT:
Also note that on page 11 it states that lower Rin values are desirable to help prevent oscillation.
With 680R to get the 32x gain you only need 21K Rf/Rs much more reasonable.
even with 1K (which the datasheet actually suggests) 30K Rf/Rs would work fine, just slightly less gain (31x).
100K for a feedback resistor is very high.
:EDIT:
Also note that on page 11 it states that lower Rin values are desirable to help prevent oscillation.
One of the National engineers on the design team for the LM4702 was very kind to call me this morning. I just got the low down.
In short the cap Ci and Cin will never see more than 6V for normal operation. So using a good bipolar will be very easy. Second He verfied that a bipolar with a film cap in series is precisely the sort of setup they had in mind for high end audio, which explains the demo board, which he also explained is not so high end as it uses only a bipolar alone to save cost for Cin. Ci in the demo board is a film cap as I suspected, but it can be treated exactly as Cin.
Also he stressed very strongly that the size of Rf and Rs are critical an should be kep to below 50K for best results. He confirmed that 680R Ri and Rin were good solid choices along with Rf and Rs of 21K.
He told me that because the imputs are very sensitive and high bandwidth/high slew rate that the amp is very prone to oscillate if you use high values for the gain setting resistors.
Its good to know I am on the right track. My mind is much more at ease.
In short the cap Ci and Cin will never see more than 6V for normal operation. So using a good bipolar will be very easy. Second He verfied that a bipolar with a film cap in series is precisely the sort of setup they had in mind for high end audio, which explains the demo board, which he also explained is not so high end as it uses only a bipolar alone to save cost for Cin. Ci in the demo board is a film cap as I suspected, but it can be treated exactly as Cin.
Also he stressed very strongly that the size of Rf and Rs are critical an should be kep to below 50K for best results. He confirmed that 680R Ri and Rin were good solid choices along with Rf and Rs of 21K.
He told me that because the imputs are very sensitive and high bandwidth/high slew rate that the amp is very prone to oscillate if you use high values for the gain setting resistors.
Its good to know I am on the right track.
My mind is much more at ease.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.