capslock said:
To some point i will agree with you, but still the opamp needs to see a resistance which is equal on both inputs. => in noninverting : Z - feedback (I-) = Z input (I+) and for inverting : Z - feedback = Z to ground (I+).
The reason to do this is to get optimum dc offset and for Highperformance parts like AD8610/OPA627/AD797 Better distortion perfomance.
If we do not do this it is a lak of time choosing a switch with low distortion = flat delta-resistance. Thats the reason why the SSM2404 is made performe well into 10k an 100k loads.
For what NELSON PASS says :
Listen to him. Virtuel ground or shunt to ground means that a device like the SSM2404 only has a delta-resistance in the area of ~ 20mOhm and when it has to shunt like 2k to ground => thd in the area of 100dB. With 10k this is -113dB.
The 20mOhm is a guess.. Maybe this is even lower like 5 - 10mOhm => 10k => -120dB <-> -126dB
Sonny
OPA627 and AD8610
The OP627 is a Difet input op amp with IB of 5pA max . The AD8610 is a Jfet input op amp with IB of 10pA max. They can see high resistance to ground without offset problem. Oh well one out of three is not too bad since the AD797 has an IB of 0.25 uA.
The OP627 is a Difet input op amp with IB of 5pA max . The AD8610 is a Jfet input op amp with IB of 10pA max. They can see high resistance to ground without offset problem. Oh well one out of three is not too bad since the AD797 has an IB of 0.25 uA.
Hugh,
Just a quick note, absolute phase in audio is audible.
Nelson,
I like the sound of the X volume control ... assuming the patent is issued, will Pass Labs have the usual policy toward diy efforts?
Just a quick note, absolute phase in audio is audible.
Nelson,
I like the sound of the X volume control ... assuming the patent is issued, will Pass Labs have the usual policy toward diy efforts?
Re: OPA627 and AD8610
Yes that is right, but distortion performance will drop if the input are going to see high or unbalanced source resistance... Just try make some sims on those opamps versus input resistance ... And the sim do not lie!
😉
mmmhhh .. I am going to get some direct hits today or tomorow .. i think Now i am going to post on the ongoing Tripath thread.
Sonny
HarryHaller said:
Yes that is right, but distortion performance will drop if the input are going to see high or unbalanced source resistance... Just try make some sims on those opamps versus input resistance ... And the sim do not lie!
😉
mmmhhh .. I am going to get some direct hits today or tomorow .. i think Now i am going to post on the ongoing Tripath thread.
Sonny
Jfet
A jfet amp does not require matched input impedance to get good distortion performance. The input current is only a real issue for when the switches are open. During the time switches are set to provide the actuall attenuation the input impedance should be in the realm of 10K to 100K for reasonable noise performance. The circuit for a typical single ended opamp circuit often has
differing inpedances seen by the inverting and non-inverting inputs. Trusting an op amp model for distortion measurements is self delusion of the highest order. Op amp models are usually macromodels and do not even use the models for actual transistors in the device. Also Spice circuits using identical models for the transistors in a differential pair will give much better results than in real life, due to differences parameters in the two actual devices. Spice is just not a good indicator of distortion as Mr. Pass has indicated elsewhere.
* AD797B SPICE Macro-model 10/92, Rev. A
* AAG / PMI
*
* This version of the AD797 op amp model simulates the worst case
* parameters of the 'B' grade. The worst case parameters used
* correspond to those in the data sheet.
*
* Copyright 1992 by Analog Devices, Inc.
*
* Refer to "README.DOC" file for License Statement. Use of this model
* indicates your acceptance with the terms and provisions in the License
* Statement.
*
* Node assignments
* non-inverting input
* | inverting input
* | | positive supply
* | | | negative supply
* | | | | output
* | | | | | decompensation
* | | | | | |
.SUBCKT AD797B 1 2 99 50 38 14
*
* INPUT STAGE & POLE AT 500 MHz
*
IOS 1 2 DC 100E-9
CIND 1 2 20E-12
CINC1 1 98 5E-12
GRCM1 1 98 POLY(2) 1 31 2 31 (0,5E-9,5E-9)
GN1 0 1 44 0 1E-3
CINC2 2 98 5E-12
GRCM2 2 98 POLY(2) 1 31 2 31 (0,5E-9,5E-9)
GN2 0 2 47 0 1E-3
EOS 9 3 POLY(1) 22 31 40E-6 1
EN 3 1 41 0 0.1
D1 2 9 DX
D2 9 2 DX
Q1 5 2 4 QX
Q2 6 9 4 QX
R3 97 5 0.5172
R4 97 6 0.5172
C2 5 6 3.0772E-10
I1 4 51 100E-3
EPOS 97 0 99 0 1
ENEG 51 0 50 0 1
*
* INPUT VOLTAGE NOISE GENERATOR
*
VN1 40 0 DC 2
DN1 40 41 DEN
DN2 41 42 DEN
VN2 0 42 DC 2
*
* +INPUT CURRENT NOISE GENERATOR
*
VN3 43 0 DC 2
DN3 43 44 DIN
DN4 44 45 DIN
VN4 0 45 DC 2
*
* -INPUT CURRENT NOISE GENERATOR
*
VN5 46 0 DC 2
DN5 46 47 DIN
DN6 47 48 DIN
VN6 0 48 DC 2
*
* GAIN STAGE & DOMINANT POLE AT 55 Hz
*
EREF 98 0 31 0 1
G1 98 10 5 6 10
R7 10 98 10
E1 99 11 POLY(1) 99 31 -1.209 1
D3 10 11 DX
E2 12 50 POLY(1) 31 50 -1.209 1
D4 12 10 DX
G2 98 13 10 31 1E-3
R8 13 98 10
G3 99 14 98 13 34.558E-3
G4 99 16 98 98 34.558E-3
G5 14 15 15 14 20E-3
G6 16 17 17 14 20E-3
R9 15 18 400
R10 17 18 400
E3 18 98 16 98 1
R11 16 98 5.7875E7
C5 16 98 50E-12
V1 99 19 DC 4.3208
D5 16 19 DX
V2 20 50 DC 4.3208
D6 20 16 DX
RDC 14 98 1E15
*
* COMMON-MODE GAIN NETWORK WITH ZERO AT 4.3 kHz
*
ECM 21 98 POLY(2) 1 31 2 31 (0,0.5,0.5)
RCM1 21 22 1
CCM 21 22 37.281E-6
RCM2 22 98 1E-6
*
* POLE-ZERO PAIR AT 3.9 MHz/10 MHz
*
GPZ 98 23 16 98 1
RPZ1 23 98 1
RPZ2 23 24 0.63934
CPZ 24 98 24.893E-9
*
* NEGATIVE ZERO AT -300 MHz
*
ENZ 25 98 23 31 1E6
RNZ1 25 26 1
CNZ 25 26 -5.3052E-10
RNZ2 26 98 1E-6
*
* POLE AT 300 MHz
*
GP2 98 27 26 31 1
RP2 27 98 1
CP2 27 98 5.3052E-10
*
* POLE AT 500 MHz
*
GP3 98 28 27 31 1
RP3 28 98 1
CP3 28 98 3.1831E-10
*
* POLE AT 500 MHz
*
GP4 98 29 28 31 1
RP4 29 98 1
CP4 29 98 3.1831E-10
*
* OUTPUT STAGE
*
VW 29 30 DC 0
RDC1 99 31 23.25E3
CDC 31 0 1E-6
RDC2 31 50 23.25E3
GO1 98 32 37 30 25E-3
DO1 32 33 DX
VO1 33 98 DC 0
DO2 34 32 DX
VO2 98 34 DC 0
FDC 99 50 POLY(2) VO1 VO2 9.86E-3 1 1
VSC1 35 37 0.945
DSC1 30 35 DX
VSC2 37 36 0.745
DSC2 36 30 DX
FSC1 37 0 VSC1 1
FSC2 0 37 VSC2 1
GO3 37 99 99 30 25E-3
GO4 50 37 30 50 25E-3
RO1 99 37 40
RO2 37 50 40
LO 37 38 10E-9
*
* MODELS USED
*
.MODEL QX NPN(BF=5.5556E4)
.MODEL DX D(IS=1E-15)
.MODEL DEN D(IS=1E-12 RS=6.3708E3 AF=1 KF=1.59E-15)
.MODEL DIN D(IS=1E-12 RS=474 AF=1 KF=7.816E-15)
.ENDS AD797B
Two very generic transistor models in the whole circuit!
A jfet amp does not require matched input impedance to get good distortion performance. The input current is only a real issue for when the switches are open. During the time switches are set to provide the actuall attenuation the input impedance should be in the realm of 10K to 100K for reasonable noise performance. The circuit for a typical single ended opamp circuit often has
differing inpedances seen by the inverting and non-inverting inputs. Trusting an op amp model for distortion measurements is self delusion of the highest order. Op amp models are usually macromodels and do not even use the models for actual transistors in the device. Also Spice circuits using identical models for the transistors in a differential pair will give much better results than in real life, due to differences parameters in the two actual devices. Spice is just not a good indicator of distortion as Mr. Pass has indicated elsewhere.
* AD797B SPICE Macro-model 10/92, Rev. A
* AAG / PMI
*
* This version of the AD797 op amp model simulates the worst case
* parameters of the 'B' grade. The worst case parameters used
* correspond to those in the data sheet.
*
* Copyright 1992 by Analog Devices, Inc.
*
* Refer to "README.DOC" file for License Statement. Use of this model
* indicates your acceptance with the terms and provisions in the License
* Statement.
*
* Node assignments
* non-inverting input
* | inverting input
* | | positive supply
* | | | negative supply
* | | | | output
* | | | | | decompensation
* | | | | | |
.SUBCKT AD797B 1 2 99 50 38 14
*
* INPUT STAGE & POLE AT 500 MHz
*
IOS 1 2 DC 100E-9
CIND 1 2 20E-12
CINC1 1 98 5E-12
GRCM1 1 98 POLY(2) 1 31 2 31 (0,5E-9,5E-9)
GN1 0 1 44 0 1E-3
CINC2 2 98 5E-12
GRCM2 2 98 POLY(2) 1 31 2 31 (0,5E-9,5E-9)
GN2 0 2 47 0 1E-3
EOS 9 3 POLY(1) 22 31 40E-6 1
EN 3 1 41 0 0.1
D1 2 9 DX
D2 9 2 DX
Q1 5 2 4 QX
Q2 6 9 4 QX
R3 97 5 0.5172
R4 97 6 0.5172
C2 5 6 3.0772E-10
I1 4 51 100E-3
EPOS 97 0 99 0 1
ENEG 51 0 50 0 1
*
* INPUT VOLTAGE NOISE GENERATOR
*
VN1 40 0 DC 2
DN1 40 41 DEN
DN2 41 42 DEN
VN2 0 42 DC 2
*
* +INPUT CURRENT NOISE GENERATOR
*
VN3 43 0 DC 2
DN3 43 44 DIN
DN4 44 45 DIN
VN4 0 45 DC 2
*
* -INPUT CURRENT NOISE GENERATOR
*
VN5 46 0 DC 2
DN5 46 47 DIN
DN6 47 48 DIN
VN6 0 48 DC 2
*
* GAIN STAGE & DOMINANT POLE AT 55 Hz
*
EREF 98 0 31 0 1
G1 98 10 5 6 10
R7 10 98 10
E1 99 11 POLY(1) 99 31 -1.209 1
D3 10 11 DX
E2 12 50 POLY(1) 31 50 -1.209 1
D4 12 10 DX
G2 98 13 10 31 1E-3
R8 13 98 10
G3 99 14 98 13 34.558E-3
G4 99 16 98 98 34.558E-3
G5 14 15 15 14 20E-3
G6 16 17 17 14 20E-3
R9 15 18 400
R10 17 18 400
E3 18 98 16 98 1
R11 16 98 5.7875E7
C5 16 98 50E-12
V1 99 19 DC 4.3208
D5 16 19 DX
V2 20 50 DC 4.3208
D6 20 16 DX
RDC 14 98 1E15
*
* COMMON-MODE GAIN NETWORK WITH ZERO AT 4.3 kHz
*
ECM 21 98 POLY(2) 1 31 2 31 (0,0.5,0.5)
RCM1 21 22 1
CCM 21 22 37.281E-6
RCM2 22 98 1E-6
*
* POLE-ZERO PAIR AT 3.9 MHz/10 MHz
*
GPZ 98 23 16 98 1
RPZ1 23 98 1
RPZ2 23 24 0.63934
CPZ 24 98 24.893E-9
*
* NEGATIVE ZERO AT -300 MHz
*
ENZ 25 98 23 31 1E6
RNZ1 25 26 1
CNZ 25 26 -5.3052E-10
RNZ2 26 98 1E-6
*
* POLE AT 300 MHz
*
GP2 98 27 26 31 1
RP2 27 98 1
CP2 27 98 5.3052E-10
*
* POLE AT 500 MHz
*
GP3 98 28 27 31 1
RP3 28 98 1
CP3 28 98 3.1831E-10
*
* POLE AT 500 MHz
*
GP4 98 29 28 31 1
RP4 29 98 1
CP4 29 98 3.1831E-10
*
* OUTPUT STAGE
*
VW 29 30 DC 0
RDC1 99 31 23.25E3
CDC 31 0 1E-6
RDC2 31 50 23.25E3
GO1 98 32 37 30 25E-3
DO1 32 33 DX
VO1 33 98 DC 0
DO2 34 32 DX
VO2 98 34 DC 0
FDC 99 50 POLY(2) VO1 VO2 9.86E-3 1 1
VSC1 35 37 0.945
DSC1 30 35 DX
VSC2 37 36 0.745
DSC2 36 30 DX
FSC1 37 0 VSC1 1
FSC2 0 37 VSC2 1
GO3 37 99 99 30 25E-3
GO4 50 37 30 50 25E-3
RO1 99 37 40
RO2 37 50 40
LO 37 38 10E-9
*
* MODELS USED
*
.MODEL QX NPN(BF=5.5556E4)
.MODEL DX D(IS=1E-15)
.MODEL DEN D(IS=1E-12 RS=6.3708E3 AF=1 KF=1.59E-15)
.MODEL DIN D(IS=1E-12 RS=474 AF=1 KF=7.816E-15)
.ENDS AD797B
Two very generic transistor models in the whole circuit!
Re: Jfet
my experience with SPICE is very limited (an analog design class senior year), but what little knowledge i have leads me to agree completely. performance simulation in SPICE is highly dependent on accurate device simulation; changing transistor-specific parameters (e.g. device geometry) can have huge implications on distortion performance. one of the goals of my project was to optimize overall bandwidth and distortion performance of a simple JFET opamp design by playing with device geometries, emitter degeneration, etc., and tinkering with any device parameters slightly caused large variations in simulated performance. so if the device models in SPICE are not absolutely accurate (and from the looks of AD's own model they are not) then it is useful as a macromodel only at best as Harry says.
HarryHaller said:
my experience with SPICE is very limited (an analog design class senior year), but what little knowledge i have leads me to agree completely. performance simulation in SPICE is highly dependent on accurate device simulation; changing transistor-specific parameters (e.g. device geometry) can have huge implications on distortion performance. one of the goals of my project was to optimize overall bandwidth and distortion performance of a simple JFET opamp design by playing with device geometries, emitter degeneration, etc., and tinkering with any device parameters slightly caused large variations in simulated performance. so if the device models in SPICE are not absolutely accurate (and from the looks of AD's own model they are not) then it is useful as a macromodel only at best as Harry says.
Okay okay! I get it. Back to my thing about JFET input devices like the AD8610 and OPA627 or op275,opa134,opa604,opa2604.
They all have the same problem in noninverting mode :
VOLTAGE MODULATED INPUT CAPACITANCE.
This will add distortion but a way of lowering its effect is to balance the source resistance seen by I+ and I-. And lowering the source resistance seen by these two input. Which means source resistance in the area of <10KOhm.
Every time you lower the source resistance by 10 you lower the capacitance effect by 10.
Thats why i will keep my opinion against "driving an opamp with a highimpedance source".
Off course this effect is lowered a lot in inverting mode which is rarely used for no specified reason.
The opamp will perfom a lot better in this mode.
Sonny
They all have the same problem in noninverting mode :
VOLTAGE MODULATED INPUT CAPACITANCE.
This will add distortion but a way of lowering its effect is to balance the source resistance seen by I+ and I-. And lowering the source resistance seen by these two input. Which means source resistance in the area of <10KOhm.
Every time you lower the source resistance by 10 you lower the capacitance effect by 10.
Thats why i will keep my opinion against "driving an opamp with a highimpedance source".
Off course this effect is lowered a lot in inverting mode which is rarely used for no specified reason.
The opamp will perfom a lot better in this mode.
Sonny
Digital control audio potentiometer
Folks,
Progress report, and further advice respectfully sought.
After carefully examining the market my digital man (Ben) and I have come to a few conclusions we'd like to share.
1. Most digital pots are CMOS, 3-wire (SPI) controlled. The 3 wire is fine, the CMOS is not, because of asymmetric impedance characteristics and high rds on, typically 50R.
2. Motorized pots will soon become dinosaurs as the CMOS (and even bipolar like the DS1808) chips are evolving very quickly. The best of them are cermet; these are very difficult to source (particularly in Australia) in dual gang, and in any case cermets are not easy to get in log tapers. We have decided not to use a motorized pot.
3. Resistor ladder pots are the usual approach, but sub-optimal with SS switches because of the nasty signal effects. Some, like the CS3310 and the Wolfson chip, use a combination of switched resistors in the feedback loop and the input of an opamp, which gives a range of typically 96dB, incorporating around 31dB of gain at the top end. This sounds like a purist approach, but I had hoped to avoid the opamp as their high output drive is unnecessary in this application since we are driving a tube grid.
4. Nelson Pass made the significant point some time back that a CMOS switch at ground potential cannot contribute distortion to a signal taken from a distant wiper above. The math is simple, and it's a very good idea, implemented, he said, in his X-preamp. OK, let's consider this. Now, the conjecture........
Nelson's idea lead me to the shunt attenuator, reputedly the best of them. In this approach the signal is presented along the 'hot' lead via a fixed, series resistor to the target amplifier, and a shunt resistor is taken from this point and switched to ground. By switching in different shunt resistors, we form a variable impedance voltage divider with a fixed upper resistor. All switching is then conveniently performed directly at ground potential, with unused shunt resistors left dangling at an 'off' CMOS switch. (I trialled this approach using a simple bipolar, and in the off position, with no attenuation, the sound was very audibly distorted!!)
You could therefore use an existing ladder digital pot, but there's a problem with uneven steps at low amplitudes since the step determining resistors are inside the chip and cannot be altered.
Ben hit on the idea of using a mux (multiplexer). There are plenty available, many with 3 wire control, and they are inexpensive and a well proven technology, particularly in communications. But only a few have rds on figures less than 50R, which you'd need. And 16 or 32 switch versions are hard to find. There is one from Dallas (the Maxim MAX4571) which has eleven SPST audio/video clickless switches, but the damn thing is serially controlled, which is tricky to implement.
Does anyone have any ideas they'd like to contribute? I'm trying to set up at least 32 steps, with 1.5dB increments for the first 24 steps, and 2.5dB steps for the last 7, giving a range of 53.5dB. A mute would be added externally with a simple reed relay. The control needs to be universal, of course.
Thanks in advance,
Cheers,
Hugh
Hugh R. Dean
Research/Technical Director
www.printedelectronics.com
Melbourne AUSTRALIA
Folks,
Progress report, and further advice respectfully sought.
After carefully examining the market my digital man (Ben) and I have come to a few conclusions we'd like to share.
1. Most digital pots are CMOS, 3-wire (SPI) controlled. The 3 wire is fine, the CMOS is not, because of asymmetric impedance characteristics and high rds on, typically 50R.
2. Motorized pots will soon become dinosaurs as the CMOS (and even bipolar like the DS1808) chips are evolving very quickly. The best of them are cermet; these are very difficult to source (particularly in Australia) in dual gang, and in any case cermets are not easy to get in log tapers. We have decided not to use a motorized pot.
3. Resistor ladder pots are the usual approach, but sub-optimal with SS switches because of the nasty signal effects. Some, like the CS3310 and the Wolfson chip, use a combination of switched resistors in the feedback loop and the input of an opamp, which gives a range of typically 96dB, incorporating around 31dB of gain at the top end. This sounds like a purist approach, but I had hoped to avoid the opamp as their high output drive is unnecessary in this application since we are driving a tube grid.
4. Nelson Pass made the significant point some time back that a CMOS switch at ground potential cannot contribute distortion to a signal taken from a distant wiper above. The math is simple, and it's a very good idea, implemented, he said, in his X-preamp. OK, let's consider this. Now, the conjecture........
Nelson's idea lead me to the shunt attenuator, reputedly the best of them. In this approach the signal is presented along the 'hot' lead via a fixed, series resistor to the target amplifier, and a shunt resistor is taken from this point and switched to ground. By switching in different shunt resistors, we form a variable impedance voltage divider with a fixed upper resistor. All switching is then conveniently performed directly at ground potential, with unused shunt resistors left dangling at an 'off' CMOS switch. (I trialled this approach using a simple bipolar, and in the off position, with no attenuation, the sound was very audibly distorted!!)
You could therefore use an existing ladder digital pot, but there's a problem with uneven steps at low amplitudes since the step determining resistors are inside the chip and cannot be altered.
Ben hit on the idea of using a mux (multiplexer). There are plenty available, many with 3 wire control, and they are inexpensive and a well proven technology, particularly in communications. But only a few have rds on figures less than 50R, which you'd need. And 16 or 32 switch versions are hard to find. There is one from Dallas (the Maxim MAX4571) which has eleven SPST audio/video clickless switches, but the damn thing is serially controlled, which is tricky to implement.
Does anyone have any ideas they'd like to contribute? I'm trying to set up at least 32 steps, with 1.5dB increments for the first 24 steps, and 2.5dB steps for the last 7, giving a range of 53.5dB. A mute would be added externally with a simple reed relay. The control needs to be universal, of course.
Thanks in advance,
Cheers,
Hugh
Hugh R. Dean
Research/Technical Director
www.printedelectronics.com
Melbourne AUSTRALIA
Have you been looking on the off isolation and channel - channel crosstalk....
Off course you have ... Choosing MAX4571 says it all...
If they are used as a shunt to ground (R2R : no loose resistors) you should be able to get some really good results from it..
IIC data bus (SDA+SCL) : There are plenty of sample codes floating around on the net.... Or go for a micro with build in IIC bus.
I do not think you have to worry about noise from a micro.. A device like the AT89S8252 can be turned into sleep (Oscillator OFF) and be turned on by an level sensitive interrupt in 16mSec!?. It works just fine... I could also supply a bit off code on that.
Sonny
Off course you have ... Choosing MAX4571 says it all...
If they are used as a shunt to ground (R2R : no loose resistors) you should be able to get some really good results from it..
IIC data bus (SDA+SCL) : There are plenty of sample codes floating around on the net.... Or go for a micro with build in IIC bus.
I do not think you have to worry about noise from a micro.. A device like the AT89S8252 can be turned into sleep (Oscillator OFF) and be turned on by an level sensitive interrupt in 16mSec!?. It works just fine... I could also supply a bit off code on that.
Sonny
Digital Pots and Noise
When we are discussing noise, I just want to point out again Linear Technology's Application Note #70. Although the note deals with implementation of the '1533 low noise switcher, most of the appendix, and it runs for about 50 pages, deals with techniques for measurement of noise (ground loops in your lab will lead you to the lunatic asylum!) Measurement techniques are a the top of the list (I guess that they like old Analog scopes compared to the new digital models). They also have good commentary on diodes, diode snubbers, and transformers.
In a couple old Analog Devices application notes they used the R2R ladder in a DAC as a programmable resistor.
When we are discussing noise, I just want to point out again Linear Technology's Application Note #70. Although the note deals with implementation of the '1533 low noise switcher, most of the appendix, and it runs for about 50 pages, deals with techniques for measurement of noise (ground loops in your lab will lead you to the lunatic asylum!) Measurement techniques are a the top of the list (I guess that they like old Analog scopes compared to the new digital models). They also have good commentary on diodes, diode snubbers, and transformers.
In a couple old Analog Devices application notes they used the R2R ladder in a DAC as a programmable resistor.
hugh,
you might want to check out the Son of Dork Attenuator column. i'm dealing with the same problems and i've sort of settled on a dual-stage variable shunt design probably using CMOS switches like the AD SSM2402 but i'm not sure yet.
dorkus
you might want to check out the Son of Dork Attenuator column. i'm dealing with the same problems and i've sort of settled on a dual-stage variable shunt design probably using CMOS switches like the AD SSM2402 but i'm not sure yet.
dorkus
Hello Hugh, in my experience 1.5dB and certainly 2.5dB steps are too big.
I have lived with 1dB steps and found this to be quite good, and 0.5 dB steps would have been better maybe.
Forget about old theory books saying that any less than 3dB is not audible - 1dB steps are perfectly audible.
Sorry I have no experience with selecting electronic volume controllers except for those in consumer gear, and some of these do not sound very good and some cause zipper noise.
Regards, Eric.
I have lived with 1dB steps and found this to be quite good, and 0.5 dB steps would have been better maybe.
Forget about old theory books saying that any less than 3dB is not audible - 1dB steps are perfectly audible.
Sorry I have no experience with selecting electronic volume controllers except for those in consumer gear, and some of these do not sound very good and some cause zipper noise.
Regards, Eric.
Hi Sonnya, Dorkus and Eric,
Thank you for your valued comments.
I visited the 'Attenuator' thread and sheepishly realized you were covering the very same ground, and I apologize for the duplication. My project may be useful for others here; I certainly hope so. However, as long as everyone's aware of this and other related threads, I guess it shouldn't be an issue.....
Now, at the risk of exposing my supreme ignorance on matters digital, I should explain the design brief. The attenuator will be used in the upcoming full featured AKSA hybrid preamp, which will feature full remote with five- source selection, volume, and mute. The analog design is now complete and will soon be into beta testing. This will be sold as a kitset, quite expensive, but trials reveal stratospheric performance, easily beating the big names for spatial realism, tonality and clarity.
1. The attenuator will be shunt, using a quality series pass resistor of 22K, likely a bulk foil or something similar.
2. Muxes with CMOS switches will be used to ground a range of quality resistors. Rds on MUST be less than 40R for a reasonable range. (A single 40R CMOS to ground would offer almost 55dB of attenuation with a 22K series pass, but whether it would sound good [even at this low level] is another matter. I expect we will be obliged to insert an additional resistor atop the Mux to linearise the shunt element. Listening tests are called for.)
3. Microprocessor/PIC control, while feature-rich, is deemed overkill and I wish to keep it simple with readily available logic chips. Since I have to service this beast, I don't like the added complexity of software, and logic chips are very cheap.
4. There will not be sufficient steps with say 16 switches and 55dB of range, so a split series pass will be required. A second series resistor, switchable with a reed switch, will double the steps. We plan to increase the series 22K to around 90K, with a 68K additional resistor. Using five or six multiple combinations of switches, it should be possible to easily achieve 40 steps.
Eric, I am not sure I agree with you about 0.5dB steps. I do feel 1dB steps are enough; for a dynamic range of 45dB, sufficient, I feel with a mute, it seems to me 1dB would be fine for the most used portion of the adjustment, and 2dB or even more for the very soft and very loud portions, which are little used.
So, I invite suggestions for good chips, and even for the logic circuitry, which we are investigating right now. Originally I had intended to use high quality relays in a shunt attenuator, but after careful examination of all the multiplexers in the market, it seems this is both practical and inexpensive. With care, this could be a better option to a DACT; that is the intention, with the added benefit of remote operation and lower cost.
My sincere thanks to those who have commented to date.
Cheers,
Hugh
Hugh R. Dean
Research/Technical Director
www.printedelectronics.com
Thank you for your valued comments.
I visited the 'Attenuator' thread and sheepishly realized you were covering the very same ground, and I apologize for the duplication. My project may be useful for others here; I certainly hope so. However, as long as everyone's aware of this and other related threads, I guess it shouldn't be an issue.....
Now, at the risk of exposing my supreme ignorance on matters digital, I should explain the design brief. The attenuator will be used in the upcoming full featured AKSA hybrid preamp, which will feature full remote with five- source selection, volume, and mute. The analog design is now complete and will soon be into beta testing. This will be sold as a kitset, quite expensive, but trials reveal stratospheric performance, easily beating the big names for spatial realism, tonality and clarity.
1. The attenuator will be shunt, using a quality series pass resistor of 22K, likely a bulk foil or something similar.
2. Muxes with CMOS switches will be used to ground a range of quality resistors. Rds on MUST be less than 40R for a reasonable range. (A single 40R CMOS to ground would offer almost 55dB of attenuation with a 22K series pass, but whether it would sound good [even at this low level] is another matter. I expect we will be obliged to insert an additional resistor atop the Mux to linearise the shunt element. Listening tests are called for.)
3. Microprocessor/PIC control, while feature-rich, is deemed overkill and I wish to keep it simple with readily available logic chips. Since I have to service this beast, I don't like the added complexity of software, and logic chips are very cheap.
4. There will not be sufficient steps with say 16 switches and 55dB of range, so a split series pass will be required. A second series resistor, switchable with a reed switch, will double the steps. We plan to increase the series 22K to around 90K, with a 68K additional resistor. Using five or six multiple combinations of switches, it should be possible to easily achieve 40 steps.
Eric, I am not sure I agree with you about 0.5dB steps. I do feel 1dB steps are enough; for a dynamic range of 45dB, sufficient, I feel with a mute, it seems to me 1dB would be fine for the most used portion of the adjustment, and 2dB or even more for the very soft and very loud portions, which are little used.
So, I invite suggestions for good chips, and even for the logic circuitry, which we are investigating right now. Originally I had intended to use high quality relays in a shunt attenuator, but after careful examination of all the multiplexers in the market, it seems this is both practical and inexpensive. With care, this could be a better option to a DACT; that is the intention, with the added benefit of remote operation and lower cost.
My sincere thanks to those who have commented to date.
Cheers,
Hugh
Hugh R. Dean
Research/Technical Director
www.printedelectronics.com
hi hugh,
so, you are doing a two-stage design too huh? 😉
will there be some sort of active stage between the two stages? i suppose you can calculate the required values for both attenuators but remember that they will be interdependent. also, 90k series is a lot and it will severely hamper noise performance. you should keep series resistance at 22k max or so i think.
Nelson Pass has mentioned that CMOS switches work very well when grounded. browse through this thread:
http://www.diyaudio.com/forums/showthread.php?threadid=3267&perpage=15&pagenumber=3
some good info there, including some links to people who have implemented these things.
so, you are doing a two-stage design too huh? 😉
will there be some sort of active stage between the two stages? i suppose you can calculate the required values for both attenuators but remember that they will be interdependent. also, 90k series is a lot and it will severely hamper noise performance. you should keep series resistance at 22k max or so i think.
Nelson Pass has mentioned that CMOS switches work very well when grounded. browse through this thread:
http://www.diyaudio.com/forums/showthread.php?threadid=3267&perpage=15&pagenumber=3
some good info there, including some links to people who have implemented these things.
😉AKSA said:
I would like to challenge you on that. I use like 200micro/year.
It can make you circuit a lot simpler and a lot easier to debug, when this is constructed without being a overkill. And very stable too.
Just because it is a micro it does not have to contain a lot of part. Mostly i use no more than a 5 volt supply,xtal,a reset controller (MAX707), a few caps, a few resistors, and then the mux or dac or relay driver. I do not think that you can do it a lot simpler.
Throw me a blockdiagram or a feature list and i will show you a micro circuit.
😉 Seriously
Sonny
yeah, micro is the way to go
i have tried doing even the most basic things with standard logic and it is more difficult. i made a remote control integrated amp with nothing but power, mute, processor direct input, 4-input selector, and motorized volume control, and even with a ready-made IR receiver IC i still wound up with a whole bunch of logic chips (AND, OR, inverters, counter, latches, etc. etc.) and a mess of wiring. a simple micro and a couple ancillary chips could have accomplished everything and a whole lot more with far less complexity.
i have tried doing even the most basic things with standard logic and it is more difficult. i made a remote control integrated amp with nothing but power, mute, processor direct input, 4-input selector, and motorized volume control, and even with a ready-made IR receiver IC i still wound up with a whole bunch of logic chips (AND, OR, inverters, counter, latches, etc. etc.) and a mess of wiring. a simple micro and a couple ancillary chips could have accomplished everything and a whole lot more with far less complexity.
Yep, dorkus got it right. Micros are definitely the way to go. They may be a bit intimidating at first, but once you've seen how flexible they are, and much they can simplify this sort of project, you'll never want to go back. They're not hard to use either, once you've gone past the initial learning curve (which isn't much), and certainly are far easier to debug and work with than a pile of discrete logic chips. As previously mentioned, many micros also have a "sleep" mode in which they shut off their clock oscillator and drop to a low power halted state, where they will generate zero electrical noise. Wake-up is instantaneous as far as a human is concerned, and almost all micros have built-in EPROM memory, which is useful for, say, remembering what the volume setting was the last time you powered down the preamp...
I highly recommend either the Microchip PIC or Atmel AVR series. Both are widely used and have excellent free dev. tools. As well, quite a few of these have in-circuit programming capability which doesn't require an expensive EPROM programmer - just a simple serial cable adapter you can assemble for a dollar or two.
Have fun 🙂
I highly recommend either the Microchip PIC or Atmel AVR series. Both are widely used and have excellent free dev. tools. As well, quite a few of these have in-circuit programming capability which doesn't require an expensive EPROM programmer - just a simple serial cable adapter you can assemble for a dollar or two.
Have fun 🙂
btw... i once designed a shunt attenuator like this, though i never built it. I was using reed relays instead of CMOS switches, so to save on cost i used different combinations of parallel shunt resistors. While a MUX will only let you connect one shunt resistor to ground at any given time, you may find it useful to employ a different CMOS switch which allows you to connect multiple shunt resistors in parallel. This has the added advantage that the CMOS switch resistances will also be in parallel, and thus reduced. A SIPO shift register driven by the micro is a good way to control the relay drivers / CMOS switches, as it's output will remain active and static even when the micro enters it's sleep state.
I've also used SIPO shift registers to output static parallel data for multi-segment LED displays (might be nice to display the attenuation setting). In audio equipment, this is preferable to the often-used technique of a high frequency pulse driving each LED segment in sequence, which saves pins, but is electrically noisy (and causes that annoying vibrating appearance to the display).
I've also used SIPO shift registers to output static parallel data for multi-segment LED displays (might be nice to display the attenuation setting). In audio equipment, this is preferable to the often-used technique of a high frequency pulse driving each LED segment in sequence, which saves pins, but is electrically noisy (and causes that annoying vibrating appearance to the display).
- Status
- Not open for further replies.