I'm about to build a digitally controlled (Hi-Fi!?)
preamp, and have most of the design figured out
(thanks to a lot of good internet sites).
The critical part of the design that I'm still
confused about is how to do the input switching
in a "HiFi" way, i.e. with low distortion and
noise levels.
I have three alternatives as I see it:
1. Use gold-plated relays. (As a perfectionist I'm
a bit unhappy about the clicking sound from the
relays).
2. Use extremely expensive audio switch circuits
(i.e. SSM2402:s).
3. Use analog muxes. I have AD406 chips.
The preferred alternative would be #3, but I can't
find any information about distortion for these
devices in the datasheet. The crosstalk and
isolation figures are in the DS, but not a single
figure on distortion.
Has anyone tried analog muxes in a preamp?
Opinions? What's an acceptable "HiFi" level of
distortion?
Thanx in advance,
preamp, and have most of the design figured out
(thanks to a lot of good internet sites).
The critical part of the design that I'm still
confused about is how to do the input switching
in a "HiFi" way, i.e. with low distortion and
noise levels.
I have three alternatives as I see it:
1. Use gold-plated relays. (As a perfectionist I'm
a bit unhappy about the clicking sound from the
relays).
2. Use extremely expensive audio switch circuits
(i.e. SSM2402:s).
3. Use analog muxes. I have AD406 chips.
The preferred alternative would be #3, but I can't
find any information about distortion for these
devices in the datasheet. The crosstalk and
isolation figures are in the DS, but not a single
figure on distortion.
Has anyone tried analog muxes in a preamp?
Opinions? What's an acceptable "HiFi" level of
distortion?
Thanx in advance,
Relays have a nice sound! i mean the click 
If you have your mux in a feedback chain the distortion will be very little. Think of an inverting amp and the mux signal goes through the input resistors. If you have 10 k as input to the inverting inout via the mux and you have 10k-100 k as feedback, then you will have extremly low distortion. One further advantage is that the mux have very low voltage across the switches, virtual ground you know. 4066 isn't so bad but this chip hasn't got any declicking and smooth on and off like the more expensive ones. If you have a microcontroller yoou can easily fix a small delay at on and of. The smoothness of the switching can be hard to fix though.
Consider a high performance analog switch, avoid the 4066 and similiar ones.
If you have your mux in a feedback chain the distortion will be very little. Think of an inverting amp and the mux signal goes through the input resistors. If you have 10 k as input to the inverting inout via the mux and you have 10k-100 k as feedback, then you will have extremly low distortion. One further advantage is that the mux have very low voltage across the switches, virtual ground you know. 4066 isn't so bad but this chip hasn't got any declicking and smooth on and off like the more expensive ones. If you have a microcontroller yoou can easily fix a small delay at on and of. The smoothness of the switching can be hard to fix though.
Consider a high performance analog switch, avoid the 4066 and similiar ones.
Hi peranders, and thanks for the input.
Regarding the mux:es that i already have, they're
not 4066:es but 16-way ADG406 from Analog Devices
(I missed the 'G' in ADG in my original post). The
quoted R-ON is quoted as 50 ohms typical, max 80
ohms. Typical +/- 4 ohms difference between
the different switches.
Since I'll have an input stage with an input
impedance of approx 47 kOhms, I don't think there's an issue with gain loss if using the device as a straightforward input switch (think
of using it as a replacement for a traditional mechanichal switch). What I am unsure of, however,
is if this kind of "simple" connection will give
me lots of distortion.
There is another chip from Analog, the AD8113,
that is a real crossbar matrix. This device is
the only analog mux circuit (apart from SSM2402
switches), that I have found any distortion figures for. The AD8113 is, however, rather expensive and also has a high-density pinout that
I will be having major problems with when/if I
try to make a suitable PCB.
The ADG406 runs on +/- 15V and has a R-ON of 50
ohms, but will it give me acceptable sound if I
use it in an "straight-through" connection, or do
I need some kind of feedback circuit?
Regarding the mux:es that i already have, they're
not 4066:es but 16-way ADG406 from Analog Devices
(I missed the 'G' in ADG in my original post). The
quoted R-ON is quoted as 50 ohms typical, max 80
ohms. Typical +/- 4 ohms difference between
the different switches.
Since I'll have an input stage with an input
impedance of approx 47 kOhms, I don't think there's an issue with gain loss if using the device as a straightforward input switch (think
of using it as a replacement for a traditional mechanichal switch). What I am unsure of, however,
is if this kind of "simple" connection will give
me lots of distortion.
There is another chip from Analog, the AD8113,
that is a real crossbar matrix. This device is
the only analog mux circuit (apart from SSM2402
switches), that I have found any distortion figures for. The AD8113 is, however, rather expensive and also has a high-density pinout that
I will be having major problems with when/if I
try to make a suitable PCB.
The ADG406 runs on +/- 15V and has a R-ON of 50
ohms, but will it give me acceptable sound if I
use it in an "straight-through" connection, or do
I need some kind of feedback circuit?
I'll probably get drawn and quatered for this , but my "interim" preamp uses a scheme like this one from the Hafler 915 preamp.
It's not bad, sound-wise, and you can throw it together VERY quick. I use Maxim analog switches- higher power rails, MUCH better specs all the way around . I use 2SK389 as the JFETs (I like seeing the zero for DC offset).
Here is an old netnews posting of some related discussion. Enjoy and good luck.
mlloyd1
++++++++++++++++++++++++++++++++++++++++++
Newsgroups: sci.electronics.design
Subject: Re: 4053 THD results
Date: Thu, 03 May 2001 17:28:10 GMT
On Thu, 03 May 2001 13:10:22 +0100, Lamebert <lambert@newsguy.com>
scribbled:
>wavefront@bigfoot.com (Martin Griffith) wrote:
>
>>
>>Just done a quick test of a 4053 analogue mux when used as an audio
>>switch in a virtual earth config
>>
>>opamp Philips 5532
>>mux Nat semi CD4053bcn, about 10 years old
>>i/p R 10K
>>f/bR 10K
>>opamp output load impedance >10K
>>
>>input level +18dBu @1KHz
>>THD on o/p was better than .005%
>>
>>input level 0dBu @1KHz
>>THD on o/p was better than .002%, probably lash up limited!
>
>Interesting, I wonder what the original poster was hoping for.
>
>What supply was the 4053 on?
>
>I suspect an HC4053 would be better, from memory they have substantially
>lower on resistance than 4000 series.
>
Oops!, sorry for the omission
5532 was on +-15, with 100nF to gnd from each rail
The 4053 was on+-5v derived 3 1K resistors in series between the 5532
supplies.
The opamp needed 47pF across the feedback R to calm it down, haven't
done noise or F response, but should be adequate
I'll play around a bit, and put a couple of the gates in parallel, and
see how it improves.
Martin
, but my "interim" preamp uses a scheme like this one from the Hafler 915 preamp.It's not bad, sound-wise, and you can throw it together VERY quick. I use Maxim analog switches- higher power rails, MUCH better specs all the way around . I use 2SK389 as the JFETs (I like seeing the zero for DC offset).
Here is an old netnews posting of some related discussion. Enjoy and good luck.
mlloyd1
++++++++++++++++++++++++++++++++++++++++++
Newsgroups: sci.electronics.design
Subject: Re: 4053 THD results
Date: Thu, 03 May 2001 17:28:10 GMT
On Thu, 03 May 2001 13:10:22 +0100, Lamebert <lambert@newsguy.com>
scribbled:
>wavefront@bigfoot.com (Martin Griffith) wrote:
>
>>
>>Just done a quick test of a 4053 analogue mux when used as an audio
>>switch in a virtual earth config
>>
>>opamp Philips 5532
>>mux Nat semi CD4053bcn, about 10 years old
>>i/p R 10K
>>f/bR 10K
>>opamp output load impedance >10K
>>
>>input level +18dBu @1KHz
>>THD on o/p was better than .005%
>>
>>input level 0dBu @1KHz
>>THD on o/p was better than .002%, probably lash up limited!
>
>Interesting, I wonder what the original poster was hoping for.
>
>What supply was the 4053 on?
>
>I suspect an HC4053 would be better, from memory they have substantially
>lower on resistance than 4000 series.
>
Oops!, sorry for the omission
5532 was on +-15, with 100nF to gnd from each rail
The 4053 was on+-5v derived 3 1K resistors in series between the 5532
supplies.
The opamp needed 47pF across the feedback R to calm it down, haven't
done noise or F response, but should be adequate
I'll play around a bit, and put a couple of the gates in parallel, and
see how it improves.
Martin
Attachments
This thread might be interseting for you:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=3267&highlight=cmos
NP's opinion looks quite promising !
BTW: it wouldn't be that difficult to implement declicking measures for 4066 and the like - even without a microcontroller.
Regards
Charles
http://www.diyaudio.com/forums/showthread.php?s=&threadid=3267&highlight=cmos
NP's opinion looks quite promising !
BTW: it wouldn't be that difficult to implement declicking measures for 4066 and the like - even without a microcontroller.
Regards
Charles
You can clicks anyway due to DC offsets. It's also hard to design a fast zero crossing detector capable of a few micro volts or at least under 1 mV.phase_accurate said:
It's also possible to use discrete JFET's as analog switches. Then it's very easy to get them slow. National have application notes about this somewhere....
Hi,
If you must use analogue mux's, I'd recommend DG211 / DG212. They are a world away from the first generation analogue switches, without the cost penalty of the SSM's.
I have used them for source switching by putting them at the "virtual earth" (inverting) input of an opamp. There is an input resisitor for each source.
In my experience, this scheme gives minimum crosstalk and minimum distortion.
If this isn't clear, I'll post a schematic.
With regard to silent switching, there are 2 issues here:
If you switch a waveform fast, the discontinuety will sound. Zero swiching will help. Specialised (expensive) switches are the answer.
Most analogue switches crosstalk from the control pin to the signal. Speed limiting the control signal, and good earthing practice will help to reduce this.
Cheers,
If you must use analogue mux's, I'd recommend DG211 / DG212. They are a world away from the first generation analogue switches, without the cost penalty of the SSM's.
I have used them for source switching by putting them at the "virtual earth" (inverting) input of an opamp. There is an input resisitor for each source.
In my experience, this scheme gives minimum crosstalk and minimum distortion.
If this isn't clear, I'll post a schematic.
With regard to silent switching, there are 2 issues here:
If you switch a waveform fast, the discontinuety will sound. Zero swiching will help. Specialised (expensive) switches are the answer.
Most analogue switches crosstalk from the control pin to the signal. Speed limiting the control signal, and good earthing practice will help to reduce this.
Cheers,
thd switches
The switches distort because their on-resistance varies with signal level. Assuming a sine wave across the on-switch, the on-resistance increases when the sine goes to max and vice versa.
This can be limited by 1 - operating the switch at max supply voltage (should be clear from ds), and 2 - limit the signal level across the switch.
Example: use a 50 ohms switch at virtual ground with a 10k in series. At 1vrms, the switch sees nominal 50/10000 * 1V = 5mV.
The amount of thd can be estimated by looking in the data sheet for Ron vs signal level.
You will for instance see that at signal level variation of 1V, the on res varies from 50 to 55 ohms (just making these numbers up, check the ds).
On the total res this would be a variation of 10050 to 10055 ohms or some delta of 500ppm.
In the above example, 5mV across the switch, it would probably be sub-ppm.
The crosstalk etc influences the linear distortion (freq response), and can be limited by switch construction and keeping all impedances as low as possible (which is in conflict with min thd which asks for larger impedances).
Jan Didden
The switches distort because their on-resistance varies with signal level. Assuming a sine wave across the on-switch, the on-resistance increases when the sine goes to max and vice versa.
This can be limited by 1 - operating the switch at max supply voltage (should be clear from ds), and 2 - limit the signal level across the switch.
Example: use a 50 ohms switch at virtual ground with a 10k in series. At 1vrms, the switch sees nominal 50/10000 * 1V = 5mV.
The amount of thd can be estimated by looking in the data sheet for Ron vs signal level.
You will for instance see that at signal level variation of 1V, the on res varies from 50 to 55 ohms (just making these numbers up, check the ds).
On the total res this would be a variation of 10050 to 10055 ohms or some delta of 500ppm.
In the above example, 5mV across the switch, it would probably be sub-ppm.
The crosstalk etc influences the linear distortion (freq response), and can be limited by switch construction and keeping all impedances as low as possible (which is in conflict with min thd which asks for larger impedances).
Jan Didden
Analog Switches and Muxes
If you are looking to use quality parts for will find that many of the Siliconix switches and muxes introduce no distortion and have no leakage voltage or linearity problems. While working at TI, I used many different switches and muxes made by Siliconix and Harris and found their performance to be extremely good.
You can find many of these parts isolation and cross talk greater than 100db. In fact many of their switches are great for sample and hold applications and multi channels high performance and video systems. Of course a 4066 and 4053 are cheap parts and should not be used if your expecting quality audio performance.
If you are looking to use quality parts for will find that many of the Siliconix switches and muxes introduce no distortion and have no leakage voltage or linearity problems. While working at TI, I used many different switches and muxes made by Siliconix and Harris and found their performance to be extremely good.
You can find many of these parts isolation and cross talk greater than 100db. In fact many of their switches are great for sample and hold applications and multi channels high performance and video systems. Of course a 4066 and 4053 are cheap parts and should not be used if your expecting quality audio performance.
peranders said:Is it possible to make slow on and offs with a 4066? I don't think so. Isn't the inputs buffered (= high gain)?
It's peculiar that AD not have mentioned any distortion figures at all....
Often the distortion figures are not specified in the traditional way. Rather, what you will see is a graph of Ron versus input level. That shows that the larger the input level, the larger the Ron. That again means that the Ron varies through the signal cycle. Note that the variation is lowest with highest supply voltage.
So, if you use such a switch as input switch, where the signal level can be up to several volts peak, there will be appreciable Ron modulation which can give distortion, depending on the input resistance (it is an attenuator with attenuation varying through the signal cycle which gives 3rd harmonic distortion). You can do some rough calculation to check the variation to get a feel for the distortion.
On the other hand, if you use them say as level switches at the inverting (virtual earth) input of a feedback (op)amp, the signal level is extremely small (fractional mV) so the distortion is zero for practical purposes.
Jan Didden
For Peranders: Yes, you may quote me
In addition to the resistance vs. voltage issue, there is probably an (typically unspecified) capacitance vs. voltage issue because there must be some parasitic capacitance to the substrate.
1. Connecting the switch directly to the non-inverting input of an op-amp means it sees practically no load other than the pull to ground resistor which can be > 1 Meg with a FET-input op amp, so there is no resistance issue. On the other hand, it sees a large voltage swing, so I would use this only if the switch is driven from a low (< 500 R?) source impedance.
2. Putting the switch into the inverting feedback network means minimized voltage swing but some current flows, so the resistance issue is back.
3. Using an inverting configuration and putting the switch between the virtual ground point and the inverting input is probably ideal (close to no current and no voltage), but it means you need a resistor pair for each input.
1. Connecting the switch directly to the non-inverting input of an op-amp means it sees practically no load other than the pull to ground resistor which can be > 1 Meg with a FET-input op amp, so there is no resistance issue. On the other hand, it sees a large voltage swing, so I would use this only if the switch is driven from a low (< 500 R?) source impedance.
2. Putting the switch into the inverting feedback network means minimized voltage swing but some current flows, so the resistance issue is back.
3. Using an inverting configuration and putting the switch between the virtual ground point and the inverting input is probably ideal (close to no current and no voltage), but it means you need a resistor pair for each input.
capslock:
Just doing a really quick eyeball ...
Looking at the circuits below and assuming (for the sake of discussion) :
1. the driving element has "adequately low" (i.e. less than 100ohms) Zout,
2. Op amps gain blocks have same distortion even though different configuration (yeah, I know Circuit 2 should actually have lower THD thanks to lower common mode signal, but I have to add another op amp and pair of resistors to get the correct signal polarity back )
3. For Circuit 2, I'm not going to add another switch in the feedback loop to compensate for the input leg switch's Ron
I'm thinking the switch in Circuit 1 has less current through it and less voltage across it (so better performance?) than the switch in Circuit 2. I'm also worried about hanging a parasitic capacitance at the inverting input of the op amp.
What did I miss?
mlloyd1
Just doing a really quick eyeball ...
Looking at the circuits below and assuming (for the sake of discussion) :
1. the driving element has "adequately low" (i.e. less than 100ohms) Zout,
2. Op amps gain blocks have same distortion even though different configuration (yeah, I know Circuit 2 should actually have lower THD thanks to lower common mode signal, but I have to add another op amp and pair of resistors to get the correct signal polarity back
)3. For Circuit 2, I'm not going to add another switch in the feedback loop to compensate for the input leg switch's Ron
I'm thinking the switch in Circuit 1 has less current through it and less voltage across it (so better performance?) than the switch in Circuit 2. I'm also worried about hanging a parasitic capacitance at the inverting input of the op amp.
What did I miss?
mlloyd1
capslock said:
Attachments
mux stuff
The resistance issue is a non-issue, as long as it is a constant resistance (what's 100 Ohms between friends, eh?). The resistance MODULATION is an issue, meaning the resistance varies through the signal cycle. That means distortion. The modulation comes from the voltage swing of the mux. So, fig 2, with minimal voltage swing, is the better one.
Jan Didden
The resistance issue is a non-issue, as long as it is a constant resistance (what's 100 Ohms between friends, eh?). The resistance MODULATION is an issue, meaning the resistance varies through the signal cycle. That means distortion. The modulation comes from the voltage swing of the mux. So, fig 2, with minimal voltage swing, is the better one.
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.