resolved: best way to handle switching in audio pre-amp

[Mac]

In an earlier post, I asked how to handle switching of line level signals
in an audio pre-amp.

I received a lot of good suggestions, but ultimately I think I will go
with the ADG452, a quad analog SPST switch from AD with Ron = 4 Ohms. The
logic input is TTL compatible when the logic power supply is +5V, and the
performance of the analog switch is specified under single and split
supplies of various different voltages.

Thanks to all who responded and pointed me in the right direction.

--Mac

 

[Winfield Hill]

Mac wrote...

In an earlier post, I asked how to handle switching of line level
signals in an audio pre-amp.

I received a lot of good suggestions, but ultimately I think I will go
with the ADG452, a quad analog SPST switch from AD with Ron = 4 Ohms.
The logic input is TTL compatible when the logic power supply is +5V,
and the performance of the analog switch is specified under single and
split supplies of various different voltages.

I didn't see the thread, but what's the appeal of using such a low Ron
switch? Have you examined its capacitance specs? The high capacitance
of a low-Ron switch can give trouble for the driving opamp, introduce
high-frequency noise into the audio signal path from the supply lines,
and no doubt endanger the genetic pool for your grandchildren as well.

I'd start with a high-Z opamp follower after the CMOS switch, reducing
any effect from a changing Ron with signal voltage, then evaluate my
Ron spec, and avoid overdoing it. More (or less) is not always better.

For example, a 74hc4053 often looks appropriate after such an evaluation.

 

[Mac]

Mac wrote...

I didn't see the thread, but what's the appeal of using such a low Ron
switch? Have you examined its capacitance specs? The high capacitance
of a low-Ron switch can give trouble for the driving opamp, introduce
high-frequency noise into the audio signal path from the supply lines,
and no doubt endanger the genetic pool for your grandchildren as well.

I'd start with a high-Z opamp follower after the CMOS switch, reducing
any effect from a changing Ron with signal voltage, then evaluate my
Ron spec, and avoid overdoing it. More (or less) is not always better.

For example, a 74hc4053 often looks appropriate after such an evaluation.

The ADG452 is a CMOS switch, of course. And I was planning on putting a
unity gain buffer after it.

***********74HC4052**********

I did look at the 74HC4052 (I only need 4 signal sources). The TI
datasheet lists the channel-to-channel crosstalk as a TBD. Also, the sine
wave distortion specification is 0.018% into 10 k || 50 pF. This is at
1-10 kHz. Maybe with the much higher input resistance I'll have from my
follower I'll get better performance. Not that 0.018% is BAD, mind you.

The off isolation is -67 dB measured at 1 MHz with a load of 50 Ohms || 10
pF.

The 74HC4052 can't be powered from +/- 15 Volt rails. If I use it, I'll
have to create +/- 5 Volt rails. Or convert the whole design to +/- 5 Volt
rails. The design is only on paper, so far, so this would not be a huge
problem.

*******ADG452*******
The ADG452 specs the crosstalk as -90 dB (50 Ohm, 5pF, 1MHz). The off
isolation is -65 dB (same conditions as crosstalk).

The ADG452 can run from +/- 15 V rails. I would still need +5V for the
logic power supply.

************SUMMARY**********
The 74HC4052 may actually do the job, as well as eliminate some
external logic, and at a tiny fraction of the price of the ADG452. But I
got scared off by unspecified crosstalk, and the drawback that I will have
to create a - 5V rail for it.

Performance-wise, it doesn't appear to me that the ADG452 is inferior to
the 74HC4052 in any way.

*******Further questions********
As long as I have your attention, would you individually buffer each
input, as well, or would you just connect the inputs directly to the
switch? Obviously, buffering the inputs would be a lot more expensive, but
then your switch would be between two op-amps, and it would really make
the resistance variation of the switch almost irrelevant.

Also, Ban recommended the SSM2135 for buffering of high-fidelity audio
when using +/- 5V rails. Do you have any additional recommendations? Maybe
I should just use +/- 5V, then use the 74HC4052 followed by an SSM2135,
followed by individual left and right volume controls and I'll be done.

Anyway, thanks for responding.

regards,
Mac

 

[justin]

I've been following this thread with interest having an upcoming
project with a simillar problem; how does one decide between a switch
and a mechanical relay? A relay will outperform a CMOS switch in every
aspect except for power.

Any thoughts?

j.

 

[Richard Henry]

I've been following this thread with interest having an upcoming
project with a simillar problem; how does one decide between a switch
and a mechanical relay? A relay will outperform a CMOS switch in every
aspect except for power.

Any thoughts?

Switching speed?

 

[Winfield Hill]

Mac wrote...

The ADG452 is a CMOS switch, of course. And I was planning on putting a
unity gain buffer after it.

***********74HC4052**********

I did look at the 74HC4052 (I only need 4 signal sources). The TI
datasheet lists the channel-to-channel crosstalk as a TBD. Also, the
sine wave distortion specification is 0.018% into 10 k || 50 pF. This
is at 1-10 kHz. Maybe with the much higher input resistance I'll have
from my follower I'll get better performance. Not that 0.018% is BAD,
mind you.

The off isolation is -67 dB measured at 1 MHz with a load of 50 Ohms
|| 10 pF.

This is an inadequate analysis, based on datasheet numbers for a poor
configuration. First, there's no need for a 10k load, creating that
0.02% distortion. That's because the load for an ON switch should
simply be a follower opamp input. The load for an OFF switch should
be another ON switch, providing high OFF isolation and low crosstalk,
without requiring a 50-ohm load. Finally, it's serious spec overkill
to look at 1MHz OFF isolation and crosstalk when considering an audio
application, IMHO.

A proper consideration of audio distortion, isolation and cross-talk
issues shows the HC405x family to be essentially perfect when used as
suggested. Again, IMHO.

You won't find datasheet specs showing this performance, because its
beyond the capability of the high-speed IC production test equipment.
But it's easy to calculate using datasheet capacitance and Ron curves.

BTW, the 'HC405x series can be used with +/- 7.5V or 8V supplies. If
higher supply voltage is required, then use the original '405x series.

 

[martin griffith]

I've been following this thread with interest having an upcoming
project with a simillar problem; how does one decide between a switch
and a mechanical relay? A relay will outperform a CMOS switch in every
aspect except for power.

Any thoughts?

j.

I'm doing a little remote control audio proj, maybe it will end up as
a product, but for fun at the moment , as I have far too many bits of
audio equipment, and I am torn between Omron BABT (G6A-BS) and
Pickering 102's, in conjunction with TI's PGA2310 audio attenuators.
I didn't even consider solid state switches.

IMHO relays generally outperform SSswitches for audio if used
properly. Of course they do have their disadvantages, mainly they turn
off slower than they turn on. I'm using micro's to control things so
I can delay the turn on of a relay so I dont get any momentary
back-door paths, anyway all inputs are opamp buffered with 5534's.

Relays win hands down on charge injection, generally there aint none,
but I am expecting problems with crap from the coil turning off,
probably PCB layout more than anything.

Linearity is brilliant, Rdson, whats that? Signal handling pretty
good,
(1,000 VAC, 50/60 Hz for 1 min between open contacts) so its not
limited to supply rails. EMC proof, I think so.

One of the GREAT things about relays is that they can be designed
failsafe, they come in DPCO format, so on the o/p side you can use the
unpowered state to ground the i/p to the power amp, so nasty open
circuit buzzes are avoided when the system is powered down or when you
change between amplifier systems.

Also on power up when the micro is doing a system check, everything is
muted until the appropriate destination is selected. I cant remember
what happens to a SSswitch under no power conditions

Life is still quite impresive (G6A, not spec'd for Pickering) for the
average channel zapper
Life expectancy Mechanical: 100,000,000 operations min. (at 36,000
operations/hr)
Electrical: 500,000 operations min. (at 1,800 operations/hr)

They are tougher than yer average IPOD user

Malfunction: DPDT: 500 m/s2 (approx. 50G)
Shock resistance Destruction: 1,000 m/s2 (approx. 100G)

I could go on for ever...




martin

Serious error.
All shortcuts have disappeared.
Screen. Mind. Both are blank.

 

[martin griffith]

Remote over TCP/IP?

Not yet, RS422 control of real audio, with an 8051 or maybe a msp430
in charge, the development time for a TCP/IP for me, is too great,
(and I'd discover yet more areas of incompetence)
The protocol will be based loosely on the Sony 9pin protocol.

Actually , 107 2C is the functional equivalent of G6A and a bit of
overkill for audio, also, being UK made, is more expensive on this side
of the pond.

There are probably similar quality relays your side of the pond, I
use Pickering because they are so helpful.

5534s are not unity compensated, maybe 5532 or opa2434?

Sorry, I use the terms 5534/2 as the same generic ic, since i can
never remember which one is which.

The Toff -Ton is 3 mS. Micro, really? Wouldn't it be simpler to use a
RC constant on Ton to conpensate for Toff delay?

Ton is not equal to Toff. There is a 2mS overlap. This is VERY
important in unbuffered systems. Think of backdoor circuits.
I have the micro sitting there, with a 4mS house keeping system
interrupt so its one lil procedure extra.

An inverse diode and a ceramic across the coil will take care of that.
Pickerings already come with a built in diode.

I've got a cheat for this, use the PGA2310 to fade to -90dB, change
the relay status, then PGA's back to unity. but it helps to get the
hardware correct/quiet in the first place

That is great, isn't it! I mostly exploit it as a feed through/bypass
while powered down.

martin

Serious error.
All shortcuts have disappeared.
Screen. Mind. Both are blank.

 

[martin griffith]

martin griffith wrote...

snip my stuff

CMOS switches are off, until an ON logic level is presented. One
dangerous time is when the opamp supply rails are coming up, and you
really want open switches to be open, etc. A proper CMOS logic setup
will come on as planned for supply voltages as low as 1V, so that any
potential problems in this regard can be solved. However the low Ron
and the closed path to ground possible with a relay at an amplifier's
output can't be easily matched with ordinary solid-state techniques
(I can think of a painful way to do it; the relay is an easy winner).

I lost a rail once on a oneoff (ratsnest) CMOS audio switching unit,
burnt a finger really badly on one IC, had to replace all the CMOS
switches, ouch. Thank heavens that the speaker amps were not DC
coupled

Yep, that's Martin Griffith singing the praises of mechanical relays.
He's one highly-qualified and experienced bastard in this territory.

LOL, but Nope, just been in the right place at the wrong time. Baz
Porter (RIP) was a magnitude better than me. I learnt so much from
him, a Most Excellent Geezer Allround. Many many late night techy
phone chats, cool brilliant dude.

But I forgot one extra feature that relays have. Total isolation of
logic and signal supplies.

I've seen so many discrete fet switches die, mainly on the low level
ccts in the EQ switching in several Studer A80 multitracks, for no
apparent reason. Rather unpleasant circuitry, but that was a long time
ago.

Or maybe the stereo/mono switch on a Studer B62, a fet switch was
intermittently humming nicely, in FM, close to our FM TX broadcast
frequency, took weeks to find that one.

At least Maxim dont make relays, so you will be able to get them, in a
few years time. Thats not the Maxim delivery time either. Give me
relays any day. ( Dont ask me to design a PDA or a mobile phone, you
can imagine what it would look like)

But when I see stuff like the TI PGA2310, I just drool in wondermunt.



martin

Serious error.
All shortcuts have disappeared.
Screen. Mind. Both are blank.

 

[Winfield Hill]

martin griffith wrote...

But when I see stuff like the TI PGA2310, I just drool in wondermunt.

Yes, –95.5dB to +31.5dB gain in 0.5dB steps, 120dB dynamic range,
0.0004% distortion at 1kHz, nice. Don't forget the original cs3310
version, although it has slightly higher 0.001% distortion + noise.

 

[martin griffith]

martin griffith wrote...

Yes, –95.5dB to +31.5dB gain in 0.5dB steps, 120dB dynamic range,
0.0004% distortion at 1kHz, nice. Don't forget the original cs3310
version, although it has slightly higher 0.001% distortion + noise.

5v rails against 15v rails for the TI part, ISTR


martin

Serious error.
All shortcuts have disappeared.
Screen. Mind. Both are blank.

 

[Winfield Hill]

martin griffith wrote...

5v rails against 15v rails for the TI part, ISTR

Wow, a _serious_ advantage for the PGA2310, up to +/-13.5V out.
Time to get some free samples!