Replacing 4053 With…

[Slime_Lord]

I recently purchased a stock TAC Bullet after working on a heavily modified Soundcraft Series 500 for years. The size of the Bullet is perfect for my small home studio and the additional features are really useful but sonically it’s lacking the transparency that I became used to with the Soundcraft.

I’ve been slowly recapping the channels and swapping the 2, dip-8 op amps for OPA1642s (I can get the surface mount op amps off the hybrid modules, getting them back on is a real pain…). Sonically, I’m hearing some improvement after these changes but the input channels still sound veiled to me.

Further review of the schematic shows that there is a HEF4053 analog switch (3 x SPDT) in the audio path, handling muting. I read through the below article on muting circuits and suspect the 4053 is adding noise to the signal path.

https://sound-au.com/articles/muting.html
Is it possible to switch the 4053 out for 3 small relays or is there a better solution/replacement for this chip?

I’ve cropped out a small section of the B1018 schematic that shows where the 4053 is, appreciate any guidance!

 

[Newmarket]

I recently purchased a stock TAC Bullet after working on a heavily modified Soundcraft Series 500 for years. The size of the Bullet is perfect for my small home studio and the additional features are really useful but sonically it’s lacking the transparency that I became used to with the Soundcraft.

I’ve been slowly recapping the channels and swapping the 2, dip-8 op amps for OPA1642s (I can get the surface mount op amps off the hybrid modules, getting them back on is a real pain…). Sonically, I’m hearing some improvement after these changes but the input channels still sound veiled to me.

Further review of the schematic shows that there is a HEF4053 analog switch (3 x SPDT) in the audio path, handling muting. I read through the below article on muting circuits and suspect the 4053 is adding noise to the signal path.

https://sound-au.com/articles/muting.html
Is it possible to switch the 4053 out for 3 small relays or is there a better solution/replacement for this chip?

I’ve cropped out a small section of the B1018 schematic that shows where the 4053 is, appreciate any guidance!

I'll leave it to others to comment on the degree of impact the 4053 switches in the virtual earth configuration shown.
But I'll note that the three integrated switches are simply paralleled. There's no need to do that with a mechanical relay so you could use just a single relay.
Actually, I will go back to the 4053 and note that , from the datasheet Ron decreases with Supply Voltage so I'd suggest looking at whether you could increase that first as it might be easier to do than fitting a relay in there ?

 

[JohnRoberts]

I recently purchased a stock TAC Bullet after working on a heavily modified Soundcraft Series 500 for years. The size of the Bullet is perfect for my small home studio and the additional features are really useful but sonically it’s lacking the transparency that I became used to with the Soundcraft.

I’ve been slowly recapping the channels and swapping the 2, dip-8 op amps for OPA1642s (I can get the surface mount op amps off the hybrid modules, getting them back on is a real pain…). Sonically, I’m hearing some improvement after these changes but the input channels still sound veiled to me.

Further review of the schematic shows that there is a HEF4053 analog switch (3 x SPDT) in the audio path, handling muting. I read through the below article on muting circuits and suspect the 4053 is adding noise to the signal path.

https://sound-au.com/articles/muting.html
Is it possible to switch the 4053 out for 3 small relays or is there a better solution/replacement for this chip?

I’ve cropped out a small section of the B1018 schematic that shows where the 4053 is, appreciate any guidance!

I am not a big fan of second guessing design engineers but since you asked CMOS TG (transfer gate) technology has been widely used in signal routing. The topology in the schematic you shared should be pretty low distortion. The TG is in series with the input resistor of a simple unity gain inverter, so it is passing a modest AC current. Distortion in TGs are generally a function of terminal voltage which should be small in that configuration. One criticism of the design is passing a small AC signal nominally biased at 0V with the TG - power supply also grounded. I expect the signal to be much smaller than a diode drop but still not optimal (IMO).

I would suggest first experimenting that this circuit is the source of your "veiled" sound. You could pop out the TG chip and hardwire short across the TG. Does the sound quality improve? If yes there are alternatives. Can you measure distortion caused by that circuit block, that goes away with a hardwire bypass?

Back in the early 80s I switched multiple channels of a console with a single logic signal using TG. The trick that I used to deliver very low distortion was to keep the TG inside the negative feedback loop (visualize the NF resistor connected back to the input side of the TG). I was unable to measure distorion using the bench technology of the day (Sound Tech).

JR

PS: You don't need 3 relays, the TG has connected three internal switches in parallel for lower on resistance.

 

[Slime_Lord]

Thank you both for the incredibly thoughtful responses!

John, great point regarding jumping the 4053. I’ll try that later and do an a/b comparison. I’m not 100% certain this is adding to the darker sound I’m hearing and will at least get my brain to move past this section of the strip.

Newmarket, assuming the above test is the root of the issue, I’ll try increasing the VCC going to the chip. There’s no 5.75 rail on the power supply, I think (will confirm) one of the transistors is acting as a voltage regulator. Shouldn’t be too hard to find that and modify so that VCC is getting more juice. If the supply voltage is increased, is there anything further downstream that might need to be changed to accommodate for the higher supply voltage?

 

[JohnRoberts]

Newmarket, assuming the above test is the root of the issue, I’ll try increasing the VCC going to the chip. There’s no 5.75 rail on the power supply, I think (will confirm) one of the transistors is acting as a voltage regulator. Shouldn’t be too hard to find that and modify so that VCC is getting more juice. If the supply voltage is increased, is there anything further downstream that might need to be changed to accommodate for the higher supply voltage?

I am not Newmarket, but don't assume that the design engineer didn't measure the path performance and consider that.

The 5.75v rail is probably generated with a voltage divider. Keep in mind that if you change that you need to tweak the control logic voltage too.

JR

 

[Slime_Lord]

I am not Newmarket, but don't assume that the design engineer didn't measure the path performance and consider that.

The 5.75v rail is probably generated with a voltage divider. Keep in mind that if you change that you need to tweak the control logic voltage too.

JR

Good call out - I’ll start with removing and shorting the chip foot print and go from there. Don’t want to get too ahead of myself.

 

[ccaudle]

I have never designed in one of those devices myself, but at first look I find it surprising that there is capacitor coupling in and out, but the input is not biased to half supply voltage, it just has a pull down to 0V, and the device has no negative supply, just 0V and 5.6V. Unless that thick line running off page is actually a bias voltage and not 0V.
Maybe there is no voltage swing across the device used in front of a virtual ground input like that, but I would want to check at the lowest frequencies. That cap would have about 160 Ohm impedance at 10Hz. I know that is sub-sonic, but something like "puffs" from a kick drum mike or plosives from a singer could be down at that frequency before they get filtered out downstream.

One criticism of the design is passing a small AC signal nominally biased at 0V with the TG - power supply also grounded.

It looks like you could just add another 100K resistor from the R21/R22 junction up to the 5.6V supply and bias it as half voltage, but then that would probably pop when the output got connected to 0V for muting.

the input channels still sound veiled to me.

It is very difficult to design with your ears. Any way you can make some reasonably high resolution measurements to point you in the right direction?

 

[Matt Syson]

The 4053 is fine as it is although the paralleling of the 3 sections was found to be a mistake because it trebles the charge injection 'tick'. On some later issue desks AMEK simply disconected 2 of the 3 sections. Yes increasing the supply voltage does reduce the RdsOn resistance and as the 5053 has a level translator internally increasing the supply a little would not cause a problem although whether it is actually worth the effort is debatable. The design does not produce 'noise'. As John commented it can reduce the already small distortion further if a normally made (on) gate is incorporated in the negative feedback loop (also feeding the negative input of the op amp). Again the cutting and linking involved is probably not worthwhile.

 

[Matt Syson]

The Bullet was one of the more economical desks of the range and the muting circuit was deemed suitable to meet the intended specifications of the time. Many other instances of the use of 4053 actually had the VEE supply pin taken to around 0.7 Volts negative using a diode as a 'reference' voltage and pulled to wards the negative supply rail with a resistor. Biasing the audio would produce bad clicks. There was a note on some schematics of around the time that desk was built that only crtain manufacturer of 4053 should be used so there may well be slight detail differences in the internal circuitry of the chips. i never knew WHY that requirement was made as most other implementations (the other 15 or so desk designs) used at least some negative supply rail although not necessarily symmetrical.

 

[user 133392]

U4 has a grounded substrate. When muted, the 4053 inputs are being driven into reverse bias by signal current up maybe 1 to 1.2 mA peak prior to clipping. (U3 > R22.) This could setup metal migration and degrade the device.

When un-muted, is the error voltage at the inverting input of U5 really at virtual ground during transient conditions? It's not really ground - it's "virtual ground" only as long as U5 is operating in its' linear region. The negative-going transient error voltage could also reverse bias the 4053 outputs.

Not saying any of this is happening but its worth considering.

 

[Newmarket]

I didn't have time to mention this earlier but some time ago I queried this type of implementation with audio biased around 0V with Vee also tied to 0V (rather than a negative voltage).
This was in relation to a Soundcraft Spirit desk where this switching was associated with the phones/monitor output.
Consensus seemed to be that designers were kind of uneasy about it but nothing in the published data indicated a problem.

The "wire link" test option is the best idea imo.

 

[Slime_Lord]

Thanks everyone, this is all super helpful.

I removed the 4053, shorted the connections as suggested, and A/B’d between 2 channels getting the same signal from a Buchla Music Easel (it’s fun and harmonically rich, lol).

I turned off the eq and started with the preamp set all the way down, pad on. Without any gain or EQ, both channels sound identical and look the same on a spectral analyzer.

The differences were noticeable once I started turning up the preamp. The channel with the 4053 started clipping earlier than the channel without. The sound started to fold into itself earlier as well, (not sure how to describe this in technical terms, clarity disappeared). With the same amount of gain applied to the modified channel, the clip light was flickering but the detail of the signal was retained.

I ran both channels through a free THD analyzer and from a THD perspective, they measure about the same - 0.045% was the final read out after running a 300 hz tone for 2 minutes at -10 dB. I’m not entirely sure how to interpret this, so taking this with a grain of salt.

The modified channel has gained headroom, and sounds better! With that chip removed, I thiiiiink I might be able to sneak a 3PDT relay in, just need to sort out the best way of handling by that.

 

[JohnRoberts]

TG switching can be made arbitrarily clean, but perhaps not worth the trouble for a one off DIY...

A relay will git er dun.

JR

 

[gyraf]

..when comparing, make sure to use a known-good 4053 - these are known to fail in all sorts of "dirty" ways. I've replaced hundreds of them in our UA8000 over the years - when they work, they're as close to invisible as you could ever wish for - but when failing they have an annoying tendency to half-fail making stuff sometimes vaguely more gritty or muffled, sometimes outright distorted, setting in from some random threshold level. I ended up having sockets for most of the positions for them, then just replacing at suspicion - as the part cost close to nothing, and the fault-finding got annoyingly complex..

/Jakob E.

 

[jensenmann]

Years ago I was working on a TAC Matchless which had exactly the problem Jakob described in it´s muting circuit. In a conversation with Graham Langley about the problem he mentioned to use Signetics HEF4053B exclusively. They are supposed to have a different internal circuit resulting in less noise and CV feedthrough. That´s most likely the brand Matt mentioned above.
Later Amek desks used a little timeconstant at the controlport input to reduce CV feedthrough. Maybe that is installed in OP´s desk but missing on the posted schematic fraction. If not, that might be an option to improve the given circuit.

 

[Slime_Lord]

Years ago I was working on a TAC Matchless which had exactly the problem Jakob described in it´s muting circuit. In a conversation with Graham Langley about the problem he mentioned to use Signetics HEF4053B exclusively. They are supposed to have a different internal circuit resulting in less noise and CV feedthrough. That´s most likely the brand Matt mentioned above.
Later Amek desks used a little timeconstant at the controlport input to reduce CV feedthrough. Maybe that is installed in OP´s desk but missing on the posted schematic fraction. If not, that might be an option to improve the given circuit.

The 4053’s in the Bullet appear to be Phillips HEF4053B. Do you or anyone else reading this by chance have a schematic example that shows the the time constant at the control point input?

While looking for a relay to replace the 4053s, I came across this chip. I’m not familiar with switch debouncers but curious if leveraging one for switching would work.

https://logiswitch.com/shop/nobounce-ics/ls18-s/
Obviously, this isn’t pin for pin compatible and the 5.75V feeding the 4053 exceeds the max rating for the chip (5.5V). Would an adapter board feeding the ins and out of the chip and a diode to drop the voltage feeding the debouncer work in place of a relay?

 

[jensenmann]

Phillips = Signetics

The timeconstant is mainly formed by R103 and C69. If you post a larger version of your schematic we can maybe find out if that´s already installed in your desk.

 

[gyraf]

..iirc, Calrec added the switching time constant simply by driving the switch-input of the 4053 with a large resistor in series - 1M or 2M2 - taking advantage of the chip's input capacitance for slowing things down..

 

[Slime_Lord]

I’ll see what I can do about getting the larger version posted. I only have a physical copy that I purchased from AML after getting the console. I’ll snap a pic showing more of the area around it, weary of posting the whole thing since AML is providing that service.

Any thoughts on the switch debouncer mentioned earlier? Forget the diode, a 5V regulator would probably work better and switch common would be routed to the same ground as the fader.

 

[user 133392]

You don't need 3 relay poles on one SPDT contact.

You could also use a DG411-series switch if you wanted to build a small daughter board to adapt it to the 4053 pinout. The DG411-series has connections to the +/-15V rails and is logic-compatible on the control pins. It wouldn't be reversed-biased in operation like the grounded 4053.

With 4X low-Rds switches you could undoubtedly fashion a better mute with greater "offness."