how to test relays for phono signals

[crin]

i'm looking for suggestions on how to test relays for use at phono (MM) levels: how to set up the test for repeatable results, what signals to use to detect measurable differences, what to look for etc.

background: i'm building a flat phono preamp with several of wayne's MM and MC boards, various input transformers, various loads etc. mechanical switches are not an option because they would either require carrying the phono signals across the chassis to the front panel or having the switches on the rear which i don't want because the unit is rack mounted.

i have decided on using two coil latching relays so the coil is not energized the whole time in the N/O position (relays are DPDT). i have assembled a selection of suitable candidates, all of them having 12 V coils.



from left to right: Kemet EA2-12, Panasonic TXS2-L2-12, Panasonic DS2E-SL2-12, Omron G6SK-2, Omron G6AK-234P, Teledyne RF120-12. in the back: Teledyne 712-TN-5. (except for the 712-TN, all are bi-stable latching. the two Teledyne relays normally cost an arm and a leg – the 712-TN goes for >$230 at mouser – but i got them NOS on ebay for 10 to 15 bucks each.)

the reason why i want to test them first is twofold: contact integrity is an issue at these low signal levels. most signal relays have a minimum load of 10 mV, everything lower is considered dry circuit switching and is treated only marginally in the data sheets. recommended contact materials for dry circuit switching would be Au-clad AgPd, but the only relay i could find using this material for the contacts is the Omron G6AK-234. Panasonic lists the TXS2-L2-12-1 having AgPd contacts, but nobody stocks this variant, only the "normal" one (Ag Au-clad).

the second reason is to find out whether any of these relays has an audible influence on the sound. i remember a demonstration at ETF when a participant inserted a black box between a CD player and the preamp. it contained several signal relays that could be switched into the signal path. it also had a random generator that allowed for true ABX. i found the differences between the various relays shockingly big; in his lecture, the builder of the black box explained that the materials from pin to contact, the armature, the sealing against environmental hazards, the arrangement of the contacts, the force of the armature closing the non-energized NC etc were all reasons for these audible differences.

now if this happens at line level, so my reasoning, then it must be even more obvious with signal levels of 5 mV and less.

but how to detect and measure any possible difference between the relays? using them in the actual circuit would be a PITA because not two have the same pin-out or layout (arrgh!)… and using them with DIL sockets just brings another variable into the equation (and makes the contact resistance higher).

what signals would make eventual differences obvious (and measurable), and what should i test for? distortion? spectrum? assuming that the contact integrity is not equal – what happens besides voltage drop? if the contacts are contaminated, will that change the spectrum of the harmonics? will it create distortion? what kind of distortion – plain vanilla, IM, slew, …? what signal to detect this distorion – sweep, square, two-tone, pink or white noise, dirac, …?

it seems obvious to use very low signal levels to mimic the reality. but what about impedances – source Z could either be a MM cartridge or the output of a step-up transformer fed from a MC cart; the load could be either 47k (MM) or 100k (with SUT) or 6k8 (with jensen JT-346 1:12).

does it make sense to daisy chain several of each relay to enhance any difference? (i have anyway planned to see if paralleling the contacts makes a difference – on one hand, the contact resistance should be halved, on the other hand, the current through each surface contact is halved, too. and i will check if there is a difference between the Set and Reset positions.)

i would like to include the Teledyne 712TN just to see how a $$$ RF relay handles low level audio signals. the Teledynes are the only hermetically sealed relays which is certainly an advantage in terms of contact "cleanliness". and the full metal jacket probably also helps WRT noise immunity. however, i don't fully understand how to control the relay. the data sheet shows a marginal illustration what to connect:




but what if i want to use a normal toggle instead of a logic circuit? what voltage can i apply to the base? the only data i found were the emitter-base breakdown voltage (BVebo, min. 6 Vdc) and collector-base breakdown voltage (BVcbo, 60 Vdc). the coil is 5 Vdc – can i apply the same supply to the base, too?

thanks for hints!

[JR.]

To mimic MM cart and typical termination I would feed the relay with roughly 1.5k source impedance, then terminate the return with 47k and 100pF.

It will be difficult to make bench measurements at 5mV signal level while that nominal level is for 1kHz, you will see lower levels at LF and higher at HF.

perhaps add a very high performance flat boost stage, say 40 dB to step up the test signal to measurable levels, while that boost stage may define your measurement floor.

In an ideal world you would make two boost stages to support null testing, or perhaps provide a differential input to the flat boost stage.

JR

[crin]

i have the boost stages (waynes flat MM preamps), and i could also use the inverse RIAA filter to create a RIAA emphasized signal.

but what kind of signals do i use, what kind of "error" could a relay possibly cause?

one thing is the outgassing of plastic and environmental pollution seeping in through the not-perfect sealing. this will cause a film on the contacts that reduces the contact. since this is dry circuit switching, the signal current is not strong enough to arc and burn the film away. this issue has been mentioned in literature by Panasonic, Omron and others.

• will contaminated contacts just cause a linear attenuation of the signal?

• or will / could the goo also introduce some sort of distortion?

• do the various armatures and contact materials change the very low level signal?

this is what i would like to find out. but to do so, i need to know what to look for – and how to find it (if it exists at all). so i am very open to advice on measurement set-up, signal, and signal interpretation.

BTW i managed to get the non-latching Teledyne relay working. it turned out that what Teledyne called base-emitter current was plain simple the base current…

[JR.]

what you want to look for is any nonlinearity, so distortion measurements should reveal any bad switch contacts.

Most of the bad switch contacts I have heard sound really bad, nothing very subtle.

JR

[mediatechnology]

In testing the Ultra Low Distortion Oscillator I found that "dirty" connections increased even-order distortion because the contact oxide became a low-level rectifier.

What I would try in order to measure it is connect a very high gain differential amp (THAT1510 60 dB gain) across the contacts and see what micro-potentials develop across the closed contacts under load.
It's a null test with gain.

I'm still trying to wrap my head around the Teledyne...

[crin]

i'm trying to figure your proposed null test with gain – would you place the inputs of the 1512 at C and NC and measure what comes out after 60 dB of amplification? but if there's no signal, how can it be distorted?

but the idea of a null test has a lot of appeal. i'm still trying to figure out if this idea would work (assuming the differential amp only amplifies the difference between the two inputs).



so if the relay contacts do anything with the signal, it should be amplified 1000 times, whereas everything else nulls itself? (this of course provided the THAT 1512 treats the + and - signals absolutely equal.

by connecting the second set of contacts either to NC or NO (or not at all), i could also check the RuT for differences between C/NC and C/NO, for crosstalk between poles, and for the effect of having one or two sets of contacts in the signal path.

the beauty of this brain fart would be that the signal generator doesn't have to be ultra low distortion since any distortion from the contacts would be added to only one input.

i first flirted with the idea of using a balanced output. but none of my generators and asym/sym converters put out an exactely identical signal with 0 and 180 degrees. the differences are in the microvolt region, but for a null test, that's not zero.

BTW your head must be really small if you manage to wrap it around the Teledyne… (i ended up with a 2k5 current limiting resistor between the 5 Vdc supply and base. since the coil has 50 ohms DCR and 5.8 Vdc max, Ic will be 100 mA max. the coil energizes reliably with 0.9 Vdc Ub, corresponding to Ic = 39.x mA. upping Ib to 1.2 mA, Ic keeps rising to around 78 mA. after that, any additional Ib is just dissipated, not changing Ic. so to add some slack, i ended up with 2k5 giving roughly 1.8 mA Ib.)

[mediatechnology]

You got the general idea.
There needs to be a low value load resistor on the right-hand side of the relay so that there's current flow through the contact.
The 1 K Rbias might not be enough.
With the concern being the low levels of MC switching pick 10-100Ω.

You could also ditch the coupling cap in Rg (Cg) and just let the DC gain multiply out.
(It will be Vos*Gain which will be under a volt.)
AC-couple the output with a small value C instead.

I think with only the contact resistance and with 60 dB of gain you'll still be looking in the mud for signal.
The CMRR at that gain should be about 100 dB.

[JR.]

Not that I expect there to be much there there, the phono preamp input termination is typically 47k (and 100pF), and at mV level signals any rectification from oxides or dirt may be more like on/off.

Good luck

JR

[mediatechnology]

I just did a quick calculation.

Panasonic quote 50 mΩ initial contact resistance.
20 µA of signal current will produce 1 µV across the contacts.
If the voltage gain of the differential amp is 1000 then the I*R drop of the switch will produce 1 mV output.
One can easily FFT 1 mV signals.

If the relay load resistor is 100Ω then you only need 2 mV of common mode voltage at the input to the diff amp to produce a contact current of 20 µA.
2 mV of CM is not a challenge for the diff amp.

The reality is that the contact resistance is probably a lot lower and the drive current will need to be a lot higher to see anything.
But at least you have headroom.

You probably could use the MC preamp as the diff amp but I doubt the CMRR will be anywhere near what a THAT1510 is.

[mediatechnology]

Not that I expect there to be much there there, the phono preamp input termination is typically 47k (and 100pF), and at mV level signals any rectification from oxides or dirt may be more like on/off.

Good luck

JR

As someone who has replaced thousands of ITT Schadow switches in MCI consoles I know the heartbreak of on/off rectification at line level signals.
And I've also replaced a lot of unsealed P&B and Omron relays in MCI and Otari recorders.
I can't recall however ever having to replace a sealed relay except an occasional gas-filled Aromat "ice cube" style and that was rare.
That includes low signal, tape head level, cue relays.

The concern here may be MC levels and currents.

I'm really curious to see if there's anything to measure.
The beauty of the null test is that you can also use music and listen to the residual.
Open that contact however and you'll lose your hearing.

It might be a good idea to add a diode clamp to the output of the diff amp....