Why are mixers noisy?

"Clifford Heath"

Two reasons come to mind:

1. Because they have heard other hams say it and it sounded clever to them.

1. To impress fools and make themselves sound clever.

YOU need to go ask them one simple question:

What does that even mean and does it matter in a typical receiver ?

Come back when you have a sensible answer.

I won't be holding my breath.

... Phil

Ahhh, I guess they meant the old Mixmaster !!

guess... don't mixers automatically split the tones, dropping them 6dB? Noise stays put, thus effectively making it 'look' like noise was added.

I used to know how to do that calculation, now I just simulate, but I vaguely recall, from early Garmin GPS chip days, a "passive" mixer (no gain, diode(s) driven by an LO) has about an 11dB NF. ...Jim Thompson

noise cals in a rf system are based on nf ---

depending on the system design, the mixer tends to be one of the first elements in the gain lineup. and as jim said, if there's a lossey element it adds to the nf.

that's why some systems offer low noise pre-amps for certain apps.

it's all about trade off's.

AFAICT there are two reasons: first, the conversion loss of the mixer, which is usually about 6 dB, and second, the Johnson noise of the switches. (An ideal PN diode theoretically has exactly half the Johnson noise of its small-signal resistance, but in practice you're usually hitting it hard enough that it looks like a 300 kelvin resistor.)

If you saturate the switches, AM noise on the LO doesn't contribute much.

Cheers

Phil Hobbs

The reason mixers are noisy is the cheap brush type motor in them, the gears, and the large ice cubes people drop in there for their margarita.

Oh wait ...

It's Monday morning, Joerg--you can put the cocktail shaker down now. ;)

Cheers

Phil Hobbs

A silly question. First is conversion loss the same as insertion loss? And if so why does insertion loss add to the noise? I'd figure that if I attenuate the signal by 3dB I'd also attenuate the noise by the same amount. I was thinking that mixers are noisy because they add signals from both the carrier and the image frequency... (I may be using those words wrong. The output sees the noise from both sides of the LO frequency.)

George (not an RF guy) Herold

In a diode mixer, might there be a wideband shot noise component generated by the LO current sloshing through the diodes?

Mixers are noisy because of all the loud drunks that attend them.

I think you may have misunderstood what they said. I believe they were referring to the mixer Noise Figure, not that it was contributing any additional noise.

In an HF receiver, it's very common for the 1st mixer to be the first device on the RF path, just after the front end band pass or low pass filter. This makes it the key part in determining the system noise figure and determining the receiver sensitivity. An RF stage in the front end of an HF receiver is a waste of effort because the high atmospheric (lightning) noise levels will dominate the overall sensitivity figure by covering up any weak signals. The effect decreases with frequency, so that by the time you get to about 30MHz, the atmospheric noise is sufficiently low that an RF amplifier will be useful. Ignoring atmospherics, it's mostly the mixer that determines the major performance numbers (sensitivity, 3rd order intercept, dynamic range, spurious responses, etc) on the receiver data sheet.

Phil Hobbs covered the double balanced mixer, which is excellent for dynamic range, but has a 6dB conversion loss, which translated directly into a sensitivity loss. Including internal losses, a good mixer has a conversion loss and NF (noise figure) of about 6.5dB, which may be the "noise" that the hams were discussing. It's not really noise, but noise figure.

There are other types of mixers such a D flip flop, active FET mixers, DGMOS FET mixers, image reject mixers, single diode mixers, rusty bolt, etc. Any non-linear device can act as a mixer. Active devices can add gain, but also add noise, just like an amplifier stage. Of course, some are better (or worse) than others depending on what you are trying to accomplish.

Understanding Mixers - Terms Defined, and Measuring Performance

Mixer Noise Figure (with some controversy):

Basics of RF Mixers in Radio Receivers / Mixer Tutorial / Frequency Conversion (16:31 min)

I normally use insertion loss to refer to situations where the input and output are at the same frequency, but I don't know if that restriction is standard.

And if so why does insertion loss add to the noise? I'd figure

Right. Unless you use an image-reject mixer, the signal gets spread out between the IF and image frequencies, which costs you 3 dB in signal. In general there's noise at the image frequency as well, which gets downconverted to the IF, so that the noise doesn't go down by 3 dB.

And then some of the signal gets shifted by harmonics of the LO frequency, some goes to I**2 R losses in the active devices and copper, and there are also extra switching losses due to the switches not being fully on at all times.

Cheers

Phil "former RF guy" Hobbs

Yes, that's the half Johnson I'm talking about. It's sort of a cute derivation--at zero bias, the shot noise of the forward and reverse diffusion currents add up to the full Johnson noise of the zero bias resistance (as they must), but when you apply bias, the reverse current rapidly goes away, leaving just the shot noise of the forward current, which is half of what you'd calculate using the Johnson noise formula for the differential resistance.

I tried using that as a 150 kelvin noise source for noise figure measurements, but it wasn't accurate enough with the diodes I had on hand.

Cheers

Phil Hobbs

Yup.

There are some active mixers that mix and amplify at the same time that can have lower noise figures. But that comes at the cost of circuit complexity, a need for careful treatment to get it to work right, exotic components, or loss of dynamic range.

A plain ol' diode ring mixer that magically contributed no noise or losses beyond the switching losses would have a 6dB noise figure.

This implies if you use a Hilbert transformer and produced SSB, there is no mixer noise. Something to meditate on when your navel is no longer a sufficient target.

It is interesting how the better radios are the ones with more mixers, i.e. triple conversion is praised while double conversion is lambasted for too many images. Of course there is more to the foo than just mixer noise.

Going from RF to baseband still requires mixers, and their noise, if you use the phasing method.

The triple vs. double vs. single (vs. direct conversion, for that matter) depends on what radio service you're designing for. And yes, mixer noise is not the most important consideration when you're working on such a system.

re

ise

add

ome

Hi guys.. sorry for the repeat question. (I must have a pencil stuck from ear to ear, 'cause my brain ain't workin'.) So let's stick with the 'plain' diode ring mixer. It's got 6dB (voltage goes down by 2) of conversion loss. No problem. But why is this a 6dB noise figure? Doesn't the noise at the input suffer the same conversion loss? Maybe I don't know what the noise figure is. I thought it was the S/N ratio from input to ouput. If the ratio stays the same isn't that

George H.

ttdesign.com- Hide quoted text -

How do you attenuate the noise floor? :)