From: James Meyer
Subject: Re: Undersampling and its complement?
X-Newsreader: Forte Agent 1.92/32.572
Date: Sun, 03 Nov 2002 18:01:26 GMT
NNTP-Posting-Date: Sun, 03 Nov 2002 13:01:26 EST
On 3 Nov 2002 05:08:58 -0800, Winfield Hill wroth:
>Kevin Aylward wrote...
>> The main issue is that the reflected signal contains mostly fundamental,
>> the shifted frequency component being maybe 60db - 100db down. The
>> second issue is that looking for 100hz-10kz shifts in 10Mhz is a tad
>> tricky as well.
>> The way this has been done in the past in standard commercial medical
>> imaging equipment for colour flow dopplar, is to use a very high quality
>> oscillator and mix the signal down to backband. So, this is what I
>> suggest you do. The oscillator, is of course, the same one you use for
>> the transmit signal. If you need any over specific pointers, let me
> Kevin, how does one deal with all the varying phase shifts?
> Is that quadrature mixing with some kind of combining circuit
> to capture frequency-shift info only?
> - Win
Win and Kevin,
The cheap commercial Doppler I started with is a CW type. The specs I
found said that 40 to 60 dB is about the ratio of shifted to straight reflected
return amplitudes. The little instrument simply uses a diode to mix the return
signal with its self to recover the Doppler shift. It mixes down to baseband.
The oscillator is a simple LC Colpitts. For what it was sold to do, it works
Of course, a single diode is an envelop detector as well. They simply
throw away the DC recovered with a capacitor and do minimal filtering, pass the
Doppler through a loudspeaker, before it's presented to the operator. The
recovered DC makes the system extremely noisy and subject to motion artifacts.
And multiplexing the single channel to six different transducer pairs is never
going to happen.
I did get some improvement by substituting a diode ring balanced
demodulator for the single diode and using a sample of the oscillator output for
the LO input, but it still wasn't good enough.
That's why I want to roll-my-own system. I suspect things will get more
complicated when I go on to pulsed Doppler, but I also suspect that's the only
way to get the quality of signal needed here. I need a heartbeat to heartbeat
time course of blood velocity in a radial artery. It's not like I'm looking at
vessels deep in the brain or fetal stuff.