Messages
I heavily recommend to use simply MATLAB for all numerical calculation you might need during the exam: if you are asked to supply a numerical result you will submit it as a matlab script anyway, so you can include the entire calculation that get's you the result in that script. Thus, you also document your calculation, and small errors can be identified as such and will not make you loose all points.
Additionally (as I have said in the lecture but which is not stated on the course page), you are allowed to bring an approved (!) hand calculator as listed here, but there is nothing a calculator can do that MATLAB cannot do. And if you calculate everything 'on the side' and just report the final value in the MATLAB script, if you made a small mistake along the way, your result will simply be wrong and you loose all points for that result.
So: a hand calculator is...
At the exam some questions will require hand drawings. You will need to draw on special paper sheets provided at the exam that you will need to hand in to the 'chief invigilator' in the exam room.
To fill inn the required header information on those sheets you will need a black or blue ink (ballpoint) pen. So make sure to bring one to the exam!!!
Read the information here before the exam:
- Definition of propagation delay:
Not everybody got that right. It is defined on pages 40/41 in the compendium: "one measures the propagation
delay tp as the time it takes from when the switch is flipped until the output reaches the switching threshold Vsw of the next stage. ... Often the simplifying assumption is made that
Vsw = Vdd/2"
Some groups used the 'transition time' that is actually not discussed in the compendium and only relates to a single signal, i.e. not a relation between an input and an output. Other books define it as the time it takes a digital signal to go from 10% of its transition to 90% of its transition. But again: this is not relevant here.
- Measuring input capacitance: so far, two groups were looking at the output signal of the inverter to determine input capacitance. As the name suggests: this is simply wrong. Instead, you need to observe voltage and current at the input. There is several methods you can use for this and you have found and used many of those quite correctly. For example introducing a series resistance and measure the time constant tau=RC of the input (!) signal. You can also source a voltage and measure the current and use I=C dV/dt or integrate over the current used from 0 to Vdd (which is the total charge) and use Q=C Delta V. Or vice versa, source a current an measure the voltage. One ingenious group also measured the propagation delay of the inverter with another idential inverter conected at its output and once again with the large 50pF capacitance at the output, correctly concluding that the ratio of the delays is proportional to the ratio of capacitances. Thus they could quite ni...
Please redownload the compendium here:
https://bibsys-k.alma.exlibrisgroup.com/leganto/public/47BIBSYS_UBO/citation/20985586000002204?auth=SAML
It's now version 1.4. Some limitation of the manually biased Cascode current mirror from the lecture is discussed in 1.4 that is not discussed in 1.3. Unfortunately the audio is missing from the second half of that lecture's podcast, so it's important to read the compendium section 10.3 in particular for the students that rely entirely on the podcast.
Please: if you are done for the day, remove your measurement setup from the work place and tidy everything up, so that others can use that place!!! Do not leave your stuff laying around. There is lockers outside the lab where your can store your PCBs. You can use your own padlock to lock these.
Please re-download MDO_GetWaveform.m: the offsett problem is solved in the new version. For recordings with hte old version, just state that there is an extra offset in your figure due to an earlier bug in the function. If you still have the original data, you can also correct that offset, but if not, just add a statement in the report.
MDO_GetWaveform might falsely add an offset to the y-axis data that you download from the scope. This may be the case if you move a waveform up and down on the scope-screen and its zero reference (indicated by a little arrow on the left of the screen) is not zero (i.e. in the middle of the y-axis). It looks like MDO_GetWaveform will unnecessarily subtract that deviation from the data it receives. I will verify this tomorrow and upload a corrected version. In the meantime, be aware of the issue and you may check what the vertical offset is on the scope (you can see it while you adapt it for a channel with the nob on the oscilloscope) and add that offset again from your data. So if the offset is -3V you just subtract 3V from your data ...
- Note that the phase margin is based on observing the LOOP gain!!! Here we only see the total system's (closed loop) gain and it is not really easy to identify the loop. So looking at phase shift and unity gain frequency of the total system does not give you clear clues on where the phase shift and unity gain frequency of the LOOP may be. Check figure 5.20 for an illustration that the two behave differently. It displays Aol and Acl. The Phase of Aol is the same as the lopop gains (if beta has no frequency dependency), and the frequency where Aol becomes 1/beta is the unity gain frequency of the loop.
- Show some example traces in task 2! A) to show that your bias settings produce a good result. B) to illustrate the difficulties at high frequencies, i.e. where the output amplitude becomes so small that the signal drowns more and more in noise. Do not blindly trust the high level results (here the Bode plots) derived automatically from recordings: verify the raw data and how you would manually derive the high level results for a few samples to check that the automatic derivation is not in trouble.
- In task 1 on step response: Plotting both input and output with their respective offsets (i.e. the traces far appart on the y-axis) makes it very hard to make out any detail. Show zoom ins both on the y-axis (to individual traces) and x-axis (to individual transitions)!
- Include the mandatory (!) image of your lab setup!!! It's an easy point to get and an unnecessary point to loose!
- Discuss your results and setups with others: easy to spot significant deviations (many times caused by simple mistakes that are easily fixed) that way. Do not work in isolation! Collaboration among all groups is encouraged (while you still have to get and understand your own results)!
- Describe your setup: easy in Cadence by showing your schematics, but even more important in the lab task, i.e. by drawing schematics and indicate the IC pin numbers. So if something seems off with your results, we can check if there is a problem with your setup.
- Read through the entire task text again afer a while to make sure you DID follow all instructions and have not overlooked something. Example: "For both FETs, connect the bulks to the source terminals!"
Please find a new edition of the compendium (edition 1.3) for download on the course webpage. It is a minor edition adding the deduction of small signal parameters for the CG stage in the appendix and a few more comments on the single transistor gain stages in the text.
Some general FB for lab 1 that might help improve the quality of the report for lab 2...
- Answer all explicit questions and try to answer the implicit ones as well. Example: "Are there any deviations that indicate that there are some effects that the EKV model does not model?"
- Follow all instructions! Example: "Enter the parameters into the EKV matlab function (You’ll find it under ’resources’ on the course page) and plot the result together with your measurement to verify your match." So plot the graphs together, as instructed!
- Give all necessary information to enable replication of your results. Some of you did not mention the value of V_DS for the I_D vs V_GS plots. And vice versa only specified "strong inversion" for I_D vs V_DS plots instead of also explicitly giving the value used for V_GS.
- Make figures readable! Use big enough labels (In Matlab, for example: set(gca,'fontsize', 14). Label the axes (use 'xlabe...
Yesterday night (somewhat surprising to me) the rh7login.ifi.uio.no cluster got retired. Please use nano.ifi.uio.no to run Cadence instead. The lab-guide will be updated accordingly shortly.
Note that the waveform generator cannot produce any sine waves with higher frequency than 60MHz! The script text in the lab description tries to go up to 100MHz.
If you download the latest version of MFG_SetSine.m, this is checked when you execute the function.
Please re-download the matlab script GainAndPhase.m. It's a new version. The old contained a copy-paste error that caused wrong results when input x and y had different offsets.
The automatic frequency sweep ran into trouble: you have to adapt the time base of the oscilloscope to the frequency to get a meaningfull recording to determine the phase. A new function [ok] = MDO_TimebaseScale(h,scale) is provided and inserted into the example loop given in task 2. This should save the issue.
Also note the update of a 'pause' command in the matlab script provided in the lab text!
The lab 2 PDF has just been replaced with an updated version. Please download again.
It's but a small change, mentioning the Matlab commands 'semilogx' and 'hold on/off', that can help to plot Bode plots.
Use the oportunities to talk to companies situated in Oslo that may be future employers! Check out the information here:
https://www.mn.uio.no/fysikk/studier/aktuelt/arrangementer/2025/industry-day-in-the-elite-study-program.html
Be aware that ther is a power outage planned for OJD tomorrow Thu 27-Feb 21:00 to 24:00. The building will be closed and you cannot work in the lab!!!
The power supply's 3rd channel can only supply 5V. Since you will most conveniently supply power to the IC (pin 14) from that channel, you cannot use 10V. So let's just use 5V instead.
P.S: Background explanation: 1) You need a seperate channel for pin 14 from the one providing the drain voltage, since you want to measure the current going to the drain accurately. 2) Pin 14 is also used for protecting any input against voltages higher than Vdd, so it needs to be the highest voltage on the chip too. So using 10V on the drain and only 5V on pin 14 would not work either...
All 2024 videos are now available without an extra step to approve access here:
https://uio.cloud.panopto.eu/Panopto/Pages/Sessions/List.aspx#folderID=%22925a0918-e350-49ad-ad68-b28100cae647%22
You can also navigate ther from the semester page clicking on 'Videomappe, lenke til videomappe i Panopto -> 2024_podcasts'
some of you discovered that you do not find the drive m: on the lab machines (while you DO find it on the 'Lab Net Workstation' as well as on Linux on ifi Workstation and rh7login.ifi.uio.no as /hom/<username> )
This can be rectified via a very circuitous route:
On the lab Windows computer:
1) Start the program 'Firmaportal' or 'Company Portal'
2) In 'Firmaportal' find 'UiO Legacy Home Drive' and install/run it.
3) Now you DO have access to your home drive, but seemingly not as drive m:. Instead you have to click on 'This PC' and you'll find it as 'Nettverkplasseringer -> Home Drive'
So now you should have a drive that you can access from all machines at UiO, Linux or Windows. If you work from home, you will not have access to it directly from your home machine though. For that you can open Omniss...
It appears that students from 2025 do not get automatic access to the podcasts from 2024. For now, as you try to access a 2024 podcast, I get a notification and have to approve access individually for each video and student. So there will be delay. Sorry for that!
In the meantime I am investigating how to rectify that and automatically grant access to all students and all 2024 videos. Let's hope this can be done!
Unfortunately the podcast of the 21-Jan lecture is without sound, or more precisely with just noise on the audio track. Please read the compendium on the subject of 'large signal models' if you have not been to the lecture, or have a look at last year's recording of approximately the same content (we got through a bit more stuff then) here: https://uio.cloud.panopto.eu/Panopto/Pages/Viewer.aspx?id=466b46cc-4c35-4c80-9b0f-b0f900da9b91