Ive lost the thread which this topic was originally on (a while back) but now having the ability to upload my contribution (from local library) I'll add it as a new one:
Back in the mid 1980s the QL rumour mill was that the ZX8301 video chip was meant to run at 16MHz, with the 68008 CPU running at half that rate at full 8MHz. But since the chip proved unreliable at that speed it was reduced to 15MHz. This meant that the ZX8301 final mask had to be redesigned to tap off HSYNCH and VSYNCH at the correct frequencies from the clock divider chain (15625Hz and 50/60Hz respectively). Apart from the cost incurred the worst effect was that the pixel rate was now too slow to show 512 pixels within the visible part of the scan line. The relatively recent magazine articles (and interviews) with Karlin, the chip designer, that I have read from the QL DVD, never mentioned this. Although it did mention an expensive 'minor' revision.
Some math : Each horizontal scan line is 64 uS (microseconds) long, of which it is generally accepted that only 48uS is visible. To produce 512 pixels within this area requires a pixel rate of 48e-6/512, the reciprocal gives frequency of 10.667MHz. Assuming the QL hardware performs three clock edges per pixel this requires 10.667e6*3/2 = 16Mhz time base. Working backwards from the 15MHz the ZX8301 actually runs at, it will take 51.2uS to churn out 512 pixels. Or to put it another way, running at 15MHz you will only fit 480 pixels in 48uS, or more meaningfully, 80 QL characters out of 85 at CSIZE 0,0.
Fortunately my monitors (Philips 8CM852 and V7001) can be adjusted externally to show full QL display. I also have an old low-res Tatung TM101 monitor I use with a Memotech MTX, but this also had a similar problem. It uses a TMS9929A video chip running at 10.7MHz but it still lost 1 of the possible 42 characters at the left hand side. I was able to tweak the presets inside the Tatung to get it show them all (also QL mode 8, but 0.63mm pitch is too coarse for mode 4 viewing), Although I should mention you have to be careful, too much adjustment and the HT squealed a bit, it was right on the border line of its design performance.
On the subject of clock frequencies : It was a pity the fixed microdrive data bit rates where different between the ZX Spectrum (46.3uS/byte) and QL (40uS). In one fell swoop it not only made the (relatively reliable) ZX cartridges unreadable on a QL but also made QL microdrives less reliable. It's not as if it was absolutely necessary to push them further (IMHO), 85K for a 128kb QL would have been OK. Although I have red-spotted microdrives (apparently units set-up competently) I rarely use them, and then when I do I usually get trouble, unlike ZX drives on the Spectrum.
Ironic, where one clock frequency was inadequate, another was overdone.
(Source : ZX=46.3us - Ian Logan's Microdrive book page 30, data block of 540 bytes in 25mS. QL=40uS - Minerva source file M_md_read_asm. Although I'd like to verify actual RAW1/RAW2 bit rates with storage scope. ZX IF1 ULA 8MHz/?, QL ZX8302 7.5MHz/?).